Th17 effector cytokines induce shared and distinct microglial and endothelial cell responses in a mouse model for post-streptococcal encephalitis

Abstract

Group A Streptococcus (GAS) infections cause neuropsychiatric complications in children, but the mechanisms linking peripheral infection to brain dysfunction remain unclear. Using mouse genetics, single-cell RNA sequencing, and spatial transcriptomics, we show that GAS infections induce inflammatory transcriptional programs in microglia and brain endothelial cells (BECs), accompanied by loss of blood-brain barrier (BBB) gene expression in female mice. Spatial transcriptomic analyses reveal that GAS-responsive microglia localize near infiltrating CD4⁺ T cells. Several microglial chemokines induced in mice are elevated in sera from affected patients. Deletion of GM-CSF in CD4⁺ T cells partially reduces microglial chemokine gene expression, without restoring BBB integrity. In contrast, IL-17A neutralization partially rescues BBB transcriptional changes, but not BBB dysfunction, and reduces microglial chemokine expression. Microglia-specific deletion of IL-17 receptor A partially restores BBB integrity after GAS infections. Our findings identify microglial IL-17A–IL17RA signaling as a potential mediator of BBB dysfunction and neuroinflammation after GAS infections.

Group A Streptococcus (GAS) infections can cause neuropsychiatric symptoms in children, but the underlying mechanisms remain unclear. Here, the authors show that IL-17A–IL17RA signaling in microglia mediates blood brain barrier dysfunction and neuroinflammation after GAS infection in a mouse model.

Introduction

Neuropsychiatric and cognitive disturbances can arise from bacterial or viral infections, even in the absence of direct brain infection¹. Peripheral infections with Streptococcus pyogenes, or Group A Streptococcus (GAS), can cause secondary complications that impair central nervous system (CNS) function, including movement disorders such as Sydenham’s chorea (SC), and psychiatric syndromes such as Pediatric Autoimmune Neuropsychiatric Disorders Associated with Streptococcal infections (PANDAS; reviewed in refs. ^2,3). SC is characterized by involuntary, uncoordinated movements and behavioral abnormalities, whereas PANDAS presents with a spectrum of psychiatric and fine motor symptoms, including obsessive-compulsive behaviors, vocal or motor tics, reduced appetite or anorexia nervosa, and severe separation anxiety^2,3. The mechanisms driving these neuropsychiatric complications from GAS infections remain poorly understood, but are thought to involve aberrant anti-pathogen immune responses that target the CNS (reviewed in refs. ^3–5), leading to neuroinflammation, blood-brain barrier (BBB) disruption, and neuronal circuitry dysfunction (reviewed in refs. ^5,6).

Neuroimaging studies have identified increased basal ganglia volume in both SC and PANDAS patients^7,8, and increased microglial/astrocytic activation in the basal ganglia of PANDAS patients compared to healthy controls⁹. Autoantibodies targeting basal ganglia structures have also been described in both disorders, including antibodies against dopamine D2 receptors in SC^10,11, anti-dopamine 1 receptor (D1R) antibodies in PANDAS¹², and antibodies against striatal cholinergic interneurons and other neuronal targets in PANDAS^13–16. Nevertheless, the molecular mechanisms underlying the brain pathology, here referred to as post-streptococcal basal ganglia encephalitis (post-GAS-BGE), remain incompletely defined. Elucidating these mechanisms is critical for improving diagnosis, which has been hampered by the lack of reliable biomarkers^3,17–19, and for developing effective treatments for the chronic phase of these disorders^20,21.

T helper 17 (Th17) cells producing interleukin-17A (IL-17A) play an essential role in host defense against extracellular pathogens, and intranasal GAS infections elicit robust Th17 responses in both mice and humans^22–24. Using a juvenile mouse model of intranasal GAS infection, we previously showed that CD4⁺ T cells, including Th17 and Th1 subsets, migrate from the nasal cavity into the anterior brain, where they localize predominantly to the olfactory bulb (OB), the brain’s primary relay for olfactory input. This process is accompanied by BBB disruption, microglial activation, and degradation of excitatory synapses, leading to anosmia and abnormal odor-evoked responses^23,25. Genetic elimination of Th17 cells partially rescues BBB dysfunction, microglial activation, and olfactory circuit abnormalities in this mouse model²⁵, indicating a central role for Th17 lymphocytes in brain pathology. However, the transcriptional responses of microglia, brain endothelial cells (BECs), and other CNS cell types following intranasal GAS infection remain unknown.

Another key unresolved question is how Th17 effector cytokines influence inflammatory responses in microglia and BECs after GAS infections. Th17 cells exhibit considerable phenotypic plasticity^26,27, and chronic inflammation can drive a transition from conventional IL-17A-producing Th17 cells to a population that produces interferon-γ (IFNγ) and granulocyte-macrophage colony-stimulating factor (GM-CSF), and expresses the transcription factor T-bet²⁸. These pathogenic Th17 cells (Th17^path) are required for disease development in the experimental autoimmune encephalomyelitis (EAE), a mouse model for multiple sclerosis (MS)^29,30, and are associated with disease severity in several human autoimmune disorders^31–33. Despite the established role of GM-CSF in pathogenic Th17-mediated tissue damage, it remains unclear whether GM-CSF is produced by CD4⁺ T cells infiltrating the brain after GAS infections, and how GM-CSF and IL-17A differentially contribute to transcriptional changes in microglia and BECs and to the brain pathology in a juvenile mouse model of SC/PANDAS.

To address these questions, we combined mouse genetic approaches with single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, and targeted validation studies in a mouse model of SC/PANDAS. We identify extensive transcriptional changes in both microglia and BECs following repeated GAS infections, including downregulation of microglial homeostatic genes and BBB-associated genes in BECs, alongside upregulation of interferon-response, chemokine expression, and antigen-presentation programs in both cell types. Spatial transcriptomic analyses reveal enrichment of Streptococcus-responsive microglia within the glomerular layer of the OB, in close proximity to infiltrating CD4⁺ T cells. Conditional ablation of GM-CSF in CD4⁺ T cells partially attenuates microglial chemokine gene expression, but does not restore BBB integrity. Systemic neutralization of IL-17A partially rescues BBB-associated transcriptional changes in BECs and reduces microglial chemokine expression; however, compensatory peripheral immune responses associated with persistent GAS infection exacerbate BBB disruption. In contrast, microglia/macrophage-specific deletion of IL-17 receptor A (IL17RA) partially ameliorates BBB deficits following GAS infections. Together, these findings support a role for IL-17A–IL17RA signaling in microglia and macrophages in shaping BBB dysfunction after GAS infection, and suggest that targeting this pathway may complement existing therapeutic strategies for chronic SC/PANDAS in the absence of active infection^20,21.

Results

Microglia and BECs show major transcriptional shifts after multiple GAS infections

We have shown that multiple intranasal GAS infections induce infiltration of CD4⁺ T cells in the OB, accompanied by microglial activation, BBB damage and degradation of excitatory synapses, leading to aberrant odor-evoked neural circuit responses in a mouse disease model^23,25. To investigate, at the transcriptional level, how CNS cell types respond to GAS infections, we isolated and profiled OB cells from P60 female mice 18 hours after the fifth GAS infection using scRNA-seq (Fig. 1a; Supplementary Data 1). Mice inoculated with PBS served as controls.

Fig. 1. Brain endothelial cells upregulate inflammatory signatures and downregulate BBB marker expression following GAS infections.

a Experimental workflow for single-cell RNA sequencing (scRNA-seq) and MERFISH experiments. b, c UMAP visualization of olfactory bulb (OB) scRNA-seq from PBS controls (n = 3; gray) and GAS-infected (n = 6; blue) mice, with enrichment for  CD31⁺ brain endothelial cells (BECs) and CD11b⁺ myeloid cells. d Quantification of transcriptional shifts across OB cell types between PBS and GAS conditions using L^∞ distance (cross-entropy test). e Gene set enrichment analysis (GSEA) of differentially expressed genes in BECs after GAS infection. Inflammatory and interferon-related pathways are enriched, whereas blood-brain barrier (BBB) gene sets are depleted. Bars show significance of enrichment (-log₂FDR) and blue dots indicate the number of significant genes per set. f–i Heatmaps showing normalized expression (log z-scores) of several biological processes from PBS and GAS conditions. Differential expression was determined using two-sided MAST with Benjamini–Hochberg (FDR) correction (adjusted p < 0.05). Significant genes (adjusted P < 0.05) are highlighted in black. j–m Downregulation of BBB-associated transcripts (Itm2a, Itih5 and Mfsd2a) in BECs measured by scRNAseq (left; Wilcoxon Rank Sum test, *adjusted p < 0.0001 for each gene). Representative images (center) and quantification (right) of changes in the BBB transcripts by fluorescence in situ RNA hybridization (FISH) for Itm2a, Itih5 mRNAs (red) combined with immunofluorescence for BEC marker GLUT1 (j–l; cyan), immunofluorescence staining for MFSD2A (red) and BEC marker CD31 (k; cyan) and multiplexed FISH for Lcn2 mRNAs (m; yellow) with immunofluorescence staining for BEC marker CD31 (red). Scale bars are 25 μm (main) and 10 µm (inset). Comparisons were performed with unpaired two-tailed t test with Welch’s correction (exact p values: Itm2a ***p = 0.0003, MFSD2A ***p = 0.0001, Itih5 *p = 0.0314, Lcn2 *p = 0.0263; n = 7–9 mice per condition). Each dot shows an animal in all graphs. Data are mean +/−SEM.

Following quality filtering and batch correction, OB cells were visualized using Uniform Manifold Approximation and Projection (UMAP), and cluster identities were assigned based on established cell-type markers (Fig. 1b, c; Supplementary Fig. 1a). Cluster annotations were further refined, and were consistent across biological replicates (Supplementary Fig. 1e). A cross entropy test³⁴ comparing PBS and GAS conditions revealed the largest shifts in olfactory ensheathing cells (OECs), consistent with their capacity to recruit immune cells during intranasal inflammation (Fig. 1d, Supplementary Data 2)³⁵. Microglia and BECs also exhibited significant transcriptome shifts (Fig. 1d, Supplementary Data 2). In contrast, astrocytes, key mediators of inflammation in many neuroinflammatory models³⁶, displayed only modest transcriptome changes after GAS infections (Fig. 1d, Supplementary Data 2). Given the prominent BBB dysfunction and microglial activation observed in this model^23,25, we focused subsequent analyses on BECs and microglia. To increase their representation in scRNA-seq datasets, CD31⁺ BECs and CD11b⁺ myeloid cells were enriched by fluorescence-activated cell sorting (FACS) prior to sequencing (Supplementary Fig. 1b–d).

BECs upregulate inflammatory programs and downregulate BBB-related transcripts after GAS infections

To define BEC-specific transcriptional responses to GAS infections, BECs were extracted from the scRNA-seq dataset and subclustered (Supplementary Fig. 1f, g). Differentially expressed genes (DEGs) were analyzed using gene set enrichment analysis (GSEA)^37,38 with curated gene lists (Supplementary Data 3, 4). Compared to PBS controls, BECs from GAS-infected mice showed significant upregulation of genes associated with interferon response, antigen presentation, cytokine signaling, inflammation and endothelial cell (ECs) response to lipopolysaccharide (LPS)³⁹, alongside downregulation of BBB-associated genes⁴⁰ (Fig. 1e–h; Supplementary Fig. 1h–k; Supplementary Data 3). Within the BBB transcriptome, transcripts related to adherens and tight junctions, transporters, regulators of transcytosis and brain endothelial identity were significantly reduced in the GAS condition (Supplementary Fig. 1h–k). Decreased expression of two BBB-specific genes, Itm2a and Itih5⁴⁰ as well as increased expression of lipocalin 2 (Lcn2), an IL-17A-induced inflammatory mediator^41,42, was validated by fluorescence in situ hybridization (FISH) and multiplexed error-robust fluorescence in situ hybridization (MERFISH; see below), respectively, in OB sections from GAS-infected mice (Fig. 1j, l, m). Consistent with increased serum IgG transport across the compromised BBB in GAS-infected mice^23,25, the transcytosis suppressor MFSD2A⁴³ was downregulated at both the mRNA and protein levels in BECs (Fig. 1k; Supplementary Fig. 1k). Unexpectedly, genes promoting caveolar transport (including Cav1-2, Cavin1-3) were also downregulated in BECs from GAS-infected mice (Supplementary Fig. 1j). Mixed-effects modeling analysis of BEC DEGs revealed also significant downregulation of some extracellular matrix (ECM) genes (Fig. 1i), critical for BBB integrity (reviewed in ref. ⁴⁴).

Microglia upregulate inflammatory and antigen-presentation programs after GAS infections

GSEA of microglial DEGs revealed enrichment of disease-associated microglia (DAM) signatures, antigen-presentation pathways, cytokine and growth factor signaling, and interferon response genes in GAS compared to PBS conditions (Fig. 2a–c; Supplementary Fig. 2a; Supplementary Data 3, 4). Flow cytometry analysis confirmed increased expression of the antigen-presentation protein CD74, as well as TNF and CCL5 cytokines in microglia from GAS-infected mice, accompanied by reduced expression of homeostatic receptors CX3CR1 and P2RY12 (Fig. 2d–i). Elevated levels of CCL2, CCL4, CCL5, CXCL10 and TNF proteins were also detected in whole OB lysates from GAS-infected mice using a multiplex immunoassay (Supplementary Fig. 2c). Importantly, upregulation of Ccl3 and Ccl4 transcripts in microglia was not attributable to enzymatic dissociation artifacts⁴⁵ as their expression did not correlate with an ex vivo activation gene signature (Supplementary Fig. 2d, e).

Fig. 2. Microglia upregulate inflammatory and disease-associated transcriptional programs following recurrent GAS infections.

a UMAP visualization of microglia from olfactory bulbs (OBs) of PBS and GAS-treated mice identifies six transcriptional clusters: two homeostatic (hMG1–2) and four Streptococcus-responsive (srMG1–4) populations. Cells are colored by cluster (left) or condition [right; PBS (n = 3 mice), gray; GAS (n = 6 mice), blue]. b Gene set enrichment analysis (GSEA) of differentially expressed genes in microglia after GAS infection. Bars represent –log₂(FDR); dots indicate the number of significant genes per pathway. c Heatmap of antigen presentation, cytokine/chemokine, and interferon-response genes in PBS versus GAS microglia. d–i Differential expression of activation markers (Cd74, Tnf, Ccl5) and homeostatic receptors (Cx3cr1, P2ry12) measured by scRNA-seq (left; Wilcoxon rank-sum test, all adjusted p < 0.0001), and validated by flow cytometry (center; representative histograms; right, quantification of normalized median fluorescence intensity). Statistical comparisons of flow cytometry were performed using Welch’s t tests. Exact p values are CCL5 *p = 0.003, CX3CR1 ****p < 0.0001, CD74 **p = 0.0047, P2RY12 ****p < 0.0001, TNF **p = 0.003, (n = 4 PBS, n = 7 GAS mice) with mean ± SEM. j Feature plots showing expression of selected homeostatic, disease-associated, antigen-presentation, and interferon-response genes across microglial populations. k–n Heatmaps of cytokine signaling, homeostatic, disease-associated microglia (DAM), and interferon-response gene modules across the six microglial clusters. Data represent row-normalized log(z-scores) of gene expression. Differential expression across clusters was determined using a MAST with Benjamini–Hochberg (FDR) correction. Significant (black; adjusted p < 0.05) and non-significant (gray; adjusted p ≥ 0.05) genes are indicated.

To further resolve microglial heterogeneity, scRNA-seq data from PBS and GAS conditions were subclustered revealing six distinct microglial populations. Two clusters (hMG1–2), comprised predominantly of PBS-derived cells, expressed canonical homeostatic microglial genes (Fig. 2j, l). In contrast, four Streptococcus-responsive clusters (srMG1–4) exhibited graded upregulation of cytokine signaling, DAM-associated genes, and antigen-presentation programs (Fig. 2j, k, m, n). For example, while all srMG clusters expressed Ccl3 and Ccl4 transcripts, Ccl2, Ccl5, and Tnf were most highly expressed in srMG2 and srMG4 (Fig. 2k). The srMG4 cluster showed the strongest enrichment of interferon-response genes (Fig. 2n).

Finally, the scRNA-seq analysis revealed an increased abundance of macrophages in OBs from GAS-infected mice (Fig. 1b, c, Supplementary Fig. 1c, d, Supplementary Data 2). To determine whether peripheral macrophages infiltrate the brain parenchyma, Cx3cr1^GFP transgenic mice⁴⁶ were crossed to a Tmem119^tdTomato reporter mice⁴⁷. This strategy distinguishes resident microglia (GFP⁺ tdTomato⁺) from peripheral macrophages (GFP⁺ tdTomato^–). Macrophages were confined to perivascular and meningeal regions in PBS and GAS-infected mice OBs, with minimal parenchymal infiltration (Supplementary Fig. 2f–j), although both populations were increased in GAS-infected OBs (Supplementary Fig. 2i, j). Thus, peripheral macrophages accumulate near vascular and meningeal boundaries, but do not invade the brain parenchyma following GAS infections (Supplementary Fig. 2k).

Streptococcus-responsive microglia are enriched in the glomerular OB layer in close proximity to T cells

To understand the regional distribution of “Streptococcus-responsive” microglial clusters and their relationship to infiltrating T cells, we performed spatial transcriptomics using MERFISH⁴⁸ on OB sections (Fig. 3a, b). We probed 391 genes primarily expressed by microglia and BECs (See Methods), as they showed robust transcriptional shifts by scRNA-seq (Fig. 1d). Cell-type identities were assigned using both canonical markers and spatial localization (Supplementary Fig. 3a–c). MERFISH identified cell types consistent with those detected by scRNA-seq (Fig. 3a–c), while providing improved representation of neuronal populations compared to FACS-based scRNA-seq (Fig. 3a, b; Supplementary Figs. 1c, 3a–c).

Fig. 3. Spatial transcriptomics reveals enhanced Streptococcus-responsive microglial activation near infiltrating T cells.

UMAP visualization (a) and representative spatial coordinate plot (b) of all MERFISH samples. c Spatial map of microglia annotated by olfactory bulb (OB) layer. d Representative MERFISH expression maps for homeostatic (P2ry12, Gpr34) and Streptococcus-responsive (Cd74, Ifi30, Axl) transcripts in PBS and GAS OBs (n = 4 mice per group). e Relative expression of homeostatic and GAS-responsive genes in microglia located in glomerular versus granular OB layers normalized to PBS (paired ratio t tests; P2ry12 ***p = 0.0003, Gpr34 **p = 0.003, Cd74 ***p = 0.0003, Ifi30 *p = 0.0441, Axl ***p = 0.0008; n = 4 mice per group). Representative immunofluorescence (IF) images of CD74⁺ IBA1⁺ microglia (arrows) in Cx3cr1^eGFP+ mice (f) and quantification across OB layers (g). Graphs show mean ± SEM. Significance assessed by two-way ANOVA with Šídák’s multiple comparisons (ns, p > 0.05; ***p = 0.001; ****p < 0.0001; n = 4–6 mice per group). Scale bars are 50 µm (main) and 10 µm (inset). h Representative IF images of IFITM3 (red), IBA1 (yellow; myeloid cells), and CD31 (blue; vessels) in glomerular and granular OB layers from PBS and GAS-infected mice. Scale bars are 50 µm (main) and 10 µm (inset). MERFISH-based spatial mapping (i) and quantification (j) of microglial distance to the nearest CD4⁺ T cell across OB layers in PBS and GAS conditions. Comparisons between PBS and GAS in each OB layer were done with two-sided one-way ANOVA followed by Tukey’s multiple correction; glomerular vs external plexiform ****p < 0.0001; glomerular vs granular ****p < 0.0001; external plexiform vs granular ***p = 0.0001. Data are mean ± SEM). Representative IF images of CD4⁺ T cells (red) and Cx3cr1^eGFP+ microglia (k) and quantification of mean intercellular distances (l). Scale bar = 25 µm. Comparisons were performed with one-way ANOVA (glomerular vs external plexiform **p = 0.0366, glomerular vs granular *p = 0.0156, external plexiform vs glomerular *p = 0.0031; n = 4 mice per group). Data are mean ± SEM.

MERFISH revealed a distinct spatial pattern of Streptococcus-responsive gene expression across OB layers. Transcripts for Cd74, Ifi30 and Axl (pathways upregulated in GAS microglia) were expressed at higher levels in microglia located in the glomerular compared to the granular layer of the OB (Fig. 3c–e). Immunofluorescence (IF) confirmed increased CD74 protein expression, critical for proper folding and trafficking of MHC class II molecules for antigen presentation (reviewed in ref. ⁴⁹), in glomerular-layer microglia, mirroring the spatial transcriptomic pattern (Fig. 3f, g). IFITM3, a key interferon-response protein⁵⁰, similarly showed higher expression in glomerular microglia by IF (Fig. 3h), consistent with Ifi30 mRNA distribution by MERFISH. In contrast, homeostatic genes (P2ry12, Gpr34) were modestly enriched in granular-layer microglia (Fig. 3d, e).

Given that infiltrating CD4⁺ T cells accumulate predominantly in the glomerular layer and meninges^23,25, we quantified microglia-T cell proximity. Distances between microglia and CD4⁺ T cells were significantly shorter in the glomerular and external plexiform layers compared to the granular layer, as assessed by MERFISH and IF (Fig. 3i–l). Together, these data suggest that microglia in the glomerular layer undergo more pronounced transcriptional activation after GAS infections, correlating with their proximity to infiltrating CD4⁺ T cells.

GAS infection alters predicted cell–cell communication networks in the OB

To assess whether GAS infection alters intercellular communication, we analyzed scRNA-seq data using CellChat^51,52. CD4⁺ T cells were predicted to communicate primarily with microglia, olfactory ensheathing cells (OECs), and BECs, while microglia showed predicted interactions with macrophages and CD4⁺T cells (Supplementary Fig. 3e). Inferred ligand–receptor interactions included both secreted and contact-dependent pathways. Microglia–BEC interactions were predicted to involve adhesion and inflammatory signaling molecules (e.g., Jama, Tnf, Vegfb), whereas CD4⁺ T cell–microglia communication was predicted to involve cytokines and chemokines (e.g., Ifng, Tnf, Ccl5) as well as adhesion molecules (Supplementary Fig. 3f).

PANDAS/PANS patients display elevated serum levels of inflammatory cytokines and growth factors

Previous studies have reported microglial and astrocytic activation in the basal ganglia of PANDAS patients⁹ and elevated levels of select cytokines (e.g. IL-17A, TNF) in PANDAS/PANS patients sera^53,54, but the breadth of systemic inflammatory changes during the acute disease phase remains unclear. Given the elevated cytokine and chemokine expression observed in the OB of GAS-infected mice (Fig. 2c–i; Supplementary Fig. 2c), we analyzed serum samples from PANDAS/PANS patients and controls. Serum was collected from 10 PANDAS patients enrolled in an intravenous immunoglobulin (IVIg) clinical trial at the National Institute for Mental Health (NIMH)⁵⁵, 13 PANDAS/PANS cases recruited at Columbia University Irving Medical Center (CUIMC) and 11 age-and sex-matched controls obtained from the NIMH (Supplementary Tables 1, 2). Using an unbiased multiplex immunoassay profiling 45 analytes (Supplementary Data 6), we identified significant elevations in 13 cytokines, chemokines and growth factors in patient sera (Table 1; adjusted p < 0.05). Six cytokines—CCL2, CCL3, CCL4, CCL5, CXCL10 and TNF—were elevated in patient sera, and were also robustly upregulated in microglia and OB lysates from GAS-infected mice (Fig. 2c–i; Supplementary Fig. 2c). GM-CSF was likewise significantly elevated in patient sera (Table 1). These findings indicate that the acute phase of PANDAS/PANS is associated with a systemic inflammatory signature overlapping with GAS-induced microglial responses in mice.

Table 1.

Cytokines, chemokines and growth factors upregulated in sera from acute PANDAS/PANS patients

Serum protein	Healthy controls (pg/mL)	PANDAS/PANS (pg/mL)	p-value	p-adj
CCL2	25.5 (8.475–105.5)	157 (85.5–253.5)	0.0005	*
CCL3	4 (0.525–10.1)	29 (14.5–47.5)	<0.0001	**
CCL4	81 (55.5–90.5)	131 (93–181)	0.0012	*
CCL5	30 (15.25–41)	296.5 (91.75–440.75)	<0.0001	**
CXCL8	ND	6.9 (2.1–27)	0.0011	*
CXCL10	7.2 (5.975–8.525)	59 (32–78)	0.0004	**
CXCL12	237 (176–313)	664 (385.5–809)	0.0007	*
EGF	16 (0.85–103)	87 (40–152.5)	0.014	ns
GM-CSF	ND	16 (4.075–24)	<0.0001	**
HGF	54 (31.5–443)	443 (358.5–589)	0.0024	ns
IL-1RA	ND	536 (110.75–1115.75)	0.0156	ns
IL-6	ND	7.85 (2.05–26.5)	0.0174	ns
IL-7	1 (0.475–3.4)	5.2 (3.55–7.65)	0.0004	**
IL-12A/B	ND	2.9 (1.95–3.85)	<0.0001	**
IL-15	ND	7.9 (3.525–24.5)	0.0037	ns
IL-22	ND	8.9 (4.175–30.5)	0.0183	ns
IL-23	ND	16 (5.5–44.5)	0.0244	ns
KITLG	5.2 (3.45–7.4)	14 (9–18)	0.0009	*
LIF	ND	28 (18–58.5)	0.0129	ns
PDGFB	87 (43.5–586.5)	1800 (615.5–2620)	0.0004	**
PGF	94 (77–116)	188 (96–290)	0.0187	ns
TNF	5.2 (4.275–12.1)	18 (9.975–22.5)	0.0458	ns
VEGFA	88 (31.5–152)	318 (138.5–483)	0.0003	**
VEGFD	ND	22 (1.65–32)	0.0253	ns

Forty-five proteins encompassing cytokines, chemokines and growth factors were measured by multiplex immunoassay in sera from acute PANDAS/PANS patients (n = 23) and controls (n = 11; see Supplementary Tables 1, 2). Thirteen proteins were significantly elevated in PANDAS/PANS patients compared to controls. Serum concentrations are provided as median (with interquartile range).

Statistical comparisons were performed using the two-sample Mann–Whitney test, p-values with Bonferroni correction and adjusted p values (p-adj) for 24 comparisons (ns, p > 0.05; *p < 0.05; **p < 0.01).

IFNγ derived from Th1 cells contributes to GAS-induced antigen presentation and interferon response in microglia and BECs in Th17-deficient mice

Elevated GM-CSF and other T cell-derived cytokines in patient sera (Table 1), together with prior evidence that Th17 lymphocytes drive GAS-induced brain pathology²⁵ prompted us to examine how Th17 deficiency alters microglial and BEC responses. We performed scRNA-seq on OBs from GAS-infected mice lacking the Th17 fate-specifying transcription factor RORγt^25,56 (Fig. 4a). These mice exhibit robust infiltration of IFNγ⁺ Th1 cells, but markedly reduced Th17 cells after GAS infection²⁵. BECs from GAS-infected RORγt^−/− mice showed partial restoration of BBB-associated transcripts (e.g., Mfsd2a, Tjp1 and Itih5), alongside increased expression of antigen-presentation and interferon response genes relative to wild-type (WT) mice (Fig. 4b; Supplementary Data 3). Gene ontology (GO) pathway analysis confirmed enrichment of antigen presentation, interferon signaling and lymphocyte-related pathways, with concomitant downregulation of pathways related to vascular development and cell migration (Fig. 4c; Supplementary Data 5).

Fig. 4. Th17-deficient mice show partial restoration of endothelial and microglial transcriptional responses after GAS infections.

a Schematics of effector cytokine production by T cells and brain pathology in GAS RORγt^−/− mice. Heatmaps of differentially expressed genes (DEGs) in BECs (b) and microglia (e) from GAS WT and RORγt^−/− OBs (n = 3 mice/genotype). Data represent row-normalized log(z-scores) and DEGs were analyzed using a MAST with Benjamini–Hochberg. Significant genes (adjusted p < 0.05) are in black. GO pathway enrichment in RORγt^−/− versus WT BECs (c) and microglia (f). d IFNγ concentrations in OB lysates after two or five GAS infections. Significance by one-way ANOVA (ns, p > 0.05; GAS vs RORγt^{−/− *}p = 0.0177; PBS vs RORγt^−/− *p = 0.0301; n = 3–5 mice per group). Representative flow cytometry of CD74 and MHC class II in microglia (n = 3640 and 4657 cells) (g) and quantification of surface expression (h). Significance assessed with one-way ANOVA followed by Tukey’s test (MHC II: WT PBS vs. WT GAS (p = 0.7471); WT GAS vs. RORγt^−/− GAS (**p = 0.0085); WT PBS vs. RORγt^−/− GAS (**p = 0.0066). CD74: WT PBS vs. WT GAS (*p = 0.0313); WT GAS vs. RORγt^−/− GAS (***p = 0.0002); WT PBS vs. RORγt^−/− GAS (**p = 0.0075); n = 4–11 mice per group). i–l Effects of cytokines on mouse BEC: transendothelial electrical resistance (TEER) measurements (i); area under the curve (AUC) quantification (j); transwell permeability (k) and relative transcytosis (l) of Albumin-AF594. Significance assessed by mixed-effects model (j) or repeated-measures one-way ANOVA (l) with Dunnett’s test. P values for AUC (j): IL-1β + TNF (*p = 0.0218), IFNγ (**p = 0.0035); relative transcytosis (l): LPS (*p = 0.0301), No cells (*p = 0.0181) vs. untreated. (n = 3–5). TEER measurements in human brain microvascular endothelial cells (HBMECs) (m) and AUC quantification (n). Significance determined using a mixed-effects model with Dunnett’s test; IL-1β + TNF (****p = 0.0006), IFNγ (****p = 0.0019) vs control. Data are mean ± SEM. All in vitro assays (i–n) were from three independent experiments (6 replicates each).

Similarly, microglia from GAS-infected RORγt^−/− mice displayed increased expression of antigen presentation genes, but reduced expression of chemokines and cytokines (e.g. Ccl2, Ccl3, Ccl4, Tnf) (Fig. 4e; Supplementary Data 3). GO analysis corroborated enrichment of antigen processing and immune response pathways in GAS-infected RORγt^−/− microglia (Fig. 4f; Supplementary Data 5). Flow cytometry confirmed increased surface expression of CD74 and MHC class II (I-A/I-E) in microglia from GAS-infected RORγt^−/− compared to WT mice (Fig. 4g, h). Th17 cell-dependent suppression of antigen presentation genes extended beyond microglia and BECs, with increased MHC class I expression in astrocytes, OECs and neurons (Supplementary Fig. 5d). A potential mechanism for this effect could be increased IFNγ in RORγt^−/− mice, since IL-17A suppresses IFNγ expression and Th1 cell identity^57,58. IFNγ also increases MHC gene expression in myeloid cells and neurons^59,60. Consistently, IFNγ protein levels were significantly elevated in OBs from RORγt^−/−mice after GAS infection (Fig. 4d).

To understand how Th17-derived cytokines affect BBB function, we treated confluent primary mouse BECs and human brain microvascular endothelial cells (HBMECs) with either mouse or human IL-17A, GM-CSF or IFNγ, respectively. We measured the transendothelial electrical resistance (TEER) for 18–48 hours after cytokine treatment as a readout of paracellular barrier integrity. An IL-1β/TNF combination served as a positive control since these cytokines are known to disrupt BBB integrity. In vitro, IFNγ, but not IL-17A or GM-CSF, reduced TEER in mouse and human BEC monolayers, indicating impaired paracellular barrier function, albeit to a lesser degree than IL-1β/TNF combination (Fig. 4 i, j, m, n). In addition, we also cultured mouse BECs to confluency in a transwell system and measured the transport of fluorescently-labeled albumin from the top to the bottom chambers across the BEC monolayer as a proxy for transcellular barrier permeability. Although bacterial Lipopolysaccharide (LPS) increased transcellular albumin transport across the mouse BEC monolayer, none of the Th17-derived cytokines showed any effect (Fig. 4k, l). In summary, IFNγ induces antigen presentation and interferon responses in microglia and BECs, with modest direct effects on BBB integrity (Fig. 7a).

Fig. 7. Model summarizing the putative effects of IFNγ, IL-17A, and GM-CSF on brain endothelial and microglial responses to recurrent GAS infections.

a–c Proposed model illustrating how effector cytokines produced by Th1 and Th17 lymphocytes that infiltrate the brain after repeated GAS infections may differentially shape transcriptional responses in brain endothelial cells (BECs) and microglia. a IFNγ is predicted to induce interferon-response programs and antigen-presentation pathways in both microglia and BECs. b, c IL-17A and GM-CSF exert partially overlapping effects on microglial activation, including regulation of proliferative responses and induction of disease-associated microglial (DAM) genes, cytokines, and chemokines, but also have cytokine-specific effects. IL-17A signaling may affect more BEC transcriptional alterations following GAS infections. In addition, antigen-presentation signatures remain elevated in conditions of Th17 deficiency or systemic IL-17A neutralization, suggesting that IL-17A may partially suppress these pathways in both microglia and BECs. The model highlights shared and distinct roles for Th17-associated cytokines in shaping neuroimmune and vascular responses in the brain after recurrent GAS exposure in the periphery.

GM-CSF regulates a subset of microglial and BEC transcriptomic responses after multiple GAS infections

Since Th17 lymphocytes are required to induce brain pathology and a subset of transcriptome changes in microglia and BECs in the mouse model, we investigated the effects of Th17-derived cytokine (GM-CSF or IL-17A) on these transcriptome changes. To assess the role of GM-CSF, a cytokine downstream of RORγt implicated in autoimmune pathology⁶¹, we first quantified GM-CSF–producing CD4⁺ T cells during GAS infection. The proportion of GM-CSF⁺ CD4⁺ T cells and IFNγ⁺IL-17A⁺ (Th17^path) cells increased in the OB with successive GAS infections (Supplementary Fig. 4a–c). GM-CSF⁺ CD4⁺ T cells were reduced by 2-fold in RORγt^−/− mice (Supplementary Fig. 4d).

We next generated mice with CD4⁺ T cell–specific deletion of GM-CSF (Csf2^ΔCD4) by crossing Cd4-Cre^ERT262 to Csf2^fl/fl63 mice and administering 4-OH-tamoxifen daily between P16 and P20 prior to onset of GAS infections at P28 (Supplementary Fig. 4e). This protocol resulted in a significant decrease in GM-CSF⁺ CD4⁺ T cells (Supplementary Fig. 4f). Csf2^ΔCD4 mice showed normal survival and CD4⁺ T cell infiltration after GAS infections (Supplementary Fig. 4g, h). scRNA-seq revealed that microglia from Csf2^ΔCD4 mice exhibited reduced expression of cytokine and chemokine transcripts, with minimal changes in antigen presentation genes compared to Csf2^fl/fl microglia after GAS infections (Supplementary Fig. 4i; Supplementary Data 3). GO analysis confirmed downregulation of cytokine and interferon response pathways (Supplementary Fig. 4j; Supplementary Data 5). Despite these transcriptomic changes, CD68⁺ IBA1⁺ microglia numbers were increased (Supplementary Fig. 4k, l), and CD74 protein expression was unchanged in Csf2^ΔCD4 mice (Supplementary Fig. 4m, n), although this phenotype was not accompanied by changes in cell cycle transcripts (Supplementary Data 3).

In BECs, GM-CSF deletion resulted in modest upregulation of select BBB-related genes (e.g. Itih5, Bsg) and reduced expression of antigen presentation, response to bacterium and interferon-response pathways corroborated by GO analysis (Supplementary Fig. 4o, p; Supplementary Data 3, 5). Consistent with scRNA-seq data, there was no improvement in BBB permeability to serum IgG in the Csf2^ΔCD4 OB after GAS infections (Supplementary Fig. 4q–s). Together, these findings indicate that CD4⁺ T cell-derived GM-CSF preferentially modulates microglial inflammatory programs, with limited impact on BBB integrity.

Global IL-17A blockade reverses GAS-induced transcriptome changes in BECs and microglia, but exacerbates BBB dysfunction during active infection

To assess the role of IL-17A, a major cytokine produced by Th17 cells, mice were treated with an IL-17A–neutralizing antibody or isotype control throughout GAS infection (Fig. 5a). IL-17A blockade did not alter CD4⁺ T cell infiltration, or subtype distribution in the OB (Supplementary Fig. 5a–c). In BECs, IL-17A neutralization increased expression of select BBB-associated genes, particularly regulators of transcytosis, as well as antigen presentation, and shifted GO pathways toward metabolic and maturation-associated BEC programs such as cellular respiration and oxidative phosphorylation (Fig. 5b, c, Supplementary Data 5). Conversely, inflammatory and interferon-related pathways were downregulated after IL-17A blockade (Fig. 5b, c, Supplementary Data 5). Despite these transcriptomic changes, IL-17A blockade worsened BBB permeability to serum IgG (Fig. 5d–f), accompanied by further reductions in the Itm2a mRNA (BBB transcript; Fig. 5k, l). Although, a subset of interferon response genes (e.g. Ifit1,2,3, Ifitm2,3) were rescued at the scRNA-seq level (Supplementary Data 3), there was no change in IFITM3 protein level in OB CD31⁺ blood vessels between two treatments after GAS infections (Fig. 5g–i). IL-17A is critical to fight GAS infections in the periphery⁶⁴, therefore its blockade could increase peripheral inflammatory responses. Serum cytokine levels for IFNγ, CCL4, CCL5, CXCL2, CXCL10 and GM-CSF, were elevated in IL-17A-treated mice (Fig. 5m), and survival rate after GAS infections was reduced (Fig. 5j), consistent with impaired control of the peripheral infection.

Fig. 5. IL-17A inhibition partially restores endothelial transcriptional programs after GAS infections but worsens BBB dysfunction.

a Experimental timeline. b, c Heatmaps and GO pathway enrichments of DEGs in OB BECs from GAS mice with two treatments (n = 5 GAS Isotype and n = 7 GAS α–IL-17A mAb mice). Significant genes (adjusted p < 0.05; MAST with Benjamini–Hochberg correction) are in black. d–f IF images of IgG leakage (green) outside OB vessels (GLUT1; cyan) and quantification. Significance by two-way ANOVA with Šídák’s test [(e) GAS Isotype vs α–IL-17A mAb **p = 0.0153; PBS α–IL-17A vs GAS α–IL-17A mAb ***p = 0.0012; f GAS Isotype vs α–IL-17A mAb **p = 0.0388; PBS α–IL-17A vs GAS α–IL-17A mAb ***p = 0.0034; 4–6 mice/group). g–i IF for IFITM3 (pink), IBA1 (yellow), and CD31 (blue) and quantification of IFITM3⁺ area within OB vessels. Significance assessed by two-way ANOVA with Sidak’s test. h PBS vs. GAS Isotype (*p = 0.0280); PBS vs. GAS α–IL-17A mAb (*p = 0.0470). i PBS vs. GAS Isotype (*p = 0.0455); PBS vs. GAS α–IL-17A mAb (**p = 0.013); 4–6 mice/group. Scale bars are 50 µm (main) and 10 µm (inset). j Survival curves (13–48 mice per group; Mantel–Cox log-rank test; ****p < 0.0001). k, l FISH images for Itm2a mRNA (pink) and GLUT1 (blue) and quantification. Scale bars are 50 µm (main) and 10 µm (inset). Two-way ANOVA with Sidak’s test p values: GAS Isotype vs. α–IL-17A mAb (*p = 0.0333); PBS vs. GAS α–IL-17A mAb (*p = 0.0199); 4–6 mice/group. m Serum cytokine concentrations. Significance by two-way ANOVA with Šídák’s test (GAS Isotype vs. GAS α–IL-17AmAb; and PBS vs. GAS α–IL-17A mAb, respectively are: IFNγ **p = 0.0088; **p = 0.0097), CCL4 (***p = 0.0002; ****p < 0.0001), CCL5 (***p < 0.0001; ***p = 0.0002), CXCL2 (*p = 0.0127; **p = 0.0029), CXCL10 (***p = 0.0073; **p = 0.0009), GM-CSF (***p = 0.0012; **p = 0.0008); n = 4 for PBS, 6 GAS mice). Data are mean ± SEM.

Microglial scRNA-seq and GO analysis from the IL-17A mAb-treated condition revealed partial restoration of homeostatic gene expression and reduced DAM-associated transcripts and cytokine production, but increased antigen presentation and interferon-response pathways paralleling the transcriptome phenotype of RORγt^−/− microglia (Supplementary Fig. 5d, e; Supplementary Data 3, 5). Genes related to antigen presentation by MHC class I were also upregulated in other OB cell types, including OECs and astrocytes after IL-17A blockade (Supplementary Fig. 5f). In contrast to the RORγt^−/− microglial phenotypes observed by IF²⁵, the number of activated CD68⁺ IBA1⁺ myeloid cells remained unchanged after IL-17A mAb treatment, and there were was no difference in microglial expression of CD74 and IFITM3 proteins between the two treatments after GAS infections (Supplementary Fig. 5g–l), likely reflecting heightened peripheral inflammation. These findings suggest that systemic IL-17A blockade during active infection exacerbates neurovascular pathology likely through indirect effects mediated by circulating inflammatory cytokines (Fig. 5m).

Microglia-specific ablation of IL-17 receptor A rescues BBB dysfunction without altering microglia activation

To determine whether IL-17A signaling in the brain contributes directly to pathology, we examined IL-17 receptor expression and found high IL17RA expression in microglia, macrophages and neutrophils (Supplementary Fig. 3d), suggesting a critical contribution of these cells to brain pathology. To address this hypothesis, we generated mice lacking IL17RA in microglia/macrophages (Il17ra^ΔCx3cr1) by crossing Cx3cr1^CreERT265 with Il17ra^fl/fl66 mice and administering 4-OH tamoxifen between P16 and P20 prior before the onset of GAS infections at P28 (Fig. 6a). IL17RA deletion in microglia did not significantly alter CD4⁺ T cell infiltration into the OB (Supplementary Fig. 6a, b), but significantly reduced BBB permeability to serum IgG in the granular layer (Fig. 6b–d). Consistent with a partial rescue in BBB function, MFSD2A protein levels showed a non-significant increase in the OB vasculature (Fig. 6g, h), whereas Itm2a mRNA expression was comparable between genotypes after GAS infections (Fig. 6e, f). The vascular expression of IFITM3 protein was also increased, although not significantly, in the OB glomerular layer in Il17ra^ΔCx3cr1 mice after GAS infections (Fig. 6i; Supplementary Fig. 6g). Importantly, Il17ra^ΔCx3cr1 mice did not exhibit reduced survival after GAS infection (Fig. 6j), in contrast to systemic IL-17A blockade. Finally, Il17ra^ΔCx3cr1 mice showed no difference in either microglial activation (CD68⁺ IBA1⁺), expression levels for the CD74 antigen presentation marker, or IFITM3 protein compared to Il17ra^fl/fl GAS-infected mice (Supplementary Fig. 6c–h). Therefore, IL-17A/IL17RA signaling in microglia contributes to BBB dysfunction after GAS infection, while other aspects of microglial activation are likely maintained by other T cell-derived cytokines.

Fig. 6. Myeloid-specific deletion of IL17RA partially mitigates BBB disruption following recurrent GAS infections.

a Timeline of 4-hydroxytamoxifen administration and GAS infections in Il17ra^fl/fl and Il17ra^ΔCx3cr1 mice. b–d Representative immunofluorescence (IF) images of serum IgG leakage (green) outside GLUT1⁺ OB vessels (magenta) and quantification of serum IgG extravasation in the OB from PBS or GAS-infected Il17ra^fl/fl (salmon) and Il17ra^ΔCx3cr1 (red) mice. Significance by two-way ANOVA with Šídák’s test [(c) GAS Il17ra^fl/fl vs GAS Il17ra^ΔCx3cr1 *p = 0.0135, PBS Il17ra^fl/fl vs GAS Il17ra^ΔCx3cr1 ***p = 0.0004; PBS Il17ra^ΔCx3cr1 vs GAS Il17ra^ΔCx3cr1 **p = 0.0062; d PBS Il17ra^fl/fl vs GAS Il17ra^ΔCx3cr1 *p = 0.0187; PBS Il17ra^Cx3cr11 vs GAS Il17ra^ΔCx3cr1 *p = 0.0285; n = 4–6 mice/group). Scale bar = 50 µm. e, f Representative FISH images of Itm2a mRNA (red) with GLUT1 (green, vessel marker) in the glomerular OB layer, and quantification of vascular Itm2a coverage. Significance by two-way ANOVA followed by Šídák’s test (ns p > 0.05). Scale bars are 50 µm (main) and 10 µm (inset). g, h Representative IF images of MFSD2A (green) and CD31 (purple) in the OB of GAS-infected Il17ra^fl/fl and Il17ra^ΔCx3cr1 mice, with quantification of MFSD2A mean fluorescence intensity (MFI). Scale bar = 50 µm. Significance by ordinary two-way ANOVA followed by Šídák’s test (ns p > 0.05). i Quantification of IFITM3⁺ area within CD31⁺ vasculature in the OB from GAS-infected mice of both genotypes. Comparisons were done using two-way ANOVA with Šídák’s test (Il17ra^fl/fl granular vs glomerular *p = 0.0417; Il17ra^ΔCx3cr11 granular vs glomerular ***p = 0.0046; n = 4–5 mice/group). j Survival curves of PBS or GAS-infected Il17ra^fl/fl and Il17ra^ΔCx3cr1 mice (n = 11–18 mice/group). Significance assessed by the Mantel–Cox log-rank test (ns p > 0.05). Each dot shows an animal in graphs (c, d, f, h, i). Data are mean +/− SEM.

Discussion

GAS infections are a major cause of morbidity in children and adults worldwide⁶⁷, not only due to primary infections such as pharyngitis, but also through secondary neuropsychiatric complications⁴. While aberrant cellular and humoral immune responses underlie these CNS complications (reviewed in refs. ^3–5), the molecular changes in brain cells, and the contributions of T cell-derived cytokines are not understood. Using genetic and antibody blockade approaches in mice together with scRNA-seq, spatial transcriptomics, and validation experiments, we mapped the transcriptional changes in CNS cells—particularly BECs and microglia—following multiple GAS infections. We find that microglia in close proximity to infiltrating CD4⁺ T cells exhibit the most robust “Streptococcus-responsive” gene expression. We further dissect shared and unique effects of two Th17 effector cytokines, IL-17A and GM-CSF, on microglial and BEC transcriptomes. Finally, serum from PANDAS/PANS patients displays elevated levels of multiple inflammatory cytokines, highlighting a potential role for Th17 lymphocytes in human disease progression.

scRNA-seq of over 100,000 OB cells revealed that BECs and microglia are among the CNS cell types most transcriptionally altered after GAS infections. Both upregulate inflammatory programs, including antigen presentation, interferon response, and cytokine production, while downregulating homeostatic signatures: BBB-associated genes in BECs and homeostatic genes in microglia. BEC transcriptional responses mirror patterns seen in other inflammatory conditions, such as systemic LPS administration^39,68,69, and EAE⁷⁰, with upregulation of antigen presentation, interferon response, and downregulation of BBB-associated transcripts⁴⁰. These molecular changes are also consistent with our prior findings that BBB structure and function are impaired after GAS infections^23,25. Despite increased BBB permeability to large proteins (e.g. serum IgG), genes promoting caveolar transport (e.g., Cav1, Cav2, Cavin1–2) were downregulated, suggesting that decreased MFSD2A expression contributes to elevated transcellular transport⁴³; however, junctional breakdown or bulk transcytosis may also play roles. Lcn2 transcript, strongly induced in BECs, may mediate IL-17A-dependent innate immune responses, consistent with its role in other diseases^41,42.

Microglia similarly exhibit upregulation of DAM, interferon response, antigen presentation, and cytokine/chemokine genes. Heterogeneity exists among the four Streptococcus-responsive microglia clusters, with srMG4 displaying high interferon-response genes and downregulation of some classic DAM genes (e.g., Mertk, Trem2), indicating an activation state distinct from that seen in neurodegeneration models⁷¹. T cells that infiltrate the brain after GAS infections secrete IFNγ (type II interferon); however, we could not detect IFN-α or -β (type I interferons) transcriptionally in our scRNA-seq dataset from the OB (Supplementary Data 2, 3). Microglial expression of interferon-response genes has been seen in inflammation^72,73, autoimmunity^74–76, injury^77,78, neurodegeneration^79–81 and aging⁸² and is implicated in cell morphology changes, inflammation and phagocytosis, among other microglial functions^82–84. IFNγ produced by infiltrating T cells appears to drive antigen presentation gene expression in microglia, BECs, and other CNS cells after GAS infections (Fig. 7a), consistent with known effects of type II interferons on MHC regulation. Upregulated microglial chemokines (e.g., CCL2, CCL3, CCL4, CCL5, CXCL10) may facilitate recruitment of peripheral immune cells, and combined with T cell effector cytokines, could impair BBB integrity in vivo. For example, CCL2-CCR2 signaling is required for Th17^path cell recruitment to the CNS in EAE and S. pneumoniae infections⁸⁵. Several microglia-derived cytokines (e.g. TNF, IL-1β) and chemokines (e.g. CCL2, CCL5) are known to break down the BEC barrier in vitro, and they may contribute, together with T cell effector cytokines, to impair BBB function in vivo. MERFISH analysis confirms that microglial transcriptional shifts are greatest in OB regions with dense T cell infiltration, supporting a key role for T cell-microglia crosstalk in driving CNS pathology as predicted by CellChat.

The transcriptomic analysis of microglia and BECs from RORγt^−/− mice supports our prior findings that Th17 cells are critical for brain pathology in our mouse disease model²⁵. Loss of Th17 cells partially rescues microglial chemokine/cytokine expression and BBB transcripts while paradoxically increasing antigen presentation and interferon-response genes, likely due to elevated IFNγ levels and absence of IL-17A/GM-CSF inhibition. IFNγ is known to increase MHC gene expression in myeloid cells and neurons^59,60, and its production by T cells can be inhibited by IL-17A^57,58. Circulating IFNγ levels may be responsible for upregulation of antigen presentation and the interferon response in microglia and BECs (Fig. 7a).

IL-17A and GM-CSF have overlapping and distinct effects on microglia. Both contribute to microglial DAM gene expression and cytokine/chemokine regulation, but specific chemokine subsets are preferentially regulated by one cytokine (Fig. 7b, c). IL-17A more effectively suppresses antigen presentation in microglia and BECs, as evidenced by increased expression in both RORγt^−/− and IL-17A mAb-treated mice. In contrast while interferon response genes are strongly downregulated in microglia from Csf2^ΔCD4, RORγ^−/− and IL-17A mAb-treated mice, expression of specific chemokine transcripts seems to be preferentially regulated by either IL-17A or GM-CSF (Fig. 7b, c). In contrast, the effects of IL-17A and GM-CSF are quite divergent in BECs. GM-CSF deficiency in T cells minimally affects BBB transcriptome or function, but reduces inflammatory signaling in microglia, highlighting a predominantly microglia-targeted effect. IL-17A blockade rescues transcriptomic shifts in BECs but worsens BBB permeability, likely due to high levels of circulating inflammatory cytokines from unresolved peripheral inflammation. IL-17A does not act directly on BECs, because it cannot disrupt tight junctions, nor promote albumin transcytosis in primary cultured mouse or human BECs (Fig. 4i–n). Consistent with robust IL17RA expression in microglia and macrophages after GAS infections, genetic ablation of IL17RA in these cells partially rescues BBB leakage, indicating microglia mediate IL-17A effects on the vasculature (Fig. 7d). Future studies will identify downstream effectors of IL17A/IL17RA signaling in microglia and how they affect BBB function after GAS infections.

Currently, it is contested whether PANDAS/PANS has an inflammatory origin, although neuroinflammation has been seen in the basal ganglia of PANDAS patients⁹ using a PET ligand that binds both activated microglia and astrocytes⁸⁶. PANDAS/PANS patients exhibit elevated serum cytokines, chemokines, and growth factors during acute disease, mirroring cytokine upregulation in GAS-infected mouse OBs. Some elevations, such as IL-6 and CXCL10, are shared with acute GAS pharyngitis cases⁸⁷, whereas patients with invasive GAS infections have elevated levels of IL-1β, IL-6, IL-8, IL-10 and IL-18⁸⁸. Other elevated cytokines (CCL2, CCL4, CCL5, IL-7 and TNF) appear specific to neuropsychiatric complications, supporting a distinct inflammatory profile. Distinguishing PANDAS/PANS from Tourette’s syndrome (TS) or OCD remains challenging as overlapping cytokine elevations are reported. Elevated levels of IL-17A and TNF have been found in sera from PANDAS/PANS patients^53,54 and high TNF and IL-12 serum levels correlate with symptom exacerbation in children with tic disorders and OCD^89,90. In addition, immune phenotypes linked to TS include elevated levels of IL-6, IL-12, IL-17A and TNF^89,91. Our data provide putative candidate biomarkers for acute SC/PANDAS, which may aid diagnosis and clarify the role of neuroinflammation in these disorders.

Methods

Mice

Experiments involving mice were approved by Columbia University Irving Medical center (CUIMC) Institutional Animal Care and Use Committees (AAAX3452, AABN7552). Mice were bred in the CUIMC vivarium, under 12-h light/12-h dark, pathogen-free conditions. Female mice were used for all experiments, except the time course analysis of Th17 cell subtypes by flow cytometry and Csf2 recombination confirmation flow cytometry, which used even numbers of males and females. The wild-type (C57BL/6J, strain 000664) and RORγt^eGFP mice^25,56 (B6.129P2(Cg)-Rorctm2Litt/J, strain 007572) were obtained from the Jackson Laboratory and bred in the CUIMC vivarium. The Csf2^fl/fl mouse strain⁶³ was provided by Bogoljub Ciric (Thomas Jefferson University, Philadelphia, PA). Cd4-Cre^ERT2 transgenic mice (B6(129×1)-Tg(Cd4-Cre^ERT2)11Gnri/J, strain 022356)⁶² were obtained from the Jackson Laboratory and crossed to Csf2^fl/fl mice for two generations. Cd4-CreER^T2+/− Csf2^fl/fl males were mated to Csf2^fl/fl females to generate Cd4-Cre^ERT2+/− Csf2^fl/fl (Csf2^ΔCD4) experimental mice and Csf2^fl/fl littermate controls. Cx3cr1-Cre^ERT2 transgenic mice (B6.129P2(Cg)-Cx3cr1^{tm2.1(cre/ERT2)Litt}/WganJ, strain 021160)⁶⁵ were obtained from the Jackson Laboratory and crossed to Il17ra^fl/fl mice (B6.Cg-Il17ra^tm2.1Koll/J, strain 03100)⁶⁶ for two generations. Cx3cr1-Cre^ERT2+/- Il17ra^fl/fl males were mated to Il17ra^fl/fl females to generate Cx3cr1-Cre^ERT2+/- Il17ra^fl/fl (Il17ra^ΔCx3cr1) experimental mice and Il17ra^fl/fl littermate controls. P16 pups were intraperitoneally injected daily with 100 µg of (Z)-4-Hydroxytamoxifen (Millipore Sigma, H7904), dissolved in 50 µL of corn oil (Millipore Sigma, C8267) for 5 days (P16-P20). Tmem119-tdTomato⁴⁷ and Cx3cr1-GFP⁴⁶ reporter mouse lines were provided by Dr. Wassim Elyaman (CUIMC).

Human serum studies

The experiments with human sera were approved by CUIMC Institutional Review Board (IRB# AAAQ9999). The NIMH sera used in this study were analyzed in a previous publication⁹², and were obtained from the NIMH. Informed consent/assent was obtained from all subjects both at NIMH and CUIMC.

GAS intranasal infections

All experimental mice received weekly intranasal inoculations with either a suspension of Streptococcus pyogenes [Group A Streptococcus (GAS)], or phosphate-buffered saline (PBS) control, starting at P21–P28. Recombinant GAS strain expressing a 2 W epitope-tagged M protein was used for all experiments as described^22,23,25. GAS was streaked out on new blood agar plates each week. Culture media consisted of an autoclaved solution of 3% Todd-Hewitt Broth (Bacto, 90003-430) and 2% Neopeptone (Bacto, 90000-268). Several GAS colonies were used to inoculate 10 mL of culture medium and incubated overnight at 37 °C in 5% CO₂. The following day, the culture was diluted to an OD₆₀₀ of 0.2 and grown to OD₆₀₀ 0.6, centrifuged and washed in 1 mL PBS (without Ca²⁺ and Mg²⁺) and resuspended in 110 µL of PBS. The GAS suspension was kept briefly on ice prior to intranasal infections. All intranasal infections were performed in an ABSL2 vivarium facility. Mice were immobilized with light isoflurane anesthesia and a P20 pipette was used to drip GAS suspension into nostrils. To reduce lethality due to sepsis, a smaller GAS dose was used during the first two weeks (the first GAS inoculation is 8 × 10⁷ CFU per nostril, the second is 12 × 10⁷ CFU per nostril, and the third, fourth and fifth are 2 × 10⁸ CFU per nostril). During the first two weeks of GAS infections, mice were provided with nutritional supplements (DietGel and ClearH₂O, 72-27-5022).

Neutralizing antibody treatment

Starting 24 h prior to the first GAS infection, mice were injected intraperitoneally twice weekly with 500 µg of either InVivoMAb anti-mouse IL-17A monoclonal antibody, clone 17F3 (Bio X Cell catalog, BE0173), or mouse IgG1 isotype control monoclonal antibody, clone MOPC-21 (Bio X Cell, catalog BE0083), in 100 µL of dilution buffer (Bio X Cell catalog IP0070 and IP0065, respectively).

Single-cell RNA sequencing

Mice were anesthetized with isoflurane and perfused intracardially with PBS for 3 min. Nasal associated lymphoid tissue (NALT), olfactory epithelium (OE), or olfactory bulb (OB) were dissected and placed in Hanks’ Balanced Salt Solution (HBSS) without Ca²⁺ and Mg²⁺ and cut up with a sterile scalpel blade. Two or three animals were pooled per sample. Tissue was then placed in C Tubes (Miltenyi Biotec, 130-093-237), along with dissociation reagents from the MACS Neural Tissue Dissociation Kit (P) (Miltenyi Biotec, 130-092-628). Samples were loaded onto a gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec, 130-096-427) and the 37C_NTDK_1 program was run. Following dissociation, samples were filtered through a 70 µm cell strainer, washed in HBSS and resuspended in MACS buffer with myelin removal beads (Miltenyi Biotec, 130-096-733), then purified with an LS column (Miltenyi Biotec, 130-042-401), according to manufacturer instructions. Eluent was washed twice, incubated with DRAQ5 (Bio-Legend, 424101, 1:1,000) and CD16/CD32 Fc block (BD Biosciences, 553141, 1:200) at room temperature for 15 min. Cells were washed and incubated with antibodies against CD31 (FITC, BD Biosciences, 561813, 1:200) and CD11b (BV421, BioLegend, 101235, 1:100) for 30 min on ice. Cells were washed and resuspended in FACS buffer with propidium iodide (1:10,000). Live, nucleated cells (DRAQ5⁺ PI^lo) were sorted on a FACSAria II (BD), equipped with 355 nm, 405 nm, 488 nm, 561 nm and 640 nm lasers and a 130 µm nozzle. In a subset of experiments, CD31⁺ and CD11b⁺ populations were collected to enrich for cell types of interest. Sequencing was performed by the Columbia Single Cell Core using 10X Genomics Chromium Single Cell 3’ technology, with reads aligned to the mm10-2020-A transcriptome.

Immunofluorescence

Mice were anesthetized with isoflurane and perfused intracardially with PBS for 4 min, followed by 4% paraformaldehyde (PFA) for 6 min. Brains were extracted and post-fixed in 4% PFA for 4–6 h, then washed three times in PBS, incubated overnight in 30% sucrose, embedded in Tissue Plus Optimal Cutting Temperature compound (Thermo Fisher, 4585) and stored at −80 °C. Coronal sections (12 µm) were cut on a Leica CM3050 S Cryostat and stored at −80 °C. For immunofluorescence staining, slides were washed in PBS for 10 min, incubated for 1 h at room temperature in blocking buffer (10% BSA in 1X PBS with 0.1% Triton-X-100), and with primary antibodies (Table 2) diluted in PBST (0.1% Triton-X-100 in 1X PBS) with 1% BSA overnight at 4 °C.

Table 2.

Primary and secondary antibodies used for immunofluorescence staining

Target and fluorophore	Manufacturer	Catalog	Dilution
Caveolin-1	Abcam	ab18199	1:2000
CD4	BD Pharmingen	553727	1:100
CD68	Abcam	ab53444	1:2000
GLUT1	Millipore Sigma	400060	1:1000
Iba1	WAKO	016-20001	1:500
Mfsd2a	Chenghua Gu (Harvard Medical School)	Pfau et al.103	1:200
Iba1	Synaptic Systems	234009	1:400
Iba1	Abcam	AB5076	1:500
Ifitm3	Proteintech	11714-1-AP	1:1000
CD74	BioLegend	151002	1:300
CD206	Bio-Rad	MCA2235	1:200
Mfsd2a	Cell Signaling Technologies	80302	1:1000
CD68	Abcam	AB53444	1:500
ZO-1	Invitrogen	61-7300	1:100
Claudin-5	Invitrogen	35-2500	1:100
Goat Anti-Mouse IgG (H + L) – Alexa Fluor 488	Thermo Fisher (Invitrogen)	A-11001	1:1000
Goat Anti-Mouse IgG (H + L) – Alexa Fluor 594	Thermo Fisher (Invitrogen)	A-11032	1:1000
Goat Anti-Armenian Hamster IgG (H + L) – Alexa Fluor 594	BioLegend	405512	1;1000
Goat Anti-Rabbit IgG (H + L) – Alexa Fluor 488	Thermo Fisher (Invitrogen)	A-11008	1:1000
Goat Anti-Rabbit IgG (H + L) – Alexa Fluor 594	Thermo Fisher (Invitrogen)	A-11012	1:1000
Donkey Anti-Rabbit IgG (H + L) – Alexa Fluor 594	Thermo Fisher (Invitrogen)	A-21207	1:1000
Donkey Anti-Goat IgG (H + L) – Alexa Fluor 647	Thermo Fisher (Invitrogen)	A-21447	1:500
Goat Anti-Rat IgG (H + L) – Alexa Fluor 488	Thermo Fisher (Invitrogen)	A-11006	1:1000
Goat Anti-Chicken IgY (H&L) – Alexa Fluor 647	Abcam	ab150171	1:500
Donkey Anti-Rat IgG (H + L) – Alexa Fluor 488	Thermo Fisher (Invitrogen)	A-21208	1:1000

Slides were incubated in primary antibodies at 4 °C in a humidified chamber overnight. After three 10-min washes with PBST, slides were incubated for 2 h at room temperature in secondary antibodies diluted in PBST with 1% BSA. These were conjugated to AlexaFluor 488 (1:1000), AlexaFluor 594 (1:1000), or AlexaFluor 647 (1:500) (Table 2). Following three washes with PBST and two with PBS, slides were cover slipped with Vectashield (Vector Labs, Burlingame, CA) containing the nuclear stain DAPI, then sealed with clear nail polish and stored at −20 °C.

In situ hybridization

Plasmids were obtained from Transomic Technologies, and the antisense mRNAs were synthesized using the Digoxigenin RNA Labeling Kit (SP6/T7; Roche, 11175025910). DIG RNA in situ hybridization (ISH) and fluorescent in situ hybridization (FISH) experiments were performed as described^93,94. Mice used for ISH or FISH were anesthetized with isoflurane and intracardially perfused for 4 min with PBS and brains were dissected out and embedded in Tissue Plus Optimal Cutting Temperature compound (Thermo Fisher, 4585).

Multiplexed error-robust in situ hybridization (MERFISH)

Mice were anesthetized with isoflurane and intracardially perfused with cold RNAse-free PBS for 4 min. Brains were dissected out and immediately embedded in Tissue-Plus O.C.T compound and stored at −80 °C until samples could be shipped overnight on dry ice to UMass Chan Medical School. Samples were prepared according to the Vizgen Fresh Frozen Tissue Sample Preparation protocol. Tissue was sectioned in 10 µm slices onto a functionalized coverslip covered with fluorescent beads. Each coverslip contained a section from one PBS sample and a section from one GAS sample. Tissue on coverslips was fixed for 15 min at room temperature in 4% paraformaldehyde in PBS, followed by three washes with PBS. Tissue was then permeabilized in 70% ethanol for 24 h, washed with PBS and incubated with blocking solution for 1 h, followed by 1 h of incubation with the primary antibody against vessel marker CD31 (BioLegend, 102502), diluted 1:20 in blocking solution (Vizgen). The tissue was then washed three times with PBS and incubated for 1 h with an oligo-conjugated secondary antibody diluted 1:1000 in blocking solution. The sample was washed three times with PBS and fixed for 15 min at room temperature in 4% paraformaldehyde in PBS, followed by three washes with PBS. After a 30-min wash with Formamide Wash Buffer (30% formamide in 2X saline sodium citrate, or SSC) at 37 °C, the MERFISH library mix was added and allowed to hybridize for 48 h. Sample was then washed and incubated at 47 °C with Formamide Wash Buffer twice for 30 min each and then the tissue was embedded in a polyacrylamide gel followed by incubation with tissue clearing solution (2X SSC, 2% SDS, 0.5% v/v Triton X-100, and proteinase K 1:100) overnight at 37 °C. Then, tissue was washed and hybridized for 15 min with the first hybridization buffer containing DAPI, polyT and the readout probes associated with the first round of imaging. After washing, the coverslip was assembled into the imaging chamber and placed into the microscope for imaging. MERFISH imaging was performed as previously described⁹⁵ with parameter files provided by Vizgen. Briefly, the sample was loaded into a flow chamber connected to the Vizgen Alpha Instrument. First, a low-resolution mosaic image was acquired (405 nm channel) with a low magnification objective (10×). Then the objective was switched to a high magnification objective (60×) and seven 1.5-μm z-stack images of each field of view position were generated in 749 nm, 638 nm and 546 nm channels. A single image of the fiducial beads on the surface of the coverslip was acquired and used as a spatial reference (477 nm channel). After each round of imaging, the readout probes were extinguished, and the sample was hybridized with the next set of readout probes. This process was repeated until combinatorial FISH was completed.

Raw data were decoded using the MERlin pipeline (Vizgen, v0.1.12) using the relevant library codebook. Cell boundaries were segmented in each FOV using a seeded watershed algorithm with DAPI signal as the seed and poly-T signal as the watershed channel. The volume, X position, and Y position of these cell boundaries, as well as the probe counts within each cell boundary, were output for further analysis.

Probes for the following transcripts were used: Abca7, Abcc3, Abcg2, Ablim1, Actb, Acvrl1, Adam10, Adam17, Adgrf5, Adgrl4, Adora1, Aff3, Ago4, Agt, Ahr, Akap12, Aldoc, Anxa1, Ap2m1, Aqp4, Arc, Arg1, Arhgap29, Arl15, Arpc2, Atmin, Atp10a, Axl, Baiap2l1, Bard1, Bin1, Birc5, Bmp6, Brca1, Btk, C1qa, C1qb, C1qbp, C1qc, C3, C3ar1, C4a, C5ar1, Cald1, Casp7, Casp8, Cass4, Ccl2, Ccl22, Ccl3, Ccr2, Cd14, Cd163, Cd27, Cd33, Cd3e, Cd4, Cd47, Cd68, Cd72, Cd74, Cd79a, Cd84, Cd86, Cd8a, Cdh5, Cdh9, Cemip, Cenpa, Cgnl1, Chek2, Chit1, Cldn5, Clec7a, Clic4, Clu, Cmtm8, Cobll1, Col6a3, Cotl1, Crim1, Csad, Csf1r, Csf2, Csf2ra, Csf2rb, Cspg4, Cstb, Ctgf, Ctsb, Cx3cr1, Cxcl10, Cyr61, Dach1, Dapk1, Dclre1a, Ddx58, Des, Dlc1, Dna2, Dock2, Dock9, Dusp1, E2f1, Ebf1, Ebi3, Ece1, Edn1, Edn3, Efnb2, Egfl7, Egfr, Egr1, Elovl7, Emcn, Emp1, Eng, Enpp6, Entpd1, Epas1, Epb41l4a, Epha1, Erg, Esam, Esyt2, Fancd2, Fbrs, Fbxw17, Fcer1g, Fcgr1, Fcgrt, Fcrls, Fgd2, Flcn, Flnb, Flt1, Flt3, Flt4, Fn1, Folr2, Fos, Foxj1, Foxp1, Ftsj3, Gad1, Gad2, Galnt18, Gbp2, Gfer, Gna13, Gna15, Gpi1, Gpr183, Gpr34, Gpr84, Grb10, Grn, Gusb, H2-D1, Hcar2, Heg1, Hells, Hexb, Hmgb2, Hmox1, Icam1, Ifi30, Ifih1, Igsf6, Il1a, Il1b, Il1rl2, Il1rn, Il21r, Il2rg, Il3ra, Impact, Inpp5d, Irak4, Irf7, Irf8, Itga1, Itga2, Itga6, Itgae, Itgal, Itgam, Itgax, Itgb1, Itgb3, Itgb5, Itm2a, Jcad, Jun, Kif26a, Kit, Lacc1, Lair1, Lama2, Lcn2, Ldb2, Ldlrad3, Lef1, Lfng, Lgals3, Lgmn, Lig1, Liph, Lmnb1, Lrch3, Lrp1, Lrrc3, Ly6g, Ly9, Lyn, Lyve1, Lyz2, Mb21d1, Mcm2, Mcm5, Mcm6, Mctp1, Mecom, Mef2c, Meg3, Mertk, Mfsd6l, Mgat1, Mki67, Mkl2, Mmp2, Mmp9, Mpnd, Mrc1, Ms4a1, Ms4a6d, Ms4a7, Msn, Mvp, Mybl2, Myrip, Nampt, Napsa, Ncaph, Ncf1, Nckap1l, Nebl, Nfib, Nlrp3, Nos3, Nostrin, Nrm, Ocln, Olfm2, Olfml3, Osm, Osmr, P2ry12, Palmd, Pam, Pard3, Pcna, Pde10a, Pde4d, Pdgfra, Pdgfrb, Pdlim5, Pdpn, Pdzrn3, Pecam1, Picalm, Pik3cg, Pla2g4a, Plac8, Plcb4, Plcg2, Plekha6, Plekhg1, Plp1, Plpp1, Pltp, Plxdc2, Plxna2, Podxl, Pole, Ppfibp1, Prdx5, Prickle2, Prkg1, Pros1, Psmb8, Ptk2b, Ptpn6, Ptprc, Ptprg, Ptprj, Ptprm, Pvalb, Qars, Rad23b, Rad51, Rae1, Rapgef4, Rbms3, Rngtt, Rora, Rorb, Rrm2, Rsad2, Rundc3b, Sall1, Sardh, Sash1, Sdf4, Sell, Serpina3n, Serpine1, Serpinf1, Siglecf, Slamf8, Slamf9, Slc17a6, Slc1a1, Slc39a10, Slc40a1, Slc4a4, Slc7a1, Slco2b1, Slfn8, Smc3, Snx2, Sorbs1, Sorbs2, Sox10, Sox2, Sox9, Spi1, Spp1, Sptbn1, Srgn, Sst, St8sia6, Stat1, Syk, Syne1, Syne2, Tacc1, Tead1, Tek, Tgfa, Tgfbi, Tgfbr1, Tgfbr2, Tgm2, Thsd4, Timeless, Timp2, Timp3, Tlr2, Tlr4, Tmem119, Tmem173, Tmtc1, Tnfrsf11a, Tnfrsf1a, Tnfsf13b, Top2a, Trem1, Trem2, Trim47, Tshz2, Tspan33, Ttll12, Ttr, Txnrd1, Tyrobp, Unc13b, Ung, Utrn, Vac14, Vav1, Vcl, Vegfa, Vim, Vip, Vtn, Vwf, Was, Wwtr1, Xpo1, Zbp1, Zbtb46.

Flow cytometry

Mice were anesthetized with isoflurane and intracardially perfused with PBS for 4 min. Brain, as well as combined nasal associated lymphoid tissue/olfactory epithelium (NALT/OE), were dissected out, placed in cold Dulbecco’s Modified Eagle’s Medium (DMEM) (Genesee, 25–500), and pressed through a cell strainer with the end of a sterile syringe. Samples were collected in 10 mL of a 30% Percoll (Cytiva, GE17-0891-01) suspension in DMEM, and underlaid with 1 mL of 70% Percoll. Spleen samples were suspended in 3 mL of Red Blood Cell Lysis Buffer (155 mM NH₄Cl, 10 mM KHCO₃, 0.1 mM EDTA) at room temperature for 10 min. Samples were centrifuged at 4 °C at 1300 × g for 30 min, then immune cells were collected at the interface. All samples were then filtered through a cell strainer, washed with 2 mL DMEM, and centrifuged for 10 minutes at 800 × g, and resuspended in T cell media (RPMI with fetal bovine serum (FBS) 1:10, penicillin/streptomycin 1:100, MEM-NEAA 1:100, glutamine 1:100, 55 µM β-mercaptoethanol 1:1000) with 1X Cell Stimulation Cocktail (plus protein transport inhibitors) (eBioscience, 00-4975-93). Samples were incubated at 37 °C for 4 h, washed with FACS buffer, and incubated in anti-CD16/CD32 Fc block (BD Biosciences, 553141, 1:200) for 15 min on ice. All subsequent steps were performed at 4 °C. After washing with FACS buffer, samples were incubated in cell surface stains (Table 3) with either Fixable Viability Dye 780 (Invitrogen, 65086518, 1:4000) or Live/Dead Aqua (ThermoFisher, L34965, 1:1000) in the dark for 1 h, fixed for 30 min using the Intracellular Fixation & Permeabilization kit (eBioscience, 88-8824-00) according to manufacturer instructions. Intracellular stains were diluted in 1X permeabilization buffer and incubated for 1 h, then washed in permeabilization buffer and resuspended in FACS buffer.

Table 3.

Antibodies used for flow cytometry sorting and staining

Target and fluorophore	Manufacturer	Catalog	Dilution
CCL5 PE-Cy7	BioLegend	149105	1:200
CD31 FITC	BD Biosciences	553372	1:100
CD4 BV605	BD Biosciences	563151	1:100
CD11b BV421	BioLegend	101235	1:100
CD11b PerCP-Cy5.5	BioLegend	101227	1:100
CD44 BV421	BioLegend	103040	1:200
CD45 AF700	BioLegend	103127	1:100
CD45 BUV395	BD Biosciences	564279	1:100
CD45 BV421	BD Biosciences	563890	1:100
CD62L PE-Cy7	BioLegend	104417	1:100
CD74 BV711	BD Biosciences	740748	1:400
CX3CR1 BV786	BioLegend	149029	1:400
GM-CSF FITC	BioLegend	505403	1:200
IFNγ APC	BD Biosciences	554413	1:100
IL-17A PE	BD Biosciences	559502	1:100
MHC II I-A/I-E APC-Cy7	BioLegend	107627	1:400
P2RY12 PE	BioLegend	848003	1:100

Samples were analyzed in the Columbia Stem Cell Initiative Flow Cytometry Core. Time course experiments were analyzed using a ZE-5 analyzer (Bio-Rad, Hercules, CA) equipped with 355 nm, 405 nm, 488 nm, 561 nm, and 640 nm lasers. Compensation controls used splenocytes for most cell surface markers and compensation beads (BD Biosciences, 552844 for rat hosts; Thermo Fisher, 01-3333-41 for mouse or Armenian hamster hosts) for cytokines. All other flow cytometry experiments were analyzed using a NovoCyte Penteon (Agilent, Santa Clara, CA) equipped with 349 nm, 405 nm, 488 nm, 561 nm and 637 nm lasers. Compensation controls used splenocytes for live/dead control and compensation beads for all other markers.

All flow cytometry data analysis was performed using FlowJo 10.5 (FlowJo, LLC). Gates for forward and side scatter, singlets, live cells and CD4⁺ T cells were set by eye, while all other gates were set using fluorescence minus one (FMO) controls, with a typical cutoff of <1% of the population.

Cell culture

Primary mouse brain endothelial cells (mBECs; Cell Biologics, C57-6023) and primary human microvascular endothelial cells (HBMEC; Cell System, ACBRI 376) were cultured as monolayers at 37 °C with 5% CO₂ and used to evaluate the effect of cytokines in vitro. Cells were grown to confluence on Collagen IV and Fibronectin-coated (Corning, CB-40233, 356008) dishes in endothelial cell media (Cell Biologics, M1168) for mBECs and Endothelial Cell Basal Medium MV2 (PromoCell, C-22221) supplemented with 10% FBS (Cytiva, SH30071.03) and supplements recommended by the supplier for HBMECs. One day prior to cytokine treatment, cells were switched to 2% FBS without growth factor supplements. Media containing either mouse IFNγ (R&D Systems, 485-MI-100), mouse IL-17A (R&D Systems, 7956-ML-025), mouse GM-CSF (R&D Systems, 415-ML-010) or vehicle (PBS with Ca²⁺ and Mg²⁺) was applied to corresponding wells.

Transendothelial electrical resistance (TEER) was measured in real time using an electric cell-substrate impedance sensing (ECIS) instrument (Applied BioPhysics, ZTheta 96 Well Array Station) as previously described⁹⁴. mBECs and HBMECs were plated on 96-well plates containing electrode arrays (Applied BioPhysics, 96W20idf). Cytokines (Il-17A, GM-CSF, IFNγ, CCL2) were added at a concentration of 50 ng/mL, except IL-1β (R&D Systems, 401-ML) and TNF (R&D Systems, 410-MT), which were added at 10 ng/mL. Resistance was monitored over 24 h from the start of cytokine treatment. The area under the curve (AUC) was calculated for each condition and normalized to AUC of vehicle-treated cells, with values for each independent experiment compared by one-way ANOVA.

Albumin transcytosis was measured in vitro by culturing mBECs on collagen IV-coated 3.0-µm PET membrane inserts for 24-well plates (Corning, 353096). Cytokines or vehicle were added at a concentration of 100 ng/mL, with 500 ng/mL lipopolysaccharide (LPS) used as a positive control. Cells were switched to endothelial cell media with 2% FBS and without phenol. Media added to wells contained bovine serum albumin at 400 µg/mL, while media added to inside of transwell inserts contained albumin conjugated to Alexa Fluor 647 (Thermo Fisher, A34785) at 400 µg/mL. Cells were incubated at 37 °C with flow-through samples collected from bottom wells at 30 minutes, 1, 2, 4 and 6 h). Absorbance of flow-through was quantified using the accuSkan FC plate reader (Fisher Scientific, 14-377-576). Background fluorescence was subtracted, and AUC normalized to untreated for each experiment.

Multiplex immunoassays

Mouse olfactory bulb multiplex immunoassay

Twenty-four hours after the final GAS inoculation, mice were anesthetized with isoflurane and intracardially perfused with PBS for 4 min, pairs of OBs were dissected and flash frozen in liquid nitrogen, then stored at −80 °C. Samples were pulverized on ice in cell lysis buffer (Abcam, ab152163) with protease inhibitor cocktail (Thermo Fisher, 78440) and EDTA, using an electric pestle. To remove detergents that may interfere with downstream analysis, the resulting supernatant was dialyzed overnight against PBS with a 2 kDa cassette (Thermo Fisher, 66205). After normalization to total protein concentration by Pierce bicinchoninic acid assay (Thermo Fisher, 23225), analytes were measured by the Irving Institute for Clinical and Translational Research Biomarkers Core Laboratory using a custom mouse Luminex panel (Thermo Fisher, PPX-12-MXEPUF3). Samples were run in duplicate, and standard curves were generated for each analyte. Undetectable values were replaced with half of the lower detection limit for purposes of statistical comparison.

Patient serum multiplex immunoassay

Serum protein concentrations were measured by the Irving Institute for Clinical and Translational Research Biomarkers Core Laboratory using a 45-Plex Luminex assay (Invitrogen, EPX260-26088-901). Samples were run in duplicate, and standard curves were run for each analyte. Undetectable values were replaced with half of the lower detection limit (Supplementary Data 6) for statistical comparisons.

Quantification and statistical analysis

Analysis of scRNA-seq data

Single-cell RNA sequencing data (Supplementary Data 1) was analyzed using Seurat package v4.4.1⁹⁶ in RStudio. Upon data import, genes detected in fewer than three cells, and cells with fewer than 200 genes were excluded. Cells were removed from the merged data set if they had fewer than 1000 or more than 50,000 molecules detected, or greater than 20% mitochondrial reads. Data was normalized and highly variable features identified using default parameters, then scaled, followed by linear dimensional reduction using PCA. Dimensionality of the data was selected using the Elbow plot method with 50 dimensions, and cells were clustered with a resolution of 1 for OB, and 0.4 for endothelial cells and microglia. Dimensionality reduction for visualization was performed with uniform manifold approximation and projection (UMAP). The Harmony package v1.2⁹⁷ was used for batch correction.

Cluster identity was assigned using the following cell type markers: neurons (Map2, Snap25), astrocytes (Gfap, Aqp4), olfactory ensheathing cells (Frzb), oligodendrocytes/oligodendrocyte precursor cells (Pdgfra), endothelial cells (Cldn5, Pecam1), pericytes (Pdgfrb, Atp13a5), fibroblasts (Col1a1, Fbln1), microglia (Tmem119, P2ry12), macrophages (Aif1, Plac8), neutrophils (Ly6g, Camp), dendritic cells (Xcr1, Ccr9, Cd209a), CD4 T cells (Cd4), CD8 T cells (Cd8a), NK cells (Klrb1c), and B cells (Cd19).

Differential expression analysis was performed using a mixed-effects model algorithm (MAST)⁹⁸ to avoid pseudo-replication bias⁹⁹ (Supplementary Data 2, 3). Expression from biological replicates was aggregated for heatmap visualization and significance by mixed-effects analysis displayed by formatting of gene names. Expression of ex vivo activation genes (Dusp1, Fos, Hist1h1d, Hist1h2ac, Jun, Nfkbid, Nfkbiz) were added using AddModuleScore and plotted against Ccl3 and Ccl4 using FeatureScatter. Additional analysis was performed with BB Browser 3 (BioTuring) software. Signature scores were plotted in BBrowser3 using the following pathway markers: Antigen presentation (B2m, Cd74, H2-Aa, H2-Ab1, H2-D1, H2-Eb1, H2-K1, H2-Q4, H2-Q6, H2-Q7, Tap1, Tap2), disease-associated microglia (Apoe, Axl, Cd9, Csf1, Cst7, Itgax, Lpl, Spp1, Tyrobp), homeostatic microglia (Cd33, Cst3, Cx3cr1, Fcrls, Gpr34, Olfml3, P2ry12, P2ry13, Sall1, Tmem119), and interferon signaling (Ifi30, Ifi204, Ifi211, Ifit1, Ifitm3, Irf1, Irf7, Isg15, Oas1a, Stat1, Stat2).

Gene set enrichment analysis (GSEA)^37,38 with curated pathways was performed using curated and database-derived gene lists for blood-brain barrier, response to LPS, inflammation, extracellular matrix, interferon response, antigen presentation, chemokine and cytokine signaling, endothelial cell proliferation, endothelial cell migration, disease-associated microglia, apoptosis, leukocyte chemotaxis and phagocytosis (Supplementary Data 4). Analysis was run using the GSEA desktop tool (Broad Institute, v4.1.0), using pre-ranked weighted settings. GSEA using gene ontology (GO) pathway lists was performed using the enrichplot R package v1.24.4 (Supplementary Data 5). Cross entropy analysis was performed on UMAP coordinates using the Cross-Entropy-test³⁴ in R. Receptor-ligand analysis was performed using CellChat v1.6.1^51,52.

Analysis of MERFISH data

MERFISH data was analyzed in RStudio using Seurat 4.1.0.9005, R 4.0.0 and custom-made scripts as previously described¹⁰⁰. Cell segmentations with volume <50 µm³ or <10 unique transcripts were first excluded. Cell gene expression data of each cell was then normalized to that cell’s volume and the total transcript count of that cell, then scaled. To correct for global differences in total transcript counts between coverslips (each containing one GAS sample and one PBS sample), we performed ComBat¹⁰¹ batch correction (sva 3.38.0).

To identify individual cell types, we performed principal component analysis was performed using the entire probe library (391 transcripts) as the variable features, followed by linear dimensional reduction. Dimensionality of the data was selected using the jackstraw method with 28 dimensions, and cells were clustered with a resolution of 2.4. Dimensionality reduction for visualization was performed with uniform manifold approximation and projection (UMAP). Clusters were manually annotated based on the spatial distribution of the cells in the tissue and the expression cell type-specific marker genes: neurons (Meg3, Gad1), astrocytes (Aqp4, Sox9), olfactory ensheathing cells (Plp1, Cldn5), oligodendrocytes/oligodendrocyte precursor cells (Pdgfra, Sox10), endothelial cells (Cldn5, Itm2a), pericytes (Pdgfrb), fibroblasts (Cemip), microglia (Tmem119, P2ry12), macrophages (Mrc1), neutrophils (Itgal, Mmp9), and T cells (Cd3e, Cd4, Cd8a).

Because of imperfections in cell boundary segmentation, a small fraction of cells expressed cell type markers for multiple cell types. Raw images of a subset of these cells were visually inspected using MERSCOPE Visualizer software (Vizgen, 2.1.2589.1) to confirm that these clusters were due to cell segmentation errors (typified by two distinct clusters of cell-type-specific transcripts within the same cell boundary). Clusters composed of these “hybrid” cells were removed from the analysis, and embedding and clustering analysis were iteratively repeated until all “hybrid” clusters were removed.

The glomerular, external plexiform, and granular layers of the OB for each sample were outlined using MobileFish and coordinates recorded for point-in-polygon analysis and regional assignment of microglia. Raw counts were normalized to the PBS condition for each batch and used for gene expression analysis. Endothelial cell gene expression was compared on log₂ fold change values using a one-sample t test. Microglia gene expression between the glomerular and granular layers used a ratio t test. Nearest neighbor analysis of microglial distance to T cells was calculated based the x,y coordinates of the centers of the cell segmentations using a custom python script.

Statistical analysis

Most statistical analyses were performed by GraphPad Prism 10.3.0. All tests were two-sided using a significance level α = 0.05. Outliers were identified and excluded using the ROUT method (Q = 1%). Significance was notated as ns, p > 0.05; *, p < 0.05; **, p < 0.01; ***, p < 0.001. Error bars represent mean ± SEM throughout. Due to word count limitations in the figure legends, comprehensive statistical details, including specific tests, degrees of freedom, 95% confidence intervals, effect sizes, and exact p values for all comparisons are provided in the Source Data files associated with this manuscript. Unless otherwise indicated, all statistical tests were two-sided. To ensure reproducibility and control for potential litter- and cage-dependent confounding variables, all in vivo experiments were independently repeated at least twice at different times. Mice from multiple independent litters and separate cages were randomized across all experimental groups.

Immunofluorescence quantification

Quantification of microglial number

Three OB sections, corresponding to bregmas 4.5, 4.28 and 3.92, were imaged using a Zeiss AxioImager microscope at magnification 20× for each animal. The number of IBA1⁺CD68⁺ cells (corresponding to activated microglia) in the glomerular layer was manually counted, and averaged across the three sections. Iba1⁺ cells (microglia) were considered CD68⁺ if the CD68 fluorescence occupied more than 50% of the cell surface area, as previously described^23,25.

Quantification of BBB leakage

Bregma 4.28 sections were imaged using a Zeiss AxioImager microscope at magnification 10×. Using ImageJ¹⁰², 20 small, rectangular regions of interest (ROIs) were placed around the glomerular layer of the OB, avoiding the vasculature, and average fluorescence quantified for each animal. The same process was repeated for the granular layer.

Quantification of BEC and microglial marker expression

Tiled images of bregma 4.28 sections were taken at 20× using a Zeiss LSM700 or LSM900 confocal microscope and Zeiss AxioImager M2, and maximum intensity projections were created in the software. Using FIJI, ROIs across the whole OB were selected using Otsu thresholding on a vessel marker. Then average fluorescence intensity was measured within the ROIs. Due to neuronal expression of Itih5 in the glomerular layer, ROIs were restricted to the granular layer of the OB for quantification. For the microglial marker analyses, Iba1 positive cells were used for generating a mask and CD74, CD68, Ifitm3 were measured within Iba1⁺ microglial cell mask.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: C.R.W., T.C., and D.A.; Animal experiments: C.R.W., U.A., and D.A.; RNA sequencing experiments: C.R.W. and V.M.; MERFISH experiments and analysis: V.D.L., T.E.F., C.R.W., and D.P.S.; Bioinformatic analyses: C.R.W., T.E.F., and V.M.; Flow cytometry experiments: C.R.W., L.B., and S.J.H.; Immunofluorescence experiments: C.R.W., U.A., L.B., S.J.H., and D.A.; In situ hybridization experiments: D.A., B.A., and C.R.W.; Data analyses: C.R.W., U.A., D.J., and D.A.; Cell culture experiments: C.R.W., U.A., T.C., D.A., and N.A.; Patient history & sample collection: S.L.D., S.S., and W.V.; Processing and analysis of patient samples: T.C. and N.A.; Statistical analyses: C.R.W.; Resources: D.A., B.C., and D.P.S.; Funding acquisition: D.A., T.C., and U.A.; Supervision: D.A., T.C., and D.P.S.; Writing: C.R.W. and D.A.; Revising: C.R.W., U.A., B.C., T.C., and D.A.

Peer review

Peer review information

Nature Communications thanks Ari Waisman and other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

D.A., C.W., U.A. N.A. and T.C. disclose support for the research of this work from the following funders: NIMH [R01MH112849 and R56MH109987]; NINDS [2T32NS064928]; NHLBI [R61/R33 HL159949] and NIA [RF1AG078352], NEI [R01EY033994], International OCD Foundation, PANDAS Network, PANDAS Physician Network, the Global Lyme Alliance and National Center for Advancing Translational Sciences, National Institutes of Health [UL1TR001875]. We are grateful to the following major donors for their generous support of this research: Tom and Patti Walz, Newport Equities LLC, PANDAS Network, Northwest PANDAS/PANS Network, Steve and Wendy Swyter, the Look Foundation, The Alex Manfull Foundation. We also acknowledge the many families whose children were affected by these neuropsychiatric disorders (PANDAS/PANS) for their contributions to this study. DPS, TEF, and VDL disclose support for the research of this work from the following funders: NIMH [R01MH113743], NINDS [R01NS117533], NIA [RF1AG068281 and R01AG068281], Massachusetts Life Sciences Center, BrightFocus Foundation [A2022006F], Alzheimer’s Association [AARF-22-923219], and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. L.B., B.A., S.H., B.C., S.L.D., W.S.V., S.S., and V.M. declare no relevant funding.

Data availability

Raw sequencing data, metadata and count tables for scRNAseq samples have been made available in the Gene Expression Omnibus (GEO) under GSE221724. Counts, metadata and processed data for MERFISH samples have been made available at GEO under GSE221106. Processed MERFISH.vzg files for browsing on MERSCOPE visualizer software (Vizgen) and the MERFISH raw output files are available upon request. Source data are provided with this paper.

Code availability

The code for the analysis of scRNAseq data is provided as a Source Data with the manuscript. It is available at GitHub at http://github.com/AgalliuLab/s-pyogenes-scRNAseq.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76232-w.