Brain endothelial cells orchestrate a neuroprotective antiviral state in the central nervous system in response to peripheral viral pattern sensing

Summary

West Nile virus (WNV) and other neurotrophic arboviruses are human pathogens that can cause life-threatening encephalitis. To uncover determinants of severe disease progression, we employed a murine infection model of WNV and experimentally uncoupled peripheral pathogen sensing from active viral replication. Peripheral sensing of a viral dsRNA mimic in the footpad led to robust induction of antiviral type I interferon (IFN-I) stimulated genes and establishment of an antiviral state within the brain. Systematic approaches combining cytokine profiling, single-nuclei transcriptomics, immune-modulation and genetic perturbations revealed critical roles for systemic IFN-I and IFN-I receptor signaling in brain microvascular endothelial cells. This inter-organ antiviral crosstalk protected against severe encephalitis caused by a range of neurotropic viruses across Orthoflaviviridae, Togaviridae, and Orthoherpesviridae. Thus, these data unravel a cross-tissue antiviral signaling network that counteracts lethal encephalitis, pointing to therapeutic avenues for encephalitic disease caused by emerging viral pathogens.

eTOC Blurb

West Nile virus and other neurotrophic arboviruses can cause life-threatening encephalitis. Lewy et al. identify a cross-organ immune circuit where viral pattern sensing in the skin potently counteracts viral encephalitis. Mechanistically, establishment of this neuroprotective state depends on systemic type I interferon which is sensed and relayed into the brain by endothelial cells to mitigate encephalitis.

Graphical Abstract

Introduction

The central nervous system (CNS) is a vital organ in a wide range of metazoans. The functional units of the CNS are highly specialized neurons that form complex networks which regulate all vital functions of the body, including thermoregulation, digestion, proprioception, and mobility^1–4. The CNS also integrates the organismal immune status⁵. It coordinates locally restricted or organism-wide inflammatory responses, including sickness behavior and cachexia^6–8. The brain monitors inflammatory responses throughout the organism via peripheral sensors including the vagus^9,10, trigeminal¹¹, and glossopharyngeal nerves¹². In addition, afferent spinal fibers integrate and relay circulating cytokine signals to the CNS and are critical in detecting and restricting pathogenic microbes⁹. Within peripheral tissues, specialized nerves, such as TRPV1⁺ sensory neurons, mediate inflammatory responses towards pathogens in the skin^13–16 and lung¹² through release of neuropeptides, such as calcitonin gene-related peptide (CGRP), which modulate immune cell behavior^17,18.

Inflammatory responses are critical to fend off pathogenic disease, but can also result in detrimental immune-mediated tissue damage¹⁹. Excessive neuronal cell death elicited by neuroinflammation or encephalitis are particularly deleterious due to the limited self-renewal capacity of neurons^20,21. For this reason, the CNS is quarantined behind the highly selective blood brain barrier (BBB). Each neurovascular unit of the BBB is composed of brain microvascular endothelial cells (BMECs), pericytes, and astrocyte endfoot processes²². Together, they restrict passage of blood-borne entities, such as cytokines, antibodies, and immune cells^23,24. Without permitting diffusion, the BBB is critically involved in translating circulating cytokine signals into the brain parenchyma and regulating immune cell transcytosis^25,26. BMECs are in direct contact with the bloodstream and therefore form the initial layer of selectivity. However, some pathogens, including neurotropic viruses, can bypass the BBB and enter the CNS. Viral invasion of this immune-privileged tissue often results in detrimental neuroinflammation, leading to lethality.

Orthoflaviviruses are positive sense single stranded RNA ((+) ssRNA) viruses that are transmitted by arthropod vectors. Its neurotropic members, including West Nile virus (WNV), tick-borne encephalitis virus (TBEV), and Japanese encephalitis virus (JEV), are responsible for more than 100,000 annual encephalitic cases^27–30, with case fatality rates ranging from 5–10% worldwide³¹. Rapidly changing factors including climate, urbanization, and population growth are increasing arthropod host range and magnifying the future risk of such arboviral infections and outbreaks³².

WNV is an arboviral pathogen that is transmitted by infected mosquitoes, which deliver infectious virus via saliva into the host’s skin. The virus initially replicates in keratinocytes³³ and antigen presenting immune cells, like dermal dendritic cells and Langerhans cells³⁴, before disseminating into draining lymph nodes and the spleen. At this stage, WNV causes a self-limiting febrile illness in approximately 20% of infected individuals³⁵. Yet, in 5% of those patients, WNV enters the CNS and efficiently replicates in neurons³⁶, leading to severe neuroinflammation and lethality²⁹ with long-term post-infectious sequelae prevalent in survivors³⁷. The encephalitic determinants of WNV are unclear. WNV envelope protein glycosylation status is linked to neuroinvasion ability in mouse models³⁸, but fails to explain inter-patient variability and implies host immunity as major determinant of disease severity. In fact, viral clearance and survival of life-threatening WNV encephalitis critically depend on CD8⁺ cytotoxic T cell responses^39,40 and autoantibodies neutralizing type I interferon (IFN-I), and in particular the subtype IFN-α, are a major contributor to severe encephalitis in patients⁴¹.

The host senses WNV in the periphery via the pattern recognition receptors (PRRs) RIG-I, MDA5, TLR7⁴² and TLR3⁴³ and downstream signaling and activation of MAVS⁴⁴, IRF3⁴⁵, and IRF7⁴⁶ leads to expression, production, and secretion of the antiviral cytokine IFN-I. IFN-I and canonical IFN-I receptor (IFNAR1) signaling induces the expression of numerous interferon stimulated genes (ISGs)⁴⁷, which efficiently restrict WNV replication in the periphery and the CNS⁴⁸. Absence of IFNAR1 signaling expands peripheral tropism, viral titers, encephalitis, and mortality^48,49. Of note, the role of the dsRNA sensor TLR3 in restricting WNV is less well understood, as studies suggest both pro-survival and pro-viral functions^42,50,51. Previous work investigated and characterized the role of innate immunity and IFN-I signaling on disease outcome, both in physiologically relevant ex vivo systems to study CNS responses in isolation, as well as after the establishment of WNV-mediated encephalitis in vivo^52–54.

Here, we examined whether innate immune responses following pathogen recognition within a distal organ, like the skin after a mosquito bite, impact viral encephalitis. We found that early detection of viral pathogen-associated molecular patterns (PAMPs) in the footpad triggers a systemic IFN-I response that activates the BMECs of the BBB and establishes an antiviral state within the CNS, limiting viral encephalitis and subsequent hyperinflammatory dysregulation, thereby improving disease outcomes.

Results

Footpad WNV challenge induces CNS antiviral immunity prior to neuroinvasion.

We first asked whether peripheral WNV sensing, prior to neuroinvasion, triggers an immune response within the CNS. To model infection via an arthropod vector, we characterized footpad WNV infection in wildtype 4-week-old C57BL/6J mice. Animals started to lose weight at 6 days post infection (DPI), correlating with the expected time of neuroinvasion^39,55 (Figure 1A) and had a mean survival time (MST) of 10 days with a 25% overall survival rate (Figure 1B). We harvested brain tissue at 1 DPI, a timepoint several days before neuroinvasion, and quantified transcript levels of several antiviral ISGs and a panel of proinflammatory NF-κB target genes. As expected, virus was not detectable in the brain at this early time point (Figure 1C and Figure S1A). Yet, the CNS mounted a robust and distinct antiviral immune response, characterized by induction of antiviral ISGs, but not NF-κB response genes (Figure 1D, Figures S1B and S1C). Introducing the virus into an immunologically naïve brain through intracranial injection, even at a 100x lower dose, elicted more severe disease (MST = 6 days, 0% survival rate) (Figure 1E, 1F). Comparing these two routes, we found increased survival odds (Figure 1G) and median survival time (Figure 1H) when the virus invaded the CNS after peripheral infection. These observations led us to hypothesize that PAMP sensing in the periphery might induce a protective antiviral state in the CNS prior to neuroinvasion.

Figure 1. Footpad WNV challenge induces CNS antiviral immunity prior to neuroinvasion.

(A) Normalized weight and (B) survival of 4-week old C57BL/6 mice following footpad-administration of 1,000 PFU WNV. Each line in (A) represents an individual animal. Data shown are pooled from 2 independent experiments. n = 8. (C) Transcript levels of WNV NS3 in brain tissue as detected by qPCR. Samples with inconsistent amplification or Ct values > 35 were considered non-detected and are listed as n.d. Data shown are means +/− s.e.m. and pooled from 2–3 independent experiments. n = 3–5. (D) CNS expression levels of six canonical ISGs or NF-κB target genes at 1 DPI as determined by qPCR. Values were normalized to the housekeeping gene Gapdh and uninfected controls. Data shown are experiment means from 3 independent experiments. n = 3–6. (E) Normalized weight and (F) survival of mice intracranially challenged with 10 PFU WNV. Each line represents an individual animal. Data shown are pooled from 2 independent experiments. n= 8. (G) Comparison of survival likelihood and (H) survival duration for footpad and intracranially challenged animals. Data shown are means +/− s.e.m. Signficance of (G) survival percentage determined by Mann-Whitney test and significance of (H) survival time determined by one-tailed Fisher’s exact test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Distal administration of poly(I:C) uniquely ameliorates WNV-induced encephalitis

To explore the effect of peripheral immune activation on neurovirulent infections, we experimentally uncoupled PAMP sensing from pathogen replication. We delivered PAMPs via footpad injection and then modeled neuroinvasion by intracranially injecting a lethal WNV dose (Figure 2A). To investigate whether the CNS could discriminate between pathogenic kingdoms, we tested a broad range of both bacterial and viral PAMPs (Figure 2B). Pam3CSK4 (TLR1/2 agonist), FSL1 (TLR2/6 agonist), and ODN1826 (TLR9 agonist), did not prolong animal survival, (ΔMST = 0 days) (Figures 2C to 2E) or affect disease progression (Figures S2A to S2C). LPS (TLR4 agonist) (ΔMST = +1.5 days), FLA (TLR5 agonist) (ΔMST = +1 day), and R848 (TLR7/8 agonist) (ΔMST = +1 day) led to a slight survival increase (Figures 2F to 2H), but only resulted in a minimal difference in disease progression (Figures S2D to S2F).

Figure 2. Distal administration of poly(I:C) uniquely ameliorates encephalitis caused by WNV.

(A) Experimental schematic and (B) table detailing PAMP pretreatment with Pam3CSK4, FSL1, LPS, FLA, poly(I:C), R848 and ODN1826 and infection strategy. (C-I) Survival of animals pretreated with indicated PAMPs or vehicle control. Data shown are pooled from 2 independent experiments. n = 6–8. (J) Delta Mean Survival Time (ΔMST) and (K) percent survival of PAMP-pretreated animals relative to vehicle-only controls. Each bar represents combined data from 2–3 independent experiments. n = 6–8. (L) Transcript levels of WNV NS3 in brain tissue of naïve mice or 2 DPI intracranial WNV challenge following PBS or poly(I:C) footpad-treatment. n = 3–5. (M) Brain infectious viral load at indicated time points and treatment. Undetectable samples were assigned 0.5 times the limit of detection (200 PFU; also indicated by dashed line). Data were pooled from 3 independent experiments with n = 4–6. (N) Quantification of WNV NS1 antigen abundance in whole intact cleared brains as determined by immunohistochemistry and representative images from (O) PBS or (P) footpad-poly(I:C) pretreated animals 4 DPI intracranial WNV challenge, indicated by a square symbol in (N). Data shown are from one of two representative experiments. n = 4 mice. Images are maximum projections through the z axis of whole organs. Scale bars equal 1000μm. (Q-S) Transcript levels of (Q) Ifit1, (R) Rsad2, (S) Mx1 in brain tissue of naïve mice or 2 DPI intracranial WNV challenge following PBS or poly(I:C) footpad-treatment. n = 3–5. (T) Heatmap of quantified cytokines in brain lysate from poly(I:C) or PBS pretreated animals following intracranial WNV infection at 2 DPI and naïve animals. n = 3 (naïve) and n = 5 (treated). Data shown in (L-N, Q-S) are presented as mean +/− s.e.m. and each dot corresponds to an individual animal. Statistical testing for survival (C-I) was performed using Log-Rank (Mantel-Cox) test by comparing experimental conditions to vehicle control. Statistical testing for plaque assay (M) was performed using Mann-Whitney Test with Holm-Šídák correction for multiple comparisons. Statistical testing for IHC quantification and brain cytokines (N and T) was determined using unpaired Student’s t test with Holm-Šídák correction for multiple comparisons. Statistical testing for qPCR analyses (L, Q-S) was performed using an ordinary one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

In contrast to other tested PAMPs, poly(I:C) (TLR3 agonist) conferred a robust survival advantage (Figure 2I) and reduced body weight loss upon intracranial WNV challenge (Figure S2G). Overall, poly(I:C) pretreatment resulted in a ΔMST = +3 days and an overall 41% survival rate (Figures 2J and 2K). Likewise, we observed a comparable survival advantage when WNV was delivered via the contralateral footpad, a model which better mimics the natural arthropod-borne route of infection. (Figure S2H). Of note, all PAMPs apart from FLA and ODN1826 lead to a transient weight loss after 24 hours (Figures S2A to S2G), in line with previous reports on sickness behavior following poly(I:C)⁵⁶ or LPS⁵⁷ administration.

Footpad poly(I:C) treatment reduced viral RNA (Figure 2L) and infectious viral titers in brain tissue at 2 DPI and 4 DPI (Figure 2M). Whole mount immunofluorescent imaging analyses revealed an overall reduction of viral antigen signal across distinct brain regions at 4 DPI (Figure S2I), culminating in a significant reduction of WNV NS1 in whole brains (Figures 2N to 2P, Figures S2J and S2K and Videos S1 and S2). Reduced viremia correlated with attenuated ISG transcript levels (Figures 2Q to 2S) and changes of pro-inflammatory cytokines, such as IL-6, CCL2 and CXCL10 in brain lysates (Figure 2T), indicative of reduced hyperinflammatory dysregulation, a known determinant of severe encephalitis^58,59.

These data show that peripheral dsRNA sensing induces an antiviral state in the CNS that ameliorates viral encephalitis and hyperinflammatory dysregulation.

Footpad poly(I:C) induces antiviral transcripts in CNS independent of immune cell influx.

While we only found poly(I:C) to robustly protect from severe encephalitis, footpad administration of all tested PAMPs induced an innate immune response within the CNS (Figure S3A). To define the protective transcriptomic changes, we performed bulk RNA-sequencing (RNA-seq) analysis on brain tissue following footpad administration of the poly(I:C) (protective), LPS (semi-protective) or ODN1826 (non-protective) (Table S1). Transcriptomic profiles clustered by treatment, with poly(I:C) being the most distinct condition (Figure S3B). The majority of differentially expressed genes (DEGs; fold change > 2, mean FPKM > 1 and adjusted p value < 0.05) were shared between poly(I:C) and at least one other PAMP (poly(I:C) – LPS: 91; all PAMPs: 63; poly(I:C) – ODN1826: 4) (Figure S3C). Hierarchical clustering of DEGs across all conditions (302 genes) identified both PAMP-specific (cluster 3 – poly(I:C), cluster 2 – LPS) and non-specific gene sets (clusters 1, 4, 5, 6) (Figure 3A and Table S1). Gene ontology (GO) analyses revealed a strong enrichment for antiviral and interferon-induced processes for cluster 3 (Figure 3B), whereas clusters 1 and 2 were enriched for phagocytosis and chemotactic processes, respectively (Figures S3D and S3E and Table S1). Of note, local ISG induction in the skin and popliteal lymph node (pLN) as well as in the CNS were comparable following footpad poly(I:C) or footpad WNV infection (Figures S3F to S3H).

Figure 3. Distal poly(I:C) indirectly activates antiviral immunity in the CNS.

(A) Hierarchical clustering of DEGs in brain tissue at 16 hours following footpad PAMP treatment. n = 4. (B) Top 15 GO:BP terms enriched in gene expression cluster 3 (A). (C) NK cells and inflammatory monocytes in brain tissue at 24 hours post footpad-poly(I:C). Data shown are presented as mean +/− s.e.m. n = 5. (D, E) Survival of animals depleted of (D) NK cells (antibody-mediated, clone PK136) or (E) inflammatory monocytes (Ccr2-DTR mice) prior to poly(I:C) footpad administration and subsequent intracranial WNV challenge at 24 hours post poly(I:C) treatment. n = 7–9. (F) Expression levels of six canonical ISGs in the CNS following footpad administration of poly(I:C) as determined by qPCR. Values were normalized to the housekeeping gene Eef1a1 and uninfected controls. n = 5. (G-H) Survival of animals footpad-treated with poly(I:C) (G) prior or (H) after intracranial WNV challenge. Data are representative of 2–3 independent experiments. n = 6. Statistical testing for RNA-seq data (A) performed using the DESeq2 package with unsupervised hierarchical clustering in pheatmap. Statistical significance for GO Enrichment (B) was determined using EnrichGO. Statistical testing of immune cell profiling (C) was performed using unpaired student’s t test with Holm-Šídák correction for multiple comparisons. Statistical testing for survival (D, E, G, H) was determined using Log-Rank (Mantel-Cox) test by comparing experimental conditions to vehicle control. *p < 0.05, **p < 0.01, ***p < 0.001.

Investigating the cellular basis of peripheral poly(I:C)-mediated neuroprotection we found a striking increase of circulating NK cells (NK1.1⁺ of CD45⁺) and slight decrease of B cells (CD19⁺ of CD45⁺), while relative abundance of monocytes (CD64⁺CD11b⁺ of CD45⁺), inflammatory monocytes (Ly6C⁺CD64⁺CD11b⁺ of CD45⁺), neutrophils (Ly6G⁺CD11b⁺ of CD45⁺) and T cells (CD3⁺ of CD45⁺) were not affected (Figure S3I). In matching brain tissue, total NK cells (NK1.1⁺CD45⁺) and inflammatory monocytes (Ly6C⁺CD11b⁺P2RY12⁻CD45⁺) were significantly increased (Figure 3C), whereas monocytes (CD11b⁺P2RY12⁻CD45⁺), neutrophils (Ly6G⁺CD11b⁺CD45⁺), B cells (CD19⁺CD45⁺) and microglia (P2RY12⁺CD11b⁺CD45^int) remained unchanged (Figure S3J). We reasoned that CNS-infiltrating NK cells and inflammatory monocytes may contribute to poly(I:C)-mediated neuroprotection. To test this, we employed antibody-mediated depletion⁶⁰ and the Ccr2-DTR mouse model⁶¹ to efficiently deplete NK cells (Figure S3K) or inflammatory monocytes (Figure S3L) prior to poly(I:C) treatment, but found no impact on survival following subsequent intracranial challenge with WNV (Figures 3D and 3E).

Dosing experiments for poly(I:C) and LPS implied a correlation between brain ISG transcript levels and overall survival upon WNV challenge (Figures S3M to S3P). Poly(I:C)-induced ISGs rapidly peaked at 24 hours (Figures S3Q and S3R) before quickly starting to decline at 48 hours and returning to basal levels around 168 hours (7 days) (Figure 3F). In line, poly(I:C) administration at 24 hours, rather than 48 hours prior to WNV challenge was more potent in ameliorating disease (Figure 3G). Furthermore, poly(I:C) had a reduced impact when administered concurrently with WNV and no effect when given afterward, underscoring the necessity of initiating protective ISG responses in the brain prior to neuroinvasion (Figure 3H). Continous daily poly(I:C) treatment up to 3 days post WNV challenge conveyed only a minor beneficial effect on survival compared to a single dose administered 24 hours prior to infection (Figure S3S). These data suggest that footpad-delivered poly(I:C) elicits encephalitic protection by specifically inducing an antiviral ISG response in the CNS, akin to antiviral gene induction following footpad WNV challenge (Figure 1D).

Footpad-administered RNA mainly remains locally associated with innate immune cells.

To gauge organismal distribution of footpad-administered poly(I:C) (size: 900–5,200 kDa), we employed traceable compounds as a proxy. FITC-Dextran (size: 150 kDa) treatment resulted in significantly increased FITC signal in serum of intravenously treated animals, but not after footpad administration (Figure S4A). Complementary, we footpad-administered an in vitro generated pregenomic RNA (pgRNA) of hepatitis B virus (HBV; size: approx. 1,000 kDa)⁶². As determined by RT-qPCR, the majority of HBV pgRNA is retained in the footpad (90.5%) or the popliteal lymph node (pLN, 9.2%), and only a minimal fraction enters the circulation (0.14%), liver (0.18%) or brain (0.0013%) (Figure S4B). Finally, we performed immunohistochemistry using an antibody recognizing dsRNA⁶³, to investigate the distribution of poly(I:C) in the skin of the footpad (Figure S4C). At 24 hours post administration, poly(I:C) was readily detectible in the skin and predominantly associated with macrophages (F4/80⁺) and dendritic cells (CD11c⁺) (Figures S4D and S4E). Poly(I:C) was not detectable in matching brain tissue samples at 1, 2, 8 or 24 hours post footpad-poly(I:C) treatment (Figure S4F). Together, these data corroborate our observations using FITC-Dextran or HBV pgRNA. Importantly, there was a clear discernible difference in dsRNA signal between brain tissue of mice during acute WNV encephalitis and naïve control tissue, confirming sensitivity of this readout (Figure S4G). These data suggest that systemic distribution of footpad-administered poly(I:C) is limited.

Distal poly(I:C)-mediated protection is conferred through peripheral IFN-I signaling.

We then hypothesized that soluble factors contribute to the establishment of a neuroprotective state following footpad poly(I:C) administration. Both, WNV footpad infection and poly(I:C) induced a range of pro-inflammatory serum cytokines as indicated by π-score, a metric combining fold change and statistical significance⁶⁴ (Figure S5A). Analogous to our transcriptomic approach, we performed comparative analyses of serum cytokines following footpad administration of poly(I:C), LPS or ODN1826. Fourteen cytokines (IFN-α, CCL2, G-CSF, CCL5, CCL12, CXCL1, CCL4, IL6, IFN-ɣ, CXCL9, M-CSF, CCL3, TNFα, and CXCL10) achieved a π-score > 4 for at least one agonist (Figure 4A). While most induced cytokines were shared by all PAMPs (Figure 4B), IFN-α, TNFα and M-CSF were unique to poly(I:C).

Figure 4. Peripheral type I interferon sensing mediates distal poly(I:C) mediated encephalitic protection.

(A) π-score quantification of serum cytokines elicited by footpad PAMP treatment compared to vehicle controls. Cytokines were ordered by maximum π-score in any condition. Data were pooled from 2 independent experiments. n = 3–4. (B) Upset plot displaying overlap of cytokines with π-score ≥ 4. (C) Normalized weight and (D) survival from animals treated with isotype control antibody or anti-IFNAR1 blocking antibody. Data shown are pooled from 2 independent experiments. n = 8. For normalized weights, each line represents an individual animal. (E) CNS expression of six canonical ISGs at 24 hours following footpad-poly(I:C) combined with intravenous anti-IFNAR1 or isotype antibody pre-treatment 24 hours prior. Data are normalized to housekeeping gene Gapdh and isotype antibody control animals. Data shown are experiment means from 2 independent experiments. n = 4–5. (F, G) Serum concentrations of (F) IFN-α and (G) IFN-β at 24 hours following footpad treatment with poly(I:C) or PBS. Data shown are presented as mean +/− s.e.m. and each dot corresponds to an individual animal. n = 5. (H) Expression of six canonical ISGs normalized to housekeeping gene Gapdh and vehicle control animals at 8 hours after intravenous administration of IFN-α3 or IFN-β. Data shown are experimental means from 2 independent experiments. n = 3–5. (I) Normalized weight and (J) survival of animals following intravenous administration of IFN-α3, IFN-β or vehicle at 16 hours prior to infection. (K) Transcriptomic analyses and hierarchical clustering of DEGs in brain tissue following intravenous IFN-α3 or footpad poly(I:C) treatment after 16 hours. (L) Top 15 GO:BP terms enriched in cluster 1 (K). Statistical testing for mouse weight (C, I) and cytokine quantification (F, G) was performed using unpaired student’s t test with Holm-Šídák correction for multiple comparisons. For (C, I), stars on x-axis indicate statistically significant difference in weight between conditions on indicated day. Statistical testing for survival (D, J) was performed using Log-Rank (Mantel-Cox) test. Statistical testing for RNA-seq (K) were calculated with DESeq2 with adjusted p values calculated using Benjamini-Hochberg correction. Statistical significance for GO Enrichment (L) was determined using EnrichGO. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Perturbing these cytokine signaling pathways via blocking antibodies revealed a critical role of IFNAR1, but not TNFα or M-CSF in the context of poly(I:C)-mediated encephalitic protection (Figures 4C and 4D, Figures S5B and S5C) and correlated with ablated brain ISG responses (Figure 4E and Figure S5D). Importantly, intravenously administered antibodies cannot cross the BBB⁶⁵ and footpad poly(I:C) did not significantly impact BBB permeability as determined by Evan’s Blue dye extravasation (Figure S5E).

Intriguingly, poly(I:C) administration predominantly increased serum IFN-α, rather than IFN-β after 24 hours (Figures 4F and 4G) and intravenous administration of IFN-α3 elicited a more potent CNS ISG response than IFN-β (Figure 4H). In line, a single intravenous dose of IFN-α3, but not IFN-β, protected against severe WNV disease (Figures 4I and 4J).

We then examined transcriptomic changes in brain tissue associated with intravenous IFN-α3, footpad poly(I:C) and matching vehicle treatments. Hierarchical clustering of DEGs (fold change > 2, average FPKM > 1, adjusted p value < 0.05) identified a gene set (cluster 1) that was robustly induced by both, IFN-α3 and poly(I:C) (Figure 4K). This putative protective gene set predominantly consisted of ISGs and GO enrichment analyses confirmed a strong association with antiviral IFN-I responses (Figure 4L), whereas poly(I:C)-specific genes (cluster 2) were associated with more generalized innate immune responses (Figure S5F, Table S2). Taken together, systemic IFNAR1 signaling and IFN-α3 are critical for robust CNS ISG induction post footpad administration of poly(I:C).

ISG responses in BMECs following peripheral IFN-I precede glial effector responses.

We performed single-nuclei RNA-seq (snRNA-seq) of CNS cortical tissue at 2 or 16 hours post intravenous IFN-α3 or PBS-treatment to identify IFN-I responsive cell types and transcriptomic changes across all CNS cell populations over time. Using uniform Manifold Approximation and Projection (UMAP) and clustering of a total of 69,003 nuclei (Figure 5A), identified cell types via common marker gene expression (Figure 5B and Figure S6A). As expected, most nuclei were annotated as neurons (excitatory, inhibitory, and interneurons), followed by glia (oligodendrocytes, astrocytes, microglia, and oligodendrocyte precursor cells [OPCs]) and structural cells (fibroblasts, BMECs, pericytes, and ependymal cells [ECs]) (Figure S6B). Each cell cluster consisted of approximately equal nuclei counts per condition (Figure S6C). Gene module enrichment analyses of the GO modules “GO:000954 Immune Response”, “GO:0006954 Inflammatory Response”, and “GO:0034340 Response to Type I Interferon” highlighted structural cells as the immediate responders to systemic IFN-α3 (Figures S6D to S6F). Strikingly, enrichment analyses based on our identified protective gene-set (Figures 4K and 4L) highlighted BMECs as the primary responder cell type at 2 hours post treatment (Figure 5C). Interestingly, the temporal dynamics of CNS responses following this initial BMEC-driven response revealed a more even expression of this protective gene set across CNS-resident glial cells and in particular microglia at 16 hours post treatment (Figures 5C and 5D). Of note, innate immune responses in neurons appeared restricted with only a minor induction at 16 hours (Figures 5C and 5D and Figures S6D to S6F).

Figure 5. Immediate ISG induction in BMECs preceeds effector responses in glia cells.

(A) Experimental setup, (B) UMAP clustering of snRNA-seq data and annotated cell clusters of the CNS at 2 hours or 16 hours post intravenous IFN-α3 treatment or control treatment. (C-D) Module score expression of CNS protective gene signature (Figure 4K) across CNS cell types depicted as (C) heatmap and (D) UMAP projection. (E) Number of DEGs per cell cluster at 2 and 16 hours post IFN-α3 treatment compared to controls. (F) Top 10 enriched GO terms of ranked DEGs in microglia at 16 hours post IFN-α3 treatment. (G) CellChat analyses of increased microglia-derived cell-cell communication signals to other cell clusters and (H) top 7 increased pathways of microglia to neuron communication pathways at 16 hours post IFN-α3 treatment. “Relative” denotes the proportional contribution of each sample to the indicated pathway whereas “strength” represents the total inferred communication probability of all cells in the samples across all pathways. (I) Expression of Nrg1, Nrg2, Nrg3 and Nrg4 in microglia upon IFN-α3 treatment. (J) TF activity prediction analyses of CREB1, ELK1, STAT3, STAT5a and STAT5b in neuron subsets upon IFN-α3 treatment. Data of each time point are derived from tissue samples isolated from three individual mice that were pooled at equal weights of brain tissue. Statistical testing for snRNA-seq DEGs (E, F) was done via the Seurat R package. Statistical testing for GO enrichment was done using the clusterProfiler R package and corrected for multiple testing using the Benjamini-Hochberg method. Statistical testing for ligand-receptor cell interaction prediction analyses (G, H) was done using the CellChat R package.

Strikingly, microglia dominated this response at 16 hours compared to other cell types with more than 1,000 DEGs (absolute log₂ fold change > 0.5, adjusted p value < 0.05) (Figure 5E and Table S3). Ranked GO enrichment analyses revealed that these genes were associated with terms related to immunity and neuronal function (Figure 5F and Table S3). We then investigated cell-cell communication networks using the computational tool CellChat, which infers cellular interactions from single cell expression data via known receptor-ligand interactions^66,67. These analyses predicted increased microglia-neuron interaction at 16 hours (Figure 5G), primarily driven by pathways related to neuron maintenance, including neuregulin-ErbB signaling (NRG), glutamate transport, neurexin interaction (NRXN), neuronal growth regulator signaling (NEGR), receptor protein tyrosine phosphatase signaling (PTPR), adhesion G protein coupled receptor latophilin (ADGRL) and GABA-A signaling (GABA-A) (Figure 5H and Figure S6G). Of these, NRG-ErbB in particular is thought to counteract apoptosis and increase neuron health via AKT and MAPK/ERK signaling pathways^68–72. Indeed, microglia induced transcription of the ligands Nrg1, Nrg2, Nrg3 and Nrg4 upon IFN-α3 treatment (Figure 5I) and pseudobulk analyses of predicted transcription factor (TF) activity via VIPER⁷³ and DoRothEA⁷⁴ revealed increased activity of the downstream TFs STAT3, STAT5, ELK1, and CREB1 across all neuronal cell populations (Figure 5J and Table S3).

This unbiased approach identified BMECs as immediate responders to peripheral IFN-I followed by a sustained glial cell response that might facilitate neuroprotective communication between microglia and neurons.

IFNAR1 on endothelial cells is critical to relay peripheral antiviral signals into the CNS.

To complement our snRNA-seq data, we performed bulk RNA-seq analysis of purified BMECs at 6 hours post footpad poly(I:C) or PBS treatment and compared transcriptomic changes to matching whole brain tissue (Table S4). BMEC purity was confirmed by strongly increased expression of BMEC-associated tight junction⁷⁵ and cell adhesion genes⁷⁶ Cldn5, Tjp1, Pecam-1 and Icam-1 compared to bulk brain (Figures S7A to S7D). Hierarchical clustering of the top 500 variable DEGs (Figure 6A) and GO analysis confirmed enrichment for neuronal processes in brain specific genes (cluster 2) and endothelial pathways for BMEC-specific genes (cluster 3) ( Figures S7E and S7F, Table S4). Cluster 1 contained genes that were hyper-induced by BMECs and were strongly enriched for antiviral and IFN response genes (Figure 6B). This suggested that BMECs are highly responsive to footpad poly(I:C) and more robustly induce antiviral genes compared to whole brain tissue. We then generated a complementary RNA-seq dataset of BMECs isolated at 2 hours following intravenous IFN-α3 treatment (Table S5). Integration and hierarchical clustering of DEGs (fold change > 2, average FPKM > 1, adjusted p value < 0.05) identified 545 induced genes in BMECs upon both, footpad poly(I:C) or intravenous IFN-α3 administration (Figure 6C, Figure S7G and Table S5). Inference of upstream TF activity implicated increased transcription via IRF9, STAT1 and STAT2 (Figure 6D). Together these TFs form the Interferon-Stimulated Gene Factor (ISGF3) complex, the core transcriptional effector of canonical IFN-I signaling via the IFNAR1 receptor⁷⁷. This correlated with robust ISG induction by BMECs (Figures 5C and Figure 6A), but also with increased transcription of cytokines and chemokines (Figure S7H) and secretion of the IFN-I-induced cytokines CXCL9 and CXCL10 (Figures S7I and S7J).

Figure 6. IFNAR1 signaling in BMECs is necessary for distal poly(I:C)-mediated encephalitic protection.

(A) Top 500 variable DEGs in whole brain tissue or isolated BMECs at 6 hours post footpad poly(I:C) or vehicle treatment. Each column represents a single animal, and each row represents a single gene. n = 3–6. (B) Top 15 significant GO:BP enrichment terms from gene cluster 1 identified in (A). (C) Venn diagram of genes induced in the CNS at 6 hours after footpad poly(I:C) and 2 hours after intravenous IFN-α3 treatment. n = 3–4. (D) Inferred transcription factor activity based on 545 genes commonly induced by footpad poly(I:C) and intravenous IFN-α3 (C). (E) Experimental setup for testing footpad-mediated poly(I:C) protection in BMEC-specific IFNAR1 KO mice. (F) Quantification of serum IFN-α as determined. Data shown are mean +/− s.e.m. Pooled data from 2 independent experiments. n = 3–5. (G) Expression of six canonical ISGs, at 16 hours post footpad poly(I:C) administration normalized to housekeeping gene Gapdh and wildtype vehicle controls. Each column represents the experimental mean of an independent experiment with n = 3–5. (H-K) Normalized weight (H and J) and survival (I and K) of BMEC-IFNAR1-CTRL (GFP-AAV) or BMEC-IFNAR1-KO (Cre-AAV) animals footpad treated with (H-I) vehicle control or (J-K) poly(I:C) prior to intracranial WNV challenge. For weights, each line represents an individual mouse. Data shown are combined from 2 independent experiments. n= 7–8. Statistical testing for RNA-seq data (A) performed using DESeq2 package with unsupervised hierarchical clustering in pheatmap. Statistical significance for GO Enrichment (B) was determined using EnrichGO. Statistical testing for inferred transcription factor activity (D) was performed using the limma package. Statistical testing for serum ELISA (E) was performed using two-way ANOVA and Uncorrected Fisher’s LSD for individual comparisons. Statistical testing for mouse weight (H, J) was determined using unpaired student’s t test with Holm-Šídák correction for multiple comparisons. Stars on x-axis indicate statistically significant difference in weight between conditions on indicated day. Statistical testing for survival (I, K) was determined using Log-Rank (Mantel-Cox) test. *p < 0.05, **p < 0.01, ***p < 0.001.

To test the necessity of BMEC IFNAR1 signaling to ameliorate WNV disease, we genetically perturbed this signaling axis via the Adeno-Associated Virus (AAV) BI30 capsid, which preferentially targets endothelial cells in the brain⁷⁸. Using an AAV-GFP construct, we confirmed BMEC-specific targeting via co-localization of the endothelial cell marker CD31 and GFP in brain tissue (Figure S7K), as well as a drastic increase of Gfp transcript levels in CD31⁺ vs. CD31⁻ brain cells (Figure S7L). We then used AAV BI30 to deliver Cre recombinase (AAV-Cre, KO) or GFP-NLS (AAV-GFP, control) into Ifnar1^fl/fl animals to specifically ablate IFNAR1 in BMECs 2 weeks prior to footpad poly(I:C) administration (Figure 6E). While serum levels of IFN-α were comparable between BMEC-IFNAR1-deficient and respective control animals (Figure 6F), their CNS ISG response was significantly blunted (Figure 6G and Figure S7M). Importantly, AAV-Cre or AAV-GFP delivery itself did not affect weight loss or survival when pretreated with vehicle (Figures 6H and 6I). Strikingly though, BMEC IFNAR1-deficiency negated poly(I:C)-induced protection against WNV disease severity (Figures 6J and 6K), highlighting a critical role of endothelial IFNAR1 signaling to mediate a prophylactic antiviral effect in the CNS.

Footpad poly(I:C) broadly protects against viral encephalitis.

Lastly, we tested the breadth of distal poly(I:C) mediated protection against multiple encephalitic viruses. Indeed, footpad poly(I:C) effectively protected against severe encephalitis caused by the related (+) ssRNA Orthoflavivirus, Powassan virus (33% survival), the (+) ssRNA Togavirus, Sindbis virus (53% survival), and the dsDNA Orthoherpesvirus, Herpes Simplex virus 1 (100% survival) (Figures 7A and 7B).

Figure 7. Footpad poly(I:C) protects against diverse encephalitic viral challenge.

(A) Normalized weight and (B) survival of animals pretreated with footpad poly(I:C) and challenged with the indicated encephalitic virus. For normalized weight each line represents a single animal. All values were normalized to animal weight on day of poly(I:C) administration (−1 DPI). Data shown are 2–3 independent experiments. n = 3–8. Statistical testing for mouse weight (A) was performed using unpaired Student’s t test with Holm-Šídák correction for multiple comparisons. Stars on x-axis indicate statistically significant difference in weight between conditions on indicated day. Statistical testing for survival (B) was performed using Log-Rank (Mantel-Cox) test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

These findings demonstrate that peripheral viral PAMP sensing establishes a broadly protective state in the CNS against viral encephalitis and reveal a common mechanism of encephalitic protection, whereby BMECs rapidly warn the brain of peripheral immune threats.

Discussion

Pathogen encounter can result in four fundamental outcomes – (i) the host succumbs to disease, (ii) the pathogen is efficiently cleared, (iii) manifestation of chronic disease, or (iv) microbe-host co-existence via cooperative defense mechanisms. These outcomes are shaped by pathogen properties, host immune responses, and past microbial exposures, which can establish inflammatory memory and shape secondary immune responses^79–81. Dysregulated immune responses are often causal for pathological disease progression and death and are especially relevant in the context of emerging viral pathogens, where hyperinflammatory responses frequently underly severe disease^6,82–84.

Due to the strongly limited self-renewal capacity of neurons, such immune-mediated tissue damage is particularly deleterious in the CNS. Indeed, the BBB shields the CNS from most peripheral activity and threats, including systemic immune responses and pathogens, and only breaks down under severe disease or stress conditions^23,85–87. Arthropod-borne neurotropic viruses, like WNV, can readily cross the BBB and pose a global health problem with annually increasing incidence and high case fatality rate^29,31,32. Neuroinvasion of WNV is characterized by hyperinflammation and can result in post-infection sequelae^{58,59,88–90}. While previous studies have characterized innate immune responses post neuroinvasion⁵³ or in relevant ex vivo brain models^52,54, there is limited understanding of immediate innate immune responses that precede and counteract encephalitis.

WNV initially replicates in the skin before disseminating systemically to draining lymph nodes and the spleen³⁵. Our study highlights these tissues as pivotal immune sentinels that rapidly alert the CNS to viral infection. We demonstrate that peripheral viral PAMP sensing and systemic IFN-I result in rapid ISG induction in BBB-resident BMECs. This IFNAR-mediated response of BMECs licenses establishment of a potent antiviral state in the CNS prior to neuroinvasion that limits viral replication and pathological neuroinflammation. Following initial sensing of systemic IFN-I via BMECs, we provide evidence of increased microglia to neuron crosstalk via neuroprotective signaling pathways, similar to enhanced neuronal surveillance by microglia in neurodegenerative disease^91,92. Of note, the related viscerotropic orthoflaviviruses yellow fever virus (YFV) and dengue virus (DENV) cause substantially more peripheral inflammation, yet rarely induce viral encephalitis in humans. In line with this, encephalitis is common for these pathogens in murine infection models that fail to recapitulate highly inflammatory viscerotropic disease^93,94. Similarly, co-administration of mosquito salivary gland extract, which has anti-inflammatory properties^95,96, enhances WNV neuroinvasion^97,98. Finally, dampening peripheral immune activity by depleting splenic macrophages increases WNV neuroinvasion and lethality⁹⁹. These observations suggest that peripheral inflammation and encephalitic outcome may be anti-correlated. Our observation of a positive correlation of systemic IFN-I responses, brain ISG induction and reduced disease severity not only corroborates these findings, but further highlights immediate and robust peripheral IFN-I responses as a key disease determinant.

The BBB is the primary barrier through which the brain interacts with circulating cytokines. Circulating IFN-I can activate antiviral pathways in microglia without BBB breakdown¹⁰⁰ and we demonstrate that BBB-mediated detection of IFN-α is beneficial in viral encephalitis. IFN-I activity relating to BBB permeability and neuroinvasion has been previously described^101–103, but we expand this further, demonstrating a prophylactic effect of BMEC-IFN-I signaling. These findings align with human studies that link impaired systemic IFN-I responses due to neutralizing autoantibodies against IFN-α, but not IFN-β, to dramatically increase risk of developing severe viral encephalitis, without affecting overall incidence of infection^41,104,105.

Curiously, we observed that IFN-α, but not IFN-β, conveyed robust protection against intracranial WNV challenge and correlated with a 10-fold increased ISG induction in the CNS upon IFN-α dosing. Similarly, collaborative cross mouse experiments comparing asymptomatic to encephalitic mouse lines upon WNV infection found no impact of splenic IFN-β expression on encephalitic outcome¹⁰⁶. While IFN-α and IFN-β both signal via the IFNAR1-IFNAR2 complex, IFN-β has an approximately 100-fold higher receptor binding affinity than IFN-α^107,108. Expression of IFN-β is ubiquitous across most cell types upon viral PAMP recognition, whereas IFN-α expression is largely restricted to immune cells, such as plasmacytoid dendritic cells^109,110. Restricted cell type expression and reduced binding activity might increase cross-organ signaling capacity of IFN-α. The fact that IFN-α has evolved several functional subtypes in humans and mice, compared to one IFN-β subtype, might lend support to this notion. In fact, IFN-I subtypes have non-redundant roles in WNV¹¹¹ and lymphocytic choriomeningitis virus (LCMV) infection¹¹², although the precise underlying mechanisms remain unclear.

Pathogen-host interactions are inherently complex. In the context of arthropod-borne viral encephalitis, we identified rapid induction of peripheral IFN-I preceeding neuroinvasion as a critical disease determinant. Systemic IFN-α rapidly induced an antiviral and neuroprotective state in the CNS via IFNAR1 signaling on BMECs that potently ameliorated encephalitic disease outcomes. The mapping of such basic principles of neurotropic viral infections and identification of key mediators of immune pathology in this study therefore contribute to a more holistic understanding of viral encephalitis and inform much-needed therapeutic intervention options for this re-emerging disease.

Limitations of the Study

We demonstrate that peripheral IFN-I is necessary to induce an antiviral, neuroprotective state within the CNS by acting specifically through IFNAR1 on BMECs. We observed that both footpad poly(I:C) administration and footpad WNV infection induced systemic IFN-α, but poly(I:C) was more potent. This correlated with ~50% average survival in our footpad poly(I:C) and subsequent intracranial WNV infection model system compared to ~25% average survival upon WNV footpad infection. Although this supports the notion of peripheral PAMP sensing and the robustness of the resulting systemic IFN-α levels being the main determinant of neuroprotection, these data are mainly correlative. Future research involving real-time monitoring of immediate ISG responses in footpad WNV infection prior to neuroinvasion and linking them to disease severity and lethality of individual animals would provide critical additional evidence.

Peripheral viral PAMP sensing establishes a neuroprotective state that is independent of BBB leakage or CNS-infiltrating immune cells and suggests the involvement of a secondary mediator produced by the BBB. Due to intrinsic technological and biological hurdles, we only captured limited cytokine activity in our single-nuclei transcriptomic datasets, reducing our capacity to infer cytokine signaling networks in the CNS between individual cell types. While we confirmed increased transcription and secretion of cytokines by BMECs upon IFN-I sensing, further investigation is needed to determine the degree to which they contribute to altered cell-cell communication within the CNS and resulting neuroprotection.

Additionally, while we provide multiple lines of evidence of no systemic dissemination of footpad-administered poly(I:C), these experiments cannot completely exclude some contribution of direct poly(I:C) sensing by BMECs and subsequent localized IFN-I production. BMEC-specific Tlr3 depletion could help to definitively address this point and further deepen our insights into the novel phenomenon of cross-organ neuroprotection. Nonetheless, our experimental evidence supports the notion of systemic IFN-I being indeed necessary and sufficient to establish a neuroprotective state following footpad poly(I:C) sensing.

Lastly, we chose the highly pathogenic NY99 strain of WNV for our study, which is widely used in preclinical research, as it represents an isolate obtained during the inital WNV epidemic in 1999. Corroborating our findings using recent strains like the now dominant NA/WN02 lineage would be an interesting line of research with translational value given its clinical relevance. Yet, given the broad applicability of our phenotype to other encephalitic viruses of distant clades, we believe the impact of peripheral IFN-I signaling on WNV encephalitis is likely conserved across different WNV strains.

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Charles M. Rice (ricec@rockefeller.edu).

Materials Availability

This study did not generate unique reagents.

Data and Code Availability

Single-nuclei RNA-seq and bulk RNA-seq data have been deposited at Gene Expression Omnibus (GEO) and are publicly available via accession numbers GSE318728 (RNA-seq of brain tissue upon footpad PAMPs, related to Figure 3), GSE318714 (RNA-seq of brain tissue following intravenous IFN-α treatment, related to Figure 4), GSE318766 (10X Genomics single-nuclei RNA-seq, related to Figure 5), GSE318729 (RNA-seq of BMECs and brain tissue upon footpad poly(I:C), related to Figure 6) and GSE318714 (RNA-seq of BMECs following intravenous IFN-α treatment, related to Figure 6). Code used to analyze the data is available on GitHub (https://github.com/alexlercher/Lewy_etal_2026).

STAR Methods

EXPERIMENTAL MODEL DETAILS

Mice

Wildtype (JAX Strain #000664), IFNAR1^−/− (B6(Cg)-Ifnar1^tm1.2Ees/J [Jax Strain# 028288]), and IFNAR1^{fl/fl (}B6(Cg)-Ifnar1^tm1.1Ees [JAX Strain# 028256]) C57Bl/6J mice were originally purchased from Jackson Laboratories and colonies were maintained in house. Ccr2-DTR depletion mice were generously provided by Dr. T. Hohl (MSKCC)⁶¹. To ensure age synchronized cohorts, in vitro fertilization was used to amplify the colony prior to experimentation. Briefly, wildtype C57BL6/J egg donors (JAX Strain #000664) were purchased from the Jackson lab, hormonally primed with PMSG (Prospec Protein Specialists, Cat# HOR-272-b), then hCG (Sigma-Aldrich, Cat# C1063) 48 hours later. Eggs retrieved from these mice were fertilized with sperm from wildtype colony males. Two-cell embryos were surgically transferred to Charles River CD-1 females (Charles River Strain #022) for gestation. Pups were reared by CD-1 mothers until weaning at 3 weeks of age. All mouse experiments were performed in accordance with the NIH Guide for Care and Use of Laboratory Animals under protocols approved by the Institutional Animal Care and Use Committee of Rockefeller University. Four- to five-week-old animals of both sexes were used for all mouse experiments unless otherwise specified. Body weight loss greater than 20% of initial weight, seizure, or paralysis were defined as the humane endpoint in all survival studies. All infectious animal experiments and downstream processing of live cells and non-inactivated tissues were conducted under biosafety level containment in compliance with institutional and federal guidelines.

Cell Lines

BHK-21 cells (Mesocricetus auratus, sex: male, fibroblast) were obtained from ATCC (#CCL-34). Vero E6 cells (Chlorocebus sabaeus, sex: female, kidney epithelial) were obtained from ATCC (CRL-1586). Huh7.5 cells (Homo sapiens, sex: male, hepatoma) were previously derived in house from Huh7 cells¹¹⁵. Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, ThermoFisher Scientific, Cat# 11995065) supplemented to contain 1% non-essential amino acids (NEAA, ThermoFisher Scientific, Cat# 11140076), 1% Penicillin-Streptomycin (ThermoFisher Scientific #15140122), and 10% fetal bovine serum (FBS, HyClone Laboratories, Lot # AUJ35777) at 37°C and 5% CO₂ unless otherwise noted. All cell lines were confirmed to be negative for mycoplasma contamination.

Viruses

West Nile Virus

West Nile virus (WNV; Strain: NY99) stocks were derived from an infectious clone of WNV NY99 generously provided by Dr. Pei-Yong Shi¹¹⁴. RNA was generated from linearized plasmid using T7 in vitro transcription (mMessage mMachine High Yield Capped RNA Transcription kit, ThermoFisher, Cat# AM1344) and electroporated into BHK-21 cells (BTX model ECM830 electroporator). Virus was harvested 48h later from supernatant and clarified by centrifugation. The P₀ stock was used to generate a P₁ stock by infecting BHK-21 cells with a multiplicity of infection (MOI) = 0.15. Virus was harvested 48h later from supernatant, clarified by centrifugation, and stored as single use aliquots at −80°C. Final viral titer was determined by plaque assay on VeroE6 cells. P₁ stock was used for all experiments.

Powassan Virus

Powassan virus (POWV; lineage II [Deer Tick Virus]; strain: Spooner) was generated as previously described¹²⁸. Briefly, in vitro transcribed RNA was electroporated into Huh-7.5 cells. Virus-containing medium was collected seven days post-electroporation, clarified by centrifugation to remove cell debris, and stored in single-use aliquots at −80°C. Viral stock titer was determined by plaque assay on Huh7.5 cells as described¹²⁸.

Sindbis Virus

Sindbis virus (SINV; strain: SVN) stocks were generated from cDNA infectious clone plasmid pSVN¹²⁹. In brief, the plasmid was linearized, ethanol precipitated and concentrated, used as template for in vitro transcription (mMessage mMachine SP6 Kit, Invitrogen, Cat# AM1340), and electroporated into BHK-J cells. Cell supernatants were harvested at 24 hours post electroporation, clarified by centrifugation, and stored at −80°C in single use aliquots. Stocks were titrated via plaque assay on BHK-J cells as previously described¹³⁰.

Herpes Simplex Virus 1

Herpes simplex virus 1 (HSV-1; strain: KOS), generously provided by David Leib (UMass Dartmouth), was amplified in VeroE6 cells and stored in single use aliquots at −80°C. Titer was determined by plaque assay on VeroE6 cells.

METHOD DETAILS

WNV Infections

Mice were anesthetized with 4% isoflurane until unresponsive to pain pinch test. Virus was diluted to the indicated dose in 1X D-PBS supplemented with 1% FBS. For footpad infections, 29G insulin syringes were used to deliver 1,000 PFU in 20μL inoculum. For intracranial infections, Hamilton gastight syringes (Hamilton, Cat# 81001) were used to deliver 10 PFU in 10μL inoculum into right cortex.

POWV Infections

Mice were anesthetized with 4% isoflurane until unresponsive to pain pinch test. Virus was diluted to 10PFU/10μL in 1X D-PBS supplemented with 1% FBS. Hamilton gastight syringes (Hamilton, Cat# 81001) were used to deliver 10μL inoculum into right cortex.

SINV Infections

Mice were anesthetized with 4% isoflurane until unresponsive to pain pinch test. Virus was diluted to 100PFU/10μL in 1X D-PBS supplemented with 1% FBS. Hamilton gastight syringes (Hamilton, Cat# 81001) were used to deliver 10μL inoculum into right cortex.

HSV-1 Infections

Mice were anesthetized with 4% isoflurane until unresponsive to pain pinch test. Virus was diluted to 1000PFU/10μL in 1X D-PBS supplemented with 1% FBS. Hamilton gastight syringes (Hamilton, Cat# 81001) were used to deliver 10μL inoculum into right cortex.

WNV Plaque Assay

VeroE6 cells were seeded into wells at a density of 10⁵ cells/well. After 24h, media was removed, and cells were inoculated with a known dilution of viral sample diluted in OptiMeM (ThermoFisher Scientific, Cat# 31985062). Plates were incubated at 37°C, 5% CO₂ for 1h with occasional rocking. Inocula were removed and cells were overlaid with semi-solid media consisting of Dulbecco’s Modified Eagle Medium (DMEM, ThermoFisher Scientific #11995065) supplemented to contain 1% non-essential amino acids (NEA, ThermoFisher Scientific #11140076), 1% Penicillin-Streptomycin (ThermoFisher Scientific #15140122), 10% fetal bovine serum (FBS, HyClone Laboratories, Lot #AUJ35777), and 1% methylcellulose (ThermoFisher Scientific, Cat# 044779-A1). Cells were incubated at 37°C, 5% CO₂ for 96h. Monolayers were fixed using 8% paraformaldehyde [Electron Microscopy Sciences, Cat# 157–8] in 1X Dulbecco’s phosphate buffered saline (D-PBS, ThermoFisher Scientific, Cat# 14190144) and stained with crystal violet solution (1.25 % crystal violet, 20 % methanol in distilled water). Plates were rinsed in tap water and air dried before enumerating plaques by hand.

PAMP Footpad Treatments

Mice were anesthetized with 4% isoflurane (Pivetal, Cat# 78949580) until unresponsive to pain pinch test. For PAMP treatments, PAMP was diluted to indicated concentration in 1X D-PBS and injected in a volume of 20μL. PAMPs used were PAM3CSK4 (Invivogen, Cat# tlrl-pms, 10μg per dose), HMW poly(I:C) (Invivogen, Cat# tlrl-pic, 50μg per dose), LPS-EK (Invivogen, Cat# tlrl-eklps, 10μg per dose), FLA-ST (Invivogen, Cat# tlrl-stfla, 10μg per dose), FSL1 (Invivogen, Cat# tlrl-fsl, 5μg per dose), R848 Resiquimod (Invivogen, Cat# tlrl-r848–1, 10μg per dose), and ODN1826 (Invivogen, Cat# tlrl1826, 3.2 μg per dose).

Brain Harvest

Mice were humanely euthanized by CO₂ asphyxiation. Animals were cardiac perfused with ≥20mL cold D-PBS to clear any residual blood in the brain. For immunohistochemistry samples, brain was additionally perfused with 20mL cold 4% PFA/D-PBS. The head was removed from the animal and the brain dissected from the skull. For immunohistochemistry samples, brains were additionally drop fixed in 4% PFA/D-PBS overnight at 4°C. For RNA isolations, brains were minced and transferred to tubes containing 500μL Trizol Reagent (Thermo Fisher Scientific Cat#15596026) and 500μL 1.0mm glass beads (BioSpec Products, Cat#11079110). For infectious virus quantification, brains were minced and transferred to tubes containing 500μL D-PBS and 500μL 1.0mm glass beads (Biospec, Cat# 11079110). RNA and infectious virus samples were homogenized at 6000rpm for 60s (Biospec Mini-Beadbeater-16). Samples were stored at −80°C until further analysis. For cytokine analyses, whole brains were homogenized in 1mL D-PBS containing cOMPLETE EDTA-free protease inhibitor (Millipore Sigma Cat#4693132001) as described above. Samples were then centrifuged for 10 minutes at 10,000 rpm at 4°C (Eppendorf, Centrifuge 5415 R), 500μL of supernatant were transferred into a new tube and stored at −20°C until further analyses.

Terminal Sera Harvest

Mice were humanely euthanized by CO₂ asphyxiation. Animals were cardiac punctured, and blood was directly collected. Blood was spun for 5 minutes at 13,200 rpm at 4°C (Eppendorf, Centrifuge 5415 R) to clear sera. Cleared sera were stored at −80°C until further analysis.

Flow Cytometric Analyses

For flow cytometric analyses of circulating cells, 2–5 drops of blood were collected in 1mL RPMI 1640 medium (Thermo Fisher Cat#11875093 that was supplemented to contain 10% FBS, 1% Penicillin-Streptomycin (Thermo Fisher Cat#15140122) containing 2,000 U/L heparin (Stemcell Cat#07980). Samples were centrifuged at 500G for 5 minutes at 4°C and supernatants were discarded. Cells were resuspended in 500μL RBC Lysis buffer (Biolegend Cat#420301) and incubated at room temperature for 3 minutes. RBC lysis was stopped by addition of 500μL PBS (Thermo Fisher Cat#14190144), samples were centrifuged at 500×G for 5 minutes at 4°C and supernatants were discarded. Cells were resuspended in 25μL FcR block per well (200x Biolegend Cat#101320) diluted in PBS and incubated for 15 minutes at room temperature before adding 25μL of 2x surface and viability staining mix diluted in PBS. Cells were incubated for 45 minutes at 4°C and afterwards washed twice with 1mL FACS buffer (2% FBS, 2mM EDTA, 10g/L NaN₃). Cells were fixed with 4% paraformaldehyde (Sigma Cat#F8775) at room temperature for 10 minutes, washed once with 1mL FACS buffer and resuspended in 150–250μL FACS buffer and subsequently acquired on a BD LSR Fortessa or Cytek Aurora instrument.

For flow cytometric analyses of brain-resident cells, whole brains were isolated from animals following euthanasia and whole body perfusion with 20mL cold PBS. Brains were dissociated using the Miltenyi Adult Brain Dissociation Kit for mouse and rat (Miltenyi Cat#130–107-677). In brief, brains were washed in cold D-PBS (Thermo Fisher Cat#14190144), cut into eight saggital pieces and dissociated in enzyme mix using a gentleMACS Octo Dissociator with Heaters (Miltenyi Cat#130–134-029) and C Tubes (Miltenyi Cat#130–093-237) with program 37C_ABDK_01 according to the manufacturer’s instructions. Next, debris were removed according to the manufacturers protocol and the obtained cells were stained for flow cytometry analogous to cells isolated from blood (see above).

For surface marker staining anti-CD45 (BUV395, clone 30-F11, BD Biosciences Cat#563279), anti-CD45 (AF700, clone 30-F11, Biolegend Cat#103128), anti-CD11b (FITC, clone M1/70, Biolegend Cat#101205), anti-Ly6G (APC, clone 1A8, Biolegend Cat#127613), anti-NK1.1 (APC/Cy7, clone S17016D, Biolegend Cat#156509), anti-NKp46 (PE, clone 29A1.4, Biolegend Cat#137604), anti-Ly6C (BV605, clone HK1.4, Biolegend Cat#128035), anti-Ly6C (PE/Dazzle, clone HK1.4, Biolegend Cat#128043), anti-CD3 (PE/C7, clone 17A2, Biolegend Cat#100220), anti-CD3 (BUV737, clone 17A2, BD Biosciences Cat#612803), anti-CD19 (BV510, clone 6D5, Biolegend Cat#115545), anti-F4/80 (BV421, clone BM8, Biolegend Cat#123137), anti-CD64 (PE/Fire744, clone S8017D, Biolegend Cat#161009) and anti-P2RY12 (PE, clone S16007D, Biolegend Cat#848003) were used.

For dead cell exclusion, LIVE/DEAD Fixable Blue dead Cell Stain Kit (Thermo Fisher Cat#L23105) and DAPI (Thermo Fisher Cat#62248) were used.

RNA Isolation

Total RNA was isolated according to the TRIzol/chloroform extraction protocol of the Ambion PureLink RNA Mini Kit (Thermo Fisher Scientific #12183025). RNA concentration was determined by absorbance (Biotek, Cytation 7) and RNA was stored at −80 °C.

cDNA synthesis

cDNA was synthetized using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Cat# 4368814) according to the manufacturer’s instructions.

qPCR

Quantitative PCRs (qPCR) were run and analyzed on a QuantStudio3 Real-Time PCR System (Thermo Fisher Scientific) according to the PowerUP SYBR Green (Thermo Fisher Scientific #25741) protocol with 1.5 to 15 ng cDNA input. Expression levels of Gapdh (5′-GGCAAATTCAACGGCACAGT-3′, 5′-AGATGGTGATGGGCTTCCC-3′), Eef1a1 (5′-GCAAAAACGACCCACCAATG-3′, 5′-GGCCTGGATGGTTCAGGATA-3′) Ifit1 (5′-CTGAGATGTCACTTCACATGGAA-3′, 5′-GTGCATCCCCAATGGGTTCT-3′), Rsad2 (5′-TGCTGGCTGAGAATAGCATTAGG-3′, 5′-GCTGAGTGCTGTTCCCATCT-3′), Bst2 (5′-TGTAGAGACGGGTTGCGAGC-3′, 5′-CTCCTGAAGGGTCACCACGG-3′), Ifitm3 (5′-GCCTACTCCGTGAAGTCTAGGG-3′, 5′-CCAAGGTGCTGATGTTCAGGC-3′), Isg15 (5′-ACTGTGAGAGCAAGCAGCCA-3′, 5′-TTAGGTCCCAGGCCATTGCT-3′), Mx1 (5′-AGGCCATTAACAAGGGTCAGAAC-3′, 5′-GGTCTGGGACATTTGGGGAG-3′), Tnfa (5′-GCCCACGTCGTAGCAAACC-3′, 5′-TTTGAGATCCATGCCGTTGGC-3′), S100a8 (5′-TCCTTTGTCAGCTCCGTCTTCA-3′, 5′-ACGAAAGATTTCCTTTCAAACGATG-3′, Tnfrsf1b (5′-ACATTTCTGCAACAAGACCTCGG-3′, 5′-CGGATCTCCACCTGGTCAGT-3′), Il6 (5′-CACAAGTCCGGAGAGGAGACT-3′, 5′-TGCCATTGCACAACTCTTTTC-3′), Cxcl1 (5′-CCACTGCACCCAAACCGAAG-3′, 5′-CCGTTACTTGGGGACACCTTTTAG-3′), Sod2 (5′-GAAGTTCAATGGTGGGGGACA-3′, 5′-CCTCCAGCAACTCTCCTTTGGG-3′), eGFP (5′-ACGTAAACGGCCACAAGTTC-3′, 5′-AAGTCGTGCTTCATGTG-3′), HBV C (5′-CGTTTTTGCCTTCTGACTTCTTTC -3′, 5′-CTGAGTGCAGTATGGTGAGGTG-3′) and WNV NS3 (5′-GGAACATCAGGCTCACCAATAG-3′, 5′-CATCCTTTCACCCTGCACTATC-3′) were determined using a standard qPCR protocol (step 1: 20 sec at 95 °C, step 2: 1 sec at 95 °C, step 3: 20 sec at 60 °C, step 4: go to step 2 and repeat 40x, step 5: 1 sec at 95 °C, step 6: 20 sec at 60 °C, step 7: ramp down up to 95 °C at +0.15 °C/s, step 8: 1 sec 95 °C, step 9: end).

Mouse Cytokine Assays

Undiluted serum, brain or cell culture supernatant samples (see above) were sent to Eve Technologies (Calgary, AB, Canada) for cytokine analysis on a Luminex 200 system with Bio-Plex Manager Software using Luminex xMAP technology to quantitatively and simultaneously detect up to 44 cytokines by using assays MD36 or MD44. For additional IFNα2/4 quantification, in the context of the MD44 assay, sera were diluted 1:10 in 1X D-PBS and quantified by ELISA (Invitrogen, Cat# BMS6027) according to manufacturer instructions.

Neutralizing Antibody Treatment

250μg of neutralizing antibody against IFNAR1 (BioXcell, clone MAR1–5A3, Cat# BE0241), TNFα (BioXcell, clone XT3.11, Cat# BE0058)¹³¹, M-CSF (BioXcell, clone 5A1, Cat# BE0204)¹³², or horseradish peroxidase (BioXcell, clone HRPN, Cat# BE0088) was diluted in 1X D-PBS and delivered via intraperitoneal injection 24h prior to footpad poly(I:C) treatment.

Intravenous IFN-I Treatments

Mouse IFN-α3 (PBL Assay Science, Cat# 12100–1) or IFN-β (PBL Assay Science, Cat# 12401–1) was diluted in 1X D-PBS. Mice were anesthetized with isoflurane and 25,000 IU in 100μL were intravenously delivered retro-orbitally.

Murine BMEC Isolation

Brains were dissociated into single cell suspension using adult mouse brain dissociation kit (Miltenyi Biotec, Cat# 130–107-677) using gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec, Cat# 130–096-427) and gentleMACS C Tubes (Miltenyi Biotec, Cat# 130–093-237) according to manufacturer instructions. Samples were depleted of CD45⁺ cells to minimize leukocyte contamination using CD45⁺ magnetic microbeads (Miltenyi Biotec, Cat# 130–052-301) on LS columns (Miltenyi Biotec, Cat# 130–042-401). CD31⁺ brain endothelial microvascular cells were then isolated using CD31⁺ magnetic microbeads (Miltenyi Biotec, Cat# 130–097-418). Enrichment quality was determined by RNA sequencing.

Murine BMEC Culture

Cell culture plates were coated with fibronectin one day prior to BMEC isolation and culture. Per 96 well, 50μL of 20μg/mL fibronectin solution (Miltenyi Cat#130–109-393) were used and plates were coated over night incubation at 4°C with. Wells were washed thrice with double-distilled water and dried under sterile conditions. After BMEC isolation (see above), cells were plated, stimulated and cultured in EBM-2 basal medium containing all supplements (Lonza Cat#CC-3202) for up to 48 hours post isolation.

AAV Generation and Animal Transduction

AAV vectors were generated and titered by VectorBuilder Inc. (Chicago, IL, USA) using plasmids originally generated by Ben Deverman⁷⁸. Plasmids used were kiCAP-AAV-BI30 for capsid specificity (Addgene Cat#183749), pAAV-CAG-SV40NLSf-GFP-3xmiR122-WPRE-HGHpA for GFP-NLS cargo (Addgene Cat#183775), and pAAV-CAG-Cre-3xmiR122-WPRE-HGHpA for Cre Recombinase cargo (Addgene Cat#183776). AAV stocks were diluted in D-PBS and 10¹¹ genome equivalents (GE) per dose per mouse were delivered intravenously (retro-orbital). Mice were used for indicated experiments 14 days after AAV transduction to allow protein turnover.

FITC-Dextran footpad administration

Mice were anesthetized with 4% isoflurane until unresponsive to pain pinch test. 50μg FITC-Dextran, 150kDa (Sigma Aldrich, Cat# FD150S) was diluted in 1X D-PBS and was delivered into right rear footpad (20μL) or left orbital sinus (100μL). Following 30 minutes of dissemination, animals were humanely euthanized by CO₂ asphyxiation and blood was collected by cardiac puncture. Blood was spun for 5 minutes at 13,200 rpm at 4°C (Eppendorf, Centrifuge 5415 R) to clear sera and FITC-Dextran concentration was determined by 488nm absorbance on Cytation 7 instrument.

HBV pgRNA footpad administration

HBV pgRNA was generated from a linearized plasmid using T7 in vitro transcription (mMessage mMachine High Yield Capped RNA Transcription kit, ThermoFisher, Cat# AM1344) as previously described⁶². Mice were anesthetized with 4% isoflurane until unresponsive to pain pinch test and 50μg of HBV pgRNA was injected into right rear footpad or left orbital sinus. Following 30 minutes of dissemination, animals were humanely euthanized by CO₂ asphyxiation and blood was collected by cardiac puncture. Animals were perfused with cold 1X D-PBS and foot, popliteal lymph node, liver, and brain were harvested into Trizol and stored at −80°C until further experimentation.

NK cell depletion

NK cells were depleted via intraperitoneal administration of two doses of 400μL of anti-NK1.1 antibody (clone PK136) at 48h and 24h prior to poly(I:C) treatment. Anti-NK1.1 antibody was purified from hybridoma cells and generously provided by Dr. J.C. Sun (MSKCC)⁶⁰. Isotype-treated animals received an equal amount of anti-HRP antibody (BioXCell, clone HRPN, Cat#BE0088-R005).

Diphtheria toxin administration

Diphtheria toxin (Sigma, Cat# D0564) was reconstituted in PBS at 2 mg/mL and stored as single use aliquots at −20°C. Prior to use, this stock solution was diluted with PBS to 2 μg/mL and Ccr2-DTR^+/− mice were intraperitoneally administered 100μL at 24h prior to poly(I:C) treatment. Control-treated, littermate Ccr2-DTR^+/− mice received an equal amount of PBS.

Tissue Delipidation

Fixed brains from WNV-infected 4 week old mice were harvested at 2 and 4 DPI and processed using a newly optimized protocol¹³³ to image viral infection with immunolabeling in whole mount. Perfused and 4% PFA-fixed mouse brains were washed in PBS, then delipidated through several washes (1h, 2h, 4h, overnight) with B1n buffer (0.1% Triton X-100, 2% glycine, 0.01% 10N sodium hydroxide, 0.008% sodium azide in H2O). Afterward, the brains were incubated in SBiP buffer (200 mM Na2HPO4, 0.08% sodium dodecyl sulfate, 16% 2-methyl-2-butanol, 8% 2-propanol in H2O, pH 7.4) for 1-day × 6, under nutation at room temperature (RT). Subsequently, brains were transferred to B1n buffer for blocking for 4 hours at RT. The brains were then washed three times (1h × 3) with PTxwH buffer (0.1% Triton X-100, 0.05% Tween 20, 0.002% heparin (w/v), 0.02% sodium azide in PBS) and fixed with 1% PFA at 4°C overnight. Afterward, the samples were washed again with PTxwH buffer three times (1h × 3), followed by overnight incubation with B1n. The samples were then bleached in 0.3% H₂O₂ at 4°C overnight.

Skin Immunoflourescence

Murine paw skin was fixed for 1 hour at RT in 4% paraformaldehyde (Electron Microscopy Sciences) in PBS. Tissue was washed 3 times in PBS and switched to 30% sucrose overnight at 4°C. Sections were embedded sagittaly in OCT, and cryosectioned at 20 μm. Tissue sections were then permeabilized and blocked with 0.1% Triton X-100 and 1X BD Perm/Wash solution in PBS for 45 minutes at RT, blocked with Mouse On Mouse (M.O.M.) Kit (Vector Labs, Cat# MKB-2213–1) as per manufactor’s instructions. Subsequently, samples were stained with primary antibodies (overnight at 4°C), washed and stained with secondary fluorescence conjugated antibodies (2 hours at RT). Primary antibodies used were as following: CD11c-AF488 (Biolegend, clone N418, Cat# 117311), F4/80-AF647 (Biolegend, clone BM8, Cat# 123122), ITGA6-A488 (Biolegend, clone GoH3, Cat# 313608), dsRNA Mouse mAb (Cell Signaling, clone J2, Cat# 76651). Donkey anti-mouse RR-X was used a secondary antibody for dsRNA stainings and nuclei were stained with DAPI. Images were acquired on an inverted Dragonfly 202 spinning disk confocal system (Andor Technology Inc.) using the 40x oil immersion objective and a Zyla camera. Four laser lines (405, 488, 561 and 625 nm) were used for near simultaneous excitation of DAPI, Alexa-448, RRX and Alexa-647 fluorophores. Tiled images with a 1 μm stack step were acquired using the Andor Fusion software (v 2.3), and subsequently stitched with Fusion software. The signal of dsRNA was quantified using Imaris 9.5v by generating a surface on dsRNA signal-positive cells and determining the fraction of cells expressing F4/80, CD11c or neither. In cells, where detected signal for both cell-lineage markers, cells were classified as F4/80⁺ or CD11c⁺ based on the MFI of either channel.

Whole Mount Immunolabeling

We generated and used an anti-WNV human monoclonal antibody IgG1 that reacted with WNV NS1 protein. We used a de-identified blood sample stored in the Vanderbilt Center for Antibody Therapeutics biorepository that had been collected previously from an adult with a history of symptomatic, laboratory-confirmed WNV infection during the 2012 outbreak in Dallas, Texas¹³⁴. The study was approved by the Institutional Review Board of Vanderbilt University Medical Center and the sample had been obtained following written informed consent. A hybridoma clone secreting a human monoclonal antibody to WNV NS1 protein was isolated and used to produce purified IgG, exactly as previously described¹¹³. The resulting monoclonal antibody, designated WNV-121, was diluted in PTxwH buffer (1.059 mg/ml stock concentration, 2.5 μg/brain for 2 DPI, 12.5 μg/brain for 4 DPI) and added to the brains, which were kept at 37°C with gentle rocking for 10 days. Excess antibodies were removed through four washes with PTxwH (1h, 2h, 4h, overnight). The brains were then fixed with 1% PFA at 4°C overnight, followed by three washes in PTxwH buffer and incubation in B1n buffer. Next brains were incubated with Alexa Fluor 647-conjugated Fab goat anti-human secondary antibody (Alexa Fluor^® 647 AffiniPure^® Fab Fragment Goat Anti-Human IgG (H+L) [Jackson ImmunoResearch, Cat# 109–607-003, Lot# 144879, 50 μg/brain) at RT with gentle rocking for two weeks. Afterward, the brains were washed in PTxwH (1h, 2h, 4h, overnight). The samples were then fixed in 1% PFA at 4°C overnight, washed in PTxwH, and blocked in B1n at RT overnight. Continuing the delipidation process, the samples were immersed in SBiP buffer (30 min, 1h, 2h, 4h, overnight, 1-day × 6) at RT with gentle rocking. The brains were then washed twice (2h, 4h) in 20 mM PB buffer and twice in PTS solution (25% 2,2-thiodiethanol/10 mM PB) (3h, overnight), then equilibrated in 75% histodenz buffer (Cosmo Bio USA AXS1002424) with the refractive index adjusted to 1.53 using 2,2-thiodiethanol. The samples were stored at −20°C until imaging.

The cleared brain samples were imaged horizontally with tiling using the LifeCanvas SmartSPIM lightsheet microscope. 514 nm and 639 nm lasers were used for autofluorescence and WNV immunohistochemistry imaging with the 3.6x/0.2 detection lens. Lightsheet illumination was focused using an NA 0.2 lens, and the samples were axially scanned with an electrically tunable lens coupled to a Hamamatsu Orca Fusion camera in slit mode. Camera exposure was set to fast mode (2 ms) with 16-bit imaging. The X/Y sampling rate was 1.8 μm/pixel, and the Z step was 2 μm. Individual image planes were stitched to generate TIFF stacks representing the whole brain volume. As an indirect measure of WNV fluorescent signals, the maximum intensity projection (MIP) was obtained from all MIPs generated during the imaging process as a representation of each mouse brain. A mask was generated for each brain to remove background noise using U2-Net¹³⁵. Signals were segmented using triangle thresholding¹³⁶, and the number of signal pixels was divided by the total number of brain pixels.

Thick Section Immunolabeling

Delipidated brains were bisected and washed three times in PBS. Samples were embedded in 2% agarose prepared in 1X PBS and 100μm sagittal sections were cut using a vibratome. Sections were washed in PTxwH (3 × 10min) and incubated overnight with indicated antibody (WNV NS1: WNV-12; dsRNA: anti-mouse dsRNA antibody [Cell Signaling Technology, Cat. #76651]; GFP: pGt-GFP [Rockland, Cat. #600–101-215, Lot #35577]; CD31: mHm anti-CD31 [Millipore, Cat. #MAB1398Z, Lot #4259701, clone 2H8]) at a concentration of 1.25 μg/mL. Sections were washed in PTxwH and incubated for 6h with secondary antibodies (dsRNA: pGt-Ms-IgG2a FAB Alexa Fluor 647 [Jackson ImmunoResearch, Cat. #115–607-186, Lot #148384]; GFP: Alexa Fluor^® 647 AffiniPure^® Fab Fragment Goat Anti-Mouse IgG2a, Fcγ fragment specific [Jackson ImmunoResearch, Cat. #115–607-186, Lot #148384]; CD31: pGt anti-Hms–Alexa Fluor 594 (Jackson ImmunoResearch, Cat. #127–585-160, Lot #136531] each at a concentration of 1.25 μg/mL. Samples were washed in PTxwH and mounted for imaging. Samples for dsRNA imaging were additionally counterstained with SYTO^™ 24 Green Fluorescent Nucleic Acid Dye (ThermoFisher Scientific, Cat# S7559) prior to mounting. Images were acquired using a Keyence BZ-X810 fluorescence microscope with either a 4X objective (dsRNA) or a 20X objective (CD31/GFP). Z-stack images were acquired using the 20X objective across the thickness of each section with a Z-step size of 0.4 μm. Individual image planes were stitched to generate TIFF images/stacks. Maximum intensity projections were generated from Z-stacks using ImageJ and used for quantification. Fluorescence intensity was measured on each section following projection.

Evans Blue Dye Quantification

Mice previously treated with vehicle, poly(I:C), or infectious WNV were given 500μL 1% Evan’s Blue Dye (Sigma Aldrich, Cat# E2129–50G) diluted in 1X D-PBS by intraperitoneal injection. Animals were returned to cage for 1 hour to allow dye to disseminate, following which they were humanely euthanized and perfused with 1X D-PBS. Brains were harvested into 1mL formamide (EMD Millipore, Cat# S4117) with ~500 μL 1.0mm glass beads and homogenized as above described. Homogenates were incubated 24h at 56°C. Samples were spun at 13.2k rpm for 5 min to pellet tissue debris and supernatant was collected. Absorbance was measured at 610nm on Cytation 7 instrument.

QUANTIFICATION AND STATISTICAL ANALYSES

Bulk transcriptomics

RNA was isolated as described above and submitted to Admera Health (South Plainfield, New Jersey, USA). RNA was quantified by Qubit RNA HS assay (Thermofisher) and assessed for quality using Bioanalyzer 2100 Eukaryote Total RNA Nano (Agilent Tecnhologies). All RNA samples had a RIN score ≥ 8.0. Sequencing libraries were prepped using the polyA + NEBNext Ultra II Directional Kit (NEB Cat# E7760S) or the SMART-Seq v4 Ultra Low Input RNA Kit (Takara Cat# 634440) and analyzed for quality by Tapestation High Sensitivity D1000 Assay (Agilent Technologies). 150 bp paired-end sequencing was performed on an Illumina NovoSeq X Plus instrument. Quality of the reads was assessed using FastQC¹³⁷. Raw paired-end reads were adapter- and quality-trimmed using BBduk (BBMap suite v39.38) with k-mer–based 3′ adapter removal and bidirectional quality trimming and reads shorter than 51 bp after trimming were discarded prior to downstream analysis. Reads were aligned to the Mus musculus genome mm10 GCRm39 using STAR (v2.7.11)¹¹⁶. Mapped reads were summarized into genomic features using featureCounts (v 2.1.1). Count matrices were normalized, statistically analyzed, and differentially expressed genes between conditions were identified using the DESeq2 package (v1.46.0)¹²⁶ in R Studio (v4.4.2).

Bulk RNA-seq hierarchical clustering and gene ontology enrichment analyses

For hierarchical clustering of bulk RNA transcriptomics, counts were adjusted by regularized log transformation to minimize effects of lowly expressed genes. Only genes with an absolute log2 fold change greater than 1 and an adjusted p-value lower than 0.05 were considered for subsequent analyses. Heatmap visualization of variable genes was performed using the pheatmap package (v1.0.12)¹¹⁸. Gene expression was row normalized by z-score. Upset plots were generated using the UpsetR package (v1.4.0)¹¹⁹. Gene ontology (GO) enrichment analyses for biological processes (BP) of identified genes of interest was performed using the clusterProfiler package (v4.14.4)^124,125. We visualized GO results as enrichment maps where enriched GO terms are organized as nodes with edges connecting overlapping gene sets (GO terms, nodes), generating clusters of similar GO terms. For visualization purposes we grouped GO terms into clusters and labeled by cluster ID. Analyses and hierarchical clustering was performed in R Studio (v4.4.2).

Bulk RNA-seq predicted transcription factor activity analyses

Whole transcriptomics data of samples of interest was first normalized using rlog transformation in DESeq2¹²⁶ (v1.50.2) The DoRothEA package⁷⁴ (v1.22.0) was used to select transcription factor (TF) regulons of high confidence (A, B, C confidence) and the VIPER package (v1.44.0) ⁷³ was used to infer transcription factor activity based on the extracted gene expression matrices of genes of interest. Differential TF activity analyses was then performed using the limma package¹²⁰ (v3.66.0).

Single nuclei transcriptomics

Mice were euthanized after indicated treatment and time point and perfused with 30 mL PBS (Gibco) followed by harvest of 50 mg whole brain tissue sampled from the right cerebral cortex which was immediately snap frozen in liquid nitrogen. Whole frozen tissue was submitted to Admera Health (South Plainfield, New Jersey, USA). Tissue dissociation, nuclei extraction and quality control were performed according to the provider’s standardized protocols. Libraries were prepped following the manufacturer’s instructions according to the 10x Chromium Gem-X Single Cell 3′ GEM v4 Library Prep protocol (10x Genomics) and 150 bp paired-end sequencing was done on an Illumina NovoSeq X Plus instrument, targeting 20,000 nuclei per sample with 2–2.5 billion paired-end reads per sample. Reads were aligned to the Mus musculus genome mm10 GCRm39 using Cellranger (v9.0.1). Gene-barcode count matrices produced were processed using Seurat (v5.3.0)¹²⁷. Cells from individual samples were combined into a single dataset while preserving sample identity for downstream analyses. Quality control was performed excluding cells with low complexity; excessive transcript counts and high mitochondrial content. Potential doublets were identified on a per-sample basis using scDblFinder (v1.22)¹²¹, and only cells classified as singlets were retained. Data were normalized separately for each sample using the SCTransform framework, which applies a regularized negative binomial model to correct for technical variation. Samples were integrated using the SCT-based workflow in which reciprocal principal component analysis was used to align shared transcriptional states across datasets, reducing technical batch effects while preserving biologically meaningful variation. The integrated dataset was analyzed by principal component analysis (PCA) followed by graph-based clustering, and cellular relationships were visualized using Uniform Manifold Approximation and Projection (UMAP). Cluster-specific marker genes were identified using differential expression analysis in Seurat and used to annotate transcriptionally distinct cell populations. Cell types were manually annotated using a list of marker genes (Figure S6A) and corroborated by ScType¹²² using their pretrained models and scPred (v1.9.2)¹²³ using the Allen Brain Cell (ABC) Atlas as reference.

GO score analyses

Gene set expression analyses based on GO terms were performed using R Studio (v4.5.2). Genes belonging to the respective GO terms were were extracted using the GO.db package and the gene module expression scores were calculated via the AddModuleScore() feature of the Seurat package (v5.4.0). The module scores were then visualized using the ComplexHeatmap package (v2.26.0)¹¹⁷.

GO enrichment based on ranked DEGs

Differentially expressed genes (DEGs) per cell type and time point were identified using the FindMarkers() function of the Seurat package in R Studio. DEGs were identified based on log2 fold change (absolute change > 0.5), adjusted p value (< 0.05) and ranked by calculating a score, defined as log2 fold change multiplied by the negative log10-transformed adjusted p value. These lists were split in up- and down-regulated genes and GO enrichment was performed using the gseGO() function of the clusterProfiler package (v4.18.4) ¹³⁸ in R Studio (v4.5.2). For data visualization, simplified GO terms were plotted using the simplify() function with a cutoff of 0.7 of clusterProfiler.

snRNA-seq predicted transcription factor activity analyses

Pseudobulk expression data was extracted using the AggregateExpression() function of Seurat package (v5.4.0) in R Studio (v4.5.2) and analyzed using the DESeq2¹²⁶ package (v 1.50.2). The DoRothEA package⁷⁴ (v1.22.0) was used to select transcription factor (TF) regulons of high confidence (A, B, C confidence) and the VIPER package (v1.44.0)⁷³ was used to infer transcription factor activity based on the extracted gene expression matrices of genes of interest. Transcription factor activities of interest in cell types of interest were then visualized using the pheatmap package (v1.0.13).

CellChat Ligand-Receptor Communication Inference

Intercellular communication was inferred using the CellChat R package⁶⁶ (v2.2.0) in R Studio (v4.5.0). Log normalized snRNAseq counts and cell type annotations were used as input to construct cellchat objects for each condition. Communication probabilities were inferred by cellchat using expression levels of ligand and receptor genes. Significant interactions were aggregated at the pathway level to quantify both number and strength of interactions. For comparative analysis, cellchat objects from different conditions were merged and differential interactions were assessed based on total interaction counts and weights between conditions. Visualization of results was performed using built-in CellChat functions.

Table S1: Bulk RNA sequencing analyses of murine brain tissue isolated post footpad treatment with PBS, poly(I:C), LPS or ODN1826, related to Figure 3.

Table S2: Bulk RNA sequencing analyses of murine brain tissue isolated post intravenous treatment with IFN-α3, related to Figure 4.

Table S3: Single-nuclei RNA sequencing analyses of murine brain tissue isolated post intravenous treatment with IFN-α3, related to Figure 5.

Table S4: Bulk RNA sequencing analyses of murine brain tissue and brain microvascular endothelial cells isolated post footpad treatment with poly(I:C), related to Figure 6.

Table S5: Bulk RNA sequencing analyses of murine brain microvascular endothelial cells isolated post intravenous treatment with IFN-α3, related to Figure 6.

Video S1: Video of reconstructed mouse brains following intracranial WNV challenge post footpad administration of PBS, related to Figure 2.

Video S2: Video of reconstructed mouse brains following intracranial WNV challenge post footpad administration of poly(I:C), related to Figure 2.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-CD45 (clone 30-F11) BUV395	BD Biosciences	Cat#563279
anti-CD45 (clone 30-F11) AF700	Biolegend	Cat#103128
anti-CD11b (clone M1/70) FITC	Biolegend	Cat#101205
anti-Ly6G (clone 1A8) APC	Biolegend	Cat#127613
anti-NK1.1 (clone S17016D) APC/Cy7	Biolegend	Cat#156509
anti-NKp46 (clone 29A1.4) PE	Biolegend	Cat#137604
anti-Ly6C (clone HK1.4) BV605	Biolegend	Cat#128035
anti-Ly6C (clone HK1.4) PE/Dazzle	Biolegend	Cat#128043
anti-CD3 (clone 17A2) PE/Cy7	Biolegend	Cat#100220
anti-CD3 (clone 17A2) BUV737	BD Biosciences	Cat#612803
anti-CD19 (clone 6D5) BV510	Biolegend	Cat#115545
anti-F4/80 (clone BM8) BV421	Biolegend	Cat#123137
anti-F4/80 (clone BM8) AF647	Biolegend	Cat#123122
anti-CD64 (clone S8017D) PE/Fire744	Biolegend	Cat#161009
anti-P2RY12 (clone S16007D) PE	Biolegend	Cat#848003
anti-CD11c (clone N418) AF488	Biolegend	Cat#117311
anti-ITGA6 (clone GoH3) AF488	Biolegend	Cat#313608
anti-GFP (polyclonal)	Rockland	Cat#600-101-215
anti-CD31 (clone 2H8)	Sigma Aldrich	Cat#MAB1398Z
anti-WNV NS1 (clone WNV-121)	Laboratory of James E. Crowe Jr.113	N/A
anti-IFNAR1 (clone MAR1-5A3)	BioXCell	Cat#BE0241
anti-TNFα (clone XT3.11)	BioXCell	Cat#BE0058
anti-MCSF (clone 5A1)	BioXCell	Cat#BE0204
anti-HRP (clone HRPN)	BioXCell	Cat#BE0088
anti-CD45 (clone 30F11.1) MicroBeads	Miltenyi Biotec	Cat#130-052-301
anti-CD31 MicroBeads	Miltenyi Biotec	Cat#130-097-418
anti-NK1.1 (clone PK136)	Laboratory of Joseph C. Sun60	N/A
anti-dsRNA (clone J2)	Cell Signaling Technology	Cat#76651
AffiniPure Fab Fragment Goat anti-Human IgG (polyclonal) AF647	Jackson ImmunoResearch	Cat#109-607-003
AffiniPure Fab Fragment Goat anti-Mouse IgG2a (polyclonal) AF647	Jackson ImmunoResearch	Cat#115-607-186
AffiniPure Goat anti-Armenian Hamster IgG (polyclonal) AF594	Jackson ImmunoResearch	Cat# 127-585-160
Bacterial and virus strains
Virus: West Nile Virus, strain: NY99 (WNV)	Laboratory of Pei-Yong Shi114	N/A
Virus: Powassan Virus lineage II, strain: Spooner (POWV)	Laboratory of Charles M. Rice	N/A
Virus: Sindbis Virus, strain: SVN (SINV)	Laboratory of Charles M. Rice	N/A
Virus: Herpes Simplex Virus 1, strain: KOS (HSV-1)	Laboratory of David Leib	N/A
Biological samples
Chemicals, peptides, and recombinant proteins
TRIzol Reagent	Thermo Fisher	Cat#15596026
DAPI	Thermo Fisher	Cat#62248
LIVE/DEAD Fixable Blue dead Cell Stain Kit	Thermo Fisher	Cat#L23105
TruStain FcX	Biolegend	Cat#101320
polyinosinic-polycytidylic acid (polyIC)	Invivogen	Cat#tlrl-pic
Pam3CSK4	Invivogen	Cat#tlrl-pms
ODN1826	Invivogen	Cat#tlr-it1mw
Flagellin from S.typhimurium (FLA-ST)	Invivogen	Cat#tlrl-stfla
R848	Invivogen	Cat#tlr-r848
Lipopolysaccharide (LPS)	Sigma Aldrich	Cat#L2880
FSL1	Invivogen	Cat#tlrl-fsl
ODN1826	Invivogen	Cat#tlrl1826
LPS-EK	Invivogen	Cat#tlrl-eklps
RBC Lysis Buffer	Biolegend	Cat#420301
IFN-α3	PBL Assay Science	Cat#12100-1
IFN-β	PBL Assay Science	Cat#12401-1
Human Fibronectin (Fragment)	Miltenyi Biotec	Cat#130-109-393
EBM-2 Basal Medium	Lonza	Cat#CC-3202
FITC-Dextran	Sigma Aldrich	Cat#FD150S
Evan’s Blue Dye	Sigma Aldrich	Cat#E2129-50G
Diphtheria Toxin	Sigma Aldrich	Cat#D0564
cOMPLETE ETDA-free Protease Inhibitor	Sigma Aldrich	Cat#4693132001
SYTO Green Fluorescent Nucleic Acid Stain	Sigma Aldrich	Cat#S7559
Critical commercial assays
PowerUp™ SYBR™ Green Master Mix	Thermo Fisher	Cat#25741
RevertAid First Strand cDNA Synthesis Kit	Thermo Fisher	Cat#K1622
High-Capacity cDNA Reverse Transcription Kit	Thermo Fisher	Cat#$4368814
Ambion PureLink RNA Mini Kit	Thermo Fisher	Cat#12183025
mMESSAGE mMACHINE T7 Transcription Kit	Thermo Fisher	Cat#AM1344
M.O.M. (Mouse on Mouse) Blocking Reagent	VectorLabs	Cat#MKB-2213-1
Adult Mouse Brain Dissociation Kit	Miltenyi Biotec	Cat#130-107-677
GentleMACS C Tubes	Miltenyi Biotec	Cat#130-093-237
LS Columns	Miltenyi Biotec	Cat#130-122-729
LD Columns	Miltenyi Biotec	Cat#130-042-901
GentleMACS Octo Dissociator with Heaters	Miltenyi Biotec	Cat#130-107-677
SMART-Seq v4 Ultra Low Input RNA Kit	Takara	Cat#634440
NEBNext Ultra II Directional Kit	NEB	Cat#E7760S
Deposited data
Raw data bulk RNA-seq (Figure 3)	This paper	GEO: GSE318728
Raw data bulk RNA-seq (Figure 4)	This paper	GEO: GSE31871
Raw data sn-RNA-seq (Figure 5)	This paper	GEO: GSE318766
Raw data bulk RNA-seq (Figure 6)	This paper	GEO: GSE318729
Raw data bulk RNA-seq (Figure 6)	This paper	GEO: GSE318714
Experimental models: Cell lines
Cell line: Vero E6	ATCC	Cat#CRL-1586
Cell line: BHK-21	ATCC	Cat#CCL-10
Cell line: Huh-7.5	Laboratory of Charles M. Rice115	N/A
Experimental models: Organisms/strains
Mouse: C57Bl/6J	The Jackson Laboratory	Cat#000664
Mouse: Ifnar1−/−	The Jackson Laboratory	Cat#028288
Mouse: Ifnar1fl/fl	The Jackson Laboratory	Cat#028256
Mouse: Ccr2-DTR	Laboratory of Tobias M. Hohl61	N/A
Oligonucleotides
qPCR primer: Eef1a1 fwd: 5′- GCAAAAACGACCCACCAATG-3′	IDT, Lercher et al.81	N/A
qPCR primer: Eef1a1 rev: 5′- GGCCTGGATGGTTCAGGATA-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Ifit1 fwd: 5′- TTACAGCAACCATGGGAGAGAATG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Ifit1 rev: 5′- GGAACTGGACCTGCTCTGAGATTC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Ifitm3 fwd: 5′- GCCTACTCCGTGAAGTCTAGGG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Ifitm3 rev: 5′- CCAAGGTGCTGATGTTCAGGC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Bst2 fwd: 5′- TGTAGAGACGGGTTGCGAGC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Bst2 rev: 5′- CTCCTGAAGGGTCACCACGG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Isg15 fwd: 5′- ACTGTGAGAGCAAGCAGCCA-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Isg15 rev: 5′- TTAGGTCCCAGGCCATTGCT-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Mx1 fwd: 5′- AGGCCATTAACAAGGGTCAGAAC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Mx1 rev: 5′- GGTCTGGGACATTTGGGGAG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Rsad2 fwd: 5′- CCAAGACATCCTTCGTGCTGC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Rsad2 rev: 5′- AAGTTGATCTTCTCCAAACCAGCC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Il6 fwd: 5′- CGGCTTCAGGAAGGTGATGAA-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Il6 rev: 5′-TCACTGTCAGAGTCAGCAGC- 3′	IDT, Lercher et al.81	N/A
qPCR primer: Tnfa fwd: 5′- GCCCACGTCGTAGCAAACC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Tnfa rev: 5′- TTTGAGATCCATGCCGTTGGC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Cxcl1 fwd: 5′- CCACTGCACCCAAACCGAAG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Cxcl1 rev: 5′- CCGTTACTTGGGGACACCTTTTAG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Gapdh fwd: 5′- GGCAAATTCAACGGCACAGT-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Gapdh rev: 5′- AGATGGTGATGGGCTTCCC-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: S100a8 fwd: 5′- TCCTTTGTCAGCTCCGTCTTCA-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: S100a8 rev: 5′- ACGAAAGATTTCCTTTCAAACGATG-3 ’	IDT, Lercher et al.81	N/A
qPCR primer: Sod2 fwd: 5′- GAAGTTCAATGGTGGGGGACA-3′	IDT, Lercher et al.81	N/A
qPCR primer: Sod2 rev: 5′- CCTCCAGCAACTCTCCTTTGGG-3′	IDT, Lercher et al.81	N/A
qPCR primer: eGFP fwd: 5′- ACGTAAACGGCCACAAGTTC-3′	IDT	N/A
qPCR primer: eGFP rev: 5′- AAGTCGTGCTTCATGTG- 3′	IDT	N/A
qPCR primer: HBV C fwd: 5′- CGTTTTTGCCTTCTGACTTCTTTC-3 ’	IDT, Yu et al.62	N/A
qPCR primer: HBV C rev: 5′- CTGAGTGCAGTATGGTGAGGTG-3 ’	IDT, Yu et al.62	N/A
qPCR primer: WNV NS3 fwd: 5′- GGAACATCAGGCTCACCAATAG-3 ’	IDT	N/A
qPCR primer: WNV NS3 rev: 5′- CATCCTTTCACCCTGCACTATC-3 ’	IDT	N/A
Recombinant DNA
Plasmid kiCAP-AAV-BI30	Addgene, Krolak et al.78	Cat#183749
Plasmid pAAV-CAG-SV40NLSf-GFP-3xmiR122-WPRE-HGHpA	Addgene, Krolak et al.78	Cat#183775
Plasmid pAAV-CAG-SV40NLSf-Cre-3xmiR122-WPRE-HGHpA	Addgene, Krolak et al.78	Cat#183776
Software and algorithms
FlowJo version 10.10.0	BD Biosciences	https://www.flowjo.com
SpectroFlo versions 3.0 to 4.0	Cytek Biosciences	https://cytekbio.com/pages/spectro-flo
Graphpad Prism versions 9 to 10	Graphpad	https://www.graphpad.com
R versions 4.1 to 4.5	CRAN	https://www.r-project.org
RStudio	Posit.co	https://posit.co/download/rstudio-desktop/
STAR version 2.7.11	Dobin et. al.116	https://github.com/alexdobin/STAR
ComplexHeatmap package version 2.26	Gu et al.117	https://bioconductor.org/packages/release/bioc/html/ComplexHeatmap.html
CellChat package version 2.2	Jin et al.66	https://github.com/sqjin/CellChat
pheatmap package version 1.0.12	Kolde118	https://cran.r-project.org/web/packages/pheatmap/index.html
UpsetR package version 1.4	Gehlenborg119	https://cran.r-project.org/web/packages/UpSetR/index.html
DoRothEA package version 1.22	Garcia-Alonso et al.74	https://bioconductor.org/packages/release/data/experiment/html/dorothea.html
VIPER package version 1.44	Alvarez et al.73	https://bioconductor.posit.co/packages/3.19/bioc/html/viper.html
limma package version 3.66	Ritchie et al.120	https://bioconductor.org/packages/release/bioc/html/limma.html
scDblFinder package version 1.22	Germain et. al.121	https://bioconductor.org/packages/release/bioc/html/scDblFinder.html
ScType	Ianevski et. al.122	https://sctype.app/
scPred package version 1.9.2	Alquicira-Hernandez et al.123	https://github.com/powellgenomicslab/scPred
ABC Atlan - Allan Brain Map	The Allen Institute	https://brain-map.org
ClusterProfiler package version 4.14.4	Wu et al.124 Xu et al.125	https://github.com/YuLab-SMU/clusterProfiler
DESeq2 package version 1.46	Love et al.126	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Seurat package version 4.5 to 5.3	Hao et al.127	https://github.com/satijalab/seurat
Cell Ranger version 9.0.1	10x Genomics	https://support.10xgenomics.com/single-cell-multiome-atac-gex/software/downloads/latest
Other

Highlights.

• Peripheral viral pattern sensing primes antiviral defenses in the CNS

• Serum IFN-α is necessary and sufficient for establishment of this antiviral state

• Brain endothelial cells of the CNS are the primary responders to systemic IFN-α

• IFNAR1 signaling in brain endothelial cells is necessary for neuroprotection