Vero E6 cells were obtained from the American Type Culture Collection (ATCC) and maintained in Dulbecco’s modified Eagle’s medium (DMEM, Thermo Fisher) supplemented with 10% fetal bovine serum (FBS, ATCC) supplemented with 100 units/mL penicillin and 100 μg/mL streptomycin (Thermo Fisher) at 37°C in 5% CO₂. Venezuelan equine encephalitis virus (VEEV) strain TC-83 (NR-63) was obtained from the NIH Biodefense and Emerging Infections Research Resources Repository, NIAID, NIH. VEEV stocks were propagated in Vero E6 cells and harvested via clarification of cell culture supernatant by centrifugation. Titers of viral stocks were determined by standard plaque assay consisting of a methyl crystalline cellulose overlay and crystal violet staining (16).

All animal work was approved by the Lawrence Livermore National Laboratory Institutional Animal Care and Use Committee under protocol #310. All animals were housed in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited facility. Animals are tested on a quarterly basis for pathogens chosen based on recommendations from the FELASA Working Group on Health Monitoring and guidelines provided from AALAS experts. C3H/HeN mice (strain ID 025) were obtained from Charles River. 5–10-week-old female mice were used in all experiments. Groups of mice were inoculated intranasally with 2x10⁷ PFU TC-83 while under anesthesia (4-5% isoflurane in 100% oxygen). Mice were monitored daily for signs of morbidity and animals were humanely euthanized upon signs of severe disease by CO₂ asphyxiation. For tissue harvest, animals were anesthetized under isoflurane and the whole animal was perfused with 20 mL sterile PBS containing 50,000 U/L sodium heparin via the left ventricle.

Brains for flow cytometric, scRNAseq, cytokine, and viral titer analysis were isolated from infected mice on days 2, 4, or 6 post-infection. Preparation of brain tissue was performed as previously described (17). Following euthanasia and perfusion, brains were removed and placed in 1 mL digestion buffer (PBS pH 7.4 (Thermo Fisher) + collagenase (Worthington) + DNase I (Roche) to a final concentration of 3 mg/mL and 0.5 mg/mL, respectively) on ice in a 1.5 mL tube. Brains were finely diced into 1-2 mm³ pieces with scissors and tissue was digested at 37°C for 30 min in a total of 5 mL digestion buffer. A cell suspension was generated by gentle pipetting followed by passage through a 70 μm cell strainer. The cell strainer was rinsed with PBS supplemented with 5% FBS to a total volume of 20 mL. Aliquots of this suspension were stored at -80°C for cytokine and viral titer analysis. The remaining suspension was subjected to Percoll gradient centrifugation to purify mononuclear immune cells for flow cytometric analysis or RNA sequencing as previously described (17). For flow cytometry experiments, cells were stained and analyzed from individual brains. For RNA sequencing, brains from 3 animals were pooled and sequenced as a single sample for each condition.

Spleens were harvested from uninfected mice and mice at 6-days post-infection following cardiac perfusion with PBS supplemented with heparin, as described above, and placed in 1 mL of DMEM supplemented with 2% FBS. Tissue was mechanically disrupted with a plunger through a 40-micron filter over a 50 mL conical tube, followed by two washes of PBS supplemented with 2% FBS. Suspensions were pelleted by centrifugation and red blood cells were lysed using ACK Lysis buffer (Thermo Fisher). Cells were washed twice and processed for flow cytometry as described below.

Single-cell suspensions of mononuclear immune cells from uninfected mouse brains and infected brains were prepared as described above. Cells were counted on a Countess II automated cell counter prior to single-cell sequencing preparation using Chromium Single-cell 3′ GEM, Library & Gel Bead Kit v3 (10x Genomics Cat # 1000075) on a 10× Genomics Chromium Controller following manufacturers’ protocol. Subsequently, libraries were sequenced on Illumina NextSeq 2000. The Cell Ranger Single-Cell Software Suite (version 6.1.2) (10x Genomics) was then used to perform sample demultiplexing, barcode processing, and single-cell gene counting. After demultiplexing the sequencing files with “cellranger mkfastq”, “cellranger count” was used to perform alignment to mouse reference transcriptome (mm10), barcode processing and gene counting. Further analysis was performed using Seurat (18). First, cells with fewer than 500 detected genes per cell or mitochondrial content greater than 10% and genes that were expressed by fewer than 5 cells were filtered out. Potential doublets and CD45^- cells were also removed. After pre-processing, we performed data normalization using default parameters and identified 2000 most variable features. Then, anchors for data integration were identified using the ‘FindIntegrationAnchors()’ function using default parameters. Next, these anchors were passed to the ‘IntegrateData()’ function and a new integrated matrix with all four datasets was generated. After data integration, data was scaled using ScaleData() function with default parameters, and the dimensionality of the data was reduced by principal component analysis (PCA). Subsequently, cells were grouped into an optimal number of clusters for de novo cell type discovery using Seurat’s ‘FindNeighbors()’ and ‘FindClusters()’ functions (dims = 1:30, resolution = 0.2). A non-linear dimensional reduction was then performed via uniform manifold approximation and projection (UMAP) and various cell clusters were identified and visualized. Genes differentially expressed between clusters were identified using ‘FindMarkers()’ function implemented in Seurat. Gene ontology (GO) enrichment analysis was performed using ToppGene Suite (19) and heatmaps were generated using custom R scripts. Single-cell pseudo-time trajectories of immune cell subpopulations were constructed with Monocle 2 (20) as described before (21).

PBMCs were isolated from whole blood collected from uninfected mice and mice at 6-days post-infection via cardiac puncture. Blood was combined with 0.5 M EDTA for a final concentration of 5 mM then red blood cells were removed by incubation with ACK lysis buffer. PBMCs were then washed twice with PBS. Sequencing libraries were generated using Chromium Single-cell 3′ GEM, Library & Gel Bead Kit v3 (10x Genomics Cat # 1000075) on a 10× Genomics Chromium Controller following manufacturers’ protocol. Subsequently, libraries were sequenced on Illumina NextSeq 2000. Sequencing files were demultiplexed using “cellranger mkfastq”. Subsequently, “cellranger count” was used to perform alignment to mouse reference transcriptome (mm10), barcode processing and gene counting. Further analysis was performed using Seurat (18) as described above.

We used publicly available data (22) to identify VEEV infection-induced changes in different regions of the brain. Raw counts were downloaded from Gene Expression Omnibus (GEO, accession ID: GSE213725). Data was then normalized using the Trimmed Mean of M-values (TMM) normalization method implemented in edgeR (version 3.42.2) (23). For differential expression analysis, the limma package (24) was employed in conjunction with the voom transformation (25). Heatmaps for genes of interest were generated using the pheatmap package (26), and a boxplot was created using the ggplot2 package (27), both in R (version 4.3.1) (28).

Cells were incubated for 30 min on ice in 100 μl Hank’s balanced salt solution (Thermo Fisher) + 2% FBS with Fc block (1:100 dilution, clone 2.4G2; BD Biosciences), Zombie Yellow viability dye (1:500 dilution, BioLegend), and the following antibodies, each diluted 1:500: CD45 APC-Cy7 (clone 30-F11; BD Biosciences), CD11b PE-CF594 (clone M1/70; BD Biosciences), CD3 PerCP-Cy5.5 (clone 17A2; BD Biosciences), CD4 BV650 (clone GK1.5; BD Biosciences), CD8 FITC (53-6.7; BioLegend), CD49b PE (clone DX5; BD Biosciences), and Perforin PE/Dazzle 594 (clone S16009A; BioLegend). Cells were then fixed using BD Cytofix/Cytoperm (BD Biosciences) according to manufacturer’s instructions. Flow cytometry was performed using a FACSAria Fusion and data were analyzed using FlowJo software (version 10.10.0). Microglia and infiltrating myeloid and lymphoid cells were identified as preciously described (17). Microglia were identified as CD45^int CD11b^int. To distinguish CD45^hi CD11b⁺ myeloid cells from NK cells that also express CD11b, we identified these cells as CD45^hi CD11b⁺ CD49b^-. The remaining cells were labeled as lymphocytes. Ly6C expression was also confirmed in myeloid cells. Gating was further informed by fluorescence minus one (FMO) controls previously performed during panel optimization.

Cytokines were quantified using LEGENDplex™ multiplex bead-based assay (BioLegend) using the mouse anti-virus response panel according to manufacturer’s instructions. Flow cytometry of the beads was performed using a FACSAria Fusion and data were analyzed using BioLegend’s cloud-based analysis software available at https://legendplex.qognit.com (version 2023-02-15).

Brains for immunohistochemical staining were collected from infected mice at 6 dpi. Upon euthanasia and perfusion as described above, mice were perfused with 20 mL 10% neutral buffered formalin (NBF). Isolated brains were subsequently immersed in 10% NBF at 4°C for 3 days with gentle agitation. Following fixation, brains were paraffin-embedded and cut into 5 μm thick sections. All sections were dewaxed with xylene and hydrated with alcohol. Citrate or Tris-EDTA were used for antigen retrieval, and hydrogen peroxide (ab64218, Abcam) was used to block endogenous peroxidase. After blocking non-specific sites with CAS-block (008120, Thermo Fisher Scientific), sections were incubated with Ly6c primary antibody diluted 1:200 (ab314120, Abcam) and secondary antibody diluted 1:500 (ab6720, Abcam). 3,3′-diaminobenzidine (DAB) kit (ab64238, Abcam) was used for visualization, and hematoxylin was used to stain the nuclei. All sections were rinsed with distilled water and sealed with permount (sp15-100, Fisher Scientific).

Spatial transcriptomic analysis of a VEEV-infected brain collected 6 days post-infection was conducted using 10x Visium technology (10x Genomics). 5 µm brain sections were mounted onto charged slides, deparaffinized, and stained with hematoxylin and eosin. The brain sections were imaged at 4x magnification using an ECHO Revolve microscope and the tiles were stitched together using Affinity Photo (version 2.3, Serif). The brain sections were then destained and decrosslinked to release RNA sequestered by formalin fixation. Next, the sections were incubated with mouse-specific probes targeting the whole transcriptome (PN-1000365, 10x Genomics), allowing for the hybridization and ligation of each probe pair. Gene expression probes were released from the tissue and captured by spatially barcoded oligonucleotides on the Visium slide surface using the Visium CytAssist instrument. Gene expression libraries were then prepared from each tissue section and sequenced using the Illumina NextSeq 2000.

After sequencing, the data and histology images were processed with the Space Ranger software (10x Genomics) and Seurat (18). After demultiplexing the sequencing files with “spaceranger mkfastq”, the fastq files and the brightfield image of the tissue were provided to “spaceranger count” and read alignment, tissue detection, fiducial detection, and barcode/UMI counting were performed. Seurat was used then used for filtering (nFeature_Spatial > 500 & nCount_Spatial > 500 & percent.mt < 30), data normalization, dimensionality reduction, clustering, identification of spatially variable genes, and visualization. For each sample, after filtering, the data was normalized using “SCTransform()” function implemented in Seurat, with default parameters. Dimensionality reduction and clustering analysis was performed using “RunPCA()”, “FindNeighbors()”, “FindClusters ()” and “RunUMAP()” functions (dims = 1:30, resolution=0.25). We then applied the “FindAllMarkers()” function to identify genes enriched in each cluster. We also performed integrative analysis of uninfected and infected brain samples using the “SelectIntegrationFeatures(nfeatures = 5000)”, “PrepSCTIntegration()”, “FindIntegrationAnchors()”, and “IntegrateData ()” functions, followed by dimensionality reduction using PCA and UMAP. Clusters were visualized in UMAP space using DimPlot() and overlaid on the tissue image using “SpatialDimPlot()” function.

Statistical significance was determined using Prism Version 10.2.3 (347) (GraphPad, La Jolla, CA). Experiments were performed at least twice and corresponding n, specific tests and p values are indicated in the figure legends. Assumptions of normality and variance were evaluated within Prism.

To understand the progression of the neuroinflammatory immune response to VEEV, we profiled the fluctuations of immune cell populations in the brain and their transcriptional activity using scRNAseq on immune cells isolated from mouse brains over a time course of VEEV infection. We first confirmed infection dynamics in an established mouse model of VEEV. C3H mice were infected intranasally with 2x10⁷ PFU VEEV TC-83 and monitored for signs of disease over six days, encompassing both the early lymphoid phase and central nervous system phase of the typically biphasic disease observed in mice (29, 30). Signs of disease including ruffled fur and hunched posture were observed in some mice by 4 days post infection (dpi) and by 6 dpi all mice exhibited these symptoms in addition to ataxia, a sign of neurological disease ( Figure 1A ). Weights of infected mice began to decline by 5 dpi and were significantly different versus uninfected controls by 6 dpi ( Figure 1B ). At various timepoints post infection, additional groups of mice were euthanized to measure brain viral load. High infectious viral titers in the brain were detected by 2 dpi and peaked by 4 dpi ( Figure 1C ). Histological examination of infected brains was performed at 6 dpi. Hematoxylin and eosin-stained coronal sections of infected brains revealed increased cellularity, edema, and perivascular cuffing ( Figures 1D, E ). Our results closely recapitulated previously characterized disease progression and pathological characteristics in this model of VEEV infection (30).

C3H mice were then infected as described above and at 2, 4, and 6 dpi, brain cerebral hemispheres from infected mice and uninfected controls were harvested and processed to perform scRNAseq or spatial transcriptomic analyses ( Figure 1F ). For scRNAseq, single-cell suspensions from 3 mice were pooled and mononuclear immune cells were isolated using Percoll gradient centrifugation, as previously described (17). The following number of cells were profiled for each condition: Uninfected: 2,298, TC-83 2 dpi: 1,698, TC-83 4 dpi: 5,005, and TC-83 6 dpi: 6,488. Unsupervised clustering of the data resulted in 9 clusters that were each assigned to a putative cell-type identity based on their unique profile of differentially expressed genes (DEGs) encoding cell-type specific markers ( Figures 2A, B ). Brains at each timepoint post-infection exhibited unique immune compositions as compared to the uninfected control ( Figure 2C ). In uninfected brains, microglia were the predominant immune population detected, as expected. Microglia formed 3 subclusters over the course of infection that we termed resting microglia (MG), activated MG 1, and activated MG 2 (microglia 1, 2, and 3). Activated MG 1 and 2 are representative of a deviation from a homeostatic cell state upon infection. Resting MG compose 88% of total immune cells in uninfected mice and are the most enriched in microglia homeostatic markers Cx3cr1 and Tmem119, which are downregulated in activated MG 1 and activated MG 2 ( Figure 2B ; Supplementary Figure 1A ). At 2 dpi, activated MG 1 and 2 increased in relative abundance (16% and 3%, respectively) and exhibit enrichment in additional genes indicative of an activated state, including Cd63, Cd72, and Mif ( Figures 2B, D ). Activated MG 1 continued to increase in relative abundance through 4 dpi, detected at a comparable level to resting MG, 16% and 17%, respectively, while activated MG 2 remained at approximately 3%. At 6 dpi, activated MG 1 decreased to 3% and activated MG 2 to less than 1%. Distinct antiviral response gene expression was observed across the microglia subclusters, with activated MG 1 exhibiting the highest expression of viral RNA sensors Ifih1, which encodes melanoma differentiation-associated protein 5 (MDA5) and Ddx58, which encodes retinoic acid-inducible gene I (RIG-I), as well as type I IFN gene Ifnb1 ( Supplementary Figure 1A ). Following the early shifts in microglial populations, we observed an immense infiltration of myeloid and lymphocyte populations that together comprise a larger proportion of sequenced cells at day 4 and 6 post infection than microglia ( Figures 2C, D ). We additionally confirmed this pattern of myeloid and lymphocyte infiltration using flow cytometry ( Figures 2E, F ).

Between uninfected and 2 dpi, the most notable shifts were those in the microglia subclusters, as described earlier, as well as detection of cluster 0, identified as myeloid cells, cluster 2, identified as NK cells, and cluster 6, identified as proliferating NK and T cells. Between 2 and 4 dpi, there was a large increase in myeloid cells (4% to 34%) and NK cells (4% to 16%) and a modest increase in cluster 5, identified as dendritic cells (DC) (3% to 7%). We also identified a small neutrophil cluster which accounted for less than 0.5% of cells at 4 and 6 dpi. T cells accounted for ~3% of the immune cells present in the brain at 4 dpi. Between 4 and 6 dpi, the myeloid and dendritic cell clusters remain relatively steady (40% and 7%, respectively) while there is an increase in NK cells (16% to 24%) and cluster 4, identified as T cells (3% to 18%) ( Figures 2D, E ). Notably, it is between day 4 and 6 during which neurological symptoms and severe brain pathology present and weight loss and survival begin to decline.

Williams et al. recently investigated the dynamics of the host response to VEEV infection in the mouse brain using bulk RNA sequencing (RNAseq) and histological techniques (22). We reanalyzed the bulk RNA-seq data obtained from three different regions of the brain: 1) the main olfactory bulb (MOB) where robust expression of viral protein was detected as early as 1 dpi, 2) piriform cortex (PIR), a region with robust viral protein expression by 2 dpi and 3) hippocampus (HIP), a region showing strong viral protein expression by 4 dpi (22). This data showed that Ifih1 and Ddx58 were upregulated immediately after infection in all three regions, which was followed by upregulation of type I IFN genes including Ifna2, Ifna4, Ifna5 and Ifnb1 ( Supplementary Figure 1B ). Interferon expression decreased over time in all three regions. Consistent with the decrease in resting microglia observed in our scRNAseq data, a dramatic decrease in the expression of microglia markers Cx3cr1, Tmem119, P2ry12 and Fcrls was observed while several markers of infiltrating myeloid cells including Ly6c2 and Plac8 were upregulated ( Supplementary Figure 1B ). In addition, a significant increase in the expression of NK and T cell markers over time was also observed in all three regions examined. In alignment with our data, an increase in NK cell markers was observed between 3 and 7 dpi, followed by a significant increase in T cell markers between 5 dpi and 7 dpi ( Supplementary Figure 1B ). Consistent with the lack of B cells and few neutrophils observed in our scRNAseq data, the expression of B cell and neutrophil markers was extremely low, as shown in the MOB at 6 dpi ( Supplementary Figure 1C ). We then reanalyzed the expression of interferon and markers of the key infiltrating populations in all eight parts of the brain isolated by Williams et al., (MOB, PIR, striatum (STR), motor cortex (MTX), HIP, sensory cortex (STX), thalamus (THA), and cerebral cortex (CBX). High expression of type I IFN genes was observed earliest in the MOB and PIR, at 3 dpi, followed by STR, MTX, HIP, STX, and THA across 5 and 6 dpi ( Supplementary Figure 1D ). Ly6c2 was upregulated in the MOB from 3 through 7 dpi whereas it was upregulated from 5 to 7 dpi in the PIR, STR, MTX, HIP, STX, and THA. NK cell and T cell markers were upregulated highest from 5 to 7 dpi in the MOB, and PIR, and at 7 dpi in the MTX, HIP, STX, and THA. Expression of these markers was least changed in the CBX across all timepoints. The early expression of viral sensing genes and type I IFN genes in the bulk RNAseq data set preceding high expression of markers of infiltrating cells coincides with our scRNAseq data that implicates microglia as a key early responder to VEEV infection and early producer of interferon. Both data sets then indicate a sequential recruitment of myeloid cells followed by lymphocytes at later timepoints. The bulk RNAseq data set demonstrates that in addition to temporal differences, there are unique spatial signatures of gene expression that can potentially help define the progression of the immune response through specific regions of the brain over time and the extent of infiltration in each of these regions.

To further examine the transcriptional activity of microglia and infiltrating myeloid populations, all cells from clusters 0, 1, 3, 5, 7, and 8 ( Figure 2A ) were extracted and re-clustered using an unsupervised clustering approach. This included the following number of cells for each condition: Uninfected: 2,211, TC-83 2 dpi: 1,527, TC-83 4 dpi: 3,835, and TC-83 6 dpi: 3,398. Analysis revealed 10 unique cell clusters ( Figures 3A–C ). Microglial subpopulations (cluster 0; resting MG, cluster 3; activated MG 1, and cluster 5; activated MG 2) clustered apart from various infiltrating myeloid populations and exhibited enrichment for Cx3cr1, Trem2, and Tmem119, as well as complement component genes C1qa-c. Cluster 1 and 2 highly expressed monocyte/macrophage markers Ly6c2 and Plac8, while lacking Tmem119 and were, therefore, annotated as ‘monocyte/macrophage (Mono/Mac) 1’ and ‘Mono/Mac 2,’ respectively ( Figure 3C ; Supplementary Figure 2A ) (31, 32). Clusters 4, 7, and 8 were identified as Cd209a⁺, plasmacytoid, and Ccr7⁺ DCs, respectively, based on enrichment of specific genes such as Cd209a and class II MHC genes Cd74, H2-Aa and H2-Ab1 (cluster 4), Klk1 and Mctp2 (cluster 7) and Ccr7 (cluster 8) ( Figure 3C ; Supplementary Figure 2 ) (33). Cluster 9 expressed neutrophil markers S100a8 and S100a9 ( Figure 3C ).

To understand the function of the myeloid clusters, we analyzed expression of key antiviral response genes. Type I IFN genes Ifnb1 and Ifna2 were primarily expressed by microglia, specifically activated MG 1 ( Figure 3D ). Elevated levels of interferon stimulated genes were observed across other myeloid subclusters, mainly Mono/Mac 1 and 2 and Cd209a⁺ DCs as early as 2 dpi, but more notably at 4 dpi and through 6 dpi ( Figures 3D, E ). Activated MG 1 were also enriched for inflammatory cytokines Ccl3 and Ccl4, Csf1, a key regulator of macrophage differentiation, tumor necrosis factor alpha (Tnf), and Il12b, a cytokine that acts on T and natural killer cells ( Figure 3G ). Ifnb1 and Csf1 elevation was observed by 2 dpi and through 6 dpi, in a small (< 20) percent of total cells ( Figure 3F ), while the other cytokines were observed highly expressed by greater percentages of total cells and beginning at 4 dpi, through 6 dpi. Activated MG 2 exhibited elevated Ifna2 expression, though lower than activated MG 1 and not until 6 dpi ( Figure 3D ; Supplementary Figure 1A ), and were enriched for Ccl5, Ccl7, Ccl12, Cxcl16, and Cxcl13 ( Figure 3G ). Mono/Mac 1 shared overlap with activated MG 2 with enrichment for Ccl5, Ccl7, and Cxcl16, and additionally expressed high levels of Ccl8. Mono/Mac 2 expressed high levels of chemokines including Ccl2, Ccl6, Ccl9, Cxcl1, Cxcl2 and Cxcl10, as well as Il15 and Il18, two interleukins implicated in activation of T/NK cells ( Figure 3G ) (34–36). Mono/Mac 1 also exhibited elevated expression of Il15 and Il18, but to a lesser extent. Notably, Cd209a⁺ DCs had high expression of Cxcl9, a prominent mediator of lymphocyte infiltration ( Figure 3G ). We validated our transcriptomic data by measuring protein levels of a subset of cytokines and chemokines included in our analysis in brain homogenate isolated from infected mice. Consistent with the observed increase in gene expression, IL-1β, IL-6, CXCL1, CXCL2, CXCL10, CCL5 and IFN-β were upregulated at the protein level in infected brains ( Figure 3H ).

To investigate the relationship between myeloid subclusters, we performed trajectory analysis, ordering single cells along pseudotime to reconstruct potential differentiation trajectories ( Supplementary Figure 2B ). Mono/Mac and microglia subpopulations clustered distinctly through each timepoint. Microglia clusters exhibited a shift from resting to activated subpopulations. By 4 dpi, activated MG 1 partially occupy a shared branch with Mono/Mac clusters. Meanwhile, activated MG 2 and proliferating cells occupy distinct branches. Interestingly, only a small portion of activated MG 1 highly express Ifnb1 at each timepoint ( Supplementary Figure 2C ).

Having established that microglia subclusters and infiltrating myeloid cells display distinct inflammatory profiles in response to VEEV infection, we next asked whether the spatial distribution of these immune cell subtypes also differ in the brain parenchyma using spatial transcriptomic analysis performed at 6 dpi. Unsupervised clustering of the spatial transcriptomic data from uninfected and infected brains resulted in eleven clusters. The majority of these clusters corresponded to specific anatomical regions in the brain ( Figures 4A–C ). Cluster 2 showed enrichment for myeloid markers such as Ly6c2 and Plac8 and was localized at the periphery of the cortex and within the hippocampal region ( Figures 4C–F ). The uninfected brain had significantly fewer cells from cluster 2 compared to the infected brain at 6 dpi ( Figures 4E, F ). Examination of spatial expression patterns of Ly6c2 and Plac8 further revealed a distinct localization of myeloid cells expressing these genes in the periphery of the cortex and within the hippocampal region of infected brains ( Figure 4G ). Localized Ly6c was confirmed in these regions via immunohistochemical analysis ( Figure 4I ). Minimal Ly6c expression was noted in the uninfected brain and robust signal was detected in the infected brain concentrated around the hippocampal formation ( Figure 4I ). When compared to the localization patterns of Prox1, which appears enriched at the hippocampal formation, and Nrgn1, within the cortex, the expression of the myeloid-specific markers appears to distinctly localize at a layer in between the Prox1- and Nrgn1-high regions, within the hippocampal region ( Supplementary Figure 3D , Figure 4G ). Our re-analysis of the bulk RNAseq data set generated by Williams et al. also indicated greater immune infiltration, including myeloid populations, within the hippocampus compared to the cortex ( Supplementary Figure 1D ). Expression of several upregulated chemokines in the infected brain including Ccl2, Ccl4, Ccl5, Ccl7 also localized to regions enriched in myeloid cells whereas Tnf and interleukins including Il6, Il12b, Il15 and Il18 were detected throughout the brain ( Figure 4C ; Supplementary Figures 3A, B ). Interestingly, Cxcl10 had a more robust expression throughout regions of the infected brain where we did not observe high Ly6c, suggesting that Cxcl10 is primarily expressed by cells other than the localized myeloid subpopulation ( Supplementary Figure 3C ). Consistent with the scRNAseq data, there was a significant reduction in cells expressing markers of homeostatic microglia such as Cx3cr1 and Tmem119 and an increase in cells expressing Cd72, a gene enriched in activated microglia, in infected brains ( Figure 4H ). These data indicate that by 6 dpi, there is widespread activation of microglia and regional specificity in infiltrating myeloid subpopulations.

To gain a more comprehensive understanding of lymphocyte responses to infection, all cells from lymphocyte clusters (clusters 2, 4 and 6; Figure 2B ) were extracted and reanalyzed. This included the following number of cells for each condition: Uninfected: 66, TC-83 2 dpi: 89, TC-83 4 dpi: 987, and TC-83 6 dpi: 2,783. Cells were primarily identified as one of two subclusters of NK cells, referred to as NK cells 1 and 2, CD8⁺ T cells, CD4⁺ T cells, or smaller detected clusters of γδT cells and cells expressing markers of proliferation ( Figure 5A ). At both 4 and 6 dpi, NK cell clusters composed a greater proportion of total lymphocytes than CD8⁺ and CD4⁺ T cells, which exhibited a steep increase at 6 dpi ( Figures 5B, C ). High expression of genes encoding cytotoxic proteins, including perforin, granzyme B, and FasL, were observed at both 4 and 6 dpi ( Figure 5D ). Perforin and granzyme genes (Prf1 and Gzmb, respectively) were expressed in both NK subclusters and CD8⁺ T cells at both timepoints, with NK cells exhibiting higher expression ( Figure 5E ). While expression in NK cells was comparable for both timepoints, CD8⁺ T cells exhibited higher expression of Prf1 and Gzmb at 6 dpi. Additionally, Ifng expression was detected in both NK sub clusters and CD8⁺ T cells, each exhibiting higher expression at 4 dpi. Production of perforin at 6 dpi was also observed by flow cytometry and IFN-γ by protein analysis ( Figures 5G–J ). Perforin protein was detected in NK cells and CD8⁺ T cells, but not CD4⁺ T cells ( Figure 5I ). Further, perforin protein was significantly higher in the brain than spleen in both cell types, suggesting unique cytotoxic effector expression in the brain versus systemically. An increase in the expression of NK cell inhibitory genes such as Klrc1 and Klrd1 was observed in both NK and CD8⁺ T cells at 6 dpi ( Figure 5E ). This together with the decreased Ifng expression at 6 dpi may indicate NK cell exhaustion around the later timepoint. Spatial transcriptomic analysis at 6 dpi revealed detection of NK and T cell associated genes throughout the brain with no observable specific localization patterns, though, Ifng expression sometimes formed concentrated clusters ( Figure 5F ). These results indicate a robust and ubiquitous upregulation of cytotoxic effectors throughout the brain in response to VEEV infection that coincide with timepoints of severe neurological dysfunction.

To compare the immune landscape in the brain to systemic immune changes induced by VEEV infection, we profiled peripheral blood mononuclear cells (PBMCs) from infected and uninfected mice at 6 dpi using scRNAseq. Our analysis identified twelve immune cell clusters including NK cells, T cells, B cells, mono/macs, neutrophils, DCs and plasma cells ( Figures 6A–C ). Consistent with our findings in the brain, the proportion of NK cells dramatically increased in the blood after infection ( Figure 6D ). Mono/Macs also expanded in both the infected blood and brain ( Figures 3C, D , 6D ). In contrast to what we observed in the brain, T cells 1, which encompassed CD4⁺ T cells and regulatory T cells, decreased and the proportion of CD8⁺ T cells remained similar between infected and uninfected blood ( Figure 6D ). Neutrophils, which were identified in the brain at only small relative abundance, dramatically expanded in the blood after infection and upregulated antiviral response genes, including Ddx58, Ifih1, Ifit1, Ifit2 and Isg15 ( Figure 6E ). Blood Mono/Macs and DCs also activated a large number of antiviral response genes in response to infection ( Figure 6E ).

Next, we compared infection-induced transcriptional changes in PBMCs to immune cells from the brain ( Supplementary Figures 4A, B ). Similar to the brain, blood Mono/Macs from infected mice activated Plac8, Ly6c2 and cathepsin family proteases ( Supplementary Figure 4C ). We found that cytokines such as Ccl2, Ccl5 and Ccl7 had markedly lower expression in infected blood Mono/Macs compared to brain Mono/Macs while Ccr2, a receptor involved in myeloid recruitment to the CNS, had an elevated expression in blood Mono/Macs ( Figure 6F ; Supplementary Figure 4C ) (37). Peripheral blood NK cells from infected mice expressed perforin, granzymes, cytokines and inhibitory receptors at a level comparable to NK cells from infected brains, with slightly higher Gzmb expression in the infected brain ( Figure 6G ). Interestingly, uninfected blood NK cells also had robust expression of Prf1 and Gzma but, not Gzmb. The expression levels of perforin and granzymes were also upregulated in CD8⁺ T cells after infection ( Figure 6H ). Unlike uninfected NK cells, uninfected blood CD8⁺ T cells did not express Prf1 or Gzma, suggesting that cytotoxic gene expression in CD8⁺ T cells was infection-induced ( Figure 6H ). Overall, these findings emphasize a tissue-specific upregulation of cytotoxic effectors as a potential key component of the immune landscape in the brain in response to VEEV infections.