P56 male C57BL/6N mice were obtained from Charles River Laboratories and housed in Allentown cages with corncob bedding, paper nestlets, and environmental enrichment with ad libitum access to rodent chow (Envigo, cat# 2918) and non-acidified water. Mice were maintained in a temperature- and humidity-controlled room on a 12:12 light-dark cycle. All animal procedures were conducted under approved animal use protocol 20013084 at the Division of Comparative Medicine (DCM), University of Toronto.

Mice received an intravenous tail vein injection of 3 μg anti-CD45.2-PE antibody (clone 104, BioLegend, cat# 109808) or an isotype control, PE rat IgG2b, κ antibody (Biolegend, cat# 400608) diluted in sterile PBS (Anderson et al., 2014). After 3 min to allow antibody circulation and labeling of intravascular leukocytes, mice were euthanized via CO2 asphyxiation and brain tissue was rapidly harvested and transferred into a petri dish containing cold calcium/magnesium-free DPBS (Wisent, MultiCell, cat# 311-425) prior to issue dissociation.

The olfactory bulb and cerebellum were removed, and brain tissue was washed in calcium/magnesium-free DPBS (Wisent, MultiCell, cat# 311-425), prior to dissociation with the Adult Brain Dissociation Kit (P) kit (Miltenyi Biotec, cat# 130-107-677). GentleMACS Octo Dissociator (Miltenyi Biotec, cat# 130-134-029) settings were based on those used by Liu et al. (2021). Following dissociation, sample cleanup was performed in accordance with the manufacturers protocols for: (i) Debris Removal Solution (Miltenyi Biotec, cat# 130-109-398), (ii) Myelin Removal Beads II (Miltenyi Biotec, cat#130-096-733), or (iii) Debris Removal followed by Myelin Removal. The resulting cell pellet was then processed for flow cytometry.

Cells were incubated in blocking buffer comprised of rat anti-mouse CD16/CD32 (1:100, Mouse BD Fc Block, BD Biosciences, clone 2.4G2, cat# 553141) and 0.5% BSA (BioShop, cat: ALB005.100) in calcium/magnesium free DPBS (Wisent, MultiCell, Cat: 311-425) for 20 min at 4 °C. For IV labeling experiments, cells were subsequently stained with the following antibodies diluted in blocking buffer for 15 min at 4 °C; anti-ACSA2-APC (1:200, Miltenyi Biotec, clone REA969, cat# 130-117-535) and anti-CD45-BV605 (1:200, BioLegend, clone 30-F11, cat# 103140). For microglial phenotyping, cells were subsequently stained for 15 min at 4 °C with the following antibodies diluted in blocking buffer: anti-CD45-BV605 (1:200, BioLegend, clone 30-F11, cat# 103140), anti-CD11b-VioBrightFITC (1:200, Miltenyi Biotec, Clone REA592, 130-113-805) and anti-ACSA2-APC (1:200, Miltenyi Biotec, clone REA969, cat# 130-117-535). Cells were washed twice in FACS buffer (DPBS + 0.5 % BSA) and centrifuged for 7 min at 4 °C and 300 × g. Prior to acquisition, cell viability was assessed using DAPI (Thermo Scientific, cat# 62248), and nucleated cells were identified with DRAQ5 (Thermo Scientific, cat# 62254). Flow cytometry was performed on a BD FACSDiscover S8 (Network Biology Collaborative Centre, Toronto, ON, Canada) and data were analyzed using FlowJo (v 10.8.1, BD Biosciences).

Mice were euthanized via transcardiac perfusion with 30 mL ice-cold DPBS followed by 30 mL of 4% (w/v) paraformaldehyde (PFA, BioShop, cat# PAR070) under isoflurane anesthesia. Brains were harvested and post-fixed in 4% PFA overnight, then transferred to 30% (w/v) sucrose solution for cryoprotection. Brains were sectioned in series at 20 μm thickness at −25 °C using a cryostat (Thermo Scientific, HM525 NX). Sections were collected onto SuperFrost Plus microscope slides (Fisherbrand, cat# 1255015) and stored at −20 °C until staining.

Sections were rehydrated with three washes of DPBS, then incubated in blocking buffer containing 5% BSA (BioShop, cat#ALB001) and 0.1% Tween-20 (Bioshop, cat# TWN510) for 1 h at room temperature in a humidified chamber. Slides were incubated overnight at 4 °C in primary antibody solution containing anti-TMEM119 (1:100; Abcam, cat# AB209064) in a humidified chamber. Following three DPBS washes, slides were incubated for 1 h at room temperature with donkey anti-rabbit Alexa Fluor 647 (1:400, Invitrogen, cat#A31573) and anti-ACSA2-PE (1:50; Miltenyi Biotec, cat# 130-116-244, clone REA969). Tissue was washed three times in DPBS, counterstained with DAPI (1:3000; Invitrogen, cat# D3571), and mounted with Mowiol (Sigma-Aldrich, cat# 81381) mounting medium.

Images were acquired using a 20 × objective on a Zeiss LSM 900 confocal microscope with Zen Blue software (version 3.3; Zeiss). Intensity thresholds were set based on background signal measured in secondary-only and a rat IgG2b-PE isotype control (1:50, Miltenyi Biotec, Cat# 130-123-273). Linear brightness adjustments and colocalization analysis of TMEM119 and ACSA2 were performed using Zen Blue.

Raw single cell RNA seq reads in fastq format were downloaded from SRA (BioProject: PRJNA952594, Runs: SRR27010906 and SRR27010907). Reads were aligned to the mouse mm10/GRCm38 genome, demultiplexed and counted with the 10x Genomics Cell Ranger v8.0.1 using default parameters and the Ensembl mouse transcriptome release 99. Demultiplexed counts matrices where then processed in R version 4.5.1 using Seurat v5.3.0 (Butler et al., 2018; Hao et al., 2021, 2024; Rosenberg et al., 2018; Satija et al., 2015; Stuart et al., 2019). Cells were merged as a single Seurat object. A filter specified a minimum of 200 features per cell and 10 cells per feature from each set of samples. Data was processed using Seurat’s recommended sctransform workflow: first sctransform was applied to normalize feature counts; followed by PCA and UMAP dimension reductions; followed by finding neighbors; then finding clusters and their respective markers. The log-normalized counts were scaled gene-wise from a range of 0 to 1 across cell-types, to generate heatmaps with the ComplexHeatmaps R package v2.24.1 (Gu et al., 2016) and visualize expression of astrocyte and microglia features. The following R packages were also used for visualization or processing: ggplot2 (Wickham, 2016), dplyr (Wickham et al., 2025), circlize (Gu et al., 2014) and Viridis (Garnier et al., 2024).

The MapMyCells: Cell Type Mapper v1.6.1 library from Allen Institute [RRID:SCR_024672] was used with the Hierarchical method in Python 3.13.8, with precomputed statistics to annotate cell types for each cell. 10x Whole Mouse Brain taxonomy (CCN20230722) rev 230821 was used as reference. Seurat object data was exported for the cell type mapping to maintain consistency between processing steps. The cell annotations were in turn used to count proportion of cell types (Supplementary Table 1) in a bar plot and apply categorization in a dim plot.

Module scores showing relative gene expression of astrocyte and microglial marker genes (Supplementary Table 2) were calculated using Seurat’s AddModuleScore, using default settings.

The source code is available on a Git repository¹, where the release tag published marks the revision that generated the figures in this paper. Exact versions of all used software packages can be automatically retrieved using the lock files in the code repository.

To assess the purity of ACSA2-sorted astrocyte preparations used in downstream transcriptomics analyses, we reanalyzed a previously published 10x Genomics Chromium scRNAseq dataset of ACSA2-MACS-FACS-purified astrocytes from the murine stroke-injured and naïve cortex (Figure 1A; Scott et al., 2024). Using the Allen Institute’s MapMyCells [RRID:SCR_024672], we classified astrocytes, microglia, border-associated macrophages (BAMs), oligodendrocyte progenitor cells (OPCs), endothelial cells, pericytes, dendritic cells (DCs), and vascular leptomeningeal cells (VLMCs). We next investigated the proportion of cell types across stroke and naïve datasets and observed that microglia represented 53% and 76% of cells in the dataset, respectively (Figure 1B and Supplementary Table 1). When we examined the relative gene expression of canonical cell type genes, Atp1b2, the gene encoding ACSA2, was highly enriched in astrocytes (Figures 1C, D), however, lower gene expression was still observed in microglia, suggesting that these cells are present within the ACSA2 sorted fractions.

To assess whether ACSA2 was expressed on the surface of microglia, we performed flow cytometric analysis. To distinguish microglia from circulating myeloid cells that can also express ACSA2, we performed an intravenous tail vein injection with anti-CD45.2-PE antibody 3 min prior to tissue harvest, followed by ex vivo staining with antibodies against CD45-BV605 and ACSA2-APC (Figure 2A). This approach distinguishes intravascular leukocytes (labeled with I.V. antibody; CD45.2-PE+) from tissue-resident cells (unlabeled; CD45.2-PE-), while the ex vivo CD45-BV605 staining labels all CD45⁺ cells regardless of location. Using this strategy, we identified two populations of ACSA2⁺ cells: circulating ACSA2⁺ immune cells (CD45.2-PE⁺/CD45-BV605⁺/ACSA2-APC⁺) and CNS-resident ACSA2⁺ immune cells (CD45.2-PE^–/CD45-BV605⁺/ACSA2-APC⁺) (Figure 2A). Quantification revealed that CNS-resident ACSA2⁺ immune cells comprised 26.3 ± 3% (n = 3) of all live, nucleated, ACSA2⁺ cells in the brain. Circulating ACSA2⁺ immune cells constituted 2.03 ± 0.3% (n = 3) of the total ACSA2⁺ population (Figures 2B, C), which is consistent with the observation of DCs in our scRNAseq data (Figure 1B). These findings demonstrate that CNS-resident ACSA2⁺ cells are the predominant source of contamination.

To validate the microglial identity of ACSA2⁺ CNS-resident cells, we examined the expression of CD45, CD11b, and ACSA2 by flow cytometry. Debris removal is a critical step in CNS tissue processing, and previous studies employing ACSA2-based isolation have utilized either debris removal solution or myelin removal beads (Batiuk et al., 2017; Scott et al., 2024). We therefore assessed whether these strategies contribute to different rates of microglial contamination of ACSA2⁺ fractions. Debris removal solution resulted in 15.2 ± 3.2% (n = 3) of ACSA2⁺ cells that were CD45⁺/CD11b⁺ microglia (Figures 3A,B). Myelin removal beads reduced this contamination to 3.9 ± 0.5% (n = 3). Combining these two strategies provided no additional reduction in microglia (5.5 ± 0.9%, n = 3) (Figures 3A,B). The majority of CD45⁺/CD11b⁺ cells were found in the ACSA2- fraction across all three methods (∼78%–82%; Supplementary Figure 1). To note, while debris removal solution and myelin removal beads alone yielded similar proportions of ACSA2⁺ cells (34.8 ± 2.1% and 35.8 ± 0.6%, respectively), the combined approach reduced ACSA2⁺ yield to 24.4 ± 1.6% without a corresponding reduction in microglia contamination. To confirm the microglial identity of these contaminating cells in situ, we performed immunohistochemistry on adult brain tissue and observed ACSA2⁺/TMEM119⁺ cells (Figure 3C). Altogether, these data suggest that a subset of microglia express ACSA2.