Mice

All animal experiments were performed with the approval and according to the regulations of the District Government of Upper Bavaria and reported according to the guidelines. Animals were housed at the German Center for Neurodegenerative Diseases (DZNE) in Munich in groups in Greenline IVC GM500 plastic cages in a temperature-controlled environment (21 ± 2 °C) on a 12-h light–dark cycle with food and water available ad libitum. The mouse lines used in the study include: WT C57BL/6J mice from Janvier Labs, B6.Cg-Tg(APPSwFlLon,PSEN1*M146L*L286V)6799Vas/Mmjax (5xFAD)21 and B6.129S7-Rag1tm1Mom/J (Rag1 KO)22 from the Jackson Laboratory, and Apptm3.1Tcs (APPNLGF)23. Experiments were performed with adult male mice aged 4, 6, 10 and 12 months as indicated in the figures and figure legends. Mice were randomly assigned to different treatment groups. For the depletion of CD8+ T cells, male mice, 5xFAD mice aged 6 months were injected intraperitoneally with anti-CD8 antibody (catalog no. BP0061, Bio X Cell) and their respective isotype control (catalog no. BE0119, Bio X Cell) twice a week for a total of 6 weeks. 5xFAD male mice aged 6 months were given either antibodies against PD-1 (catalog no. BP0146, Bio X Cell) and CTLA4 (catalog no. BP0131, Bio X Cell) or their isotype controls (catalog nos. BE0089 and BP0087, Bio X Cell) at a concentration of 10 mg kg−1 or 20 mg kg−1 body weight, respectively. Antibodies were injected intraperitoneally twice a week for 6 weeks. WT and 5xFAD female mice aged 6 months were fed powdered chow (ssniff V1120) mixed with baricitinib (catalog no. V0338, InvivoChem) dissolved in corn oil at a concentration of 10 mg kg−1 body weight. Alternatively, control animals were fed with powdered chow mixed with corn oil only. 5xFAD female mice aged 6 months were fed either with control AIN-76A standard chow or 0.12% PLX5622 (MedChemExpress, mixed in with AIN-76A standard chow, Research Diets) for 6 weeks. EdU (catalog no. E10187, Invitrogen) was dissolved in drinking water at a final concentration of 0.2 mg ml−1 for 2 weeks. Water was replaced every 48 h.

Human samples

Patients with a neuropathological diagnosis of AD were retrieved from the archives of the Institute of Neuropathology, University Medical Center Göttingen, Germany. Patients were eligible to participate in the study if they or their next of kin had consented to tissue donation and to the use of their tissue for research. The study was approved by the ethics committee of the University Medical Center Göttingen (39/2/19). Sampling of the hippocampus was performed during autopsy in 16 individuals; samples were formalin-fixed and paraffin-embedded. Each individual was stratified according to the ‘ABC’ score24. Patients with intermediate or high risk (n = 12) were assigned as patients with AD. For patient information, see Supplementary Table 1.

IHC

Mice were anesthetized with a solution of 10 mg ml−1 ketamine and 1 mg ml−1 xylazine injected intraperitoneally. They were then perfused transcardially with 4% paraformaldehyde (PFA) and brains were post-fixed in 4% PFA for 2 h, followed by cryopreservation in 30% sucrose in PBS. After freezing the tissue on dry ice using Tissue-Tek O.C.T. compound, 30-μm coronal sections were cut using the Leica CM 1900 cryostat. Free-floating tissue sections were collected in a solution containing 25% glycerol and 25% ethylene glycol in PBS. Tissue sections were rinsed with 1× PBS followed by permeabilization with 0.5% Triton X-100 in PBS for 30 min at room temperature. The sections were then incubated in Fab fragment goat anti-mouse IgG (1:100 dilution, Dianova) for 1 h at room temperature to block endogenous mouse tissue immunoglobulins. Sections were blocked for 1 h in a solution consisting of 2.5% FCS, 2.5% fish gelatin and 2.5 % fish gelatin in PBS. Primary antibodies were diluted in 10% blocking solution and incubated overnight at 4 °C. The following day, sections were washed in PBS and then incubated in secondary antibodies in 10% blocking solution for 2 h at room temperature. The sections were then washed with 1× PBS and stained with DAPI for 5 min and mounted. The following antibodies were used: mouse anti-APC (1:100 dilution, catalog no. OP80-100UG, Merck Millipore), rabbit anti-B2m (1:100 dilution, ab75853, Abcam), rabbit anti-STAT1 (1:100 dilution, catalog no. 14994S, Cell Signaling Technology), rat anti-CD8 (1:100 dilution, catalog no. 100702, Promega Corporation), rabbit anti-Iba1 (1:250 dilution, catalog no. 234-004, Wako), chicken anti-IBA1 (1:400 dilution, catalog no. 234009, Synaptic Systems), mouse anti-β-Amyloid1-11 (1:1,000 dilution, catalog no. 835104, BioLegend), mouse anti-myelin-associated glycoprotein (1:200 dilution, catalog no. MAB1567, Sigma-Aldrich), mouse anti-Serpin A3N (1:100 dilution, catalog no. AF4709, Bio-Techne), rabbit anti-Olig2 (1:250 dilution, catalog no. AB9610, Sigma-Aldrich), mouse anti-PDGF R alpha (1:100 dilution, catalog no. AF1062, R&D Systems), rat anti-MHC Class II (1:100 dilution, catalog no. 14-5321-82, Invitrogen), mouse anti-ISG15 (1:100 dilution, catalog no. sc-166755, Santa Cruz Biotechnology), rat anti-myelin basic protein (1:300 dilution, catalog no. ab7349, Abcam), chicken anti-MBP (1:400 dilution, catalog no. PA1-10008, Invitrogen), rabbit anti-BCAS1 (1:500 dilution, generated against amino acids 1–667 of mouse BCAS1, custom-made by Synaptic Systems), donkey anti-mouse IgG (H+L) Alexa Fluor 488 (1:500 dilution, catalog no. A-21202, Thermo Fisher Scientific), goat anti-mouse IgG (H+L) Alexa Fluor 555 (catalog no. 1:500, catalog no. A-21422, Thermo Fisher Scientific), goat anti-mouse IgG (H+L) Alexa Fluor 647 (1:500 dilution, catalog no. A-21235, Thermo Fisher Scientific), goat anti-rabbit IgG (H+L) Alexa Fluor 488 (1:500 dilution, catalog no. A-11008, Thermo Fisher Scientific), goat anti-rabbit IgG (H+L) Alexa Fluor 555 (1:500 dilution, catalog no. A-21428, Thermo Fisher Scientific), goat anti-rat IgG (H+L) Alexa Fluor 555 (1:500 dilution, catalog no. A-21434, Thermo Fisher Scientific), goat anti-mouse IgG Alexa Fluor 555 (1:500 dilution, catalog no. A-32116, Thermo Fisher Scientific), donkey anti-goat IgG (H+L) Alexa Fluor Plus 647 (1:500 dilution, catalog no. A32849, Thermo Fisher Scientific), donkey anti-rat Alexa Fluor 488 (1:500 dilution, Thermo Fisher Scientific), donkey anti-goat Alexa Fluor 555 (1:500 dilution, Thermo Fisher Scientific) and donkey anti-rabbit Alexa Fluor 647 (1:500 dilution, Thermo Fisher Scientific). For CC1, STAT1 and Serpina3n antibodies, after permeabilization tissue sections were boiled in 10 mM citrate buffer, pH 6.0, at 85 °C for 20 min. To detect plaques, after incubation with secondary antibodies, sections were stained using either ThR or Methoxy-XO4. For ThR (catalog no. S570435, Sigma-Aldrich) sections were incubated at a concentration of 1:1,000 in 1× PBS for 15 min at room temperature. For Methoxy-XO4 (catalog no. 4920, Tocris) staining, sections were incubated with 4 µg ml−1 Methoxy-XO4 in 50% ethanol and washed with 50% ethanol for 5 min. To detect EdU, the Alexa Fluor 647 Click-iT EdU Cell Proliferation Assay Kit (catalog no. C10340, Invitrogen) was used according to the manufacturer’s protocol. Briefly, after incubation in secondary antibodies, brain sections were washed with PBS and incubated in EdU developing cocktail for 45 min at room temperature, followed by post staining with DAPI before mounting on coverslips. Images were acquired using a Leica Microsystems TCS SP5 confocal microscope or with an LSM 900 microscope (ZEISS) and were processed and analyzed with the Imaris v.9.2.0 and ImageJ v.1.41 software. For IHC of human samples, we performed heat antigen retrieval after deparaffinization (steamer, citrate buffer, pH 6, for 30 min), incubation in 3% H2O2 and blocking with normal goat serum (10%) with 0.1% Triton X-100. The primary mouse anti-CD8 antibody (1:50 dilution, clone C8/144B, catalog no. M7103, Dako) was incubated overnight and developed with 3,3′-diaminobenzidine (Dako REAL EnVision, catalog no. K500711-2, Dako) until saturation of the chromogenic reaction. The second primary mouse anti-MHC-II antibody (1:50 dilution, clone CR3/43, catalog no. M0775, Dako) was again incubated overnight and developed using a goat anti-mouse IgG (1:50 dilution, catalog no. 115-055-003, Jackson ImmunoResearch) with alkaline phosphatase and fast blue as chromogen. Stained samples were digitized using a VS200 slide scanner (Evident) and manually quantified using the OMERO.iviewer software. CD8 cells were considered in the contact zone of an MHC-II+ cell if within a range of 50 µm.

Cell culture

Microglia were isolated from P15-P21 C57BL/6J WT pups using the Neuronal Tissue Dissociation Kit (Miltenyi Biotec) according to the manufacturer’s protocol. Briefly, brain tissue was dissected, cut into small pieces and dissociated using enzymatic digestion in the gentleMACS Octo Dissociator (Miltenyi Biotec). The tissue suspension was incubated with magnetic beads against CD11b and passed through a magnetic column. Microglia were flushed out of the column and plated in DMEM containing 10% FCS, 10 ng ml−1 monocytic colony-stimulating factor, 1% penicillin-streptomycin and 1% glutamate for 4–7 days before being used for the experiments.

Ex vivo myelin phagocytosis assay

Acute isolation of adult WT and 5xFAD microglia was performed as described above. In preparation for the assay, WT brains were flushed with PBS and fresh-frozen. Frozen brains were cut into 10-µm sections and transferred on Poly-l-lysine-coated coverslips and stored at −80 °C until use as described previously25. Frozen sections were temperature-adjusted for 30 min, washed with PBS and equilibrated to the microglial medium. Then, 100,000 acutely isolated microglia were plated on sections and immediately treated with 5 ng ml−1 IFNγ, 1 µM baricitinib or 5,000 U ml−1 IFN1β. Microglia were incubated on brain sections for 4 days; then, the internalized MBP+ material was analyzed.

Stereotactic injection in the corpus callosum

A solution of 10 ng μl−1 IFNγ was prepared in sterile 1× PBS. To identify the lesion area during tissue processing, Monastral blue (autoclaved and filtered, catalog no. 274011, Sigma-Aldrich) was added to a final concentration of 0.03%. Mice were anesthetized using a solution of medetomidine (concentration 0.5 mg kg−1), midazolam (concentration 5 mg kg−1) and fentanyl (concentration 0.05 mg kg−1) injected intraperitoneally. After anesthesia, the head was shaved and bepanthen cream (catalog no. 1578847, Bayer) was used to protect the eyes. The mouse was positioned on the stereotactic injection apparatus and a small incision was made in the skin to expose the skull. A small hole was drilled into the skull at the injection coordinates: X, ±0.55 mm; Y, −1.22 mm (from bregma); Z, −1.25 mm (from bregma). A glass capillary containing an IFNγ solution was lowered to the desired coordinates and 1 µl IFNy was injected at a speed of 100 nl min−1. After the injection, the mouse was injected with buprenorphine (concentration 0.05 mg per kg of body weight) and the skin was sutured. To terminate the anesthesia, the mouse was injected with an antagonist solution consisting of atipamezole (concentration 2.5 mg kg−1), naloxone (concentration 0.5 mg kg−1) and flumazenil (concentration 0.5 mg kg−1). The animals were perfused transcardially after 48 h with 4% PFA as described above.

EM

For the ultrastructural analysis, mice were perfused with 4% PFA (EM grade, Science Services) and 2 mM calcium chloride in 1× PBS, pH 7.4 (Science Services). One hemisphere was dedicated to ultrastructural analysis using immersion fixation in 4% PFA, 2.5% glutaraldehyde (EM grade, Science Services) and 2 mM calcium chloride in 0.1 M sodium cacodylate buffer for 24 h. Coronal 50–100-µm-thick vibratome sections were incubated in the same fixative for another 24 h and stored in 0.1 M sodium cacodylate buffer at 4 °C. The rOTO en bloc staining protocol included postfixation in 1% osmium tetroxide (Electron Microscopy Sciences), 1.5% potassium ferricyanide (Sigma-Aldrich) in 0.1 M sodium cacodylate (Science Services) buffer (pH 7.4)26. The contrast was enhanced by incubation in 1% thiocarbohydrazide (Sigma-Aldrich) for 45 min at 40 °C. The tissue was washed in water and incubated in 1% aqueous osmium tetroxide, washed and further contrasted using overnight incubation in 1% aqueous uranyl acetate at 4 °C and 2 h at 50 °C. Samples were dehydrated in an ascending ethanol series and infiltrated with the Araldite epoxy resin LX112 (Ladd Research). Blocks were cured for 48 h, trimmed (TRIM2, Leica Microsystems) and sectioned at 100-nm thickness using a 35° Ultra Diamond Knife (DiATOME) on an ultramicrotome (UC7, Leica Microsystems). Sections were collected onto 1× 0.5 cm carbon nanotube tape strips (Science Services) for SEM analysis. For SEM imaging, the samples on tape were attached to adhesive carbon tape (Science Services) on 4-inch silicon wafers (Siegert Wafer) and grounded using adhesive carbon tape strips (Science Services). EM micrographs were acquired on a Crossbeam Gemini 340 SEM (ZEISS) with a four-quadrant backscatter detector at 8 kV using ATLAS 5 Array Tomography (Fibics). Lateral medium-resolution images (100 nm) allowed the identification of regions of interest that were in turn reimaged at 4–10-nm lateral resolution. Image analysis was performed in Fiji27.

Mice perfusion and cell isolation for the 10X Genomics experiments

For the 10X Genomics experiments, mice were deeply anesthetized and perfused with cold PBS. Each brain was removed and individually microdissected under a dissection microscope; the cortex and hippocampus were isolated from the brain and the attached choroid plexus was removed. Cells were isolated with a previously established isolation protocol28 using gentleMACS with the Papain Neural Tissue Dissociation Kit (Miltenyi Biotec) and actinomycin D (catalog no. A1410, Sigma-Aldrich) at a final concentration of 45 mM. After dissociation, myelin debris was removed from the cell suspension using Myelin Removal Beads II (Miltenyi Biotec). Cells were resuspended in 0.04% BSA + PBS and counted using an automated cell counter (TC20, Bio-Rad Laboratories) before loading onto the Chromium Controller.

Library preparation for the 10X Genomics experiments

Single-cell suspensions were loaded on to the Chromium Single Cell Controller using the Chromium Single Cell 3′ Library & Gel Bead Kit v.3.1 (10X Genomics) chemistry according to the manufacturer’s instructions. Sample processing and library preparation were performed according to the manufacturer’s instructions using AMPure beads (Beckman Coulter). Libraries were sequenced on the DNBSEQ sequencing system (Dresden).

Preprocessing and analyses of the 10X Genomics data

The raw sequencing data of each library were processed using Cell Ranger (10X Genomics) v.3.0 to generate the gene-by-cell unique molecular identifier (UMI) count matrix. The single-cell gene expression matrix was imported into the R package Seurat (v.4.3.0). For further analysis, cells were excluded according to the total UMI count (<500 or ≥50,000); the total detected gene number (≥7,500); and the mitochondrial UMI percentage (≥10%). To prevent the influence of the technical characteristics of the downstream analyses, we performed SCTransform to normalize the clean UMI matrix. Single-cell clustering was done according to the vignette (https://github.com/satijalab/seurat/blob/HEAD/vignettes/sctransform_vignette.Rmd). Variable features with a residual variance of more than 1.4 were selected; principal component analysis (PCA) was performed to normalize the matrix. Cluster markers were identified using the FindAllMarkers function. Microglia subclusters were identified using the same parameters described above. Microglia subcluster compositional analysis was done using scCODA v.0.1.9 according to the online vignette (https://sccoda.readthedocs.io/en/latest/getting_started.html). To capture nuanced yet biologically significant alterations, the false discovery rate was set at 0.4, as described in the vignette.

Single-cell trajectory construction

The clean UMI matrix of all microglia was imported into the R package Monocle3 and processed using normalization and PCA analysis. The functions learn_graph and order_cells were used to construct the trajectory. Homeostatic microglia were set as the start point of the trajectory. The function choose_graph_segments was used to determine the path, with homeostatic microglia as the start point and DAM4 as the end point. The Moran’s I test was performed in the function graph_test to find differentially expressed genes (q < 0.05 and Moran’s index > 0.3) across the selected path. For the slingshot trajectory analysis, the clean UMI matrix of all microglia was imported into the R package SingleCellExperiment and processed using normalization and PCA analysis. A function slingshot was used to construct the trajectory without an indicated start or end point.

MERFISHTissue collection and preparation for MERFISH

Male 5xFAD and WT litter-mate mice aged 9.5 months were anesthetized using an isoflurane chamber and film container filled with isoflurane-soaked cotton fibers and perfused with RNase-free PBS (catalog no. D8537, Sigma-Aldrich). Next, brains were carefully extracted, cut in half and frozen in isopentane on dry ice. Brains were then stored at −80 °C until further processing.

Tissue sectioning

Brain sections, 10 μm in thickness, were cut coronally using a cryotome (CryoStar NX70, Thermo Fisher Scientific) and collected onto glass slides supplied by Vizgen. To ensure proper adhesion, the glass slides were coated with a solution of 0.1 mg ml−1 poly-d-lysine bromide (catalog no. P7886, Sigma-Aldrich), left to incubate at room temperature for 1 h and dried before starting the sectioning.

Gene panel

The gene panel used in this study included 496 protein-coding genes, complemented by 54 blank probes. Within this gene panel, a diverse array of genes was selected, including recognized markers associated with brain and immune cell types, such as glial cells, vascular cells, T cells and macrophages, as well as distinct subtypes of glutamatergic and GABAergic neurons. Moreover, an assortment of reactive markers for microglia, astrocytes and oligodendrocytes were included, drawing from the existing literature; genes pertaining to cholesterol metabolic pathways and AD-related pathways were incorporated into the panel. For a comprehensive listing of the entire gene panel, see Supplementary Table 2.

Hybridization

After sectioning, samples were mounted on slides and transferred to Petri dishes (60 × 15 mm), with the tissue side facing upwards. These dishes were stored at the rear of the cryostat, maintaining a temperature of −20 °C for a minimum duration of 5 min to ensure adherence. Next, 5 ml fixation buffer (4% PFA in 1× RNase-free PBS) was gently added to the Petri dishes within a fume hood, followed by a 15-min incubation period at room temperature. Afterwards, samples were washed with 5 ml RNase-free PBS, three times each for 5 min. Samples were then permeabilized in 5 ml 70% ethanol at 4 °C overnight, in parafilm-sealed dishes, and kept under the same conditions for long-term storage (up to 1 month). For hybridization with the library (‘gene panel’), samples were washed with 5 ml Vizgen Sample Prep Wash Buffer and then incubated in 5 ml Formamide Hybridization Buffer at 37 °C for 30 min in an incubator. The Formamide Hybridization Buffer was aspirated from the tissue and 50 μl of the gene panel mix was added on top of each tissue. A piece of parafilm with an approximate size of 1.5 × 1.5 cm was carefully added on top of the liquid to spread the library mix and protect it from evaporation. The dishes were sealed with parafilm and placed in a humidified incubator at 37 °C for 36–48 h. The parafilm was removed and the samples were incubated twice in 5 ml Formamide Hybridization Buffer at 47 °C for 30 min. The samples were then washed with 5 ml Sample Prep Wash Buffer for 2 min.

Gel embedding

A fresh 10% w/v ammonium persulfate solution was prepared. For each sample, 5 ml of Gel Embedding Premix was combined with 25 μl of the 10% ammonium persulfate solution and 2.5 μl of N,N,N′,N′-tetramethylethylenediamine. In parallel, one 20-mm gel-coated coverslip for each sample was cleaned with RNaseZap, followed by 70% ethanol and dried with Kimwipes. The gel-coated coverslips were then covered with 100 μl Gel Slick Solution (VWR) for 10 min and wiped dry with Kimwipes. The Sample Prep Wash Buffer was aspirated from the samples. For each sample, 100 μl of the Gel Embedding Mix was retained in a small tube, while the remainder was added to the samples and incubated for 1 min. The Gel Embedding Mix was then poured out from the samples into a waste tube but kept aside on the bench (to monitor gel formation). The slides were then aspirated dry, leaving just enough liquid to keep the tissue from drying out; 50 μl of the retained Gel Embedding Mix was added on top of the tissue and the Gel Slick-treated coverslip was placed on top of it using tweezers, with the Gel Slick-treated side facing down toward the tissue, avoiding air bubbles. Excess Gel Embedding Solution was aspirated from the sides of the coverslips. The samples were then incubated at room temperature for 1.5 h to allow the gels to form. Finally, the coverslips were removed using a hobby blade and tweezers.

Tissue clearing

For each sample, to eliminate lipids and proteins that would interfere with imaging, 5 ml of the Clearing Premix were mixed with 50 μl Proteinase K. After the coverslips were removed from the gel-embedded samples, the clearing solution was added to each sample and the dishes were sealed with parafilm. Samples were placed in a humidified incubator overnight at 37 °C. Samples were stored in the Clearing Premix solution in the 37 °C incubator before imaging for up to a week. Samples were imaged after tissue could no longer be observed in the gel.

Sample imaging

The Clearing Premix solution was aspirated from the sample and the sample was briefly washed three times with Sample Prep Wash Buffer, then again for 10 min on a rocker and then three more times briefly. The sample was incubated with 3 ml of the appropriate first hybridization buffer, including DAPI and polyT reagent, for 15 min at room temperature on a rocker, covered from light. The sample was then washed with 5 ml of the Formamide Hybridization Buffer for 10 min at room temperature on a rocker, covered from light and then transferred to 5 ml of the Sample Prep Wash Buffer. The imaging buffer was prepared by combining the iImaging buffer, the imaging buffer activator and the RNase inhibitor at a ratio of 500:2.5:1. The hybridization buffers appropriate to the gene panel, and the imaging buffers, were loaded onto the MERSCOPE system (Vizgen). The sample was placed in the flow chamber and connected to the fluidics system of the MERSCOPE system (Vizgen), taking care to get rid of air bubbles. A low-resolution mosaic was acquired using a ×10 objective; regions of interest (whole coronal sections) were selected for high-resolution imaging with a ×60 lens. For high-resolution imaging, the focus was locked to the fiducial fluorescent beads on the coverslip. Seven 1.5-μm-thick Z-planes were taken for each field of view when imaging the tissue, including for the DAPI channel.

MERFISH data analysis

Raw images were decoded to RNA spots with spatial coordinates and gene IDs using the Merlin software (Vizgen) on the MERSCOPE instrument. Cell segmentation was performed using CellPose algorithm, using the DAPI nuclear and polyT total RNA staining channels. The resulting single-cell gene expression matrices were further analyzed in R. We performed quality control for each section and excluded cells with fewer than 30 transcripts or fewer than ten detected genes. Data were then normalized using SCTransform v.2 normalization; PCA was calculated based on all 496 measured genes. Next, the dataset was harmonized for batch effect differences between sections using Harmony (with 75 principal components as the input). UMAP embedding was computed from the first 40 Harmony dimensions using the RunUMAP function. We calculated the shared nearest neighbor graph based on the first 40 Harmony dimensions (FindNeighbors function), which was then used to detect clusters using a Louvain algorithm at a range of resolutions. We annotated cells into major classes based on the coarse clustering resolution (0.1), and spatial location and regional identity. To identify T cells, we further subclustered the immune cell clusters by repeating the PCA, and using graph construction and clustering steps, to reveal a cluster marked by T cell marker genes (Ptprc, Lck, Cd3e, Cd8a, Nkg7), which we added to the cell type annotation. We used the AddModuleScore function to calculate the ISG expression score from the set of ten known ISGs included in our panel (Stat1, Isg15, Rsad2, Usp18, Ifit1, Irf7, Ifitm3, Gbp2, Bst2, Cxcl10). In the same manner, we calculated the MHC-II expression score based on Cd74, H2-Aa and H2-Ab1. To identify differentially expressed genes in the T cell neighborhoods, we calculated 50 nearest neighbors of every T cell based on two-dimensional spatial coordinates of cell centroids. We then identified differentially expressed genes between T cell neighbors and the remaining cells (nonneighbors) using the FindMarkers function with a Wilcoxon rank-sum test.

Behavioral tests

All behavioral experiments were done and analyzed by experimenters who were blinded to the treatments and the genotypes of mice.

Barnes maze

For the Barnes maze spatial memory test, spatial cues were placed around the maze and these were kept constant throughout the study. For the baricitinib experiment, all mice from four experimental groups were consecutively subjected to the Barnes maze after 2 weeks, 6 weeks and 12 weeks of baricitinib or control treatment. For the CD8+ T cell depletion experiment, all mice from four experimental groups were subjected to the Barnes maze before and after 6-week anti-CD8 antibody or control treatment. The maze consists of a circular table with circular holes around its circumference. The goal is for the animal to reach the box that is positioned beneath one of the holes with the aid of visual cues. In the experiments, the surface of the table was brightly lit, serving as an aversive stimulus that motivated the mouse to find (and hide in) the goal box. The goal box location was the same in all testing runs. In the first 2 days of rehabilitation training, all mice were trained and guided by the same researcher to locate the box. Then, in the next 4 or 7 days, all mice learned where the box was located; the time needed to find the right aperture within the test duration of 180 s was automatically evaluated using video monitoring (EthoVision XT, Noldus Information Technology). To evaluate spatial memory, a spatial probe test was administered 24 h after the last training session. During the probe test, each mouse was allowed to search for the goal box on the platform for 180 s. Each test was terminated as soon as the mouse found the goal box or failed to find it within 180 s.

Open field test

The open field test is used to evaluate the ability for locomotion and exploration, and the level of anxiety in rodents. All mice were consecutively subjected to the open field test at different ages. During the open field test, mice were placed in a brightly illuminated transparent Plexiglas arena (50 × 50 × 30 cm) inside an enclosed cupboard. After 1 min of habituation, open field exploration was evaluated using video monitoring (EthoVision XT) for 15 or 30 min.

Elevated plus maze

The elevated plus maze is used to measure any differences in anxiety-related behaviors that can alter performance in other tests. All mice were consecutively subjected to the elevated plus maze at different ages. Briefly, the elevated plus maze has two arms enclosed by walls and two open arms without walls. Entries in both the closed and open arms, and the percentage of time spent in the open arms, were evaluated using video monitoring (EthoVision XT). Each mouse was placed in the left closed arm of the maze; after 1 min of habituation, exploratory activity was recorded for 5 min.

Statistics

For the IHC analysis, 2–3 sections from each animal were analyzed. Data are shown as the mean ± s.e.m. Each dot represents one animal. The normal distribution of the samples was tested using the Shapiro–Wilk test. For the statistical analysis, a paired or unpaired Student’s t-test or Mann–Whitney U-test was used to compare two groups. A two-sided, one-way ANOVA followed by a Tukey’s post hoc test was used for multiple comparisons. Tests were chosen according to the distribution. In all tests, a P < 0.05 was considered significant, with *P < 0.05, **P < 0.01 and ***P < 0.001. The statistical analyses were done using Prism version 9.3.1 (GraphPad Software). Data acquisition and analysis were performed in a blinded manner. No animals were excluded from the analyses.

Reporting summary

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