. Author manuscript; available in PMC: 2026 May 28.

Published in final edited form as: Nature. 2025 May 28;643(8071):499–508. doi: 10.1038/s41586-025-09050-7

Subfornical organ is a nucleus for gut-derived T cells regulating behavior

Tomomi M Yoshida ^1,^2,³, Mytien Nguyen ¹, Le Zhang ^3,⁵, Benjamin Y Lu ^1,³, Biqing Zhu ⁶, Katie Murray ³, Yann S Mineur ⁷, Cuiling Zhang ², Di Xu ^1,², Elizabeth Lin ^1,^2,³, Joseph Luchsinger ⁷, Sagar Bhatta ⁴, Daniel A Waizman ^1,², Mackenzie Coden ^1,², Yifan Ma ^1,², Kavita Israni-Winger ^1,², Anthony Russo ³, Haowei Wang ³, Wenzhi Song ^1,², Jafar Al Souz ^1,², Hongyu Zhao ^6,⁸, Joseph E Craft ^1,², Marina R Picciotto ⁷, Jaime Grutzendler ³, Marcello Distasio ⁴, Noah W Palm ¹, David A Hafler ^1,³, Andrew Wang ^1,²

PMCID: PMC12768464 NIHMSID: NIHMS2123216 PMID: 40437096

Abstract

Specialized immune cells residing in tissues orchestrate diverse biological functions through communication with parenchymal cells¹. The contribution of the innate immune compartment in the meninges and central nervous system (CNS) is well-characterized; however, whether cells of the adaptive immune system reside in the brain and are involved in maintaining homeostasis is poorly understood^2–4. Here, we identify that the subfornical organ (SFO) of the brain is a nucleus for parenchymal α T cells in the steady-state brain in both mice and humans. Using unbiased transcriptomics, we found that these extravascular T cells in the brain are distinct from meningeal T cells and more robustly secrete IFNγ and express tissue residence proteins, including CXCR6, which are required for their retention in the brain and for normal adaptive behaviour. These T cells are primed in the periphery by the microbiome and traffic from the white adipose and gastrointestinal tissues to the brain. Once established, their numbers can be modulated by either altering the gut microbiota or adipose tissue composition. Thus, we find that CD4 T cells reside in the brain at steady state and are anatomically concentrated in the SFO in mouse and man, are transcriptionally and functionally distinct from meningeal T cells, and secrete IFNγ to maintain CNS homeostasis via a homeostatic fat-brain and gut-brain axis.

One-Sentence Summary:

CD4 T cells residing in specific regions of the steady-state brain are programmed in the periphery to coordinate adaptive behaviour.

Virtually all organs in the body are home to specialized immune cells that are transcriptionally adapted to reside in the tissue and orchestrate diverse biologic functions through communication with parenchymal and non-parenchymal cells. Immune cells, including T cells, can be programmed to coordinate specific responses through a combination of their T cell receptor (TCR) and contextualizing signals encountered in their environment to communicate what they sensed by relaying soluble factors (cytokines) or by direct cell-cell signaling. Gut-educated CD4 T cells have been implicated in multiple neurological disease^5–10, but whether a homeostatic gut-immune-brain axis exists remains underexplored. While some groups have reported T cells in central nervous system (CNS) tissue under steady-state conditions^2–4, it is still widely assumed that the presence of lymphocytes in the brain indicates underlying neuropathology due to the concept of immune privilege^11,12. The meningeal lymphatic network surrounding the brain parenchyma has thus been considered to be the main brain-adjacent tissue where immune cells reside at homeostasis. Previous studies suggested that meningeal T cell-derived cytokines^13–18 are key factors that are required for normal mouse behaviours at steady-state. However, whether and how these meningeal T cells or their effector molecules enter the CNS and correctly find their intended cellular targets to coordinate organismal responses are not understood. Indeed, whether and where T cells are present in the mammalian brain at steady remains a topic of controversy^2–4.

Brain T cells are distinct and localized in the SFO

To assess whether extravascular CD4 T cells exist in the steady-state mouse brain, we first utilized immunohistochemistry-based fluorescent microscopy. After surveying sections throughout the entire brain with positive signal, we focused on nine different brain regions identified in this screen. While relatively rare, we found extravascular CD3⁺CD4⁺ cells in the brain which were spatially separated from blood vessels (Fig. 1a, Supplementary Video 1 and Extended Data Fig. 1a). For the dura, we additionally stained for Lyve-1 to identify lymphatic vessels (Fig. 1b, Extended Data. Fig 1b and Supplementary Video 2). After quantifying the density in each brain region, we found that extravascular CD3⁺CD4⁺ cells were most densely found in the subfornical organ (SFO) (Fig. 1b–d and Supplementary Videos 3,4). The SFO is a sensory circumventricular organ (CVO)¹⁹ and regulates important physiological functions such as feeding and drinking-related behaviour and autonomic function^19,20 and is implicated in sensing systemic inflammation^21,22. The areas with the next highest density of extravascular T cells were the dura, brainstem medulla, corpus callosum and stria terminalis (Fig. 1b). In addition to the clearly extravascular cells, we also observed extravascular T cells that were touching or directly adjacent to the blood vessel, which we called “vascular adjacent” (Extended Data Fig. 1c and Supplementary Video 5). Vascular adjacent T cells were also most highly present in the SFO, and secondly the choroid plexus (Extended Data Fig. 1c). The choroid plexus and SFO also had a higher number of intravascular T cells despite intracardiac perfusion prior to fixation (Extended Data Fig. 1d–e and Supplementary Video 6). Collectively, when accounting for both extravascular and vascular adjacent T cells, we found that they were most abundant in the SFO, although they were also observed highly in the choroid plexus (Fig. 1c). As a complementary approach, we employed tissue clearing and whole brain imaging using a CD4 fate-map reporter mouse (Ai 14 CD4-Cre). Widespread TdTomato expression was found in many morphologically neuronal- and glial-like cells (Extended Data Fig. 1f–i and Supplementary Video 7), which has been reported²³. Despite this limitation, upon overlaying reporter expression onto the Allen Mouse Brain Atlas, we still found that the SFO harbored one of the densest CD4 expression areas, even when compared amongst other CVOs (Extended Data Fig. 1j). Unlike in the SFO, where TdTomato marked cells morphologically consistent with T cells, in the other dense areas, TdTomato also marked cells morphologically consistent with neuronal and glial cells (Supplementary Video 7).

Fig. 1: — a, Coronal brain sections from 8–10-week-old C57BL/6 mouse stained with CD31/PECAM for blood vessels, CD4 and CD3 for identification of CD4 T cells. Representative image of extravascular CD4 T cells in the brain in the corpus callosum white matter. b, Quantification of vessel-distal extravascular CD3⁺ CD4⁺ T cells by brain region including dura meninges. c, and total extravascular and vascular adjacent CD3⁺ CD4⁺ T cells by brain region including dura meninges. SSS = superior sagittal sinus (n= 3 per brain area; pooled experiments from 8 female and 7 male mice). ****=p<0.0001 one-way ANOVA with Tukey’s multiple comparison’s test for **b & c. d,** Representative image of CD3⁺ CD4⁺ T cells in the subfornical organ by 2D (ImageJ) and 3D (LASX) projection. e, Deidentified post-mortem human brain samples were obtained from subjects without acute neurologic disease, structures of interest were dissected and submitted for histology following formalin fixation. After tissue processing with paraffin-embedding, samples were sectioned into 5μm slices and stained with antibodies against CD4, CD8 with DAB chromogen. f, same as e, except that sections were stained with antibodies against CD3, and DAB chromogen and slides were scanned on a whole slide imager and digital images were analyzed with custom scripts for the QuPath software. *=p<0.05, **=p<0.01 one-way ANOVA test (n=5; Patients 1–5 from Table 1). g, UMAP generated from single-cell-RNA-sequencing (scRNA-seq) of T cells (FACS: CD45-IV label⁻, CD11b⁻, CD19⁻, B220⁻ and RNA: *Cd3e* > 1) sorted from brain and meninges of 5 pooled 10-week-old mice. h, Comparison of gene expression between brain and meningeal T cells representing various T cell differentiation states.

In order to validate our T cell localization findings in human brain, we acquired samples from rapid autopsy cases (post mortem 3–6 hours) through the Yale Legacy Tissue Donation Program, excluding patients with underlying neurological or inflammatory disease (Table 1). We found a robust population of T cells distal from vasculature in the SFO via immunohistochemistry (Fig. 1e and Extended Data Fig. 1k). We then quantified the density of T cells across neighboring brain areas and found that the SFO contains the highest density of T cells, even compared to the choroid plexus, across five other donors who had no underlying neurological pathology (Fig. 1f, Table 1). Thus, the SFO is enriched for T cells in mouse and human brain.

Table 1:

Subject Demographic and Clinical Characteristics

Subject ID	Age	Sex	CNS Diagnoses	Systemic Diagnoses	Assay
1	66	M	None	Myocardial Infarction	IHC
2	78	M	None	Myocardial Ischemia	IHC
3	88	M	None	Myocardial Ischemia	IHC
4	60	M	None	Ventricular Arrhythmia	IHC
5	37	F	None	Aortic Aneurysm	IHC
6	62	F	None	Breast LCIS	Flow Cytometry
7	73	F	Non-AD Dementia	Breast Cancer	Flow Cytometry

Open in a new tab

We next investigated if brain T cells were transcriptionally distinct from meningeal T cells. We isolated murine brain and meningeal extravascular T cells and performed single-cell RNA-sequencing (RNA-seq). T cells separated into 16 clusters by differential gene expression (Fig. 1g, and Supplementary Table 1). Upon re-labeling the UMAP by tissue of origin, we found that some clusters contain both meningeal and brain-derived cells, but there was a clear separation between the two groups by UMAP (Fig. 1g). Upon comparing the expression of genes associated with canonical helper T cell subsets and tissue residency programs, we found that brain T cells more highly express genes associated with tissue-residency (Cd69, Runx3, Cxcr6), T_H1 polarization (Tbx21, Cxcr3, Ifng) and genes associated with T_H2 (Gata3, Il4) and T_H17 (Rorc) polarization, while meningeal T cells more highly express genes indicating naïve or stem-like memory T cell lineage (Sell, Tcf1, Lef1, Ccr7) (Fig. 1h). We did not detect many cells expressing canonical T_reg genes such as Il10, Tgfb, or Foxp3. These results demonstrate that brain T-cells are nonrandomly distributed in mouse and human brain at steady state and are transcriptionally distinct from meningeal T cells.

Brain T cells express CXCR6 and IFNγ at steady state

We next confirmed the protein expression of differentially expressed genes between brain and meningeal T cells (IFNγ, CXCR3, CD69, CXCR6) using flow cytometry. Within the brain, the extravascular CD4 T cells are predominantly CD44^hi memory T cells and highly express T-bet, CXCR3, CD69 and CXCR6, unlike the intravascular T cell fraction, which are predominantly CD62L^hi naïve T cells (Fig. 2a,b and Extended Data Fig. 2a–d). Accordingly, mice that lack the IL15 receptor have fewer CD4 T cells in the brain (Extended Data Fig. 2e). To measure and determine whether brain T cells were unique from other parenchymal T cells with regards to cytokine expression capacity, we stimulated cells across a wide range of tissues with PMA-ionomycin ex vivo. Consistent with our transcriptomic results, brain CD4 T cells were most capable of secreting IFNγ at steady-state compared to CD4 T cells isolated from other tissues, with the notable exception of white adipose tissue (white fat; Fig. 2c), a result we validated with an IFNγ reporter mouse (Extended Fig. 2f–i). IL-17A was also more significantly expressed in brain CD4 T cells, and a subpopulation of brain extravascular CD4 T cells had the capacity to secrete both IFNγ and IL-17A (Extended Data Fig. 3a–b). Brain Foxp3⁺ T_regs²⁴ represented a comparable fraction between the brain, white fat and mesenteric lymph node as measured by both a Foxp3-fate-mapping reporter mouse model and flow cytometric protein expression (Fig. 2d and Extended Data Fig. 3c). We could not detect the secretion of type II cytokines (Extended Data Fig. 3d). Mouse brain T cells also uniquely co-express multiple co-inhibitory receptors (Extended Data Fig. 3e) compared to other tissue T cells, consistent with healthy human CSF T cells³.

Fig. 2: — a, Intravascular labeling strategy to exclude circulating leukocytes. b, Representative FACS analysis of IFNγ, CXCR3, CD69 and CXCR6 on IV-label negative CD4 T cells. c, FACS analysis of IFNγ on CD4 T cells in brain, meninges, white fat, colon, brown fat, liver, lung, ovaries, spleen, and lymph nodes (mesenteric (m)LN, inguinal (i)LN and deep cervical (dc)LN) after a four-hour incubation with PMA-ionomycin and golgi-plug at 37°C. ****= p<0.0001 one-way ANOVA with Tukey’s multiple comparison’s (n=4 per group pooled from 3 experiments) d, Schematic of fate-map labeling. FACS analysis of T_regs in brain, white fat and mLN. (n=2 per group for Cre⁻, n=7 per group for Cre⁺) e, FACS analysis on fresh brain, blood and cervical lymph node (cLN) obtained from patient 6 (Table 1) 4–5 hours post-mortem. The density of T cells was calculated by dividing absolute counts with total tissue weight. (n=1; repeated twice). f, UMAP generated from single cell-RNA-sequencing of T cells (FACS: CD45-IV label⁻, CD11b⁻, CD19⁻, B220⁻ and RNA: *Cd3e* > 1) sorted from brain, meninges, gonadal white fat, colon-lamina propria, lung and blood of 5 pooled 10-week-old mice. g, T cell differentiation states across tissues. h, FACS analysis of brain, meningeal and white fat CD4 T cell numbers in mice placed on a high-fat or normal diet and then subjected to a 48-hour food deprivation. (n=4 per group pooled from 2 experiments) i, Schematic of photoconversion. FACS analysis of the proportion of RFP-expressing fat-photoconverted T cells out of the total GFP-expressing T cells present in the colon, adipose tissue, brain, mLN and spleen. 24-hours post-photolabeling or sham surgery. (UV: n=6 per group. Sham: n=4 per group pooled from 3 experiments) *=p<0.05, ***=p<0.01 by two-way ANOVA with Bonferroni’s multiple comparisons test for **h & i.**

To validate our findings in human, we obtained fresh human brain samples from a different set of rapid autopsy cases (Table 1) for flow cytometry. We collected multiple brain areas to compare T cell density and phenotypes across two donors. Since we could not intravascularly label postmortem humans, we classified T cells using surface markers. As in murine brain T cells, we found that T cells in the human brain highly express CXCR3 and CXCR6 compared to blood or lymph nodes (Fig. 2e and Extended Data Fig. 4a). By normalizing T cell counts to tissue weight, we found that the SFO and choroid plexus were the most densely populated with CXCR3⁺ or CXCR6⁺ T cells, but not when accounting for total T cells, which is likely due to blood contamination (Fig 2e and Extended Data Fig. 4a–d). Altogether, our data suggest that human brain T cells are also localized densely in the SFO and phenotypically resemble murine extravascular brain T cells.

There is a fat-brain T cells axis at steady state

From our tissue screen, we unexpectedly observed that white fat T cells also had a high frequency of IFNγ T cells. Thus, to examine the similarity of T cells across these organs, we sorted extravascular T cells from the brain, meninges, colon-lamina propria, gonadal white fat, lung and blood and performed single-cell RNA-seq (Fig 2f, Supplementary Table 2). When we compared the expression of genes associated with tissue residency and canonical helper T cell subset differentiation across all tissue T cells, we found that brain T cells most highly expressed genes associated with tissue residency (Cd69, Runx3, Cxcr6) and T_H1 subset (Tbx21, Cxcr3, Ifng), compared to the other tissues (Fig. 2g). These observations were repeated in a second 10X single-cell RNA-seq (Extended Fig. 5a, Supplementary Table 3).

In our efforts to compare extravascular T cells from the brain with other tissues, we were surprised to find how similar T cells from the brain and white fat were by single-cell RNA-seq and by protein expression of IFNγ, CXCR3, T-bet and CXCR6. We thus explored what might drive similarities between T cells from brain and fat, and performed bulk-RNA-sequencing on extravascular CD4 T cells isolated from brain, meninges and gonadal white fat. Upon performing PCA analysis, we found that all three tissues clustered closely together, and brain CD4 T cells clustered in between the meningeal and white fat CD4 T cells (Extended Data Fig. 5b–c). This transcriptional similarity amongst the brain, meningeal and fat CD4 T cells is unique, as brain CD4 T cells demonstrated highly distinct transcriptional profiles when compared to another conventionally sterile non-lymphoid tissue, the lung (Extended Data Fig. 5d and Supplementary Tables 4, 5). We found many pathways related to lipid sensing or metabolism significantly upregulated in the brain and white fat CD4 T cells (Extended Data Fig. 5e, Supplementary Tables 6, 7, 8). Comparative IPA analysis identified these same pathways as both statistically significantly and positively associated with brain and white fat CD4 T cells compared to meningeal CD4 T cells (Extended Data Fig. 5e). Given these similarities, we next examined whether alterations in adipose tissue would affect brain T cell populations. To increase adipose tissue mass, we placed mice on a high fat diet (HFD), which resulted in expected changes in fat mass (Extended Data Fig. 6a). We observed increased extravascular CD4 T cell numbers in the fat and the brain, but not in the meninges, in HFD-fed animals (Extended Data Fig. 6b). We then examined if fasting would produce the opposite change in T cell numbers in fat and brain. After a 48-hour fast (Extended Data Fig. 6c), T cells in the fat were greatly reduced, but we observed an increase not only in T cell numbers in the bone marrow as previously reported²⁵, but also in the brain (Extended Data Fig. 6d). We thus hypothesized that fasting-induced T cell migration into the brain is tunable by expanding the baseline white fat T cell compartment. Indeed, subjecting mice to HFD followed by fasting resulted in increased CD4 T cell numbers in the HFD-fasted mouse brains compared to the normal chow-fasted brains (Fig. 2h and Extended Data Fig. 6e–f). Finally, to determine if fasting-induced increases in brain T cell numbers is due to T cell trafficking from white fat to brain, we utilized a photoconvertible mouse model to label T cells resident in white fat and found the highest proportion of photoconverted T cells in the white fat and in the brain (Fig. 2i and Extended Data Fig. 6g). Collectively, our findings demonstrate a previously undescribed fat-brain T cell axis.

There is a gut-brain T cells axis at steady state

One major caveat regarding use of HFD and fasting models to modulate adipose tissue is that these dietary modulations can affect the microbiome^26,27 among other aspects of physiology, and the microbiome, in turn, can regulate T cell differentiation and maintenance at steady state²⁶. Concurrent with the expansion of the gut microbiome during weaning, the immune system also undergoes rapid changes, and an IFNγ signature has been reported to appear in the colon of mice at weaning²⁸. Thus, we examined whether the establishment of brain T cells is correlated with microbiome colonization of the gastrointestinal tract as a function of weaning²⁸. We found that IFNγ and IL-17A-secreting brain CD4 T cells established immediately succeeding the weaning period (Fig. 3a and Extended Data Fig. 6g). Using germ free mice, we found that this establishment required the microbiome (Fig. 3b and Extended Data Fig. 6h). We thus examined if the microbiome composition could regulate the brain T cell population. Mono-colonizing GF mice with the T_H17 cell-skewing commensal, Segmented Filamentous Bacteria (SFB)²⁹, led to more IFNγ and IL-17A-secreting brain extravascular CD4 T cells compared to germ free controls (Fig. 3c). To examine whether the microbiome was required for maintenance of brain T cells, we depleted the microbiota with antibiotics, which resulted in a reduction of CD4 T cells generally, including the brain (Extended Data Fig. 6i–j). Together, these data demonstrate that, as with dietary manipulations, the microbiome is also sufficient to dynamically tune brain T cell populations. To directly assess if T cells primed in the gut can traffic to the brain, we again utilized the photoconvertible mouse model and photoconverted cells in the entire GI tract. We found gut-photoconverted T cells broadly across all tissues surveyed in adolescent and weaning-aged mice as has been described^30,31, including the brain (Fig. 3d and Extended Data Fig. 6k–l). Weaning-induced CD4 T cell trafficking into the brain could be blocked both by VLA-4 and anti-LFA-1 blockade and S1P receptor agonism (Extended Data Fig. 6m–n). Together, these data suggest that brain CD4 T cells migrate into the brain from the periphery, namely the GI and adipose tissues via blood vessels.

Fig. 3: — a, Absolute counts and proportion of CD45 IV-label negative brain CD4 T cells capable of secreting IFNγ upon stimulation in various aged mice via FACS. **=p<0.01, ****=p<0.0001 by one-way ANOVA (n=3 per group; repeated twice) b, FACS analysis comparing total and IFNγ and IL-17A-secreting brain CD4 T cell numbers in 3, 4- and 8-week-old germ free (GF) and SPF mice. (GF: n= 3 per group. SPF: n=3 per group; pooled 2 experiments) c, Same FACS analysis as b except in GF mice mono-colonized with SFB bacteria and GF controls. (First graph: n=4 per group; repeated once. Second & Third graph: n=8 per group; repeated once) d, Same FACS analysis as in Fig. 2i. (n=8 per group pooled 3 experiments) *=p<0.05, ****=p<0.0001 by two-way ANOVA with Bonferroni’s multiple comparisons test for **b, c. e,** Degree of T cell receptor (TCR) clonal expansion and Shannon Diversity Index and ACE index per tissue. f, TCR clonal sharing as represented by the alluvial graph and Jaccard index. g, FACS analysis of CD45 IV-negative CD4 T cell numbers in Smarta TCR (Vα2 Vβ8) transgenic mouse. (n= 4 per group; repeated once). h, FACS analysis of brain, meningeal and mLN CD4 T cell numbers in adult Smarta Rag1^−/− and Rag1^+/− control mice. (left: n=12 per group pooled 3 experiments; right: n=4 per group; repeated once). i, Schematic of adoptive transfer of naïve 7B8 T cells into GF or SFB-treated mice. The absolute number and frequency of T_H17 7B8 T cells from brain measured by FACS. (n=5 per group; repeated two times). j, same as i except recipient mice were Jackson Laboratory (no SFB) or Taconic Biosciences (have SFB) C57BL/6 mice. (n=4 per group). *=p<0.05, **=p<0.01, ***=p<0.001 by unpaired parametric t-test for c, **h, i, j.**

After establishing that the microbiome primes T cells in the gut and white fat that can traffic to the brain at steady-state, we wished to determine whether these T cells share TCR clonal overlap. We ran paired TCR-seq with our single-cell RNA-seq and employed hashtag labeling and CITE-seq to track TCR clones from each mouse from brain, meninges, colon, mLN and blood. We then compared TCR clonal expansion and overlap amongst the remaining tissues (brain, meninges, colon, lung, and blood). We found that the brain and colon TCRs were significantly expanded and had restricted clonal diversity compared to other tissues (Fig. 3e and Extended Data Fig. 7a). TCR clonal overlap was observed amongst all tissues (Fig. 3f), and after demultiplexing by mouse identity, overlapping TCR clones were found to be shared within individual mice as opposed to public sharing of clones across mice (Extended Data Fig. 7b) Consistent with our functional data, we found TCR overlap between the brain-fat and brain-colon (Fig. 3f and Extended Data Fig. 7c). Collectively, these data demonstrate that the brain harbors clonally overlapping T cells with meninges, fat, and colon respectively at steady-state. While we demonstrate using orthogonal methods for gut-brain and fat-brain axes, it is also possible that there is also trafficking from other organs like the lung to the brain as had been previously suggested in pathologic conditions³².

Since brain and peripheral T cells have a high rate of TCR overlap, we wished to determine if TCR specificity was required. We utilized a fixed CD4 TCR transgenic mouse model (Smarta), which has T cells that recognize glycoproteins (GP) 61–80 of the lymphocytic choriomeningitis virus (LCMV). We observed that Smarta mice do not have LCMV-specific Vα2⁺/Vβ8⁺ T cells in the brain; rather, the brain is enriched with Vα2⁻Vβ8⁻ endogenous T cells that escaped thymic selection (Fig. 3g). When we crossed Smarta mice with Rag1 deficient mice (Smarta Rag1^−/−), we eliminated brain and white fat, but not meningeal, extravascular CD4 T cells generally (Fig. 3h, Extended Data Fig. 7d), suggesting that the residence in the brain, but not meninges, is TCR dependent. To determine whether TCR activation in the gut is sufficient to prime T cells in the brain, we utilized the Smarta Rag1^−/− mice and immunized them with GP61–80 peptide and lipopolysaccharide by oral gavage, which resulted in a mild, trending increase in CD44⁺ Smarta T cells in the brain with a boost (Extended Data Fig. 7e–f). As the immunization models induce temporally discreet and transient immune activation, we deployed a gnotobiotic model and utilized the 7B8 mouse model, a CD4 TCR transgenic model specific for SFB. We adoptively co-transferred naive 7B8 T cells into weaning-age germ free wildtype C57BL/6 mice immediately followed by an oral gavage of either germ-free or SFB-containing fecal slurry. We found that gnotobiotic mice gavaged with SFB had a statistically significantly higher proportion of T_H17 7B8 cells compared to control mice (Fig. 3i, Extended Data Fig. 7g). Similar effects were seen by transferring 7B8 T cells into Taconic Biosciences (SFB+) compared to Jackson Laboratory (SFB-) mice (Fig. 3j, Extended Data Fig.7h). Collectively, these data demonstrate that antigen specificity is important for establishing brain T cells, which are programmed in the fat and gut.

A gut-brain T cell axis in a healthy living human

To test if these findings in a living healthy human, we examined a healthy human donor who contributed blood, CSF, colon and duodenum biopsies for single-cell RNA-seq and paired TCR-seq. We compared expression of IFNγ, CXCR6, and other genes that were highly expressed in both mouse brain and human CSF T cells to cross compare across organisms. Upon conducting unbiased clustering, we annotated the clusters by tissue of origin (Fig. 4a). By examining the expression of IFNγ and CXCR6, we noticed that they were highly expressed in clusters associated with duodenum and CSF (Fig. 4b). This demonstrates that CNS T cells in mice and humans are similar in respect to certain CNS-associated genes. We then examined the extent of TCR clonal expansion and overlap across tissues. We found that gut T cells had more expanded clones, followed by the CSF and then blood T cells (Extended Data Fig. 8a). We then examined whether there was TCR clonal overlap between T cells from the gut and the CSF. We found overlapping clones across all tissues and specifically two different TCR clones shared between the gut and CSF tissues (Fig. 4c–d, and Extended Data Fig. 8b). To understand what makes the CSF T cell clone transcriptionally unique from the duodenum or blood T cell clone, we extracted the UMAP cluster that contains the shared T cell clone from each tissue and conducted multiple two-way DEG analyses. When examining DEGs that are more highly expressed in the CSF T cell clone compared to duodenum or blood T cells, we noted that Cxcr6 was highly enriched in CSF T cells across both comparisons (Fig. 4e, Extended Data Fig. 8c and Supplementary Table 9, 10), like what was observed in the steady state mouse (Fig. 2b,f). In line with DEG analysis conducted on overlapping T cell clones in the human dataset, we went back to our mouse single-cell dataset and extracted T cells with clonally overlapping TCRs across gut (colon and mesenteric lymph node), CNS (brain, meninges), and blood and conducted trajectory analysis to infer dynamic changes in T cell transcriptional states across tissues. The analysis indicated that CXCR6 is differentially expressed along the trajectory, with higher expression in the brain and a subset of colon T cells at the endpoint (Extended Data Fig. 8d).

Fig. 4: — a, Blood, cerebrospinal fluid (CSF), colon and duodenum samples were collected from a healthy human donor and processed for scRNA-seq and paired TCR-seq. T cells were from each dataset were merged and subjected to integrated analysis for batch correction and represented in a UMAP. b, IFNγ and CXCR6 RNA expression plotted on top of the integrated UMAP graph. c, Clonotype sharing was observed across all tissue. Alluvial graph representing two different TCR clones shared between CSF, duodenum, and blood or CSF, colon, and blood. d, Highlighted one shared TCR clone between CSF, duodenum and blood. e, DEG analysis comparing clusters containing the shared TCR clone from each tissue. f, FACS analysis of CD4 T cells from CXCR6 deficient mice from brain, white fat, meninges and spleen. *=p<0.05 by unpaired parametric t-test (n=4 per group; repeated twice). g, Schematic of an adoptive co-transfer of ~1 million CD45.1 wildtype and ~1 million CD45.2 CXCR6 KO CD4 T cells mixed at a 50:50 ratio into a T cell deficient mouse (TCRβ). FACS analysis of CD4 T cell ratio in brain, meninges, white fat and mesenteric lymph node (mLN) 5-week after transfer. ****=p<0.0001 by one-way ANOVA with Tukey’s multiple comparison’s test (n=5; repeated twice). h, Freshly perfused C57BL/6 wildtype mouse brain was acutely sliced (250um) with a vibratome and 2mm² punches were obtained from various brain regions. The punches were homogenized, underwent RNA extraction and cDNA synthesis for qPCR analysis. *Rpl13a* was used as the housekeeping gene. ***=p<0.001 by one-way ANOVA with Tukey’s multiple comparison’s test (n=5; pooled from 3 experiments).

Thus, we examined whether CXCR6 is a chemokine receptor important for brain residence. CXCR6 deficient mice had greatly reduced brain and white fat, but not meningeal, CD4 T cells, and notably in the IFNγ-secreting compartment (Fig. 4f, Extended Data Fig. 8e). On the other hand, deficiency in CXCL10, one major chemokine for CXCR3 and the canonical T_H1-associated chemokine receptor did not affect total brain CD4 T cell numbers (Extended Data Fig. 8f). To confirm whether CXCR6-deficiency confers a disadvantage in T cell establishment in the brain but not the meninges, we performed adoptive transfer experiment and found that transferred CXCR6-deficient CD4 T cells were significantly less able to populate the brain compared to wildtype CD4 T cells (Fig. 4g). Since we saw a higher density of T cells in the SFO region of the brain, we wished to investigate whether the expression of CXCR6 ligand CXCL16 is higher in the SFO compared to other brain regions. We compared mRNA expression of Cxcl16 across various brain areas by acutely slicing mouse brain and punching out different regions to perform qPCR (Fig. 4h). We found that Cxcl16 is most abundant in the SFO compared to other sampled brain areas (Fig. 4h). Our data collectively suggest that CXCR6 is required for the establishment of brain, but not meningeal, CD4 T cells, in the SFO which highly expresses CXCL16.

CNS CD4 T cells are required for homeostasis

We next examined if brain CD4 T cell were required for adaptive organismal responses. Given their trafficking from gut and fat, we interrogated all experimental models in which we could eliminate or perturb brain CD4 T cells using a standard short-term fasting behavioural paradigm, the novelty-suppressed feeding test (NSFT)³³ (Fig. 5a), in addition to a battery of other standard related behavioural assays, including the tail suspension, forced swim, open field, and the light-dark box tests. Mice genetically deficient in αβT cells (Fig. 5b, Extended Data Fig. 9a–c); IFNγ (which had significantly lower brain CD4 T cells as previously described^34,35, Fig. 5c, Extended Data Fig. 9e–g); T-bet, which is required for CD4 but not CD8 IFNγ expression^36,37 (Fig. 5d, Extended Data Fig. 9h–l); but not IL-17A (Extended Data Fig. 9m–q) all had defects in the NSFT. To specifically test the role of CD4-derived IFNγ, we performed adoptive transfer experiments (Fig. 5e). Transferring wildtype CD4 T cells, but not IFNγ KO CD4 T cells, which were unable to reside in the brain, rescued the latency to approach food for TCRβ KO mice in the NSFT (Fig. 5e, Extended Data Fig. 9r–u). In order to test whether IFNγ from CD4 T cells is sufficient to rescue NSFT behaviour, we transferred wildtype or IFNγ KO CD4 T cells into IFNγ KO recipient mice. However, we found that neither wildtype nor IFNγ KO CD4 T cells could establish in the IFNγ KO mouse brain and there were no behavioural changes in any tests performed (Extended Data Fig. 9v–z). This suggests that IFNγ expression in both the CD4 T cells intrinsically, and extrinsically in other cell types is necessary for these cells to establish in the brain and induce behavioural changes. Our TCR transgenic model largely phenocopied TCRβ, IFNγ, and T-bet KOs, but notable sexual dimorphism was observed (Extended Data Fig. 10a–d). Lastly, similar to TCRβ, IFNγ, Smarta Rag1, T-bet and TCRβ KOs and mice transferred with IFNγ KO CD4 T cells, CXCR6 KO mice also had a delayed time to approach food in the NSFT paradigm (Extended Data Fig. 10e–h). Thus, we found that IFNγ secreting CD4 T cells in the brain are required for normal behaviour.

Fig. 5: — a, Schematic for novelty-suppressed feeding behavior test (NSFT). Mice were fasted for 24-hours prior to being placed in an open-field with food at the center. The time it took the mouse to reach and taste the food was measured. To control for the physiological effects of food deprivation, % weight loss post-fast and food consumption in the home cage after completing the NSFT were measured as a proxy for motivation to consume food. b, % weight loss post-fast, food consumption in the home cage and latency to approach food in the NSFT in male and female TCRβ KO and wildtype controls. (Pooled 2 experiments (males) and 3 experiments (females)) c, Same as b, except in IFNγ KO mice. (Pooled 2 experiments (males) and 5 experiments (females)) d, Same as **b, c** except in T-bet KO mice. (Pooled 3 experiments (males) and 5 experiments (females)) e, Same as **b-d** except in TCRβ KO mice with wildtype (CD45.1) CD4 T cells or IFNγ KO CD4 T cells transferred. (Pooled 4 experiments (males) and 5 experiments (females)) f, Same as **b-e** except in IFNGR1^f/f and IFNGR1^f/f CX3CR1-Cre mice. (Pooled 4 experiments for both males and females) g, Same as **b-f** except in IFNGR1^f/f and IFNGR1^f/f Nestin-Cre mice. (Pooled 4 experiments (males) and 3 experiments (females)) h, Same as **b-g** except in IFNGR1^f/f GFAP-Cre mice. (Pooled 3 experiments (males) and 4 experiments (females)) For all behavioral tests, outliers were calculated using the ROUT Outlier Test (Q=1%) and excluded. *=p<0.05, **=p<0.01 by two-way ANOVA with Bonferroni’s multiple comparisons test for **b-h. i,** Graphical abstract

To determine the parenchymal cell responding to IFNγ signals, we first conditionally deleted IFNγ receptor from microglia (IFNGR1^f/f CX3CR1-Cre). Upon performing behavioural assays on these mice, we did not find statistically significant changes (Fig. 3f and Extended Data Fig. 10i–k). However, when we FACS-sorted microglia from wildtype and T-bet KO mice for bulk-RNA-seq and ran PCA analysis, we found that female, but not male T-bet KO microglia clustered distinctly by PC1 (Extended Data Fig. 10l), a number of the DEGs that were downregulated in T-bet KO microglia compared to wildtype microglia included genes that are essential in homeostatic microglial function^38–41 (Extended Data Fig. 10m). Upon applying IPA to the statistically significant (FDR < 0.05) DEGs generated between T-bet KO and wildtype microglia, we found that T-bet KO microglia specifically downregulate RXR pathways (Extended Data Fig. 10n and Supplementary Table 11), which have previously been implicated in the alteration of fatty acid transport, lipid metabolism and disturbances of mitochondria function which are essential biological processes that affect microglial state and function⁴². These data suggests that there could still be sex-dependent effects of IFNγ-signaling on the state of microglia, although this is not required for behavioural differences.

Since there were no behavioural differences in the microglia-specific IFNGR knockout, we determined whether IFNγ receptor signaling was required in neurons^16,43–45 (IFNGR1^f/f Nestin-Cre) or astrocytes (IFNGR1^f/f GFAP-Cre)^46–48. There were also no appreciable behavioural differences in the IFNGR1^f/f Nestin-Cre mice compared to their floxed littermates (Fig. 5g and Extended Data Fig. 10n–p). Surprisingly, we found that male IFNGR1^f/f GFAP-Cre behaviourally phenocopied the various transgenic mouse models lacking brain T cells (Fig. 5h, Extended Data Fig. 10q–s). This was not observed in female mice, suggesting a sex-difference in IFNGR1 signaling on astrocytes in modulating NSFT behaviour (Fig. 5h). Using ASCA-2 antibody as a surface marker to identify astrocytes⁴⁹, we found that IFNγ KO mice also showed corresponding sex differences in astrocyte composition, as male IFNγ KO mice had a significantly lower frequency of astrocytes across the whole brain compared to wildtype mice, whereas females had a similar overall frequency of astrocytes among all live cells in the whole brain of IFNγ KO and wildtype mice, but had differences in microglia frequency (Extended Data Fig. 10t). This suggests that while brain CD4 T cell-derived IFNγ is required for neuronal homeostasis in both sexes, there are sex-dependent differences in the mechanisms by which IFNγ drives these effects, consistent with known sexually-dimorphic mechanisms of neuronal and astroglial development^50–52.

In summary, our data demonstrate that there are bona fide CD4 T cells in the steady state brain that are not randomly distributed, but rather localized in the SFO, a sensory CVO and hotspot for neuroendocrine communication to coordinate physiology. They are distinct from meningeal T cells, derived in the periphery, exhibit gut- and adipose-brain crosstalk, and are required at steady-state for CNS homeostasis. We propose that trafficking specialized T cells licensed to enter the brain based on specific TCRs and transcriptional states represents another mechanism distinct from neuroendocrine mechanisms to communicate peripheral tissue states to the CNS to coordinate physiology, in the same way that dendritic cells traffic to lymph nodes to coordinate immune responses. Going forward, it will be necessary to understand the identity of antigens recognized by brain CD4 T cells and the precise mechanistic contribution of the gut microbiota and fat to shaping brain CD4 T cells. Moreover, the precise mechanisms by which brain CD4 T cells (and to what extent other extravascular brain lymphocytes) participate in maintaining CNS homeostasis in different physiologic states, including sex, will need to be elucidated to serve as the foundation for understanding the potential role lymphocytes might play in diseases of the CNS.

Material and Methods:

Laboratory animals

Mice used in this study were derived from a C57BL/6 mouse background, maintained on a standard chow diet and all procedures were performed in accordance with the guidelines set forth by the Institutional Animal Care and Use Committees (IACUC) at the institution. Wildtype C57BL/6 mice used in this study were bred in-house and the original C57BL/6 mice strain was obtained from the Jackson laboratory. Germ Free and Gnotobiotic C57BL/6 mouse studies were conducted with the help from the Noah Palm lab. CXCR6 KO mice¹ and Smarta mice² were obtained from the Joseph Craft lab. 7B8 mice³ were a kind gift of Dan Littman (NYU Langone). For experiments performed using IL-15Ra knockout (KO) mice⁴ (Jax stock #003723), IP10 KO mice⁵ (Jax stock #006087), photoconvertible KikGR mice⁶ (Jax stock #035393), TCRβ-KO mice⁷ (Jax stock #002118), IFNγ KO mice⁸ (Jax stock #002287), T-bet-KO mice⁹ (Jax stock #004648), IL-17A/F double KO¹⁰ (Jax stock #034140), IFNGRf/f mice¹¹ (Jax stock #025394), CX3CR1-Cre mice¹² (Jax stock #025524), GFAP-Cre mice¹³ (Jax stock #024098) and Nestin-Cre¹⁴ (Jax stock #003771) mice were obtained from the Jackson Laboratory. Mice were habituated for more than one week before use. At Yale, all mice were housed in the The Anlyan Center Animal Research Facility or the Sterling Hall of Medicine Animal Research Facility, and their care was monitored by the Veterinary Service Center at Yale University under IACUC protocols.

High Fat Diet

3-week-old mice were placed on either a control diet (6% fat; Bioserv cat# F4031) or a high-fat diet (36% fat; Bioserv cat#F3282) for 5–6-weeks prior to sacrifice. Body fat percentage changes were calculated from magnetic resonance imaging (MRI) data of fat mass divided by total body weight.

Integrin blockade and FTY720 treatment

For integrin blockade, 3-week-old mice were treated intraperitoneally with anti-VLA4 and anti-LFA1 antibodies or isotype control antibodies every other day for one week. For FTY720, 3-week-old mice were treated intraperitoneally with 1mg/kg FTY720 dissolved in 0.9% saline water or vehicle (0.9% saline water) daily for one week. For both experiments, tissues were harvested and analyzed via FACS one-week later.

Photoconversion in mice

Our protocol is modified from PMID: 34875227¹⁵. Photoconvertible mice (KiKGR-iUBC mice) was anesthetized via ketamine/xylazine and maintained under conscious sedation throughout the procedure. A 1–2 cm mid-line laparotomy was performed using a scalpel and the gonadal white fat or entire gastrointestinal tract was pulled out of the incision site, exposed and reflected onto a saline-soaked surgical sponge to prevent desiccation. The tissues were exposed to 365 nm (non-damaging) UV light to photo convert cells within the light-field for 5 minutes. The tissues completely covered the incision site, preventing the access of UV light into the peritoneal cavity. The tissue was then placed back into the abdominal cavity and the peritoneum and abdominal musculature was sutured with 4–0 chromic gut (Ethicon; cat#1654G). The skin was closed with 4–0 vicryl suture (Ethicon; cat#736G). Animals recovered on a heating pad and were sacrificed 24 hours post operation. After FACS analysis, mice with no photoconverted cells at the photoconverted tissue site were excluded.

Gnotobiotic Studies

Germ-free C57BL/B6 mice were maintained in flexible film isolators and fed Tekland 2018S (Envigo) chow. Prior to gnotobiotic studies, mice were transferred to positive-pressure ventilated microisolator cages (Techniplast #ISO72P). Mice were gavage with 200uL of frozen stocks of bacteria suspended in 20% glycerol in GIFU media (HyServe #05422). Segmented filamentous bacterium (SFB) stock was prepared by collecting fecal supernatant of SFB monocolonized mice in anaerobic conditions (gas composition 4% H2, 10% CO2, 86% N2). Colonization of SFB was confirmed with SFB-specific primers (5’-TGAGCGGAGATATATGGAGC-3’ and 5’-CATGCAACTATATAGCTATATGCGG-3’).

Antibiotic Treatment

Adapted the treatment from PMID: 30902637¹⁶. Broad spectrum antibiotic cocktail was administered in mice drinking water for 4.5–5 weeks. The water was changed 2 times a week. The antibiotic cocktail was comprised of 0.5mg/mL Vancomycin (Sigma-Aldrich; cat# V2002), 1mg/mL Ampicillin (Sigma-Aldrich; cat#A0166), 0.5mg/mL Metronidazole (Sigma-Aldrich; cat#M1547), and 2.5mg/mL Streptomycin (Sigma-Aldrich; cat#S9137) and 1 pack of Splenda per 250mL autoclaved water. The control water also had 1 pack of Splenda per 250mL of water.

Immunostaining of brain sections

All immunostainings were performed on C57BL/6 mice obtained from the Wang lab colony between 8 to 10-weeks-old. Mice were subjected to intracardiac perfusion with 15ml of PBS containing 2mM EDTA and 5 mL of 4% paraformaldehyde (PFA) followed by overnight fixation. They were then sectioned in 60-μm coronal sections using a vibratome (Leica VT1000S) and preserved as free-floating slices at 4C. To perform immunofluorescent staining, sections were first subjected to a 30-minute antigen retrieval incubation in 1X citrate buffer (Life Technologies, 20X stock, product no. 005000). The citrate buffer was then washed off and sections were blocked with 2% normal donkey serum (Jackson Laboratories, product no. 017-000-121), 0.5% BSA (Sigma-Aldrich, product no. A9647) and 0.3% Triton-X (American Bio) in PBS for 4 hours at room temperature. Primary antibody staining was performed overnight at 4 °C in 0.5% BSA and 0.3% Triton-X in PBS. Primary antibodies used were the following: CD4 (abcam, Clone: EPR19514, product no. ab183685 (1:1200)), CD3 (Novus Biologicals, Clone: 17A2, (1:200), product no. MAB4841), CD31 (Novus Biologicals, product no. AF3628, (1:200)). Primary antibodies were washed off sections with PBS 3 times for 10 min at room temperature. Secondary antibody staining was performed at room temperature for 3 hours in 0.5% BSA and 0.3% Triton-X in PBS. Secondary antibodies used were the following: donkey anti rabbit-AF555 (Thermo Fisher (1:500), product no. A-31572), donkey anti rat-AF488 (Thermo Fisher (1:500), product no. A-21208) and donkey anti goat-AF647 (Thermo Fisher (1:500), product no. A-21447). Secondary antibodies were washed off and sections were stained with DAPI (Thermo Fisher (1:2000)) at 1 μg ml⁻¹ for 20 min at room temperature. DAPI was washed off and sections were mounted with either ProLong Gold (Thermo Fisher) or Lab Vision^™ PermaFluor^™ Aqueous Mounting Medium (Ephredia). Samples were visualized with a Lecia Stellaris confocal microscope using the 63X oil objective. We surveyed nine different brain regions including grey matter, white matter and whole mount dura meninges with positive signal and acquired images, making sure to capture a similar total area for each region. For visualization of images displayed in this manuscript, brightness and contrast were adjusted in Fiji to enhance visualization (these adjustments were not performed before the quantification described below). The same settings were applied to all images shown for each experiment.

Immunostaining of whole mount dura meninges

Mice were anesthetized with isoflurane, transcardially perfused as described above and the skullcaps adhered with intact dura mater were collected and immersion fixed in 4% PFA overnight at 4°C. Whole mount dura was peeled from the skullcap and washed with PBS to prepare for immunostaining. The free-floating dura was blocked and permeabilized in staining buffer (1X PBS with 0.2% Triton X-100, 0.05% Tween and 2% BSA) for 1 hour at room temperature. Primary antibodies, anti-Lyve1 (Invitrogen/ThermoFisher, Clone: ALY7, product no. 14-0443-82 (1:400)), anti-CD31 (Novus Biologicals, product no. AF3628, (1:200)), anti-CD4 (abcam, Clone: EPR19514, product no. ab183685 (1:1200)), diluted in the staining buffer, was added to the whole mount dura and incubated at 4°C overnight. The dura was washed and incubated with secondary antibodies diluted in the staining buffer for 2–3 hours at room temperature with gentle agitation. Then the dura was washed in PBS and incubated with DAPI for 15 minutes at room temperature. The whole mounts were washed in PBS and flattened onto slides and mounted using ProlongGold Antifade reagent (ThermoFisher).

Confocal microscopy image analysis and quantification

Extravascular CD4 T cell quantification

Brain regions were imaged using a tiling approach using 63X objective (1024×1024 pixel resolution) and z-stack spanning the entire 60-μm coronal section. Upon identifying a CD3+ CD4+ DAPI+ cell, approximately 4–6 tiles per brain area were obtained. The Allen Mouse Brain Coronal Reference Atlas Brain was used as a reference. Extravascular CD4 T cells were identified as CD3+ CD4+ DAPI+ that are distinct from the CD31+ vasculature. Vascular adjacent CD4 T cells were identified as CD3+ CD4+ DAPI+ that touching or directly adjacent to the CD31+ vasculature. Each CD3+ CD4+ T cell from each tiled brain area was manually identified and counted using a combination of 3D and Slice feature on Imaris version 9 and 10. Image J was used to add the scale bar and adjust color balance to generate the images used in figures and the z-stack projection videos for Supplementary Videos 1, 2, 3, 5, 6. LASX software and iMovie was used to create the 3D projected video of the subfornical organ for Supplementary Video 4. A total of six female and five male wildtype mice were quantified.

Brain clearing and whole-brain imaging (performed by LifeCanvas Technologies)

The mouse was anesthesized with isofluane and trancardially perfused with ice cold 1X PBS with 10U/mL heparin, followed by 4% PFA solution. The brain was then dissected out and fixed in 4% PFA for 24 hours at 4°C. The following day, the brain was washed twice with PBS and stored in PBS with 0.02% sodium azide prior to shipping. Upon arriving at LifeCanvas, mouse brain tissue underwent the LifeCanvas SHIELD (Stabilization under Harsh conditions via Intramolecular Epoxide Linkagaes to prevent Degradation) post-fix protocol with LCT SHIELD reagents¹⁷. Then, it was cleared for 7 days in Clear+ delipidation buffer and blocked with LCT blocking buffer. Whole-brain images were acquired at 3.6x with a pixel size of 1.8μm x 1.8μm at z-step size of 4μm. CD4-TdTomato expression was detected using the 561nm laser and YoPro dye was used to register the brain to the Allen Mouse Brain Atlas with the 488nm laser. Upon acquiring images, LifeCanvas performed analysis on TdTomato counts and calculated density of reporter-protein expressing cells per brain region.

Immunohistochemistry-based imaging of human brain

Deidentified post-mortem human brain tissue samples were obtained from subjects without acute neurologic disease through the Yale Legacy Tissue Donation Program and Yale autopsy service. Donor characteristics are listed in Table1. The structures of interest were dissected by a neuropathologist (MD) and submitted fresh for flow cytometry (Subjects 6 and 7) or submitted for histology (Subjects 1–5) following formalin fixation. After standard tissue processing, paraffin-embedding, and sectioning at 5μm thickness, tissue sections were applied to glass slides and immunohistochemistry was performed using primary antibodies against CD3 (Clone 2GV6, Roche cat. #790–4341), CD4 (Clone SP35, Roche cat. #790–4423), CD8 (Cone CD8/144B, Dako, cat. #IR623), or FactorVIIIra (Polyclonal, Dako, cat. #A0082), with DAB chromogen. Slides were scanned on a whole slide imager (EasyScan, Motic, Hong Kong), and resulting digital images were analyzed with custom scripts for the QuPath software package¹⁸. Images were reviewed and anatomic structures annotated as ROIs, followed by positive staining nuclei detection and ROI area calculations. Cell counts and area data were aggregated using custom python scripts, and summary statistical analysis performed in Prism.

Obtaining single-cell suspensions from brain and meninges

Mice were injected with 2ug of fluorophore-conjugated CD45 antibody (Biolegend; Clone 30-F11; Isotype: Rat IgG2b, κ) via intravenous injection at least 5 minutes prior to sacrifice and perfusion. Following anesthesia induction with isoflurane, mice were transcardially perfused with 20ml of PBS (Gibco; cat#10010023) supplemented with 2mM EDTA (Sigma-Aldrich; cat#03690) before sacrifice. Resected brains were homogenized with a razor blade and digested for 30-minutes at 37°C under constant shaking with 1mg/mL of collagenase IV (Worthington; cat#LS004189) and 100ug/mL DNase I (Roche; grade II, from bovine pancreas; cat#10104159001) in digestion buffer: RPMI (Gibco; cat# 11875101) with 2mM MgCl2 (ThermoFisher; cat#J62411.AD), 2mM CaCl2 (Sigma-Aldrich; cat#21115), 10%FBS (Sigma-Aldrich; cat#12306C), 2mM EDTA (Sigma-Aldrich) and 2mM HEPES (Thermofisher; cat#15630080). The dura meninges were peeled off of the skull cap and also digested under the same conditions as the brain. Digested cells were washed with RPMI and filtered using a 100-μm nylon mesh strainer. Only for the brain, cells were subjected to four Percoll gradients (GE Healthcare; cat#17-0891-01) (90, 60, 40%, 0%- HBSS) centrifugation at 1700rpm for 20 minutes with no brake or acceleration at 4°C. Single cell suspension was stained with flow cytometry antibodies.

Obtaining single-cell suspensions from white adipose tissue and lung

After intravenous labeling blood leukocytes with 2ug of fluorophore-conjugated CD45 antibody (Biolegend; Clone 30-F11; Isotype: Rat IgG2b, κ) and transcardial perfusion, the resected gonadal fat and lung was homogenized with a razor blade. The fat was digested for 1 hour at 37 °C with 0.8mg/mL of collagenase II (Worthington; cat#LS004176) and 2% FBS Sigma-Aldrich; cat#12306C) in PBS with constant shaking. The lung was digested for 1 hour at 37 °C with 2mg/mL of collagenase IV (Worthington) and 200ug/mL DNase (Roche) in PBS. Digested cells were washed with PBS and filtered using a 70-μm nylon mesh strainer. The resulting single cell suspension was stained with flow cytometry antibodies.

Obtaining single-cell Suspension from colon lamina propria tissue

After intravenous labeling blood leukocytes with 2ug of fluorophore-conjugated CD45 antibody (Biolegend; Clone 30-F11; Isotype: Rat IgG2b, κ) and transcardial perfusion, the mesenteric lymph node and colon with the cecum were resected and kept in ice cold RPMI (Gibco) supplemented with 2% FBS (Sigma-Aldrich). For both samples, excess fat tissue was removed. Feces were cleaned out of the colon and cut laterally to expose the intestinal lumen. After cutting the colon into two pieces, it was submerged in 15mL of pre-warmed EDTA-DTT buffer: HBSS w/o Ca2+, Mg2+ (Gibco; cat#14170112) supplemented with 10mM EDTA (Sigma-Aldrich), 10mM DTT (Thermofisher; cat#R0861), and 2% FBS (Sigma-Aldrich). Samples were incubated in a shaker at 37°C, 250rpm for 15min, vortexed rapidly for 10 seconds, and washed with another 15mL of EDTA-DTT buffer. Tissue was rinsed in media and cut into small pieces before transferring to digestion buffer: RPMI (Gibco) supplemented with 2mg/mL Collagenase D (Worthington), 0.2mg/mL DNase I (Roche) and 2% FBS (Sigma-Aldrich). Samples were incubated shaking at 250rpm for 50 minutes at 37°C, washed with RPMI+2%FBS to stop reaction and centrifuged at 1000g for 10m at 4°C. Pellet was resuspended in 4mL 40% Percoll (GE Healthcare) and bottom layered with 2mL 80% Percoll and centrifuged at 800x g for 20 min at 22°C without brake. After the spin, the lymphocyte layer in the interphase was collected and washed off with ~10mL ice cold RMPI+2%FBS at 500g for 10minutes, 4°C. The pellet was resuspended in PBS supplemented with 2% FBS until all samples were ready for flow cytometric or Hashtag antibody incubation. The mesenteric lymph node was dissociated using scissors and resuspended in 5mL of digestion buffer: RPMI (Gibco) supplemented with 2mg/mL Collagenase D (Worthington), 0.2mg/mL DNase I (Roche) and 2% FBS (Gibco). Samples were incubated and shaken at 37°C, 250rpm for 15min, filtered through 70-μm nylon mesh strainer and centrifuged at 500g for 10 minutes at 4°C. The resulting single cell suspension was stained with flow cytometry antibodies for 30 minutes at 4°C or 15 minutes at room temperature.

Flow cytometry preparation (mouse)

Flow cytometry antibodies were stained in FACS buffer (1% BSA in PBS) with Fc block (Purified anti-mouse CD16/32 Antibody: Biolegend; cat#101302) for 15 minutes at room temperature or 30 minutes at 4°C and washed with FACS buffer two times subsequently. Flow antibodies used for surface staining include: FITC anti-mouse CD45 Antibody (Biolegend; cat#103108), Alexa Fluor 700 anti-mouse CD45 Antibody (Biolegend; cat#103128), PE anti-mouse CD45 Antibody (Biolegend; cat#103128), Alexa Fluor^® 700 anti-mouse CD45.1 Antibody (Biolegend; cat# 110723), Brilliant Violet 421^™ anti-mouse CD45.2 Antibody (Biolegend; cat# 109831), BUV737 Rat Anti-CD11b (BD; cat#612800), Brilliant Violet 421^™ anti-mouse/human CD11b Antibody (Biolegend; cat#101235), Brilliant Violet 421^™ anti-mouse CD4 Antibody (Biolegend; cat#100443), FITC anti-mouse CD4 Antibody (Biolegend; cat# 100406), PE/Cyanine7 anti-mouse CD4 Antibody (Biolegend; cat#100528), APC/Cyanine7 anti-mouse CD4 Antibody (Biolegend; cat#100414), Brilliant Violet 711^™ anti-mouse CD3 Antibody (Biolegend; cat#100241), PE anti-mouse CD3 Antibody (Biolegend; cat# 100206), PE/Cyanine7 anti-mouse CD3 Antibody (Biolegend; cat#100219), Brilliant Violet 605^™ anti-mouse CD3 Antibody (Biolegend; cat# 100237), PE anti-mouse TCR β chain Antibody (Biolegend; cat#109208), Brilliant Violet 650^™ anti-mouse TCR β chain Antibody (Biolegend; cat#109251), Brilliant Violet 711^™ anti-mouse TCR β chain Antibody (Biolegend; cat#109243), Brilliant Violet 421^™ anti-mouse TCR γ/δ Antibody (Biolegend; cat#118120), APC Anti-mouse TCR γ/δ Antibody (Biolegend; cat#118116), BUV395 Rat Anti-Mouse CD8a (BD; cat#565968), Brilliant Violet 785^™ anti-mouse CD8a Antibody (Biolegend; cat#100749), FITC anti-mouse CD183 (CXCR3) Antibody (Biolegend; cat#126536), PerCP/Cyanine5.5 anti-mouse CD186 (CXCR6) Antibody (Biolegend; cat#151121), APC anti-mouse CD183 (CXCR3) Antibody (Biolegend; cat#126511), Brilliant Violet 605^™ anti-mouse CX3CR1 Antibody (Biolegend; cat#149027), BV510 anti-mouse CD62L antibody (Biolegend; cat#104441), APC anti-mouse CD62L Antibody (Biolegend; cat#104412), APC/Cyanine7 anti-mouse/human CD44 Antibody (Biolegend; cat#103028), PerCP/Cyanine5.5 anti-mouse/human CD44 (Biolegend; cat#103031), BD Horizon^™ BUV395 Rat Anti-Mouse CD45 (BD; cat#564279), Brilliant Violet 421^™ anti-mouse CD69 Antibody (Biolegend; cat#104527), Brilliant Violet 421^™ anti-mouse NK-1.1 Antibody (Biolegend; cat#108731), Brilliant Violet 650^™ anti-mouse CD223 (LAG-3) Antibody (Biolegend; cat#125227), Brilliant Violet 711^™ anti-mouse CD279 (PD-1) (Biolegend; cat#135231), APC anti-mouse TIGIT (Vstm3) Antibody (Biolegend; cat# 142105), FITC anti-mouse TCRvβ8.3 Antibody (Biolegend; cat#156306), APC anti-mouse TCR Vα2 Antibody (Biolegend; cat#127810), Zombie NIR^™ Fixable Viability Kit (Biolegend; cat# 423105), Zombie UV^™ Fixable Viability Kit (Biolegend; cat#423107),. For intracellular staining of transcription factors, the Foxp3/Transcription Factor Staining Buffer set was used and the manufacturer supplied manual for preparing samples was followed (eBioscience; cat#00-5523-00). Flow antibodies used for transcription factor staining include: Brilliant Violet 421^™ anti-mouse FOXP3 Antibody (Biolegend; cat# 126419), Brilliant Violet 605^™ anti-T-bet Antibody (Biolegend; cat#644817), Brilliant Violet 421^™ anti-T-bet Antibody (Biolegend; cat# 644815). For cytokine staining, single cell suspensions from each tissue were subjected to 4 hours of cell stimulation using cell activation cocktail (Biolegend; cat#423301) and Golgiplug (BD; cat#555029) diluted in RPMI (Gibco) with 10% FBS (Sigma-Aldrich). Flow antibodies used for cytokine staining include: FITC anti-mouse IFN-gamma (BD; cat#562019), PE/Cyanine7 anti-mouse IFN-γ Antibody (Biolegend; cat#505825), Brilliant Violet 785^™ anti-mouse IL-17A Antibody (Biolegend; cat#506928), Brilliant Violet 711^™ anti-mouse IL-17A Antibody (Biolegend; cat#506941), APC/Cyanine7 anti-mouse IL-10 Antibody (Biolegend; cat#505036), APC anti-mouse IL-4 Antibody (Biolegend; cat# 504105), Brilliant Violet 605^™ anti-mouse IL-4 Antibody (Biolegend; cat#504125), FITC Rat Anti-Mouse Vβ 14 T-Cell Receptor (BD; cat#553258). Precision count beads (Biolegend; cat# 424902) were used to calculate absolute numbers of cell types.

Flow cytometric preparation and analysis on human autopsy samples

Brain samples were collected and transported to the laboratory 4–5 hours postmortem on ice. Upon arrival, each tissue was weighed on a scale to record starting tissue volume. Then, tissues were placed in Miltenyi gentleMACS C tube with 5–10mL of brain digestion buffer (RPMI 1640 supplemented with 2mM MgCl2, 2mM CaCl2, 10% FBS, 2mM EDTA, 2mM HEPES, collagenase D (1mg/ml) and DNase I (100ug/mL)) or 10mL of lymph node digestion buffer (RPMI containing 10% FBS, L-glutamate, sodium pyruvate, non-essential amino acids, collagenase D (1mg/ml) and DNase I (50–100μg/ml)) and set to run the GentleMACS program 37C_Multi_F (45 minutes). After digestion, liquid contents were gently poured into a 50mL conical tube lined with a 100-μm nylon mesh strainer. Strainers and gentleMACS C tubes were washed using ice cold RPMI and a syringe plunger was used to gently mash any solid leftover pieces. Samples were centrifuged at 1500 RPM for 10 minutes at 4°C, supernatant decanted and resuspended with 30% Percoll. The gradients were centrifuged at 2000 RPM for 20 minutes at 4°C with 4 accel 2 brake settings, resulting in separation of debris and the cell pellet. The debris floating on the top layer was discarded and the cell pellet was transferred to a new tube and filled up with 1X PBS to wash off the remaining Percoll. After the wash, the cell pellet was transferred to a round-bottom 96-well plate where it was prepared for flow cytometric analysis as described above in the mouse flow cytometry preparation method. Antibodies used for human samples include Fixable Viability Dye eFluor 780 (Invitrogen/eBioscience), Brilliant Violet 510 anti-human CD3 Antibody (Biolegend; cat#300448), Brilliant Violet 785 anti-human CD4 Antibody (Biolegend; cat#317442), BUV737 Mouse Anti-Human CD8 Antibody (BD; cat#741850), APC anti-human CD11b Antibody (Biolegend; cat#301309), BV650 Mouse Anti-Human CD11b (BD, cat#742640), FITC anti-human CD45RA Antibody (Biolegend; cat#304106), Alexa Fluor 700 anti-human CD45RO Antibody (Biolegend; cat# 304217), APC anti-human CD45RO Antibody (Biolegend; cat#304210), Brilliant Violet 650 anti-human CD197 (CCR7) Antibody (Biolegend; cat# 353233), FITC anti-human CD186 (CXCR6) Antibody (Biolegend, cat#356019), Brilliant Violet 421 anti-human CD186 (CXCR6) Antibody (Biolegend; cat#356014), Brilliant Violet 421 anti-human CD183/CXCR3 Antibody (Biolegend; cat#353715), PE-Cy7 Mouse Anti-Human CD183/CXCR3 (BD; cat#560831). The density of CD4 and CD8 T cells in the brain were generated from absolute counts using counting beads divided by weight of the original tissue sample.

Bulk-RNA-sequencing preparation and parameters

Single-cell suspension was isolated from tissues as described and samples were stained with flow cytometry antibodies to distinguish cell populations. CD4 T cells were enriched using a CD4 T cell enrichment kit (StemCell; Cat#19852) prior to FACS sorting CD4 T cells. CD4 T cells (CD45 IV- CD11b- CD45+ CD4+ CD3+ TCRβ+) or microglia (CD45-mid CD11b-mid CX3CR1+) were sorted using the Thermofisher Bigfoot Cell sorter or the BD FACSAriaII and cells were collected directly into RLT Lysis Buffer as part of the Qiagen rNeasy Micro kit (Qiagen; cat#74004). RNA was isolated from the sorted sample according to the manufacturer’s instructions. Gene expression library preparation and low-input poly(A)RNA-sequencing was conducted through the Yale Center of Genomic Analysis. 7–8 mice were pooled per sample to obtain sufficient RNA for downstream processing and sequencing of CD4 T cells from brain, meninges and white fat samples. 25 million reads per sample was sequenced using the Illumina NovaSeq6000 S4 sequencer.

Bulk-RNA-sequencing sample analysis

The raw sequencing data of fastq files for gene expression were aligned to mouse genome reference mm10 using hisat2. Upon obtaining aligned reads, PCA analysis and differential gene (DEG) analysis was conducted using the DESeq2 package on R. Plots were generated using ggplot on R. Volcano plots depicting top DEGs were generated using EnhancedVolacano package on R. Upon obtaining statistically significant (FDR <0.05) DEGs, they were subjected to Qiagen’s Ingenuity Pathway Analysis using the Desktop application. Results from the pathway analysis were graphed on Python using seaborn, pandas and matplotlib packages.

Single-cell RNA-seq using 10x Genomics Chromium

We performed single-cell RNA-seq and paired TCR-seq and CITE-seq for Hashtag labeling on live T cells isolated from the brain, meninges, gonadal white adipose tissue, colon lamina propria, mesenteric lymph node and peripheral blood. We used three 10-week-old adult male and two female C57BL/6 mice for a total of five mice. First, each mouse was sedated and approximately 100ul of blood was collected via retroorbital bleeding. The blood was collected in heparin-coated tubes and stored on ice until processing. The blood was resuspended in 1mL of ACK lysis buffer (Gibco; cat#A10492–01) and incubated at room temperature for 10 minutes followed by a 5-minute spin at 2000rpm. The pellet was resuspended in PBS with 2% FBS and enriched for T cells using a T cell enrichment kit (StemCell; Cat#19851). After blood collection, the mouse was intravenously injected with a PE-conjugated CD45 antibody (Biolegend; cat#103128) to label blood circulating leukocytes. 5 minutes later, the mouse was sacrificed and perfused with 20 ml of PBS (Gibco) with EDTA (2mM) (Sigma-Aldrich) to remove the blood, and the aforementioned tissues were immediately collected. Single-cell suspensions from brain, meninges, lamina propria of the colon and mesenteric lymph node were prepared as previously described. The single-cell suspensions were incubated with Fc block (Purified anti-mouse CD16/32 Antibody: Biolegend; cat#101302) and stained with Zombie NIR^™ Fixable Viability Kit (Biolegend; cat# 423105), Alexa Fluor 700 anti-mouse CD45 Antibody (Biolegend; cat#103128), BUV737 Rat Anti-CD11b (BD; cat#612800) and dump (Brilliant Violet 711^™ anti-mouse CD19 Antibody; Biolegend cat#115555 & Brilliant Violet 711^™ anti-mouse/human CD45R/B220 Antibody; Biolegend cat#103255). The Zombie UV-negative, CD45 IV-negative, CD45-positive, CD11b-negative and dump-negative cells were sorted on the Thermofisher Bigfoot Cell sorter using a 100-μm nozzle. The TotalSeq^™-C0301 anti-mouse Hashtag 1–5 Antibodies (Biolegend; cat#155861, 155863, 155865, 155867, 155869) were prepared according to manufacturer’s protocol and stained along with the FACS antibodies pre-sorting. Cells were sorted into RPMI with 5% FBS (Sigma-Aldrich), spun down at 300g for 5 min at 4 °C and resuspended in catching medium at a concentration of 300 cells per μl counted using a hemacytometer. T cells were sorted from each tissue based off the following markers: Negative for CD45-intravenous label, CD11b, CD19, and B220.

10x single-cell gene expression and VDJ library construction and sequencing

The samples were submitted to the Yale Center of Genomic Analysis to get loaded onto a 10x Genomics Chromium chip per factory recommendations. The single-cell libraries were prepared by the Chromium Next GEM Single Cell 5′ Reagent Kit v2 and V(D)J reagent chemistry according to the manufacturer’s instructions (10x Genomics) and sequenced using Illumina NovaSeq6000 S4 sequencer, at a sequencing depth of 100 million reads for brain, meningeal and fat samples and 200 million reads for blood, mLN and colon samples for gene expression library, targeting an average sequencing depth of 20,000 reads per cell. 25 million reads for brain, meningeal and fat samples for each TCR library. The raw sequencing data of fastq files for gene expression were aligned to mouse genome reference mm10 and V(D)J reference vdj_GRCm38_alts_ensembl-4.0.0 using Cell Ranger version 7.1.0 (10x Genomics).

Quality control of 10x Genomics single-cell RNA-seq

For mapping, sequences obtained from sequencing using the 10x Genomics single-cell RNA-seq platform were demultiplexed and mapped to the mm10 transcriptome using the Cell Ranger package (10x Genomics). Cells were removed if they expressed fewer than 200 unique genes, more than 4,000 unique genes or greater than 5% mitochondrial reads. Genes not detected in any cell were removed from subsequent analysis. The levels of mitochondrial reads and numbers of unique molecular identifiers (UMIs) were similar across the tissues. The average gene detection in each cell across samples was 1085. There were 5011 cells from the brain, 8496 cells from the meninges, 7134 T cells from white fat, 17,209 cells from colon lamina propria, 17,703 cells from blood and 18,138 cells from mesenteric lymph node before quality control. The final analyses after quality control and filtering for cells that highly express CD3 (Cd3e >1), there were 3242 T cells from the brain, 2568 T cells from the meninges, 1399 T cells from white fat, 9440 T cells from colon lamina propria, 13,845 T cells from blood and 14,375 T cells from mesenteric lymph node. The T cells from the brain and meninges samples were combined into one Seurat object for analysis displayed in Figure 1 and T cells from all tissues were combined into another Seurat object for analysis displayed in Figure 2.

t-SNE analysis of single-cell RNA-seq datasets and identification of cell clusters

For the single-cell RNA-seq analysis, we used the Seurat R Package (v. 4.3.0.1). In the combined brain and meningeal T cell dataset we performed t-SNE clustering using the first 5 principal components, after performing principal component analysis (PCA) on the 2000 most variable genes. The number of principal components to use for t-SNE clustering was based off of the Elbowplot after PCA analysis. This identified 16 clusters of cells in the brain and meningeal T cell combined dataset. Identification of significant clusters was performed using the FindClusters() algorithm in the Seurat package which uses a shared nearest neighbour modularity optimization-based clustering algorithm. Marker genes for each significant cluster were found using the Seurat function FindAllMarkers(). Cell types were determined using a combination of marker genes identified from the literature and gene ontology for cell types using the web-based tool Enrichr (http://amp.pharm.mssm.edu/Enrichr/).

For the dataset containing T cells from all tissues combined (Dataset 1: brain, meninges, white fat, colon lamina propria, lung and blood), and (Dataset 2: brain, meninges, white fat, colon lamina propria, mesenteric lymph node and blood), we filtered T cells using RNA expression cutoff of Cd3e > 1. For dataset 1, we utilized the first 8 principal components on the 2000 most variable genes, which yielded 16 clusters. For dataset 2, we utilized the first 12 principal components on the 2000 most variable genes, which yielded 23 clusters. For the dataset containing all tissue T cells, 4 clusters containing erythrocyte contamination and B-cell-like transcript were eliminated and t-SNE clustering was redone using the first 9 principal components. This analysis identified 16 clusters in the all tissue combined T cell dataset 2.

Demultiplexing (de-hashing) CITE-seq samples

To assign cells to each biological samples, hashing antibody-derived tag (ADT) counts were normalized using centered log ratio transformation. A negative distribution of the normalized counts was then determined by performing k-medoids clustering, fitting a negative binomial distribution, and using the q = 0.99 quantile threshold to assign HTO barcodes, as previously described¹⁹. Only cells with a single HTO were selected for downstream analysis.

T cell receptor repertoire analysis

Single-cell T cell receptor repertoire analysis was performed using the scRepertoire package (version 1.11.0)²⁰. Filtered annotated contig files generated by the CellRanger (version 7.1.0) VDJ pipeline were used as input files. Cells with paired single productive alpha- and beta-chains were selected for downstream analysis. Clonotypes were defined by using the amino acid sequences. The proportion of the total repertoire occupied by each clonotype was calculated using the clonalProportion function. Indices of diversity (Shannon entropy, Inverted Simpson) and species richness (Chao, ACE) were calculated using clonalDiversity function.

Trajectory analyses

To perform trajectory inference analysis, we fit a minimum spanning spanning tree (MST) to the UMAP dimensionality reduction plot using the Slingshot package (version 2.7.0)²¹. The piecewise linear trajectory was smoothed using simultaneous principal curves to derive trajectory and pseudotime values. To determine associations between gene expression and pseudotime values, we fit a negative binomial general additive model using the tradeSeq package (version 1.10.0)²². Statistical significance between the start and end points of the lineage was determined using the Wald test (p < 0.05), as applied by the startVsEndTest function.

Human Subject Gut Tissue Sample Collection

Subject was enrolled in the Intestinal Immunity in Neurologic Disease Study, IRB protocol #2000033081, at Yale School of Medicine. Informed consent was obtained from the subject in accordance with the respective protocol and sequencing and data storage and publication plans were approved by the Yale IRB. Biopsies were obtained during an upper endoscopy and colonoscopy for standard of care, using jumbo biopsy forceps, to sample the intestinal mucosa. Biopsies were sampled from the duodenum, terminal ileum, cecum, right colon, and left colon. The number of biopsies taken were in accordance with that accepted to be safe under standard of care. Biopsies were immediately stored in cryovials containing DMEM (Gibco, cat. no 11965092) and placed on wet ice for transport.

Human gut Biopsy Epithelial Layer Dissociation

After arrival, tissue biopsies were washed two times in cold PBS, then three times in cold PBS/ 10mM EDTA (Promega, cat. no V2341). Samples were then placed in 25mL of pre-warmed epithelial strip buffer:1X PBS (Gibco, cat. no 14190144), 5mM EDTA, 1mM DTT (Sigma, cat. no 646563), 5% FBS (Gibco, cat. no A4766801), 15mM HEPES (Gibco, cat. no 15630106) and set in a shaking incubator at 37°C for 30 minutes. After incubation, samples were placed on ice for 10 minutes and then mixed well for 10 seconds using a vortex. Supernatant was collected in a new tube and the tissue was washed in 25mL epithelial strip buffer. Tissue was then kept on ice in a small amount of epithelial strip buffer for later lamina propria digestion. Supernatant was spun down at 300g for 5 minutes at 4°C, supernatant was removed and the pellet was resuspended in 5mL of rinse buffer (1X RPMI 1640 (Gibco, cat. no 11875093), 10mM HEPES, 1% Pen/Strep (Gibco, cat. no 15140122), 5% FBS). After the pellet was spun down and supernatant was removed, the pellet was transferred to a 1.5mL microcentrifuge tube, spun down and ultimately resuspended in 50–200µL 0.4% BSA/PBS (Sigma-Aldrich, cat. no A9647) at the concentration of 700–1200 cells/µL for 10x Genomics Chromium single cell loading immediately.

Human Gut Biopsy Lamina Propria Layer Dissociation

Tissue samples saved on ice from epithelial layer dissociation were transferred to a 5mL snap-closure tube with 5mL of digestion buffer:1X RPMI 1640, 2% FBS, 1% DNase I (Roche, cat. no 1010459001, 10mg/mL in H2O), 4% Liberase TM (Roche, cat. no 5401119001, 2.5mg/mL in H2O) and placed in a spinning incubator for 45 minutes. After incubation, 500µL of 100% FBS was added to the tube and vortexed for 20 seconds. Tissue and buffer were poured over a 70µM filter into a new tube. The bottom of a small syringe was used to gently mash the tissue through the filter. The filter was then rinsed with 2% FBS/RPMI to bring the total volume of the cell suspension in the tube to 30mL. Sample was spun down at 450g for 3 minutes. Supernatant was removed and pellet was resuspended in 1mL of 0.4% BSA/PBS and transferred to a 1.5mL microcentrifuge tube. Sample was spun down at 300g for 3 minutes. Supernatant was removed, then pellet was resuspended in 1mL of ACK Lysing Buffer (Gibco, cat. no A1049201) and incubated at room temperature for 1 minute. After incubation, the sample was spun down, supernatant was decanted and ultimately the pellet was resuspended in 50–200µL 0.4% BSA/PBS at the concentration of 700–1200 cells/µL for 10x Genomics Chromium single cell loading.

Human Subject Blood and CSF Tissue Sample Collection and Processing

Subjects for paired blood draw and lumbar puncture (LP) were recruited as part of the IRB-approved study, and written consent was obtained. LPs were performed under sterile conditions. There was no randomization or blinding in our study. Peripheral blood mononuclear cells (PBMCs) were isolated from fresh whole blood via density gradient by layering blood diluted 1:1 in phosphate-buffered saline (PBS) over Ficoll-Paque. Red blood cells were lysed using ACK lysing buffer (Gibco) per the manufacturer’s protocol and filtered through a 40-µm filter. Cells were resuspended in 1× Dulbecco’s PBS with 0.04% bovine serum albumin for loading into 10x Genomics Chromium at suspensions ranging from 700 to 1200 cells/µL. Cerebrospinal fluid (CSF) was kept on ice until processing. Total CSF was spun for 10 min at 300g 4°C with a swing bucket centrifuge, and supernatant was aspirated. The pellet was then transferred to a 1.5-ml tube, spun for 10 min at 300g 4°C with supernatant, and then aspirated to leave 50 to 100 µL for cell counting and loading into 10x Genomics Chromium at suspensions ranging from 700 to 1200 cells/µL.

10x single-cell gene expression and VDJ library construction and sequencing for human samples

The single-cell libraries were prepared by the Chromium Next GEM Single Cell 5′ Reagent Kit v2 and V(D)J reagent chemistry according to the manufacturer’s instructions (10x Genomics) and sequenced using Illumina NovaSeq6000 S4 sequencer, at a sequencing depth of 300 million reads per sample for gene expression library, targeting an average sequencing depth of 50,000 reads per cell, and 50 million reads per sample for each TCR library. The raw sequencing data of fastq files were aligned to human genome reference GRCh38 (GENCODE v32/Ensembl 98) using Cell Ranger version 7.0.1 (10x Genomics).

T cell receptor repertoire and single-cell RNA-seq clustering analysis for human dataset

The single-cell data for each tissue were processed under the standard scanpy pipeline, and then the T cells were identified by canonical markers. The data integration was conducted via Harmony, and the clusters were obtained by leiden clustering. The marker genes for each cluster were inferred using Wilcoxon rank-sum test. The TCR clonal analysis was conducted using the functions clonalHomeostasis, clonalDiversity, clonalOverlap, and compareClonotypes from R package scRepertoire.

Adoptive co-transfer of CD4 T cells into T cell deficient mice

Adult wildtype, CXCR6-deficient or IFNγ-deficient mice were sacrificed and spleens and lymph nodes were harvested. In the biosafety hood, the spleen and lymph nodes were mechanically dissociated and passed through a 70um strainer into 50mL conical tubes, which were then centrifuged at 500g for 5 minutes at 4oC. the pellet was resuspended in 1mL of PBS supplemented with 2% FBS. For the spleen, the pellet was resuspended and incubated in 1mL of ACK lysis buffer for 3 minutes at room temperature, washed with PBS, centrifuged at 500g for 5 minutes at 4oC and then resuspended in 1mL of PBS supplemented with 2% FBS. CD4 T cells were enriched from the single-cell suspensions using the EasySep^™ Mouse CD4+ T Cell Isolation Kit (StemCell; cat#19852) according to manufacturer’s instructions. CD4 T cells were counted using a hemacytometer and 50:50 mix validated using CD45.1 and CD45.2 staining on the flow cytometer. Adjustments were made until the transfer mix is an equal mix of both genotypes. The adoptive transfer cell mix was transferred to TCRβ KO mice via retroorbital injection under anesthesia.

Vibratome sectioning and punch biopsy for qPCR

Protocol was adapted from PMID: 34095867²³. C57BL/6 wildtype mice were transcardially perfused with ice cold 1X PBS with 2mM EDTA. Brains carefully harvested and stored in ice-cold artificial CSF (120mM NaCl, 3.5mM KCl, 1mM CaCl2, 1mM MgCl2, 0.4mM KH2PO4, 5mM HEPES, 994mL dH20, 10mM glucose) until ready for slicing. When ready, the brain was cut in a way to enable stable mounting on the stage using superglue (dorsum down) and a small agarose block (3% agarose in artificial CSF) was glued behind the brain to provide structural support during slicing. The vibratome tray was filled with ice cold artificial CSF and surrounded by ice to maintain cold temperature while slicing. The sectioning parameters were set on the Leica VT 1000S so that the section thickness is 250 μm and cutting speed is at dial level 2. Free floating sections were transferred to a 24-well plate filled with ice cold artificial CSF and various brain areas were punched using a 2mm biopsy punch (Integra Miltex). The punched brain slices were transferred to RNA-bee (Tel-Test, Inc CS-502) and homogenized using Omni International Bead Ruptor Elite. RNA was purified from tissues using Qiagen RNAEasy Kit (Cat: 74106) according to manufacturer’s instructions. cDNA was generating using oligodT (Sigma-Aldrich) and SuperScript IV Reverse Transcriptase (Invitrogen, cat#18090200) according to manufacturer’s instructions. Quantitative PCR was performed on a CX96 Real-Time System (Bio-rad) with PerfeCTa SYBR Green SuperMix (Quanta/VWR, cat#101414-162). Relative expression was calculated using Rpl13a as a housekeeping gene. Sequences of the CXCL16 primers are Cxcl16 Forward: CGTTGTCCATTCTTTATCAGGTTCC and Cxcl16 Reverse: TTGCGCTCAAAGCAGTCCA.

Behavioural Assays

Prior to all behavioural assays, mice were acclimated in the testing room for 30 minutes. The Biosafety Hood in the room was turned on to act as white noise for the duration of acclimation and testing. In studies where both males and female mice were tested, the male mice were tested prior to females to prevent residual female scent from influencing the results of the experiment. The cage genotypes were blinded for the experimenter using plain numbered paper covers and revealed at the end of the test. In the case where the same animals were tested for all the behavioural assays, each assay was separated by 24 hours and were tested in the following order: Open field test > Light-dark box > Tail suspension test > Forced swim test > Novelty-suppressed feeding test

Open Field Test (OFT)

Mice were placed facing a corner in an open field (59.6 × 47.3 × 31.1cm) with fresh bedding and were allowed to roam freely in the space for 10 minutes. Their movements were recorded by a GoPro camera. The video recording was used to calculate total distance traveled and frequency spent in the center using ezTrack package on Python²⁴. Less distance traveled and lower frequency spent in the center was used as a readout for anxiety-like behaviour.

Light-dark box

The light-dark box apparatus is a rectangular box (18 × 10 × 13cm) consisting of a dark chamber and a bright chamber and mice are allowed to move freely between the two through a narrow opening. The mouse is first introduced into the bright chamber facing the corner so they don’t immediately run towards the dark chamber upon placement. The time it takes the mouse to cross into the dark side the first time is recorded. Immediately upon crossing into the dark chamber, a 6-minute timer is started. During the 6 minutes, the number of entries or transitions into the bright chamber was tallied and the cumulative time spent in the aversive bright chamber was measured. The time to first cross, number of transitions and cumulative time spent in the bright chamber were used as a proxy of anxiety-like behaviour.

Tail Suspension Test (TST)

Mice were suspended approximately 15cm above the ground using a designated ring stand. Tape was used to securely attach the first 2mm of the mouse tail to the device that hooks onto the ring stand adapter. The mice were suspended for a duration of 6 minutes and a GoPro camera was used to record the experiment. During the assay, the cumulative time that the mouse spent immobile was recorded. Mice that escaped the suspension by climbing up its tail were excluded from the experiment.

Forced Swim Test (FST)

In two sessions separated by 24 hours, mice were introduced into a 3L beaker of 37°C water for a duration of 5 minutes. The cumulative time spent immobile was recorded. Two mice were tested simultaneously in beakers separated by an opaque screen. A GoPro camera was used to capture a video recording. Immobility was defined by any time the mice did not move its paws intentionally to swim towards an intended direction.

Novelty-Suppressed Feeding Test (NSFT)

24 hours before the test, each mouse was weighed and tails were labeled with a marker to easily identify each mouse. Mice were then placed in a wire-bottom cage (to prevent them from eating feces or bedding) with no food for 24 hours prior to behavioural testing. The test chamber (59.6 × 47.3 × 31.1cm) is prepared with fresh bedding and a piece of food is placed on top of a small Whatman paper at the center of the open field. Each mouse is placed in the open field facing the corner and the time it takes for the mouse to pick up the piece of food and nibble on it is recorded. Mice were given a maximum of 10 minutes to reach the food; any mouse that took longer was automatically assigned a value of 600 seconds. Once the mouse reaches the food, it is removed from the open field and weighed. After testing, each mouse is placed back in its home cage with food pellets and the weight of food consumed in 5 minutes is measured. If a mouse consumed less than 0.05g of food after 5 minutes, it was excluded due to low food motivation.

Measuring circulating metabolites

The following kits were used to measure circulating β-hydroxybutyrate (BHOB) (Cayman Chemical Company, cat#700190), triglycerides (TG) (Fujifilm; cat#996–02895) and non-esterified fatty acids (NEFA) (Fujifilm; cat#991–34793) as per manufacturer’s instructions.

Isolating microglia and astrocytes

Protocol was adapted from PMID: 34233165^c. Mice were injected with 2ug of PE-conjugated CD45 antibody (Biolegend; Clone 30-F11; cat#103128) via intravenous injection at least 5 minutes prior to sacrifice and perfusion. Following anesthesia induction with isoflurane, mice were transcardially perfused with 20 ml of PBS (Gibco; cat#10010023) before sacrifice, brains were harvested and kept on ice. Brains were dounce homogenized in ice cold HBSS w/o Ca2+, Mg2+ (Gibco; cat#14170112) 20 strokes with the ‘A’ pestle or 25–30 stokes with the ‘B’ pestle and pass through a 70um cell strainer into a 50mL conical, spun down at spin 500 g 5 min at 4C and resuspended in 5.5 ml of HBSS with 4.5 ml of the 90% Percoll (GE Healthcare; cat#17-0891-01) solution in a 15 ml tube. Samples were spun for 1hr at 500 g with full acceleration and braking. The top 1mL was removed and Percoll was washed off with PBS by centrifuging for 10 min at 800 g at 4°C. The supernatant was taken off and the remaining pellet was transferred to a 96-well round-bottom plate. The sample was stained with Fc block and flow cytometry antibodies for 30 minutes at 4°C and washed with PBS two times prior to FACS sorting. The CD45-IV (Biolegend; cat#103128) negative, Zombie NIR (Zombie NIR^™ Fixable Viability Kit; Biolegend; cat# 423105) negative, CD45 (FITC anti-mouse CD45 Antibody; Biolegend; cat#103108)-low-intermediate, CD11b (BUV737 Rat Anti-CD11b; BD; cat#612800))-high and CX3CR1 (Brilliant Violet 605^™ anti-mouse CX3CR1 Antibody; Biolegend; cat#149027)-high population was FACS sorted for bulk-RNA-sequencing. ASCA-2 (ACSA-2 Antibody, anti-mouse, PE, REAfinity^™; Miltenyi Biotetec; cat#130-116-244) was used to positively stain for astrocytes.

Statistical analysis and reproducibility

Each N represents a biological replicate. Statistical tests were performed using GraphPad Prism 10 and R. Bar graphs and dot plots represent means and error bars displayed in graphs represent standard error of means. For pooled analysis of results from different independent repeats, all mice from the same experimental group were pooled and a new statistical comparison was made for the entire pooled experiment, as performed for individual repeats. All measurements were taken from distinct samples. For all datasets, normality was calculated using the Shapiro–Wilk test and parametric or non-parametric tests were used accordingly. For comparisons between two groups with normal distribution, a two-sided unpaired t-test was performed; for comparisons between two groups with a normal distribution with different variance across groups, a two-sided unpaired t-test with Welch’s correction was used; for comparisons between two groups with non-normal distribution, a two-sided Mann–Whitney U-test was used; for comparison of more than two groups, ANOVA was used with correction for multiple comparisons using Tukey’s correction for one-way ANOVA and Bonferroni’s correction for two-way ANOVA test for parametric-distributed datasets and the Kruskal–Wallis test with Dunn’s correction for non-parametric-distributed datasets. P < 0.05 was considered significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. All experiments were repeated unless noted. Grubbs’ Outlier Test (p<0.05) and ROUT Outlier Test (Q = 1%) was used to determine outliers, which were excluded in certain experiments as denoted in the figure legends.

Extended Data

Extended Data Fig. 3: — a, Single-cell suspensions isolated from brain, meninges, white fat and spleen were stimulated with PMA-ionomycin and incubated with Brefeldin A at 37°C for four hours. Representative flow cytometric plots of IFNγ and IL-17A double producing cells present in the brain at steady-state. b, Representative flow cytometric expression of IL-17A from brain, meninges, white fat and spleen. Single-cell suspensions isolated from brain, meninges, white fat, colon, brown fat, liver, lung, ovaries, spleen, and lymph nodes (mesenteric (m)LN, inguinal (i)LN and deep cervical (dc)LN) were stimulated with PMA-ionomycin and incubated with golgi-plug at 37°C for four hours. **** = p<0.0001 one-way ANOVA with Tukey’s multiple comparison’s test (n=4 per group pooled from 3 experiments). c, Representative flow cytometric expression of transcription factor Foxp3 and cytokine IL-10 from steady-state brain, meninges, white fat and spleen. d, Representative flow cytometric expression of transcription factor Gata3 and cytokine IL-4 staining in the steady-state brain, spleen and lung. e, Representative flow cytometric expression of co-inhibitory receptors PD-1, Tigit and LAG3 in the steady-state brain, spleen and lung.

Extended Data Fig. 4: — a, Flow cytometric analysis was performed on single-cell suspensions isolated from fresh brain samples (n=2) obtained 4–5 hours post-mortem. Representative flow gating strategy to isolate CD4 and CD8 T cells from the human autopsy samples; the CSF data was used as the example. CXCR3 and CXCR6 gating on CD4 and CD8 T cells were displayed for CSF and Blood samples. Gating strategy was kept the same across all organs harvested for the experiment. b, The frequency of CXCR3 of CXCR6-expressing CD8 T cells from patient 6 (Table 1) was graphed. The density of CD4 and CD8 T cells in the brain were generated from absolute counts using counting beads divided by weight of the original tissue sample. (n=1) c, The frequency and density of CXCR3 or CXCR6-expressing CD4 T cells from patient 7 (Table 1) were graphed. The density of CD4 and CD8 T cells in the brain were generated from absolute counts using counting beads divided by weight of the original tissue sample. (n=1) d, The frequency and density of CXCR3 or CXCR6-expressing CD4 T cells from patient 7 (Table 1) was graphed. The density of CD4 and CD8 T cells in the brain were generated from absolute counts using counting beads divided by weight of the original tissue sample. (n=1)

Extended Data Fig. 5: — a, UMAP generated from single cell-RNA-sequencing (scRNA-seq) of T cells (FACS: CD45-intravenous label⁻, CD11b⁻, CD19⁻, B220⁻ and RNA: *Cd3e* > 1) sorted from brain, meninges, gonadal white fat, colon lamina propria, mLN and blood of 5 pooled 10-week-old C57BL/6 mice. Comparison of gene expression across all tissue T cells representing various T cell differentiation states. b, Schematic of bulk-RNA-sequencing experimental setup. c, PCA analysis of the CD4 T cells from brain, fat and meninges. (n=8 per biological replicate represented as a dot in the PCA) Differentially expressed genes (DEGs) (cutoff FDR < 0.05) from comparing brain/white fat and brain/meningeal CD4 T cells were utilized for Qiagen Ingenuity Pathway Analysis (IPA). Volcano plots represent DEGs (cutoff FDR < 0.05) from comparing brain/white fat, brain/meninges and white fat/meninges. d, PCA analysis of the CD4 T cells from brain, iLN and lung. (n=8 per biological replicate represented as a dot in the PCA) The circles in the PCA plot represent the 95% confidence interval. Volcano plot representing DEGs (cutoff FDR < 0.05) from bulk-RNA-sequencing analysis of brain/LN and brain/lung CD4 T cells. e, DEGs (cutoff FDR < 0.05) from comparing brain/white fat, brain/meningeal and white fat/meningeal CD4 T cells were utilized for Qiagen Ingenuity Pathway Analysis (IPA).

Extended Data Fig. 6: — a, Body fat percentage of mice fed a high-fat diet (HFD). (n=3 per group; pooled 2 experiments). b, Tissue CD4 T cell numbers HFD-fed mice. (n=7 per group; pooled 2 experiments). c, Body fat percentage of mice fasted for 48 hours. (n=9 per group; pooled 2 experiments). d, CD4 T cell numbers in mice fasted for 48 hours. (n=3 per group; brain, white-fat, brown-fat pooled 2 experiments; repeated twice). e, Body fat percentage changes in mice fed a HFD and then fasted. (n=3 per group) f, Spleen and bone marrow CD4 T cell numbers as in e. (n=3 per group; spleen=pooled 2 experiments). g, Representative FACS plots from fat-photoconverted mice. h, Brain IFNγ+ and total CD8 T cell numbers at various ages. (n=3 per group; repeated once) i, Total CD4 and CD8 T cells and IFNγ-secreting CD8 T cell numbers in brains of various-aged germ free and SPF mice. (n=3 per group; repeated once). j, FACS analysis of brain CD4 T cells post antibiotic (ABX)-treatment. k, Same as j, except for total CD4 T cells in other tissues. (n=4 per group; repeated once). l Representative FACS plots of adult and m, newly-weaned mice that underwent gut or fat photoconversion. Proportion of photoconverted T cells quantified. (n=3 per group; pooled 2 experiments) n, Brain CD4 T cell numbers in weaning-aged mice treated with anti-VLA4 and anti-LFA1. (n=4 per group) o, Brain CD4 T cell numbers in weaning-aged mice treated with FTY720. (n=4 per group). Statistical tests: Unpaired t-test for **a, b, c, d, k, n, o.** Two-way ANOVA with Bonferroni’s multiple comparisons test for **e, f, i.** One-way ANOVA with Tukey’s multiple comparisons test for **h, m.** For all graphs, *= p<0.05, **=p<0.01, ***=p<0.001, ****= p<0.0001.

Extended Data Fig. 7: — a, Percent of unique T cell clones per tissue and line graph representing raw number of unique TCR clones in each tissue from the mouse single-cell TCR-seq dataset. b, Demultiplexed brain sample according to unique Hashtag antibody barcode (TotalSeq-C Hashtag 1–6) labeling each mouse. c, TCR clonal sharing as measured by the Morisita index (n=5). d, FACS analysis of CD4 T cells in Smarta TCR (Vα2 Vβ8) transgenic mouse. *=p<0.01 by unpaired t-test (n=4 per group, pooled 2 experiments in white fat and mesenteric lymph node (mLN); repeated once). e, FACS analysis of Smarta Rag1^−/− mice that were immunized using lipopolysaccharide (LPS) as an adjuvant and co-administered with vehicle or GP_61–80 peptide by oral route to prime antigens through the gut. (n=4 per group) f, Same as e, with an additional boost of vehicle, LPS or LPS and GP_61–80 peptide (n=3 per group). g, 50,000 naive 7B8 T cells were transferred into gnotobiotic mice administered germ-free or SFB fecal slurry. FACS analysis of the absolute number and frequency of TH17 7B8 T cells from white fat, mLN, small intestine and spleen 5-weeks later. *=p<0.05, **=p<0.01 by unpaired parametric t-test (n=5 per group; repeated twice). h, same as g except 50,000 naive 7B8 T cells were transferred into Jackson Laboratory (no SFB) or Taconic Biosciences (have SFB) mice. *=p<0.05 by unpaired parametric t-test *=p<0.05 by unpaired parametric t-test (n=4 per group).

Extended Data Fig. 8: — a, Degree of TCR clonal expansion from T cell clones in the human single-cell RNA-seq dataset (CSF, Blood_1&2, Colon, Duodenum) was analyzed using ScRepertoire. Inverse Simpson Index, Chao Index, and Inverse Pielou Index applied to TCR clonal expansion. **=p<0.01, ***=p<0.001 and ****=p<0.0001 by one-way ANOVA (n=5) b, Highlighted the T cells with shared TCR clone between CSF, colon and blood in each tissue’s UMAP graph. c, Volcano plot of differentially expressed gene (DEG) after comparing clusters containing the shared TCR clone from each tissue (Colon vs Blood_2 and Colon vs CSF). d, Trajectory analysis on mouse scRNA-seq dataset from Extended Data Fig. 5 to identify genes highly expressed in brain T cells along the Slingshot pseudotime trajectory. e, FACS analysis on CXCR6 deficient mice for IFNγ and IL-17A-secreting CD4 T cells in the brain, white fat, and meninges and spleen. **=p<0.01 and ****=p<0.0001 by one-way ANOVA test (n=3 per group) f, FACS analysis on IP10 (CXCL10) deficient mice for total CD4 T cell numbers and proportion of CXCR3+ IFNγ+ CD4 T cells in the brain. **=p<0.01 by unpaired t-test (n=5 per group, outliers were calculated using Grubbs’ Outlier Test and excluded.)

Extended Data Fig. 9: — a, TCRβ (knockout) KO mice subjected to the light-dark box test (LDT), b, open field test (OFT), c, tail suspension test (TST) and two days of forced swim test (FST). (Pooled 3 experiments (males) and 3 experiments (females)) d, Brain CD4 and CD8 T cell numbers in IFNγ KO mice. (n=4 per group pooled; repeated twice). e, IFNγ KO mice subjected to LDT, f, OFT g, TST and FST. (Pooled 3 experiments (males) and 4 experiments (females)). h, Total and IFNγ+ brain CD4 and CD8 T cell numbers in T-bet KO mice (n=3 per group; repeated twice). i, subjected to LDT, j, OFT, k, TST and FST (Pooled 3 experiments (males) and 4 experiments (females)). l, β-hydroxybutyrate (BHOB), triglycerides (TG) and non-esterified fatty acids (NEFA) measured in the serum. (n=5 per group). m, CD4 and CD8 T cell numbers in IL17A/F double knockout (dKO) mice (n=3 per group) n, subjected to the novelty suppressed feeding test (NSFT), o, LDT, p, OFT, q, TST and FST (Pooled 2 experiments (males and females), except OFT males). r, Representative FACS plot and bar graph of TCRβ KO mice adoptively transferred with wildtype (CD45.1) or IFNγ KO CD4 T cells for 5-weeks. (n=12 (males), 6 (females) per group; pooled 2 experiments) s, LDT, t, OFT u, TST and FST. (Pooled 4 experiments (males) and 5 experiments (females)). v, same as r except with IFNγ KO recipient mice. (n=5 per group) w, LDT, x, NSFT, y, OFT, z TST and FST. (Pooled 3 experiments (males) and 2 experiments (females)). Statistical tests: Unpaired t-test for **d,h,m. T**wo-way ANOVA with Bonferroni’s multiple comparisons test for **a-c, e-g, i-l, n-z**. *=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001. Outliers were calculated using the ROUT Outlier Test (Q = 1%) and excluded.

Extended Data Fig. 10: — a, Smarta Rag knockout (KO) or heterozygous (het) mice were subjected to the novelty-suppressed feeding test (NSFT), b, light-dark box test (LDT), c, open field test (OFT), d, tail suspension test (TST) and two days of forced swim test (FST). (Pooled 3 experiments for males and females). e, CXCR6 KO mice were subjected to the NSFT, f, LDT, g, OFT, h, TST and FST. (Pooled 3 experiments for males and females). i, IFNGR1^f/f CX3CR1-Cre mice were subjected to the LDT, j, OFT, k, TST and FST. (Pooled 4 experiments for males and females) l, Bulk-RNA-sequencing of FACS sorted microglia from wildtype and T-bet KO mice. Separate PCA plot by sex. m, Volcano plot of differentially expressed genes (DEGs). DEGs (cutoff FDR < 0.05) were utilized for Qiagen Ingenuity Pathway Analysis (IPA). n, IFNGR1^f/f Nestin-Cre mice were subjected to the LDT, o, OFT, p, TST and FST. q, IFNGR1^f/f GFAP-Cre mice were subjected to the LDT, r, OFT, s, TST and FST. t, Single-cell suspensions were isolated from the brains of IFNγ KO mice and prepared for FACS to identify microglia and astrocytes. (n=3 per group; pooled 2 experiments for females) Statistical tests: *=p<0.05, **=p<0.01 and ***p<0.001 by two-way ANOVA with Bonferroni’s multiple comparisons test for **a-k, n-t**. Outliers were calculated using the ROUT Outlier Test (Q = 1%) and excluded.

Supplementary Material

Supp Table 1

NIHMS2123216-supplement-Supp_Table_1.csv^{(26.3KB, csv)}

Supp Table 3

NIHMS2123216-supplement-Supp_Table_3.csv^{(17KB, csv)}

Supp Table 2

NIHMS2123216-supplement-Supp_Table_2.csv^{(23.4KB, csv)}

Supp Table 6

NIHMS2123216-supplement-Supp_Table_6.csv^{(133.3KB, csv)}

Supp Table 4

NIHMS2123216-supplement-Supp_Table_4.csv^{(2.7MB, csv)}

Supp Table 7

NIHMS2123216-supplement-Supp_Table_7.csv^{(2.2MB, csv)}

Supp Table 5

NIHMS2123216-supplement-Supp_Table_5.csv^{(3.7MB, csv)}

Supp Table 9

NIHMS2123216-supplement-Supp_Table_9.csv^{(18.7KB, csv)}

Supp Table 10

NIHMS2123216-supplement-Supp_Table_10.csv^{(15.3KB, csv)}

Supp Table 11

NIHMS2123216-supplement-Supp_Table_11.csv^{(24.1KB, csv)}

Supp Table 8

NIHMS2123216-supplement-Supp_Table_8.csv^{(2.2MB, csv)}

Supp Video 1

Download video file^{(1.9MB, mp4)}

Supp Video 2

Download video file^{(11.2MB, mp4)}

Supp Video 3

Download video file^{(9.1MB, mp4)}

Supp Video 4

Download video file^{(12.7MB, mp4)}

Supp Video 6

Download video file^{(14.4MB, mp4)}

Supp Video 5

Download video file^{(9.8MB, mov)}

Supp Video 7

Download video file^{(27MB, mov)}

Acknowledgments:

We would like to thank the Yale Legacy Tissue Donation Program and Center for Human Brain Discovery for the human autopsy tissue samples. Dr. David Wolfsohn for performing the colonoscopy for our human study collection. Dr. Dan Littman (NYU Langone) for sharing 7B8 mice and an aliquot of FITC Rat Anti-Mouse Vβ 14 T-Cell Receptor antibody. Elsie Schaubeck and others from LifeCanvas Technologies. Ewa Menet, Brittany Shonts, Ashraf Khalil, Jesse Bonin and Diane Trotta for cell sorting through the Yale Flow Cytometry Core Facility. Ivan Klenovskiy for helping with Python coding-related tasks.

We thank the Yale Center for Cellular & Molecular Imaging Confocal Facility for usage of their Lecia Stellaris microscope. We thank Yale Flow Cytometry for their assistance with cell sorting and instrument maintenance. The Core is supported in part by an NCI Cancer Center Support Grant # NIH P30 CA016359. The BD Symphony was funded by shared instrument grant # NIH S10 OD026996. We thank the Yale Center for Genome Analysis for their assistance with library prep and genomic sequencing services. YCGA-related research reported in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number 1S10OD030363-01A1. We would like to acknowledge Biorender for their platform to generate schematics.

Funding:

AW was supported by National Institutes of Health (NIH) grants R01 AI162645 and R01 AR080104, the Smith Family Foundation, the Colton Center for Autoimmunity at Yale, and the Food and Allergy Science Initiative; an award from the PEW Charitable Trusts; and gifts from the Mathers Family Foundation, the Ludwig Family Foundation and the Knights of Columbus. DAH was supported by NIH grants (P01 AI073748, R01 AI22220, UM 1HG009390, P01 AI039671, P50 CA121974, R01 CA227473), Race to Erase MS and National MS Society. TMY was supported by the Smith Family Foundation and NIH grant 1F31NS130957-01A1. YSM and MRP were supported by DP1 DA050986. JEC was supported by NIH grant R37 AR40072.

Competing interests:

DAH has received research funding from Bristol-Myers Squibb, Novartis, Sanofi, and Genentech. He has been a consultant for Bayer Pharmaceuticals, Repertoire Inc, Bristol Myers Squibb, Compass Therapeutics, EMD Serono, Genentech, Juno therapeutics, Novartis Pharmaceuticals, Proclara Biosciences, Sage Therapeutics, and Sanofi Genzyme. AW has received research funding from NGM Biopharmaceuticals and Astra Zeneca. He has been a consultant for Seraova Bio, The Column Group, and Soleil Labs. He is on the scientific advisory board for NGM Biopharmaceuticals.

Data and materials availability:

The data that support the findings of this study are available on request from the corresponding authors, DAH and AW. Sequence files and metadata for all samples used in this study has been deposited in the NCBI Sequence Read Archive (SRA). Single-cell RNA-sequencing data generated on T cells isolated from steady-state adult mouse brain, meninges, white fat, colon-lamina propria, lung and blood referred in Main Figures 1, 2 and 3 are available in the NCBI SRA under PRJNA1240821. Single-cell RNA-sequencing data generated on T cells isolated from steady-state adult mouse brain, meninges, white fat, colon-lamina propria, mesenteric lymph node and blood referred in Extended Data Figure 5 are available in the NCBI SRA under PRJNA1074276. Bulk RNA-sequencing data generated on CD4 T cells FACS isolated from steady-state adult mouse brain, meninges and white fat referred in Extended Data Figure 5 are available in the NCBI SRA under PRJNA1073984. Bulk RNA-sequencing data generated on CD4 T cells isolated from steady-state adult mouse brain, lungs and lymph nodes referred to in Extended Data Figure 5 are available in the NCBI SRA under PRJNA1074381. Single-cell RNA-sequencing data generated from healthy human adult CSF, colon, duodenum and blood are referred to in Main Figure 4 and Extended Data Figure 8 are available in the NCBI SRA under PRJNA1074056. Bulk RNA-sequencing data generated from microglia isolated steady-state adult wildtype and T-bet KO mouse brain referred to in Extended Data Figure 10 are available in the NCBI SRA under PRJNA1073982.

Main Text References:

1.Jaschke NP & Wang A Integrated control of leukocyte compartments as a feature of adaptive physiology. Immunity 58, 279–294, doi: 10.1016/j.immuni.2025.01.013 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Pasciuto E et al. Microglia Require CD4 T Cells to Complete the Fetal-to-Adult Transition. Cell 182, 625–640.e624, doi: 10.1016/j.cell.2020.06.026 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Pappalardo JL et al. Transcriptomic and clonal characterization of T cells in the human central nervous system. Science Immunology 5, eabb8786, doi:doi: 10.1126/sciimmunol.abb8786 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Smolders J et al. Tissue-resident memory T cells populate the human brain. Nat Commun 9, 4593, doi: 10.1038/s41467-018-07053-9 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Schnell A et al. Stem-like intestinal Th17 cells give rise to pathogenic effector T cells during autoimmunity. Cell 184, 6281–6298.e6223, doi: 10.1016/j.cell.2021.11.018 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Berer K et al. Commensal microbiota and myelin autoantigen cooperate to trigger autoimmune demyelination. Nature 479, 538–541, doi: 10.1038/nature10554 (2011). [DOI] [PubMed] [Google Scholar]
7.Hsiao EY et al. Microbiota modulate behavioral and physiological abnormalities associated with neurodevelopmental disorders. Cell 155, 1451–1463, doi: 10.1016/j.cell.2013.11.024 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Seki D et al. Aberrant gut-microbiota-immune-brain axis development in premature neonates with brain damage. Cell Host & Microbe 29, 1558–1572.e1556, doi: 10.1016/j.chom.2021.08.004 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Kim S et al. Maternal gut bacteria promote neurodevelopmental abnormalities in mouse offspring. Nature 549, 528–532, doi: 10.1038/nature23910 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Sampson TR et al. Gut Microbiota Regulate Motor Deficits and Neuroinflammation in a Model of Parkinson’s Disease. Cell 167, 1469–1480.e1412, doi: 10.1016/j.cell.2016.11.018 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Medawar PB Immunity to homologous grafted skin; the fate of skin homografts transplanted to the brain, to subcutaneous tissue, and to the anterior chamber of the eye. Br J Exp Pathol 29, 58–69 (1948). [PMC free article] [PubMed] [Google Scholar]
12.Carson MJ, Doose JM, Melchior B, Schmid CD & Ploix CC CNS immune privilege: hiding in plain sight. Immunol Rev 213, 48–65, doi: 10.1111/j.1600-065X.2006.00441.x (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Radjavi A, Smirnov I & Kipnis J Brain antigen-reactive CD4+ T cells are sufficient to support learning behavior in mice with limited T cell repertoire. Brain, Behavior, and Immunity 35, 58–63, doi: 10.1016/j.bbi.2013.08.013 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Rattazzi L et al. CD4+ but not CD8+ T cells revert the impaired emotional behavior of immunocompromised RAG-1-deficient mice. Translational Psychiatry 3, e280–e280, doi: 10.1038/tp.2013.54 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Derecki NC et al. Regulation of learning and memory by meningeal immunity: a key role for IL-4. J Exp Med 207, 1067–1080, doi: 10.1084/jem.20091419 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Filiano AJ et al. Unexpected role of interferon-γ in regulating neuronal connectivity and social behaviour. Nature 535, 425–429, doi: 10.1038/nature18626 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Ribeiro M et al. Meningeal γδ T cell-derived IL-17 controls synaptic plasticity and short-term memory. Sci Immunol 4, doi: 10.1126/sciimmunol.aay5199 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Alves de Lima K et al. Meningeal γδ T cells regulate anxiety-like behavior via IL-17a signaling in neurons. Nat Immunol 21, 1421–1429, doi: 10.1038/s41590-020-0776-4 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Hindmarch CC & Ferguson AV Physiological roles for the subfornical organ: a dynamic transcriptome shaped by autonomic state. J Physiol 594, 1581–1589, doi: 10.1113/jp270726 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Pulman KJ, Fry WM, Cottrell GT & Ferguson AV The subfornical organ: a central target for circulating feeding signals. J Neurosci 26, 2022–2030, doi: 10.1523/jneurosci.3218-05.2006 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Takahashi Y, Smith P, Ferguson A & Pittman QJ Circumventricular organs and fever. Am J Physiol 273, R1690–1695, doi: 10.1152/ajpregu.1997.273.5.R1690 (1997). [DOI] [PubMed] [Google Scholar]
22.Zhang J et al. Projections from subfornical organ to bed nucleus of the stria terminalis modulate inflammation-induced anxiety-like behaviors in mice. Science Advances 10, eadp9413, doi: 10.1126/sciadv.adp9413. [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Omri B et al. CD4 expression in neurons of the central nervous system. Int Immunol 6, 377–385, doi: 10.1093/intimm/6.3.377 (1994). [DOI] [PubMed] [Google Scholar]
24.Pasciuto E et al. Microglia Require CD4 T Cells to Complete the Fetal-to-Adult Transition. Cell 182, 625–640 e624, doi: 10.1016/j.cell.2020.06.026 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Collins N et al. The Bone Marrow Protects and Optimizes Immunological Memory during Dietary Restriction. Cell 178, 1088–1101.e1015, doi: 10.1016/j.cell.2019.07.049 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Zheng D, Liwinski T & Elinav E Interaction between microbiota and immunity in health and disease. Cell Res 30, 492–506, doi: 10.1038/s41422-020-0332-7 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Wang B et al. A High-Fat Diet Increases Gut Microbiota Biodiversity and Energy Expenditure Due to Nutrient Difference. Nutrients 12, doi: 10.3390/nu12103197 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Al Nabhani Z et al. A Weaning Reaction to Microbiota Is Required for Resistance to Immunopathologies in the Adult. Immunity 50, 1276–1288.e1275, doi: 10.1016/j.immuni.2019.02.014 (2019). [DOI] [PubMed] [Google Scholar]
29.Ivanov II et al. Induction of intestinal Th17 cells by segmented filamentous bacteria. Cell 139, 485–498, doi: 10.1016/j.cell.2009.09.033 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Morton AM et al. Endoscopic photoconversion reveals unexpectedly broad leukocyte trafficking to and from the gut. Proc Natl Acad Sci U S A 111, 6696–6701, doi: 10.1073/pnas.1405634111 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Galván-Peña S, Zhu Y, Hanna BS, Mathis D & Benoist C A dynamic atlas of immunocyte migration from the gut. Sci Immunol 9, eadi0672, doi: 10.1126/sciimmunol.adi0672 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Odoardi F et al. T cells become licensed in the lung to enter the central nervous system. Nature 488, 675–679, doi: 10.1038/nature11337 (2012). [DOI] [PubMed] [Google Scholar]
33.Samuels BA & Hen R in Mood and Anxiety Related Phenotypes in Mice: Characterization Using Behavioral Tests, Volume II (ed Gould Todd D.) 107–121 (Humana Press, 2011). [Google Scholar]
34.Kunis G et al. IFN-γ-dependent activation of the brain’s choroid plexus for CNS immune surveillance and repair. Brain 136, 3427–3440, doi: 10.1093/brain/awt259 (2013). [DOI] [PubMed] [Google Scholar]
35.Zheng Y et al. Blocking the IFN-gamma signal in the choroid plexus confers resistance to experimental autoimmune encephalomyelitis. Faseb j 37, e22833, doi: 10.1096/fj.202201767R (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Szabo SJ et al. A novel transcription factor, T-bet, directs Th1 lineage commitment. Cell 100, 655–669, doi: 10.1016/s0092-8674(00)80702-3 (2000). [DOI] [PubMed] [Google Scholar]
37.Szabo SJ et al. Distinct effects of T-bet in TH1 lineage commitment and IFN-gamma production in CD4 and CD8 T cells. Science 295, 338–342, doi: 10.1126/science.1065543 (2002). [DOI] [PubMed] [Google Scholar]
38.Wang Y et al. TREM2-mediated early microglial response limits diffusion and toxicity of amyloid plaques. J Exp Med 213, 667–675, doi: 10.1084/jem.20151948 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Lee S et al. APOE modulates microglial immunometabolism in response to age, amyloid pathology, and inflammatory challenge. Cell Rep 42, 112196, doi: 10.1016/j.celrep.2023.112196 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Lituma PJ et al. Altered synaptic connectivity and brain function in mice lacking microglial adapter protein Iba1. Proc Natl Acad Sci U S A 118, doi: 10.1073/pnas.2115539118 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Hong S et al. Complement and microglia mediate early synapse loss in Alzheimer mouse models. Science 352, 712–716, doi: 10.1126/science.aad8373 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Wójtowicz S, Strosznajder AK, Jeżyna M & Strosznajder JB The Novel Role of PPAR Alpha in the Brain: Promising Target in Therapy of Alzheimer’s Disease and Other Neurodegenerative Disorders. Neurochem Res 45, 972–988, doi: 10.1007/s11064-020-02993-5 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Clark DN et al. Prolonged STAT1 activation in neurons drives a pathological transcriptional response. J Neuroimmunol 382, 578168, doi: 10.1016/j.jneuroim.2023.578168 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Clark DN et al. Prolonged STAT1 signaling in neurons causes hyperactive behavior. Brain Behav Immun 124, 1–8, doi: 10.1016/j.bbi.2024.11.018 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Clark DN, Begg LR & Filiano AJ Unique aspects of IFN-γ/STAT1 signaling in neurons. Immunol Rev 311, 187–204, doi: 10.1111/imr.13092 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Smith BC et al. Astrocyte interferon-gamma signaling dampens inflammation during chronic central nervous system autoimmunity via PD-L1. Journal of Neuroinflammation 20, 234, doi: 10.1186/s12974-023-02917-4 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Hindinger C et al. IFN-γ signaling to astrocytes protects from autoimmune mediated neurological disability. PLoS One 7, e42088, doi: 10.1371/journal.pone.0042088 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Clark IC et al. Barcoded viral tracing of single-cell interactions in central nervous system inflammation. Science 372, eabf1230, doi:doi: 10.1126/science.abf1230 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Kantzer CG et al. Anti-ACSA-2 defines a novel monoclonal antibody for prospective isolation of living neonatal and adult astrocytes. Glia 65, 990–1004, doi: 10.1002/glia.23140 (2017). [DOI] [PubMed] [Google Scholar]
50.Giedd JN, Castellanos FX, Rajapakse JC, Vaituzis AC & Rapoport JL Sexual dimorphism of the developing human brain. Prog Neuropsychopharmacol Biol Psychiatry 21, 1185–1201, doi: 10.1016/s0278-5846(97)00158-9 (1997). [DOI] [PubMed] [Google Scholar]
51.Rurak GM et al. Sex differences in developmental patterns of neocortical astroglia: A mouse translatome database. Cell Reports 38, 110310, doi: 10.1016/j.celrep.2022.110310 (2022). [DOI] [PubMed] [Google Scholar]
52.Barr MM, García LR & Portman DS Sexual Dimorphism and Sex Differences in Caenorhabditis elegans Neuronal Development and Behavior. Genetics 208, 909–935, doi: 10.1534/genetics.117.300294 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

Methods References:

53.Unutmaz D et al. The primate lentiviral receptor Bonzo/STRL33 is coordinately regulated with CCR5 and its expression pattern is conserved between human and mouse. J Immunol 165, 3284–3292, doi: 10.4049/jimmunol.165.6.3284 (2000). [DOI] [PubMed] [Google Scholar]
54.Oxenius A, Bachmann MF, Zinkernagel RM & Hengartner H Virus-specific MHC-class II-restricted TCR-transgenic mice: effects on humoral and cellular immune responses after viral infection. Eur J Immunol 28, 390–400, doi: 10.1002/(sici)1521-4141(199801)28:01<390::Aid-immu390>3.0.Co;2-o (1998). [DOI] [PubMed] [Google Scholar]
55.Yang Y et al. Focused specificity of intestinal TH17 cells towards commensal bacterial antigens. Nature 510, 152–156 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Lodolce JP et al. IL-15 receptor maintains lymphoid homeostasis by supporting lymphocyte homing and proliferation. Immunity 9, 669–676, doi: 10.1016/s1074-7613(00)80664-0 (1998). [DOI] [PubMed] [Google Scholar]
57.Dufour JH et al. IFN-gamma-inducible protein 10 (IP-10; CXCL10)-deficient mice reveal a role for IP-10 in effector T cell generation and trafficking. J Immunol 168, 3195–3204, doi: 10.4049/jimmunol.168.7.3195 (2002). [DOI] [PubMed] [Google Scholar]
58.Griswold SL, Sajja KC, Jang CW & Behringer RR Generation and characterization of iUBC-KikGR photoconvertible transgenic mice for live time-lapse imaging during development. Genesis 49, 591–598, doi: 10.1002/dvg.20718 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Mombaerts P et al. Mutations in T-cell antigen receptor genes alpha and beta block thymocyte development at different stages. Nature 360, 225–231, doi: 10.1038/360225a0 (1992). [DOI] [PubMed] [Google Scholar]
60.Dalton DK et al. Multiple defects of immune cell function in mice with disrupted interferon-gamma genes. Science 259, 1739–1742, doi: 10.1126/science.8456300 (1993). [DOI] [PubMed] [Google Scholar]
61.Finotto S et al. Development of spontaneous airway changes consistent with human asthma in mice lacking T-bet. Science 295, 336–338, doi: 10.1126/science.1065544 (2002). [DOI] [PubMed] [Google Scholar]
62.Haas JD et al. Development of interleukin-17-producing γδ T cells is restricted to a functional embryonic wave. Immunity 37, 48–59, doi: 10.1016/j.immuni.2012.06.003 (2012). [DOI] [PubMed] [Google Scholar]
63.Lee SH et al. Identifying the initiating events of anti-Listeria responses using mice with conditional loss of IFN-γ receptor subunit 1 (IFNGR1). J Immunol 191, 4223–4234, doi: 10.4049/jimmunol.1300910 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Yona S et al. Fate mapping reveals origins and dynamics of monocytes and tissue macrophages under homeostasis. Immunity 38, 79–91, doi: 10.1016/j.immuni.2012.12.001 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Gregorian C et al. Pten deletion in adult neural stem/progenitor cells enhances constitutive neurogenesis. J Neurosci 29, 1874–1886, doi: 10.1523/jneurosci.3095-08.2009 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Tronche F et al. Disruption of the glucocorticoid receptor gene in the nervous system results in reduced anxiety. Nat Genet 23, 99–103, doi: 10.1038/12703 (1999). [DOI] [PubMed] [Google Scholar]
67.Schnell A et al. Stem-like intestinal Th17 cells give rise to pathogenic effector T cells during autoimmunity. Cell 184, 6281–6298.e6223, doi: 10.1016/j.cell.2021.11.018 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Al Nabhani Z et al. A Weaning Reaction to Microbiota Is Required for Resistance to Immunopathologies in the Adult. Immunity 50, 1276–1288.e1275, doi: 10.1016/j.immuni.2019.02.014 (2019). [DOI] [PubMed] [Google Scholar]
69.Park YG et al. Protection of tissue physicochemical properties using polyfunctional crosslinkers. Nat Biotechnol, doi: 10.1038/nbt.4281 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Bankhead P et al. QuPath: Open source software for digital pathology image analysis. Sci Rep 7, 16878, doi: 10.1038/s41598-017-17204-5 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Stoeckius M et al. Cell Hashing with barcoded antibodies enables multiplexing and doublet detection for single cell genomics. Genome Biol 19, 224, doi: 10.1186/s13059-018-1603-1 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Borcherding N, Bormann NL & Kraus G scRepertoire: An R-based toolkit for single-cell immune receptor analysis. F1000Res 9, 47, doi: 10.12688/f1000research.22139.2 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Street K et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics 19, 477, doi: 10.1186/s12864-018-4772-0 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Van den Berge K et al. Trajectory-based differential expression analysis for single-cell sequencing data. Nat Commun 11, 1201, doi: 10.1038/s41467-020-14766-3 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Qi G, Mi Y & Yin F Characterizing brain metabolic function ex vivo with acute mouse slice punches. STAR Protoc 2, 100559, doi: 10.1016/j.xpro.2021.100559 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Pennington ZT et al. ezTrack: An open-source video analysis pipeline for the investigation of animal behavior. Sci Rep 9, 19979, doi: 10.1038/s41598-019-56408-9 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Favuzzi E et al. GABA-receptive microglia selectively sculpt developing inhibitory circuits. Cell 184, 4048–4063.e4032, doi: 10.1016/j.cell.2021.06.018 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials