Infiltrating monocyte-derived macrophages does not survive long term in stroke brain despite their dominance in the acute ischemic core and myeloid derived IGF-1 play dichotomous roles in stroke recovery Authors: Hongli Ma, Nicolas Hugues, Kirsten Voice, Joshua Peter, Aleksandr Taranov, Christina Thapa, Krishna Roskin, Yu Luo Published: 2026 ABSTRACT BackgroundIschemic stroke elicits a strong neuroinflammatory response characterized by activation of microglia and infiltration of monocytes-derived macrophages (iMM), which have been speculated to have differential functions in stroke. However, because the gene expression profiles of microglia and iMM overlap in the injured brain, distinguishing these two cell populations has posed a challenge for the field.MethodsUsing a recently characterized microglia-specific TMEM119CreER-Ai14 tdTomato reporter mouse model and single cell RNA sequencing (scRNA-seq) analysis, we prelabeled microglia in the brain prior to stroke, enabling detailed characterization of the transcriptomic and spatial distributions of microglia vs. iMM across different stages of post-stroke. ScRNA-seq findings were validated through immunohistochemistry, RNAscope, and animal models.ResultsHere, we report that microglia and iMM are enriched at distinct locations and exhibit differential temporal dynamics in the stroke brain. Our genetic tracing data reveals that iMM does not survive in the chronic stroke brain past 30 days in our model. Additionally, scRNA-seq further revealed distinct transcriptomic states between microglia and iMM at D7 and D14 in the stroke brain. Genetic tracing also allowed us to identify novel markers for activated microglia vs. iMM in the acute stroke brain such as Gm21188. Finally, our data also showed that Igf1 is up-regulated in both microglia and iMM after stroke. Cx3cr1ERT2 Igf1fl/fl mediated myeloid-specific Igf1 gene deletion at 5–9 Days after stroke lead to decreased CD8+ T-cell infiltration in the stroke brain. Functionally, myeloid specific Igf1 iKO mice show expedited sensorimotor function recovery but worsened anxiety-like behavior.ConclusionsIn summary, Tmem119CreER-Ai14 tdTomato reporter mouse line is a useful tool for the field to clearly delineate the cellular dynamics and transcriptomic profiles of microglia vs. iMM in a variety of central nervous system pathological conditions, which could have broad implications beyond stroke studies. Moreover, we have characterized novel subtypes and markers for activated microglia and iMM, as well as a possible novel role for myeloid cell-derived IGF-1 in driving T-cell infiltration and/or survival in the post-stroke brain. This study provides valuable insights for future investigations aimed at modulating microglia vs. iMM to promote stroke resolution and functional recovery in vivo.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13073-026-01687-x. Background Despite the recent advancements in extending the stroke treatment window, the number of patients eligible to receive acute recombinant tissue plasminogen activator (rtPA) or endovascular surgical treatment is significantly limited (9.3% and 3.5%) [1]. Additionally, secondary neuronal damage following the initial neuronal death is starting to be recognized in the field [2–4]. Therefore, there is an urgent need for novel therapeutic strategies to halt or slow the ongoing secondary neuronal damage beyond the initial hours/days of ischemic onset. Recently, there has also been increased evidence supporting the role of inflammation and glial cell activation in the evolution of the pathophysiology of stroke. Significant data suggest that extensive and consistent neuroinflammation is an important component in the pathogenesis of stroke, specifically in relation to both the acute neuronal death and the subsequent sub-chronic and chronic pathology resolution and recovery [5, 6]. Specifically, neuroinflammation has been implicated as a critical factor that may contribute to such secondary injury [7, 8]. Additionally, immune cells are speculated to play important roles in the transition to a neurorepair process after stroke [9]. Resident microglia and infiltrating monocytes/macrophages (iMM) are two critical components of the neuroinflammatory response after stroke. Microglia are the innate component of the brain’s immune system, and their role in maintaining homeostasis of neuronal function under physiological conditions and mixed dual effects under pathological conditions has become an area of great interest recently [10]. Ischemic stroke elicits a strong neuroinflammatory response, characterized by the activation of immune cells in the brain from the very acute stages. The neuroinflammatory-induced activation of adult microglia, inducing a switch from homeostatic to dyshomeostatic profiles, enables microglial cells to respond to acute tissue damage, clear debris, and regulate inflammation [11, 12]. Such spontaneous processes are tightly regulated by microglial production of cytokines and trophic factors under pathophysiological conditions [13]. Thus, the selective elimination of microglia leads to an increased neuronal death and to a 60% increase in the infarct size, while microglia repopulation can reverse this adverse phenotype [14]. Although microglia participate in many critical aspects of stroke pathology, their exact role at different stages of ischemic injury remains unclear and debated [15, 16]. Similarly, infiltrating peripheral monocytes, particularly those of splenic origin, have also been implicated in determining the lesion size during the acute primary and chronic secondary neuronal damage [17]. Specifically, it has been proposed that infiltrating monocytes might have a more robust phagocytotic function [18] or direct the angiogenesis process via the secretion of factors such as IL-6 [19]. Despite increasing recognition of the roles of microglia and iMM in central nervous system (CNS) injury and neurodegeneration, a major unresolved challenge in the field is the inability to reliably distinguish these two populations in the injured adult brain. Under physiological conditions, microglial homeostatic signature genes have been utilized to differentiate microglia vs. iMM. Additionally, CD45 expression levels have also been used to distinguish microglia (CD45 low/medium) vs. iMM (CD45 high). This limitation arises because activated microglia downregulate homeostatic markers (e.g., P2ry12, Tmem119, Cx3cr1) while upregulating macrophage-associated genes (e.g., Cd45, Arg1), leading to substantial overlap in transcriptomic profiles with iMM. Therefore, distinguishing these two cell populations has presented a challenge for the field and has led to conflicting results in various neurological conditions [20–25]. A previous study suggested that iMM can survive into the chronic stage (at least over 60 days) in neonatal stroke and adopt a microglial-like profile [26]. However, whether it is true in the adult stroke brain is unknown. Additionally, given the rapid turnover of monocytes relative to microglia, iMM lineage tracing can label and track only a temporal cohort of monocytes at the time of tamoxifen administration. A detailed understanding of the immune cell landscape and the differences in transcriptomic profiles and functional roles between microglia and iMM in post-stroke brains at various stages is still lacking. Based on our recent comprehensive evaluations of the major lines of current myeloid and microglia-specific CreER driver mouse models [27, 28], we address this gap by using TMEM119CreER/CreER-Ai14 mouse line [29, 30], which is capable of pre-labeling almost 100% of microglia in the CNS at the basal level, with minimal blood and splenic labeling of immune cells [27, 31]. Pre-labeling of microglia enabled us to accurately and independently track microglia and iMM dynamics following stroke in the present study. By integrating this approach with single cell RNA sequencing (scRNA-seq), we systematically delineate the temporal dynamics, transcriptomic profiles, and functional roles of microglia vs. iMM across acute and subacute stages after stroke. Using this mouse model, we observed that, in contrast to neonatal models, iMM do not persist long-term in the adult stroke brain but instead exhibit a transient presence, peaking at day 7 and largely disappearing by one month. ScRNA-seq further revealed distinct transcriptomic states and functional pathways of microglia and iMM at D7 and D14. Furthermore, although iMM transiently upregulate certain microglial signature genes such as P2ry12, they fail to acquire canonical microglial lineage identity such as Sall1. In addition, microglia and iMM exhibited distinct cell–cell interactions and signaling crosstalk with other infiltrating immune cells. We also identify Gm21188, which could serve as a specific gene marker of activated iMM after stroke. Finally, functional interrogation of IGF-1 produced by microglia and iMM in the recovery stage of stroke reveals enhanced sensorimotor function recovery but increased anxiety levels by deleting Igf1 in microglia and iMM at 5 days post-stroke (Cx3cr1ERT2 Igf1 KO). Together, in contrast to a neonatal stroke study showing that iMM persist for over 60 days and progressively acquire a ramified morphology with microglial marker expression (Sall1, Tmem119, P2ry12) [26], our data reveal that in the adult brain these iMM fail to survive long-term after stroke and only transiently upregulate select microglial signature genes (such as P2ry12) without expressing the canonical microglial lineage gene Sall1. This study offered useful tools for tracking and distinguishing microglia from iMM and provided insights to address critical gaps in our knowledge of stroke pathology, with broad implications and potentially applications in other CNS diseases as well. Methods Animals All animal experiments were performed with approval from the Institutional Animal Care and Use Committee (IACUC) office in accordance with institutional and national regulations. All animals were maintained in the University of Cincinnati’s animal facility under a 14-hour light and 10-hour dark cycle and were provided with standard food and water ad libitum. The Cx3cr1CreER(Jung) (JAX: 020940) and Tmem119CreER (JAX: 031820) transgenic mouse lines, in which Cre recombinase expression is driven by the Cx3cr1 promoter (myeloid cells) or the Tmem119 promoter (microglia) were obtained from the Jackson Laboratory. Ai14 mice (JAX:007914), which display robust tdTomato expression following Cre-mediated recombination, were crossed with Tmem119CreER mice to specifically label parenchymal microglia. The Cre-loxP recombination system also was employed to generate a cell type–specific deletion of the Igf1 gene. The Cx3cr1CreER(Jung) were crossed with floxed Igf1 mouse line (JAX: 016831). All the Cre driver lines used in this study carry the CreERT2, enabling tamoxifen-inducible control of Cre recombinase activity. For euthanasia, mice were administered avertin followed by transcardial perfusion. Study cohorts and experimental design Tmem119CreER-Ai14 mice were used for histological analyses, including immunohistochemistry (IHC) and RNAscope. Animals were assigned to predefined post-injury cohorts (sham, D3, D7, D14, and D30) for tissue collection and imaging-based quantification. Sample sizes for each cohort were as follows: sham (n = 3), D3 (n = 4), D7 (n = 5), D14 (n = 3), and D30 (n = 5). Animals were randomly allocated to experimental groups and analyzed in a blinded manner where applicable. A separate cohort of mice was used for tissue dissociation and single-cell RNA sequencing (scRNA-seq) analysis. Sample sizes for each cohort were as follows: sham (n = 3), D7 (n = 3), D14 (n = 3). Cx3cr1CreER Igf1fl/fl conditional knockout mice and littermate wild-type controls were used to generate a 2 × 2 experimental design consisting of genotype (WT vs. Igf1 conditional knockout) and injury condition (sham vs. stroke). Animals were assigned to four groups: WT, WT+stroke, Igf1 conditional knockout sham (Igf1 iKO), and Igf1 conditional knockout stroke (Igf1 iKO+stroke). Sample sizes for each cohort were as follows: WT (n = 7), WT+stroke (n = 5), Igf1 iKO (n = 13), and Igf1 iKO+stroke (n = 13). The brain samples were collected following transcardial perfusion. All computational analyses were performed on scRNA-seq datasets generated in this study (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1374719). Standard preprocessing, normalization, clustering, and differential expression analysis were conducted using Seurat. Cell-type annotation was performed using SingleR with the Brain Immune Atlas as the reference dataset [32, 33]. Ligand–receptor interaction analysis was performed using ligand–receptor interactions analysis (LIANA) to infer intercellular communication between identified cell populations [34]. Tamoxifen (TAM) administration Both male and female Tmem119CreER-Ai14 mice were treated with TAM solution (Sigma, T5648, 180 mg/kg) by oral gavage for five consecutive days. 30 mg tamoxifen powder was dissolved in 900 µL sunflower seed oil and 100 µL 100% ethanol. The mice were used for middle cerebral artery occlusion (MCAo) surgery or sham 3 weeks after treatment. Cx3cr1CreERIgf1fl/fl mice were treated with TAM solution on day 5 after MCAo surgery or sham by oral gavage for five consecutive days. Middle cerebral artery occlusion (MCAo) Transient cerebral ischemia was performed using the intraluminal filament model of MCAo according to our previously established procedure [35]. In brief, the MCA was blocked for 45 min using a 22 monofilament under isoflurane anesthesia (induction 4–5%; maintenance 1.5–2%) and then reperfusion. Reperfusion was verified by monitoring cerebral blood flow (CBF) in the middle cerebral artery territory—specifically over the right squamosal bone—using transcranial laser Doppler flowmetry (PeriFlux System 5000, Perimed, Järfälla-Stockholm, Sweden). The success of ischemic stroke induction was confirmed by magnetic resonance imaging (MRI) three days after MCAo. During both the surgical procedure and the recovery period, mice were placed on a heating pad to maintain body temperature. After surgery, animals were kept in an incubator maintained at 30–33 °C to promote recovery. Immunofluorescence and microscopy For immunofluorescence staining, mice were transcardially perfused with ice-cold PB followed by 4% paraformaldehyde (PFA). After dehydration in 20% and 30% sucrose, brains were sectioned at 30 μm. Primary antibody incubation was performed overnight at 4 °C in blocking buffer (4% BSA containing 0.3% Triton X-100) using the following antibodies: IBA1 (Abcam, 1:500; Wako, 1:1000), ApoE (Abcam, 1:250), Ki67 (Invitrogen, 1:300), CD93 (R&D Systems, 1:400), P2RY12 (BioLegend, 1:500), CD4 (Invitrogen, 1:400), CD8 (Abcam, 1:400). After washing, sections were incubated with appropriate Alexa Fluor–conjugated secondary antibodies (Invitrogen; 488, 555, 594, 647, or 790 nm; 1:500) in the dark. Fluorescence images were captured using a Leica TCS SP8 confocal microscope or Zeiss Axio Imager. Quantitative analyses were performed using ImageJ software. Immunofluorescence images were acquired at 40× magnification on a confocal microscope to ensure sufficient spatial resolution for accurate cell identification, particularly in densely packed regions following stroke. Individual cells were defined based on DAPI staining, which enabled clear discrimination of closely adjacent cells even in regions of high cellular density. For marker quantification, positive cells were identified by co-localization of the marker signal with DAPI-defined nuclei, ensuring that signals arising from overlapping or non-cellular regions were excluded. To ensure robustness and minimize observer bias, cell counting was performed independently by two investigators in a blinded manner. Cell isolation and sorting Mice were transcardially perfused with cold 1× HBSS within 3 min. The whole brain was removed and immersed in cold 1× HBSS with transcriptional and translational inhibitors, actinomycin, anisomycin, and typtolide, which prevent activation of microglia during the process. Then, ischemic hemispheres were separated by micro-dissecting scissors under a fluorescent microscope and mechanically dissociated with a scalpel. We use the papain dissociation kit (9001-73-4, Worthington Biochemical Corporation, Lakewood, NJ, USA) for subsequent dissociation following the manufacturer’s instructions. Following tissue dissociation, cells were filtered through a 70 μm filter and spun at 300 g for 5 min. Pallet was suspended in a 37% Percoll solution and centrifuged at 800 × g for 20 min to remove myelin and cellular debris. The resulting cell pellet was collected and resuspended in fluorescence-activated cell sorting (FACS) buffer (PBS supplemented with 1% (v/v) fetal bovine serum) for FACS staining and cell sorting. For cell sorting, single-cell suspensions were incubated with an anti-CD16/CD32 antibody to block Fc receptors at 4 °C for 15 min. Cells were then stained with CD45 (clone 30-F11) at 4 °C in the dark for 10 min. After washing with FACS buffer, DAPI was added to stain cell nuclei, and the cells were resuspended in FACS buffer for flow cytometric analysis. CD45⁺ cells were sorted using an Aria II cytometer (BD Biosciences, San Jose, CA, USA) and collected into 0.5 ml of 0.01% BSA-PBS for scRNA-seq. Single-cell RNA sequencing (scRNA‐seq) analysis Sample preparation and sequencing Single-cell transcriptomes of sorted CD45+ cells in brains were performed by Single Cell Genomics Facility in Cincinnati Children’s Hospital Medical Center following established protocols. The final concentration of sorted single cells was about 1,500 cells/uL. Single-cell suspensions were processed using the 10x Genomics Chromium platform (10x Genomics) to generate Gel Beads-in-Emulsion (GEMs), enabling the encapsulation of individual cells with barcoded gel beads and reverse transcription reagents. GEM formation and reverse transcription were performed according to the manufacturer’s instructions using GEM-X Chromium Single Cell 5’ Library & Gel Bead Kit v3. Sequencing libraries were generated through enzymatic fragmentation, end repair, A-tailing, adapter ligation, and sample index PCR, following the standard 10x Genomics protocol. The final Illumina-ready libraries were sequenced on a NovaSeq 6000 system (Illumina, San Diego, CA, USA), targeting approximately 50,000 reads per cell. Primary processing and reference genome construction Raw sequencing data were processed using the 10x Genomics Cell Ranger pipeline (v9.0.0) for demultiplexing, alignment, and generation of gene-barcode matrices [36]. To allow detection of the tdTomato transgene, a custom mouse reference genome (based on mm10-2020-A) was built by appending the tdTomato sequence to the Ensembl GTF and FASTA files using the cellranger mkref command. Reads were then mapped to this modified reference, and unique molecular identifiers (UMIs) were quantified to obtain expression matrices for downstream analysis. Quality control and preprocessing Download analyses were conducted in R using Seurat (v5.0.0). Ambient RNA contamination was corrected using SoupX (v1.6.2) for each sample [37]. To identify and exclude potential multiplets, DoubletFinder (v2.0.6) was applied independently to each dataset [38]. Cells predicted as doublets were removed prior to downstream analysis (doublet cells of sham:1169; doublet cells of D7:1665; doublet cells of D14: 1519). In addition, we further inspected clusters based on the distribution of nUMI and nFeature values, as well as the co-expression of mutually exclusive lineage markers. Clusters showing mixed transcriptional signatures inconsistent with known immune cell identities were excluded from downstream analyses. After quality control (% mitochondrial expression < 15, 300 < nFeature < 9000, 600 < nUMI(count) for all samples), only high-confidence singlets were retained. Normalization, integration, and clustering The filtered objects were normalized and variance-stabilized with the SCTransform function. To mitigate inter-sample variability, datasets were integrated using the Harmony algorithm (v1.2.3) [39]. The integrated data were then subjected to linear dimensional reduction (PCA) and non-linear embedding using Uniform Manifold Approximation and Projection (UMAP). Clustering was performed using the FindNeighbors and FindClusters functions with optimized resolutions (0.4 for whole samples; 0.6 for tdTomato+ re-cluster; 0.4 for iMM re-cluster) based on the elbow plot. Cell type annotation and differential expression Cell type annotation was achieved using the SingleR package (v2.4.1) with reference to the Brain Immune Atlas to assign high-confidence and un-bias immune identities to each cluster [32, 33]. For differential gene expression analysis, we used Seurat FindMarkers with the Wilcoxon rank-sum test. Genes were considered significantly differentially expressed if they met the following criteria: adjusted p-value (Benjamini–Hochberg correction) < 0.05 and absolute log2 fold change > 0.25. Only genes expressed in at least 10% of cells in either group were retained to reduce sparsity-driven artifacts. To evaluate reproducibility, we assessed consistency of major cell populations and transcriptional signatures across biological replicates within each experimental group, each consisting of 2–3 individual mice. We observed that major cluster identities and canonical marker gene expression were highly consistent across individual animals within the same group. In addition, key differentially expressed genes showed consistent directionality across mice, supporting robustness of the observed transcriptional changes at the biological replicate level. Ligand-receptor interaction analysis (LIANA) Intercellular signaling was profiled using LIANA (v0.2.0) and the Mouse Consensus ligand–receptor resource integrated within LIANA, capturing ligand–receptor interactions within annotated immune compartments and, separately, interactions linking tdTomato⁺ or tdTomato⁻ cells to the annotated immune cell populations [34]. LIANA aggregated predictions from multiple computational methods (including CellPhoneDB, NATMI, SCA and LogFc) and ranked interactions based on the consensus mean rank score. The Mouse Consensus ligand–receptor database was used as the reference resource. Data visualization Visualization of results, including UMAP plots, violin plots, and heatmaps, was performed using ggplot2 (v3.4.3) and ComplexHeatmap (v2.10.0). RNAscope in situ hybridization Mice were transcardially perfused with ice-cold PB within 3 min followed by 4% PFA. After dehydration in 20% and 30% sucrose, brains were embedded in O.C.T. and sectioned at 16 μm thickness using a Leica cryostat. RNAscope fluorescent in situ hybridization was performed using the ACD RNAscope Multiplex Fluorescent v2 Assay Kit (ACD Biosciences, Newark, CA, USA) according to the manufacturer’s guidelines. Probe hybridization was carried out using Mm-Tspan7 (1846781, ACD Biosciences, Newark, CA, USA) and Mm-Gm21188 (461231, ACD Biosciences, Newark, CA, USA) target probes. Immunohistochemistry was performed after the RNAscope hybridization and nuclei were counterstained with DAPI. Confocal images were acquired with a Leica SP8 confocal microscope at the University of Cincinnati Imaging Core. Behavioral assays Locomotion Spontaneous locomotor activity was monitored in a 42 × 42 × 31 cm transparent automated chamber for either 1–23 h with free access to food and water. The side-mounted laser sensor was used to monitor the behavioral activity of the mouse inside the chamber (Omnitech Electronics INC, Columbus, OH). Lighting conditions were set to match those of the animals’ regular housing environment, following a 14/10 h light–dark cycle. Behavioral data were automatically recorded and analyzed using Fusion software (Accuscan, Columbus, OH, USA). Rotarod Motor coordination and balance of the mice were measured by rotarod test using the Roto-rod Series 8 (IITC Life Science Inc., Woodland Hills, CA). The rotarod test latency was recorded as the time until the animal fell from the rotating rod. The rod rotation for mouse was 25 rpm. Each animal performed three trials, with a minimum inter-trial interval of 15 min in their home cage before the next trial. Tape removal The adhesive tape removal test was used to assess sensorimotor function. A small piece of adhesive tape (about 2–3 mm of diameter) was placed on the plantar surface of each forepaw-green on the left paw and orange on the right paw—to allow visual distinction. Mice were individually placed on a transparent platform, and the latency to remove each piece of tape was recorded. Prior to testing, animals underwent three pre-trial sessions to minimize stress and ensure task familiarity. Elevated Plus Maze (EPM) Anxiety-related behavior was evaluated in the Elevated Plus Maze (EPM), consisting of two open and two closed arms (30 × 5 cm each) elevated 50 cm above the floor. Before testing, each mouse was habituated to the testing room for 30 min. Each mouse was placed at the maze center facing an open arm and allowed to explore for 5 min. The apparatus was cleaned with 70% ethanol between trials. Time spent and entries in open and closed arms were recorded using AnyMaze software (Stoelting Co., Wood Dale, IL). Conditional fear contest Conditioned fear responses were performed in a sound-attenuated chamber (FreezeScan, CleverSys, Inc., Reston, VA, USA). Mice were habituated for 2 min, then exposed to a mild foot shock (0.5 mA, 2 s). Contextual memory was evaluated 24 h later by re-exposing animals to the same chamber for 5 min without shock. Freezing responses were quantified using FreezeScan software. Statistical analysis All statistical analyses were performed using GraphPad Prism 11. Statistical significance between two groups was assessed using unpaired Student’s t-test, while one-way or two-way ANOVA followed by Tukey’s post hoc test was used for multiple-group comparisons. Data are presented as mean ± SEM. A p value < 0.05 was considered statistically significant. Results Novel reporter mouse model enabling fate mapping and tracking of CNS microglia versus infiltrating monocytes/macrophages (iMM) in the MCAo animal model To visualize infiltrating monocytes/macrophages (iMM) after stroke, we used the Tmem119CreER-Ai14 tdTomato reporter mouse line, which selectively labels microglia but does not recombine in circulating monocytes [28, 30]. Specifically, we and others have previously characterized that TMEM119CreER/CreER [29, 30]-Ai14 mouse line is capable of pre-labeling almost 100% of microglia in the CNS at the basal level, with minimal blood and splenic labeling of immune cells [27, 31]. We hypothesized that if CNS microglia are pre-labeled with tdTomato reporter before injury, this feature would enable a clear distinction between resident microglia and invading monocytes under conditions of injury or disease. To confirm successful tdTomato pre-labeling of parenchymal microglia, TMEM119CreER/CreER-Ai14 mice were pretreated with tamoxifen (180 mg/kg, oral gavage once daily for 5 consecutive days) at three weeks prior to middle cerebral artery occlusion (MCAo) surgery (Fig. 1A). We then carried out immunohistochemistry (IHC) staining to identify microglia (tdTomato+, IBA1+) and iMM (tdTomato−, IBA1+) in sham mice or stroke mice at different time points. First, our data confirms that parenchymal microglia in the Sham group were successfully labeled, with > 99.93% of microglia co-expressing tdTomato and IBA1 (Fig. 1A, B). It has been previously reported that iMM are detected in post stroke day 3 or 7 in rodent brains [40, 41], therefore, we first analyzed on day 3 (D3) and 7 (D7) after stroke in TMEM119CreER/CreER-Ai14 mice. IHC analysis reveals numerous tdTomato⁻/IBA1⁺ cells within the infarct core (Fig. 1B) at both time points, validating successful tracing of resident microglia and iMM (especially when the contralateral hemisphere can serve as a control for the recombination efficiency of tdTomato reporter). We also note that tdTomato labels certain peri-blood vessel cells, as reported previously [27, 31]. However, these perivascular tdTomato+ cells are IBA1 negative, therefore, they do not interfere with our identification of resident microglia vs. iMM. To further assess the dynamic changes and spatial distribution of microglia and iMM in the stroke brain, we quantified their numbers at day 3, 7, 14, and 30 post-stroke within the infarct core, ipsilateral cortex (CTX), the lateral and medial border region surrounding the stroke core region (Fig. 1C, D and Additional file 1: Fig. S1a, b). Our data demonstrates that iMM is distributed within the infarct core, peri-infarct border, and the medial striatal peri-infarct region, but is absent from the cortex region in the ipsilateral hemisphere (Fig. 1D). In the infarct core, the average number of iMM significantly increased beginning on D3 (270.5 vs. 2.3 cells/mm² in controls, p < 0.0001), peaked at D7 (635.4 vs. 0.91 cells/mm² in controls, p < 0.0001), and then declined rapidly by D14 (42.42 cells/mm², n.s. vs. control), with nearly complete disappearance by D30 (12.73 cells/mm², n.s. vs. control) (Fig. 1C). In the lateral border region, iMM sharply increases and reaches a peak at D3 (273.5 vs. 4.55 cells/mm² in controls, p < 0.0001), followed by a gradual decline at D7 (139.4 vs. 3.64 cells/mm² in controls, p < 0.0001) and D14 (15.15 cells/mm², n.s. vs. control), and is also nearly undetectable at D30 (12.12 cells/mm², n.s. vs. control) (Additional file 1: Fig. S1a). In the medial border region, iMM began to increase on D3 (383.3 vs. 0 cells/mm² in controls, p < 0.0001), peaked at D7 (481.2 vs. 0 cells/mm² in controls, p < 0.0001), and then declined rapidly and nearly complete disappearance by D14 (3.03 cells/mm², n.s. vs. control), and D30 (9.09 cells/mm², n.s. vs. control) (Additional file 1: Fig. S1b). In contrast, microglia number in the infarct core exhibits a substantial expansion at later stages (such as D14 and D30 post-stroke) (Fig. 1C). These differential dynamics for microglia and iMM in the core and the peri-core areas suggest that iMM might have a more prominent role in the early-stage post-stroke (3–7 days), especially in the core and peri-core areas, while microglia numbers amplify and have a more predominant role in the subacute (D14) and chronic (D30) stage after stroke when iMM decreases and eventually are nearly non-detectable at D30 post stroke. In the CTX, lateral border, and medial border region, the average number of microglia increase and peak at D7 (CTX: 1404 vs. 278.2 cells/mm² in controls, p < 0.0001; lateral border: 4387 vs. 385.5 cells/mm² in controls, p < 0.0001; medial border: 3955 vs. 401.8 cells/mm² in controls, p < 0.0001), and then decline by D14 (CTX: 534.3 cells/mm², n.s. vs. control; lateral border: 2972 vs. 315.2 cells/mm² in controls, p < 0.0001; medial border: 1190 vs. 306.1 cells/mm² in controls, p = 0.0152) and D30 (CTX: 382.4 cells/mm², n.s. vs. control; lateral border: 1944 vs. 360 cells/mm² in controls, p = 0.0001; medial border: 1472 vs. 325.5 cells/mm² in controls, p = 0.0002) (Fig. 1C, D and Additional file 1: Fig. S1a, b). Notably, although microglia numbers declined by D30 across regions, their abundance remained higher than that observed in the Sham group (Additional file 2: Table S1). This profile indicates that iMM and microglia display distinct dynamics and spatial distributions within the stroke brain, and further demonstrates that iMM do not survive long-term in the post-stroke environment—contradicting prior reports in neonatal stroke or adult stroke study without fate mapping [26, 42]. Fig. 1The Tmem119CreER-Ai14 mouse enables precise fate mapping and tracking of CNS-resident microglia and infiltrating monocytes/macrophages (iMM) in the MCAo model. A Schematic of the mouse model used to label parenchymal microglia and the experimental timeline. B Representative whole-brain immunohistochemistry (IHC) images stained for IBA1 in Sham and 7-day post-stroke groups. Scale bar = 1–20 μm. C–D Representative IHC images of IBA1 staining and tdTomato fluorescent marker within the infarct core or ipsilateral cortex (CTX) at days 3, 7, 14, and 30 following stroke. n = 3–5 mice per group. Scale bar = 20 μm. The yellow arrow marks the tdTomato+/IBA1+ microglia; the white arrow marks the tdTomato−/IBA1+ iMM. Data shown as mean ± SE. Each data point represents the average of an individual animal. Two-way ANOVA with Tukey post hoc pairwise analysis The Tmem119CreER-Ai14 mouse enables precise fate mapping and tracking of CNS-resident microglia and infiltrating monocytes/macrophages (iMM) in the MCAo model. A Schematic of the mouse model used to label parenchymal microglia and the experimental timeline. B Representative whole-brain immunohistochemistry (IHC) images stained for IBA1 in Sham and 7-day post-stroke groups. Scale bar = 1–20 μm. C–D Representative IHC images of IBA1 staining and tdTomato fluorescent marker within the infarct core or ipsilateral cortex (CTX) at days 3, 7, 14, and 30 following stroke. n = 3–5 mice per group. Scale bar = 20 μm. The yellow arrow marks the tdTomato+/IBA1+ microglia; the white arrow marks the tdTomato−/IBA1+ iMM. Data shown as mean ± SE. Each data point represents the average of an individual animal. Two-way ANOVA with Tukey post hoc pairwise analysis Transcriptomic profiles of microglia and iMM after ischemic stroke Previous studies showed that immune cells displayed substantial molecular heterogeneity across various time points [43, 44]. To explore differences in transcriptomic profiles and functional roles between microglia and iMM in post-stroke brains at various stages, we performed single cell RNA sequencing (scRNA-seq) analysis in Sham, D7, and D14 post-stroke brains using CD45+ fluorescence-activated cell sorting (FACS) enrichment (Fig. 2A). We chose the D7 and D14 time points because they have been proposed as critical transition points for transcriptomic profiles in immune cells to better support neurorepair and regeneration in stroke, yet recent scRNA-seq data have mostly focused on the hyperacute and acute stages of stroke pathology [43, 44]. Additionally, our genetic tracing IHC analysis shows that iMM and microglia exhibit distinct dynamic and spatial distributions at these time points, suggesting potential functional transitions in the two cell populations (Fig. 2B). Following quality control (QC) filtering, removal of ambient RNA, and elimination of doublets (Additional file 1: Fig. S2a, b), we resolved 16 major clusters in 50,369 cells using unsupervised clustering followed by Uniform Manifold Approximation and Projection (UMAP, Fig. 2C, Additional file 1: Fig. S2c and Additional file 2: Table S2). Based on the expression of established cell type–specific markers together with unsupervised cell type annotation (Brain Immune Atlas), these clusters were classified as microglia, border-associated macrophages (BAM), infiltrating monocytes/macrophages (iMM), neutrophils, classical dendritic cells (cDC), T cells, NK/NKT cells, B cells, and endothelial-associated cells (EA) (Fig. 2C, D). Consistent with the IHC analysis, iMM exhibits a marked expansion at D7 compared to sham sample followed by a reduction in numbers at D14 (Fig. 2C, E). For microglia cluster, we observed a reduction in the relative proportion of homeostatic microglia at D7 compared to the Sham group, which show a partial recovery at D14. By D14, microglia show a significant increase, whereas iMM is markedly diminished. These findings are consistent with the IHC results. Additionally, our data shows successful labeling of microglia using tdTomato mRNA expression, which identifies both the homeostatic microglia in sham brains and the reactive microglia profiles on D7 and D14 in stroke brains (Fig. 2F and Additional file 1: Fig. S2d). Importantly, BAM cluster in sham brains and the iMM cluster in the stroke brains are mostly negative for tdTomato (Fig. 2F), further validating our strategy to utilize the TMEM119CreER/CreER-Ai14 mouseline to pre-label microglia. We noted that there is a small number of tdTomato positive cells (%) in the automatically annotated iMM cluster in stroke groups, which are likely to reflect microglia cells that are substantially activated and upregulate genes that are common markers for activated iMM. We further tested this hypothesis and showed a detailed analysis of this cluster later in this paper. Importantly, using the Ligand-receptor Analysis framework (LIANA) method for cell-cell interaction analysis, our data show that iMM and microglia display both overlapping and distinct cell-cell interactions with other immune cell types and blood vessel-associated cells via differential ligand-receptor signaling pathways (Fig. 2G). For example, both iMM and microglia interact with cDC through Cd74 and with neutrophils through Cd44. However, microglia show closer interactions with T cells via Itga4 and Cd3g, whereas iMM primarily interact through Ptprc and Itgab1, highlighting distinct communication patterns between the two populations. Fig. 2Single-cell RNA sequencing (scRNA-seq) analysis of CD45⁺ leukocytes in stroke-affected brains of Tmem119CreER-Ai14 mice. A Scheme of experimental design for scRNA-seq analysis. B FACS sorting of CD45+ cells from brain for single cell immune RNA-seq. C UMAP visualization of CD45⁺ immune cell populations isolated from the brains of Tmem119CreER-Ai14 mice. Colors denote distinct immune clusters. Female and male mice were pooled per group for Sham, D7, and D14 post-stroke samples. D Scaled average expression and percentage of cells expressing key phenotypic markers across identified populations. E Relative proportions of brain leukocyte populations in Sham, D7, and D14 post-stroke conditions. F UMAP visualization highlighting tdTomato expression enriched in microglia clusters but mostly absent in iMM cluster. G LIANA analysis depicting ligand–receptor interaction pairs between microglia or iMM and other immune cell types. Dot color reflects interaction strength (logFC), and dot size indicates interaction specificity Single-cell RNA sequencing (scRNA-seq) analysis of CD45⁺ leukocytes in stroke-affected brains of Tmem119CreER-Ai14 mice. A Scheme of experimental design for scRNA-seq analysis. B FACS sorting of CD45+ cells from brain for single cell immune RNA-seq. C UMAP visualization of CD45⁺ immune cell populations isolated from the brains of Tmem119CreER-Ai14 mice. Colors denote distinct immune clusters. Female and male mice were pooled per group for Sham, D7, and D14 post-stroke samples. D Scaled average expression and percentage of cells expressing key phenotypic markers across identified populations. E Relative proportions of brain leukocyte populations in Sham, D7, and D14 post-stroke conditions. F UMAP visualization highlighting tdTomato expression enriched in microglia clusters but mostly absent in iMM cluster. G LIANA analysis depicting ligand–receptor interaction pairs between microglia or iMM and other immune cell types. Dot color reflects interaction strength (logFC), and dot size indicates interaction specificity Cerebral ischemia elicits distinct transcriptomic states in microglia at days 7 and 14 The majority of current scRNA-seq analysis of microglia profiles in stroke studies focus on the hyperacute stages [43–45]. Resolution of neuroinflammation and shifting of microglia profile from acute injury response to repair-supporting profile might be critical for reducing secondary neuronal injury and restoring the homeostasis of the brain microenvironment [46, 47]. In this study, to investigate whether microglia adopt distinct transcriptional programs on D7 and D14 after stroke, we performed re-clustering of tdTomato⁺/Cldn5− (filter out tdTomato⁺ endothelial cells) microglia (Additional file 2: Table S3). This analysis resulted in 12 microglial clusters, all showing strong expression of canonical microglial markers such as Hexb, P2ry12, Olfml3, and Fcrls (Fig. 3A, B and Additional file 1: Fig. S3a, b). Notably, cells in cluster Mg11 lack expression of microglial markers but instead exhibit strong expression of endothelial-associated cell genes (e.g., Ptn and Nedd4), suggesting an origin from the vessels (Fig. 3C). Thus, we focused on clusters Mg0–10 in the following analyses. In the Sham group, microglia are predominantly distributed in cluster Mg0, which shows robust expression of homeostatic microglial markers such as P2ry12, Tmem119, and Sall1 (Fig. 3B). In line with recent reports, our data also show that these homeostatic genes are downregulated at 7 and 14 days after stroke (Fig. 3C, D). Cluster Mg8-10 do not show an obvious difference in both sham and stroke samples, and we detect these clusters with high expression of microglial (P2ry12, Tmem119 and Cx3cr1) marker genes coexisting with either neutrophil (S100a8 and S100a9), DC (Flt3, Cd209a, H2-Eb1, H2-Aa), T cells (Skap1, Gimap3) (Fig. 3C). We opted to exclude these cells from further analysis because they might be doublets that were not removed by Doublet Finder. Thus, while cluster Mg0 mainly represents homeostatic microglia, clusters Mg1–7 represent stroke-related microglial clusters and the heatmap showed the top 10 upregulated genes in these clusters (Fig. 3C and Additional file 1: Fig. S3c). Among the stroke-induced microglia clusters, clusters Mg1, Mg5, and Mg6 are mostly enriched in D7 stroke samples and clusters Mg2 and Mg3 are mostly enriched in D14 stroke samples. Apoe gene, which has been associated with microglial responses to ischemic injury, is upregulated in multiple microglia clusters at both day 7 and day 14 (Fig. 3E). IF staining confirms ApoE expression in tdTomato+ microglia on D3, D7, D14 and D30 (Fig. 3F). At D3 (Fig. 3F), ApoE signal showed strong colocalization with tdTomato⁺ microglia. However, starting from D7 and D14, a substantial fraction of ApoE signal no longer colocalized with tdTomato. To further investigate the cellular source of ApoE, we performed additional immunofluorescence staining at D7 and D14 using ApoE, IBA1, and tdTomato. We observed that some of the ApoE⁺ signals do not colocalize with either tdTomato or IBA1. Notably, these ApoE⁺ signals also lacked DAPI colocalization, suggesting that they are not associated with intact nucleated cells. This shift is consistent with the evolving microenvironment of the infarct core, where blood–brain barrier disruption, degeneration of axon, and accumulation of extracellular debris become more prominent during the subacute phase after ischemic stroke. Therefore, ApoE signals could originate from degenerated of axon or cellular debris within the infarct core (Additional file 1: Fig. S3d). Fig. 3Sub-clustering analysis of microglia reveals differential transcriptional profiles at days 7 and 14 after ischemic stroke. A (Left) UMAP analysis of merged Sham, D7, and D14 microglial single-cell transcriptomes reveals 12 subclusters; (Right) UMAP colored by time points shows clear segregation of microglial clusters across Sham, D7, and D14 groups. B UMAP displaying microglial signature gene expression. C Dot plot showing scaled expression of selected signature genes across distinct clusters; dot size represents the proportion of cells within each cluster, and color intensity reflects average expression levels. D Relative proportions of microglial subsets in Sham, D7, and D14 post-stroke conditions. E UMAP visualization of Apoe expression across Sham, D7, and D14 groups. F Representative IHC images of ApoE staining in the infarct core and border regions at days 3, 7, 14, and 30 post-stroke. Scale bar = 100 μm. White arrows highlight ApoE+ microglial cells Sub-clustering analysis of microglia reveals differential transcriptional profiles at days 7 and 14 after ischemic stroke. A (Left) UMAP analysis of merged Sham, D7, and D14 microglial single-cell transcriptomes reveals 12 subclusters; (Right) UMAP colored by time points shows clear segregation of microglial clusters across Sham, D7, and D14 groups. B UMAP displaying microglial signature gene expression. C Dot plot showing scaled expression of selected signature genes across distinct clusters; dot size represents the proportion of cells within each cluster, and color intensity reflects average expression levels. D Relative proportions of microglial subsets in Sham, D7, and D14 post-stroke conditions. E UMAP visualization of Apoe expression across Sham, D7, and D14 groups. F Representative IHC images of ApoE staining in the infarct core and border regions at days 3, 7, 14, and 30 post-stroke. Scale bar = 100 μm. White arrows highlight ApoE+ microglial cells Next, we further investigate the distinct microglial clusters induced by stroke on D7 or D14 post-stroke. D7 microglia are primarily distributed within clusters Mg1, Mg5, and Mg6, whereas clusters Mg2 and Mg3 are most prominent at D14 (Fig. 3A). Within the D7 population, cluster Mg1 displays elevated expression of genes associated with the disease-associated microglia (DAM) profile, including Apoe, Spp1, Lgals3, Lpl, Apoc2, Fabp5, and Gpnmb, as well as metabolite-energy-associated genes (Fig. 4A and Additional file 1: Fig. S4a). GO enrichment analysis shows that this cluster is enriched in cellular metabolism and respiration (Fig. 4B). In contrast, clusters Mg5 and Mg6 are enriched for mitotic genes (Ki67, Ccnb2, Mcm5, Mcm2), indicating the presence of proliferative microglia during this phase of stroke (Fig. 4C and Additional file 1: Fig. S4b). Our data further confirms colocalization of the Ki67 signal with tdTomato at D3 after stroke, analyzed by immunofluorescence (Fig. 4D). We also found that a substantial fraction of Ki67⁺ cells did not colocalize with tdTomato-labeled microglia on D7 and D14 after stroke. We performed immunofluorescence staining at D7 and D14 using Ki67 and IBA1. We observed that a large proportion of Ki67⁺ cells colocalize with IBA1⁺ tdTomato⁻ cells on D7, which suggested these Ki67⁺ IBA1⁺ tdTomato⁻ cells represent iMM (Additional file 1: Fig. S4c). At D14, in addition to Ki67⁺ IBA1⁺ tdTomato⁻ cells, we also observed a subset of Ki67⁺ cells that did not colocalize with either IBA1 or tdTomato (Additional file 1: Fig. S4c). Based on previously study, these cells are consistent with proliferating non-myeloid populations within the injured brain, including astrocytes, NG2 glia/oligodendrocyte precursor cells (OPCs), and vascular-associated cells, which are known to undergo proliferation during the subacute phase after ischemic stroke and contribute to tissue repair and remodeling [48–52]. We observed that Igf1 expression in microglia was elevated in clusters Mg1, Mg5, and Mg6 but declined by D14 (Fig. 4E), suggesting that microglial igf1 might play an important role in the D7 to D14 profile transition after stroke. Among D14-enriched clusters, cluster Mg2 was characterized by upregulation of chemokine genes (Ccl4, Ccl3, and Ccl6) and TNF-associated genes (Fig. 4F), indicating a pro-inflammatory phenotype described in ischemic stroke [45, 53, 54]. Pathway enrichment analysis shows that genes are enriched in the immune response signaling pathway and TNF superfamily cytokine production, supporting the idea that this microglial cluster could be responsible for the prolonged inflammatory response after stroke (Fig. 4G). In contrast, cluster Mg3 exhibited re-emergence of homeostatic markers (P2ry12, Tmem119) accompanied by downregulation of Spp1 and Lpl, suggesting a shift toward restoration of the homeostatic microglial state (Fig. 3C and Additional file 1: Fig. S4d). Both D7 and D14 microglia contained substantial proportions of clusters Mg4 and Mg7. Cluster Mg4, in agreement with previous studies, is characterized by strong upregulation of interferon-stimulated genes (Isg15, Ifit3, Ifit2) and interferon regulatory factors (Irf7, Stat1, Stat2) (Fig. 4H). Pathway enrichment confirmed activation of interferon-related processes, consistent with type I IFN response module scoring (Fig. 4I, J). In addition, we also found that a specific cluster (cluster Mg7) showed enriched genes of phagocytosis (Cd93, Ms4a4a, Anxa1) [55–58] and lipid transport (Apoe, Apoc2, Ldlrad3, Psap, Pltp) [59] (Fig. 4K and Additional file 1: Fig. S4e). Pathway enrichment analysis shows genes enriched in lipid transport, endocytosis, and phagocytosis (Fig. 4L). We next examined the spatial distribution of CD93⁺ microglia and found that CD93 is predominantly expressed in blood vessel endothelium and CD93⁺/tdTomato⁺ microglia were positioned in close association with the vasculature (Fig. 4M). However, the overlap between CD93 and tdTomato-positive microglia may reflect microglial processes contacting blood vessels rather than true cellular co-localization. To further address this point, we have analyzed IHC samples using confocal microscopy and 3D reconstruction to enable precise spatial analysis. Our results show both true CD93+/tdTomato+ microglia as well as tdTomato+ microglia contacting CD93+ blood vessels (Fig. 4N). These data allow clearer visualization of both scenarios, helping to distinguish CD93⁺ tdTomato⁺ cells from microglial processes contacting blood vessels. Pseudotime analysis revealed a trajectory in which sham microglia (Mg0) transition toward Mg1 (D7 enriched) or Mg2/Mg3 (D14 enriched), while proliferating microglia (Mg5 and Mg6) also converge toward the Mg1 state (Fig. 4O). Fig. 4Cerebral ischemia elicits distinct functional subclusters of microglia at days 7 and 14. A (Left) UMAP colored by time points demonstrates distinct segregation of microglial clusters across Sham, D7, and D14 groups. (Right) Feature plots showing representative marker gene expression for cluster Mg1. B Gene ontology (GO) terms enriched in the marker genes of cluster Mg1. C (Left) UMAP visualization highlighting clusters Mg5 and Mg6. (Right) Feature plots showing representative gene expression of Mg5 and Mg6. D Representative IHC images of Ki67 staining within the infarct core and border regions at days 3, 7, 14, and 30 post-stroke. Scale bar = 100 μm. White arrows highlight Ki67+ microglial cells. E (Left) UMAP showing Igf1 expression. (Right) Boxplot comparing Igf1 expression among Sham, D7, and D14 groups. Data shown as mean ± SE. F (Left) UMAP visualization highlighting cluster Mg2. (Right) Feature plots showing representative gene expression of Mg2. (G) GO terms enriched in the marker genes of Mg2. H (Left) UMAP visualization highlighting cluster Mg4. (Right) Violin plot showing representative gene expression of Mg4. (I) IFN module score analysis highlighting Mg4 cluster as a heightened IFN-responsive cluster. J GO biological processes enriched in Mg4. K (Left) UMAP visualization highlighting cluster Mg7. (Right) Violin plot showing representative gene expression of Mg7, especially highlighting enriched Cd93, Ms4a4a, Pltp and Psap expression. L GO biological processes enriched in Mg7. M Representative IHC images of CD93 immunostaining in the infarct core and border regions at days 3, 7, 14, and 30 post-stroke. Scale bar = 100 μm. White arrows highlight CD93+ microglial cells. N Representative 3D IHC images of CD93 immunostaining in the infarct core at D7 post-stroke. Scale bar = 10 μm. O Pseudotime trajectory analysis showing predicted microglial state transitions. Statistical analysis was performed using one-way ANOVA with Tukey post hoc tests Cerebral ischemia elicits distinct functional subclusters of microglia at days 7 and 14. A (Left) UMAP colored by time points demonstrates distinct segregation of microglial clusters across Sham, D7, and D14 groups. (Right) Feature plots showing representative marker gene expression for cluster Mg1. B Gene ontology (GO) terms enriched in the marker genes of cluster Mg1. C (Left) UMAP visualization highlighting clusters Mg5 and Mg6. (Right) Feature plots showing representative gene expression of Mg5 and Mg6. D Representative IHC images of Ki67 staining within the infarct core and border regions at days 3, 7, 14, and 30 post-stroke. Scale bar = 100 μm. White arrows highlight Ki67+ microglial cells. E (Left) UMAP showing Igf1 expression. (Right) Boxplot comparing Igf1 expression among Sham, D7, and D14 groups. Data shown as mean ± SE. F (Left) UMAP visualization highlighting cluster Mg2. (Right) Feature plots showing representative gene expression of Mg2. (G) GO terms enriched in the marker genes of Mg2. H (Left) UMAP visualization highlighting cluster Mg4. (Right) Violin plot showing representative gene expression of Mg4. (I) IFN module score analysis highlighting Mg4 cluster as a heightened IFN-responsive cluster. J GO biological processes enriched in Mg4. K (Left) UMAP visualization highlighting cluster Mg7. (Right) Violin plot showing representative gene expression of Mg7, especially highlighting enriched Cd93, Ms4a4a, Pltp and Psap expression. L GO biological processes enriched in Mg7. M Representative IHC images of CD93 immunostaining in the infarct core and border regions at days 3, 7, 14, and 30 post-stroke. Scale bar = 100 μm. White arrows highlight CD93+ microglial cells. N Representative 3D IHC images of CD93 immunostaining in the infarct core at D7 post-stroke. Scale bar = 10 μm. O Pseudotime trajectory analysis showing predicted microglial state transitions. Statistical analysis was performed using one-way ANOVA with Tukey post hoc tests iMM showed distinct transcriptomic states on D7 and D14 after stroke After annotation, a cluster of cells is annotated as iMM. Importantly, this cluster of cells is mostly absent in the sham brain but increases in the D7 stroke sample and decreases in the D14 stroke brain (Fig. 5A). This dynamic is consistent with our IHC data using tdTomato reporter tracing, showing that iMM increases in day 7 stroke brain, but the numbers diminish at day 14 and eventually distinguish at D30 after stroke. These data together suggest that iMM is not long-living in the stroke brain, contradicting previous speculations [26, 42, 60]. All iMM expressed high levels of Lyz2 and Apoe (Fig. 5A), indicating that they represent highly activated phagocytic cells transitioning toward a macrophage-like state. We further resolved 8 iMM clusters (Fig. 5B, Additional file 2: Table S4). Cells in cluster iMM4 did not express typical iMM markers but instead showed strong expression of epithelial-associated genes (e.g., Ptn and Nedd4), suggesting an origin from vascular-associated cells (iMM4 was excluded from downstream analysis) (Fig. 5C). Cluster iMM5 exhibited high tdTomato expression together with canonical microglia lineage marker genes such as Sall1 [61], indicating that a subset of strongly activated microglia may be automatically annotated as iMM (Fig. 5C and Additional file 1: Fig. S5a). In the sham brain, iMM were scarce, and 63% of them grouped into cluster iMM5, which is closely associated with blood monocyte signatures (Ly6c2, Ccr2, Plac8) [44] (Additional file 1: Fig. S5b). Clusters iMM0, iMM1, iMM2, and iMM3 were predominant found in D7 samples, whereas cluster iMM7 was the major population at D14. Clusters iMM0 and iMM2 shared gene signatures (Cd68, Lgals3, Lgals1, Gpnmb, Fabp5, Ctsb, Ctsd, Spp1) previously described in stroke-associated monocyte-derived cells at D2 [44] (Fig. 5C, D). GO enrichment analysis showed that both cluster iMM0 and iMM2 cells exhibited high metabolic and activation states, whereas cluster iMM2 was additionally enriched for oxidative response and lipid transport pathways (Fig. 5D and Additional file 1: Fig. S5c). Clusters iMM1 and iMM3 shared signatures, including Tmem176a, Tmem176b, and Clec4n, associated with lymphocyte and myeloid differentiation (Fig. 5E). However, cluster iMM3 also showed strong expression of interferon-stimulated genes (ISG), including Ifit2, Ifit3, and Isg15, together with interferon regulatory factors such as Irf7 and Stat2 (Fig. 5E and Additional file 1: Fig. S5d). Pathway analysis confirmed enrichment of interferon-related processes (Additional file 1: Fig. S5e). Cluster iMM7 was the dominant cluster at D14. Cxcl10 emerged as a specific marker of cluster iMM7, accompanied by high ISG expression and blood monocyte–associated genes (Sell, Plac8), indicating that these cells remain in an activated state supporting residual inflammation (Fig. 5C, F). Volcano plot results and GO enrichment analysis were consistent with these observations: iMM at D14 exhibited elevated expression of interferon-associated genes, whereas D7 iMM were characterized by enhanced metabolic activity (Angptl4, Npl, Cdk1) (Fig. 5G and Additional file 1: Fig. S5f, g). Trajectory analysis revealed a sequential transition beginning with peripheral monocytes (cluster iMM5), followed by progression into clusters iMM0, iMM2, and iMM1, and culminating in differentiation into cluster iMM3. Cluster iMM7 appeared as the only population dissociated from this trajectory (Fig. 5H). These findings suggest that stroke creates a niche that allows monocytes to enter the brain but prevents long-term retention, with a small set of monocytes remaining in an infiltrating state in the inflamed brain, particularly at D14 (cluster iMM7). The upregulation of Cxcl10 in a subset of iMM at this sub-chronic stage could contribute to T-cell infiltration at later time points in the stroke brain via CXCL10-CXCR3 signaling. Fig. 5Transcriptional changes of iMM on D7 and D14 after stroke. A (Left) UMAP highlighting the iMM cluster (cluster 3). (Right) Feature plots showing Lyz2 and Apoe expression within the CD45⁺ immune cell single-cell RNAseq dataset. B (Left) UMAP analysis of merged Sham, D7, and D14 iMM transcriptomes reveals eight subclusters. (Right) UMAP colored by time points demonstrates clear segregation of iMM populations across Sham, D7, and D14 groups. C Heatmap displaying the top genes for each iMM subcluster. Note that the automatically annotated iMM Cluster 4 is tdTomato+ microglia that are substantially activated and that resemble the iMM profile. D (Left) Feature plots showing representative gene expression profiles of Mg0 and Mg2. (Right) GO terms enriched in the marker genes of Mg0. E (Left) Feature plots showing representative gene expression profiles of Mg1 and Mg3. (Right) GO terms enriched in the marker genes of Mg1. F (Left) Violin plot showing representative gene expression of Mg7. (Right) GO biological processes enriched in Mg7. G Volcano plots comparing D14 versus D7 iMM transcriptomes. H Pseudotime trajectory analysis showing predicted transitions in iMM states. I (Left) UMAP visualization of Igf1 expression. (Right) Boxplot comparing Igf1 expression in iMM among Sham, D7, and D14 groups. Data shown as mean ± SE. J (Left) UMAP showing Cd93 expression enriched in iMM. (Right) Representative IHC images of CD93 staining in the infarct core on D7 post-stroke. Scale bar = 100 μm. Red arrows highlight CD93+/IBA1+ iMM. K Representative IHC images for CD93, IBA1, and P2RY12 in the infarct core and contralateral region. Scale bar = 20 μm. Yellow arrows highlight CD93+/tdTomato+/IBA1+/P2RY12+ microglial cells; White arrows highlight CD93+/tdTomato−/IBA1+/P2RY12+ iMM. Statistical analysis was performed using one-way ANOVA with Tukey post hoc tests Transcriptional changes of iMM on D7 and D14 after stroke. A (Left) UMAP highlighting the iMM cluster (cluster 3). (Right) Feature plots showing Lyz2 and Apoe expression within the CD45⁺ immune cell single-cell RNAseq dataset. B (Left) UMAP analysis of merged Sham, D7, and D14 iMM transcriptomes reveals eight subclusters. (Right) UMAP colored by time points demonstrates clear segregation of iMM populations across Sham, D7, and D14 groups. C Heatmap displaying the top genes for each iMM subcluster. Note that the automatically annotated iMM Cluster 4 is tdTomato+ microglia that are substantially activated and that resemble the iMM profile. D (Left) Feature plots showing representative gene expression profiles of Mg0 and Mg2. (Right) GO terms enriched in the marker genes of Mg0. E (Left) Feature plots showing representative gene expression profiles of Mg1 and Mg3. (Right) GO terms enriched in the marker genes of Mg1. F (Left) Violin plot showing representative gene expression of Mg7. (Right) GO biological processes enriched in Mg7. G Volcano plots comparing D14 versus D7 iMM transcriptomes. H Pseudotime trajectory analysis showing predicted transitions in iMM states. I (Left) UMAP visualization of Igf1 expression. (Right) Boxplot comparing Igf1 expression in iMM among Sham, D7, and D14 groups. Data shown as mean ± SE. J (Left) UMAP showing Cd93 expression enriched in iMM. (Right) Representative IHC images of CD93 staining in the infarct core on D7 post-stroke. Scale bar = 100 μm. Red arrows highlight CD93+/IBA1+ iMM. K Representative IHC images for CD93, IBA1, and P2RY12 in the infarct core and contralateral region. Scale bar = 20 μm. Yellow arrows highlight CD93+/tdTomato+/IBA1+/P2RY12+ microglial cells; White arrows highlight CD93+/tdTomato−/IBA1+/P2RY12+ iMM. Statistical analysis was performed using one-way ANOVA with Tukey post hoc tests We also observed that Igf1 expression in iMM was elevated at D7 but declined by D14, consistent with the temporal dynamics of iMM abundance. In Fig. 5I, iMM exhibited high expression of Cd93, and CD93⁺/IBA1⁺/tdTomato⁻ iMM were markedly increased in the infarct core on D7 after stroke, suggesting that CD93⁺ iMM accumulates after stroke (Fig. 5J). CD93⁺ cells exhibited P2RY12 immunoreactivity, with some originating from activated microglia (tdTomato+), but others lacking tdTomato expression, suggesting an iMM origin. Indeed, immunofluorescence analysis revealed CD93⁺/P2RY12⁺/tdTomato⁻ cells in the infarct core on D7 and D14 but disappeared on D30 (Fig. 5K; Additional file 1: Fig. S5h), supporting that CD93+ iMM temporarily adopt certain microglia-like genes such as P2ry12 after infiltrating into the stroke brain. TMEM119CreER-Ai14 lineage tracing in stroke mice identified the profiles of a subset of highly activated microglia that resembled those of iMM but retained distinct microglial makers and show cell–cell interactions with other infiltrating immune cells Interestingly, as previously discussed above, in the automatically annotated iMM cluster, we detect a subgroup of tdTomato+ cells (Fig. 6A), demonstrating they were microglial cells pre-labeled before stroke onset. When comparing tdTomato+ (activated microglia that upregulate some iMM markers) vs. true iMM (tdTomato− iMM cells) in this cluster, we observe distinct differentially expressed genes (DEGs) in tdTomato+ vs. tdTomato− cells (Fig. 6B). The tdTomato+ cells exhibited higher expression of canonical microglial genes (Sall1, Lrrc3, Siglech) and gliogenesis genes (Sparc, Tgfa) [62, 63], whereas the tdTomato− cells showed increased expression of activated iMM genes (Ms4a7, Ccl12, Rpl24) (Fig. 6B, Additional file 2: Table S5). These DEGs demonstrate that ontology markers are preserved in microglia vs. iMM even when they share many activated genes and are annotated into the same cluster by automatic annotation. This also suggests that there may be different functions between tdTomato+ microglia and tdTomato− iMM, even if they are annotated in the same cluster. Indeed, GO analysis shows that activated microglia in this cluster are enriched for functional pathways for gliogenesis and crosstalk with leukocytes, while iMM is functionally enriched for metabolic remodeling (Fig. 6C). This exciting result suggests that, despite certain similar gene expression profiles resembling previously described monocytes markers, activated microglia and iMM still adopt distinct transcriptomic profiles and functional pathways, indicating they may carry out distinct functions as well. This is corroborated with differential cell-cell interaction of microglia or iMM with other immune cells and EAs (Fig. 6D). Additionally, using this system, comparing DEGs in tdTomato+ vs. tdTomato− cells, we validate our animal model and strategy to be useful in identifying canonical microglia lineage markers such as Sall1 [61] and Sparc [62, 63]. Additionally, we identified additional promising novel cell markers that could distinguish activated microglia vs. iMM. Tspan7 gene is enriched in activated microglia (tdTomato+) vs. iMM (tdTomato−). Conversely, Gm21188 is enriched in tdTomato− iMMs. We compared the expression patterns of these novel microglia or iMM markers across several independent brain scRNA-seq datasets (Additional file 1: Fig. S6a and b, showing a representative stroke dataset) [44] and found striking consistency in the unique enrichment of these markers in either microglia or iMM across all datasets examined [44, 64, 65]. Building on this interesting data, we then carry out the RNAscope validation at different time points after stroke (D3 and D7). Our data demonstrated that Gm21188 mRNA is localized within iMM, whereas Tspan7 expression is predominantly observed in microglia at the stroke core (Fig. 6E, F). This result demonstrates that the genetic lineage tracing strategy can be a useful tool to distinguish activated resident vs. iMM in many disease conditions and to identify novel markers for these two cell populations in various pathological conditions. Fig. 6The transcriptional profiles of a subset of tdTomato⁺ microglia within cluster 3 resembled those of iMM while retaining distinct microglial features. A UMAP visualization of cluster 3 with tdTomato⁺ cells highlighted. B Volcano plots comparing the transcriptomes of tdTomato⁺ versus tdTomato⁻ cells within cluster 3. C GO terms enriched in the marker genes of tdTomato⁺ and tdTomato⁻ cells, respectively. D Bubble plot illustrating ligand–receptor interaction pairs between tdTomato⁺ or tdTomato⁻ cells and other immune cell types, where dot color indicates interaction strength (logFC) and dot size reflects interaction specificity. E (Top) UMAP visualization of Gm21188 and Ms4a7 expression. (Bottom) Representative image showing combined IHC staining for IBA1 with Gm21188 RNA-scope hybridization. Scale bar = 20 μm. The yellow arrow marks the tdTomato+/IBA1+ microglia; the white arrow marks the tdTomato−/IBA1+ iMM. F (Top) UMAP visualization of Tspan7 and Sall1 expression. (Bottom) Representative image showing combined IHC staining for IBA1 with Tspan7 RNA-scope hybridization. Scale bar = 20 μm. The yellow arrow and yellow dotted line marks the tdTomato+/IBA1+ microglia; the white arrow and white dotted line marks the tdTomato−/IBA1+ iMM The transcriptional profiles of a subset of tdTomato⁺ microglia within cluster 3 resembled those of iMM while retaining distinct microglial features. A UMAP visualization of cluster 3 with tdTomato⁺ cells highlighted. B Volcano plots comparing the transcriptomes of tdTomato⁺ versus tdTomato⁻ cells within cluster 3. C GO terms enriched in the marker genes of tdTomato⁺ and tdTomato⁻ cells, respectively. D Bubble plot illustrating ligand–receptor interaction pairs between tdTomato⁺ or tdTomato⁻ cells and other immune cell types, where dot color indicates interaction strength (logFC) and dot size reflects interaction specificity. E (Top) UMAP visualization of Gm21188 and Ms4a7 expression. (Bottom) Representative image showing combined IHC staining for IBA1 with Gm21188 RNA-scope hybridization. Scale bar = 20 μm. The yellow arrow marks the tdTomato+/IBA1+ microglia; the white arrow marks the tdTomato−/IBA1+ iMM. F (Top) UMAP visualization of Tspan7 and Sall1 expression. (Bottom) Representative image showing combined IHC staining for IBA1 with Tspan7 RNA-scope hybridization. Scale bar = 20 μm. The yellow arrow and yellow dotted line marks the tdTomato+/IBA1+ microglia; the white arrow and white dotted line marks the tdTomato−/IBA1+ iMM Abolishing post-stroke Igf1 upregulation in microglia and iMM accelerated recovery of sensorimotor function but increased anxiety level in stroke mice Our results show that microglia and iMM both show high levels of Igf1 upregulation on D7 and D14 post stroke, which supports a potential role of microglial and iMM Igf1 during the recovery stage of stroke (Fig. 7A). Previous animal study has suggested that exogenous IGF-1 delivery in the hyperacute stage of stroke is neuroprotective [66, 67]. However, the role of endogenous Igf1 upregulation, especially in microglia and iMM, remains poorly understood, especially in the subacute and chronic stages of stroke. Given the recent promising results of systemic IGF-1 delivery (intravenously) in the acute stage of human stroke patients, which reduced the initial infarct volume, alleviated stroke severity, and enhanced 90-day functional outcomes, it is even more critical to understand the role of this growth factor in the later stage of stroke recovery to facilitate optimal design of treatment window and strategy [68]. To investigate the role of Igf1 in both resident microglia and iMM in the subacute to chronic stage of stroke recovery, we generated Cx3cr1CreERIgf1fl/fl conditional knockout mice, in which Cre recombinase is expressed under the control of the Cx3cr1 promoter. Upon tamoxifen administration at post-stroke day 5–10, Igf1 is selectively ablated in all Cx3cr1-expressing cells, including CNS-resident microglia as well as peripheral monocytes that infiltrate the brain following stroke. Cx3cr1CreERIgf1fl/fl (termed Igf1 iKO) mice received tamoxifen treatment (180 mg/kg, oral gavage once daily for 5 consecutive days) starting on D5 after MCAo surgery (Fig. 7B, C). A comprehensive behavioral test battery was employed to assess general locomotor activity, motor coordination, anxiety levels, and cognitive function related to learning and memory. All mice were pre-trained (locomotion, rotorod and tape removal) prior to surgery, and individuals that exhibited poor baseline performance were excluded from subsequent analyses. First, to evaluate voluntary locomotor activity, an automated open field tracking system was used to test the mice at pre, D3, D7, D14, D21 and D30 post stroke. We did not observe any change in total distance and the mean velocity among non-stroke WT or Igf1 iKO mice at any of the time points either during the exploratory or during the subsequent light and dark cycles after stroke (Additional file 1: Fig. S7a, b), suggesting that microglia-Igf1 iKO does not affect general locomotion function either at baseline or in stroke mice. Similarly, Igf1 iKO or WT stroke do not show difference in general locomotion at D3, D7, D14 and D21 (1 h after the initial naïve exposure to the chamber, Fig. 7D), but Igf1 iKO stroke mice showed reduced average speed at D30 compared to WT stroke mice (5.334 vs. 6.485, p = 0.0278) (Fig. 7D). Then, we performed a rotarod test to evaluate motor coordination and motor learning in the mice. Rotarod testing revealed that motor coordination and motor learning were transiently impaired at D14 after stroke, which recovered by D21 and D30. However, there is no difference between the WT or Igf1 iKO mice (Fig. 7E and Additional file 1: Fig. S7c). This finding suggests that the loss of Igf1 in Cx3cr1-expressing cells post stroke does not exacerbate gross motor dysfunction following stroke, indicating that microglia- or iMM-derived Igf1 may not be essential for the recovery of general motor performance. Next, the tape-removal test was performed to assess fine sensorimotor function. Our data shows the average latency to remove the tape of left forepaw (affected paw) is significantly increased from D3 after stroke compared with Sham (209.5 vs. 11.29 s, p < 0.0001). Igf1 iKO stroke mice showed better performance at D14 (59.44 vs. 149.7 s, p = 0.0282) and D21 (24.78 vs. 118.2 s, p = 0.0215) compared with WT stroke mice, which suggests that deleting Igf1 in microglia and iMM helps accelerate recovery of sensorimotor function (Fig. 7F). The right forepaw (unaffected side) shows a transient deficit following the surgical procedure but recovers quickly and similarly in both WT and Igf1 iKO mice (Fig. 7F). We next performed the elevated plus maze (EPM) to evaluate anxiety-like behavior. The Igf1 iKO stroke mice spent less time in the open arms compared with the WT stroke mice (111.2 vs. 31.01 s, p = 0.0227) (Fig. 7G and Additional file 1: Fig. S7d), suggesting that Igf1 deletion in microglia and iMM may exacerbate anxiety-like behavior after stroke. Finally, we use the contextual fear conditioning (CFC) test to measure cognitive function related to learning and memory. Our data showed that stroke significantly reduced fear memory-related freezing compared with non-stroke mice (average 49.15% vs. 72.48%, p = 0.0013); however, there was no significant difference between WT and Igf1 iKO stroke mice, suggesting that deletion of Igf1 in microglia and iMM does not affect cognitive function in this test (Fig. 7H). In summary, our behavioral data show post-stroke Igf1 iKO in microglia and iMM help recovery of sensorimotor function but increase anxiety level after stroke. Fig. 7Conditional Igf1 knockout in microglia and iMM accelerates post-stroke sensorimotor recovery but increases anxiety-like behavior and reduces CD8⁺ T-cell infiltration. (A) Boxplot comparing Igf1 expression between microglia and iMM on D7 and D14. (B) Schematic of the mouse model used to induce Igf1 knockout in microglia and iMM. (C) Experimental timeline, showing the order of behavioral measurements. (D) Behavioral measurements in TAM-treated Cx3cr1CreERIgf1WT/WT or Cx3cr1CreERIgf1fl/fl after stroke showing total distance and average speed in the open field test (OFT) of the first hour in the locomotion chamber. (E) Accelerated rotarod learning test. (F) The latency of tape removal. (G) Elevated plus maze test. (H) Contextual fear test (CFC) showing freezing episode and %freezing during testing. n = 5–11 per group. (I) Representative IHC images for CD8 and CD4 in the infarct core at days 3, 7, 14, and 30 following stroke. The yellow arrow marks the CD8+ T cells; the white arrow marks the CD4+ T cells. (J) Quantification of CD8+ T cells number at days 3, 7, 14, and 30 following stroke. n = 3–4 mice per group. (K) Quantification of CD4+ T cells number at days 3, 7, 14, and 30 following stroke. n = 3–4 mice per group. (L) Representative IHC images for CD8 and CD4 in the infarct core after stroke in TAM-treated Cx3cr1CreERIgf1WT/WT or Cx3cr1CreERIgf1fl/fl. The yellow arrow marks the CD8+ T cells; the white arrow marks the CD4+ T cells. (M) Quantification of CD8+ T cells number. n = 5–6 mice per group. (N) Quantification of CD4+ T cells number. n = 5–6 mice per group. Data shown as mean ± SE. (O, P) Bubble plot illustrating ligand–receptor interaction pairs between Igf⁺ or Igf1⁻ cells microglia/iMM and T cells, where dot color indicates interaction strength (logFC) and dot size reflects interaction specificity. Statistical analysis was performed using the Student’s t test (G, M, N), and one-way (H) or two-way (A, D, E, F, J, K) ANOVA, as appropriate, with Tukey post hoc tests Conditional Igf1 knockout in microglia and iMM accelerates post-stroke sensorimotor recovery but increases anxiety-like behavior and reduces CD8⁺ T-cell infiltration. (A) Boxplot comparing Igf1 expression between microglia and iMM on D7 and D14. (B) Schematic of the mouse model used to induce Igf1 knockout in microglia and iMM. (C) Experimental timeline, showing the order of behavioral measurements. (D) Behavioral measurements in TAM-treated Cx3cr1CreERIgf1WT/WT or Cx3cr1CreERIgf1fl/fl after stroke showing total distance and average speed in the open field test (OFT) of the first hour in the locomotion chamber. (E) Accelerated rotarod learning test. (F) The latency of tape removal. (G) Elevated plus maze test. (H) Contextual fear test (CFC) showing freezing episode and %freezing during testing. n = 5–11 per group. (I) Representative IHC images for CD8 and CD4 in the infarct core at days 3, 7, 14, and 30 following stroke. The yellow arrow marks the CD8+ T cells; the white arrow marks the CD4+ T cells. (J) Quantification of CD8+ T cells number at days 3, 7, 14, and 30 following stroke. n = 3–4 mice per group. (K) Quantification of CD4+ T cells number at days 3, 7, 14, and 30 following stroke. n = 3–4 mice per group. (L) Representative IHC images for CD8 and CD4 in the infarct core after stroke in TAM-treated Cx3cr1CreERIgf1WT/WT or Cx3cr1CreERIgf1fl/fl. The yellow arrow marks the CD8+ T cells; the white arrow marks the CD4+ T cells. (M) Quantification of CD8+ T cells number. n = 5–6 mice per group. (N) Quantification of CD4+ T cells number. n = 5–6 mice per group. Data shown as mean ± SE. (O, P) Bubble plot illustrating ligand–receptor interaction pairs between Igf⁺ or Igf1⁻ cells microglia/iMM and T cells, where dot color indicates interaction strength (logFC) and dot size reflects interaction specificity. Statistical analysis was performed using the Student’s t test (G, M, N), and one-way (H) or two-way (A, D, E, F, J, K) ANOVA, as appropriate, with Tukey post hoc tests Igf1 iKO in microglia and iMM decreased CD8+T cell infiltration after stroke To further elucidate the mechanism by which Igf1 deletion in microglia and iMM facilitates recovery of sensorimotor function but increases anxiety level, we performed doublecortin (DCX) immunostaining to assess neurogenesis. Adult hippocampal neurogenesis has been implicated in the regulation of anxiety-like behavior in rodents [69]. However, no significant difference was observed among the four groups, suggesting that microglial and iMM Igf1 deletion does not affect hippocampal neurogenesis under homeostasis or after stroke in mice (Additional file 1: Fig. S7e, f), corroborating recent findings in rodent studies by us and others [70, 71]. We next sought to explore whether Igf1 deletion influences behavior by modulating inflammation and cell-cell interactions with other immune cells. A recent study also showed that CD8+ T cell depletion improved motor recovery on the rotarod but decreased crossing time in the open field after stroke [72]. This phenotype resembles what we observed in Igf1 iKO mice following stroke. Therefore, we performed CD8 and CD4 IHC staining to measure the dynamics of infiltrating T cells at D3, D7, D14, and D30 in WT or Igf1 iKO stroke mice (Additional file 2: Table S6). Our data showed that the average number of CD8⁺ and CD4⁺ T cells exhibited a slight increase at D3 (CD8+: 3.75 cells/mm², n.s. vs. control; CD4+: 2.83 cells/mm², n.s. vs. control)and D7 (CD8+: 3.75 cells/mm², n.s. vs. control; CD4+: 4.83 cells/mm², n.s. vs. control) within the infarct core, followed by a sharp rise at D14 (CD8+: 39.56 vs. 0 cells/mm² in controls, p < 0.0001; CD4+: 15.67 vs. 0 cells/mm² in controls, p < 0.0001 ) and D30 (CD8+: 31.17 vs. 0.5 cells/mm² in controls, p < 0.0001; CD4+: 12.0 vs. 0.25 cells/mm² in controls, p = 0.0005 ) in WT stroke mice (Fig. 7I-K). Interestingly, as hypothesized, Igf1 iKO stroke mice show fewer numbers of CD8⁺ T cells at D30 within the infarct core (0.33 vs. 1.0 in controls, p = 0.001). CD4⁺ T cells show a reduced trend in the Igf1 iKO stroke group, but did not reach statistical significance (0.59 vs. 1.0 in controls, p = 0.0939) (Fig. 7L-N). Additionally, non-stroke WT or Igf1 iKO mice both show minimal infiltration of CD8⁺ and CD4⁺ T cells in the brain. To further dissect how Igf1 influences T-cell infiltration, we compared ligand–receptor interactions between Igf1⁺ and Igf1⁻ microglia/iMM and T cells using LIANA (Fig. 7O, P). Igf1⁺ myeloid populations exhibit a more coordinated and functionally diverse interaction landscape with T cells. First, Igf1⁺ microglia uniquely exhibit chemokine signaling via the Cxcl16–Cxcr6 chemotactic axis, which is known to regulate recruitment of CXCR6⁺ T cells to inflamed tissues (Fig. 7O). Second, Igf1⁺ microglia show enhanced MHC-I–mediated interactions (e.g., H2-K1 and H2-M3 with CD8 receptors), suggesting increased capacity for CD8⁺ T-cell engagement and activation, whereas Igf1⁻ microglia display a more limited antigen presentation profile (Fig. 7O). Notably, Igf1⁺ iMM displays a strongly enriched Spp1-centered interaction network, engaging multiple integrin and CD44 receptors, which have been implicated in immune cell migration and extracellular matrix interactions (Fig. 7P). This network is largely absent in Igf1⁻ iMM. In contrast, Igf1⁻ populations display more limited chemotactic and migratory signaling and are relatively enriched in classical co-stimulatory interactions (e.g., Cd86–Cd28) and cell adhesion (e.g., Itgav–Thy1), suggesting functional divergence. Together, these findings suggest that Igf1⁺ myeloid cells promote CD8⁺ T-cell infiltration not through a single pathway, but via coordinated enhancement of chemotaxis, antigen presentation, adhesion, and extracellular matrix interactions. Thus, our results suggest that Igf1 upregulation in microglia and iMM may be critical for attracting CD8+T cell infiltration in the later stages of stroke. Functionally, this reduced CD8+T cell infiltration might lead to complex behavioral results with Cx3cr1ERT2 Igf1 iKO mice showing improved sensorimotor function recovery but increased anxiety levels at 1 month after stroke. Discussion Using TMEM119CreER-Ai14 model to pre-label resident microglia before stroke, we observed that iMM markedly increased in the infarct core on D3 after stroke, peaked at D7, and almost disappeared by 1 month. Our data support that iMM can transiently upregulate certain microglia signature genes such as P2ry12, however, they do not induce lineage signature genes that are exclusively expressed in resident microglia, such as Sall1 [61]. This contrasts with a neonatal stroke study reporting that infiltrating monocytes persist for more than 60 days and gradually acquire a ramified morphology with microglial marker expression (Sall1, Tmem119, P2ry12) [26]. The difference between the long-term survival and transcriptomic profile of iMM in neonatal vs. adult brain have been reported in non-stroke animals [61] and our result further highlight the reduced potential of adult iMM to colonize the CNS after infiltration compared with their neonatal counterparts. At different stroke locations, microglia and iMM have distinct dynamics. In the stroke core, iMM peaks at D7 and then rapidly declines, becoming nearly absent by D30. In contrast, microglia number in the infarct core exhibits a substantial expansion at later stages (such as D14 and D30 post-stroke). This suggests a potential role switch from the acute stage to the subacute stage in the core between iMM and microglia. scRNA-seq further revealed distinct transcriptomic and functional states of microglia and iMM at D7 and D14. We also identify Gm21188, which could be a specific gene maker of activated iMM after stroke. Finally, our data showed that Igf1 was up-regulated in both microglia and iMM after stroke, and Cx3cr1ERT2 Igf1 iKO showed sensorimotor function recovery, suggesting a potential mechanism involved in decreased CD8+ T-cell infiltration after stroke. By utilizing FACS-isolated CD45⁺ cells from Tmem119CreER-Ai14 mice, where microglia are genetically pre-labeled with tdTomato, we achieved higher-resolution characterization of microglial populations within the brain immune landscape. We identified 12 clusters of microglia based on tdTomato pre-labeling. The inclusion of two post-stroke time points (D7 and D14) in our scRNA-seq analysis provided a temporal framework to examine the dynamic progression of microglial activation and phenotypic transitions following stroke, especially in subacute to chronic stage, which is understudied in stroke models. Our data demonstrated that there are distinct cellular transcriptomic responses of microglia on D7 and D14 after stroke. D7 displayed elevated expression of genes associated with DAM and proliferation (Ki67), which is in line with previous reports [43]. Our data also revealed transcriptionally distinct microglial clusters on D14 after stroke. One cluster (Mg3) shows recovery in the expression of homeostatic genes (P2ry12, Tmem119), with concomitant suppression of Spp1 and Lpl, suggesting an inflammation-resolving microglial profile on D14. Another microglial cluster (Mg2) was characterized by the upregulation of chemokine genes (Ccl4, Ccl3, and Ccl6) together with TNF-associated genes. A previous scRNA-seq study at post-MCAo days 30 and 90 also reported substantial upregulation of chemokine genes, indicating sustained microglial inflammatory programming during the chronic post-stroke phase [54]. Another previous study identified a similar pro-inflammatory microglial cluster at the acute stage (24 h post-MCAo) [45], suggesting that this inflammatory phenotype emerges early after stroke and persists into the chronic phase, potentially contributing to prolonged neuroinflammation. Furthermore, consistent with previous findings, microglia exhibited robust interferon-stimulated gene (ISG) signatures at both D7 and D14 after stroke [43, 44]. Interestingly, we noted that cells in cluster 5 were annotated as microglia but did not exhibit tdTomato labeling. This cluster accounted for 3.9% of the total population and was present in both sham and stroke groups. The abundance of cluster 5 decreased at D7 but recovered by D14 and most of cells in cluster 5 are in Sham group, suggesting that these cells represent a homeostatic microglial subset. It is not clear why this cluster of microglia shows low tdTomato transcript counts while almost 99.93% of IBA1+ microglia shows tdTomato immunoreactivity in the brain. The low tdTomato mRNA transcripts in this cluster may be due to the overall low number of transcripts sequenced in this cluster. Leveraging the Tmem119CreER-Ai14 mouse model enabled the identification of a distinct Cd93-expressing microglial cluster (Mg 7), which emerged specifically after stroke (D7 and D14) but not in sham controls. This population has been largely unexplored in the context of stroke in previous research. CD93⁺ iMM were markedly accumulated in the infarct core on D7 after stroke. We also found CD93⁺/P2RY12⁺/tdTomato⁻ cells in the infarct core on D3, D7, and D14 but disappeared on D30, suggesting that CD93⁺ iMM might transiently upregulate P2RY12 but does not survive for a prolonged time after stroke. A recent study demonstrated that monocyte-derived microglia exhibited high levels of CD93 expression in microglia-ablated neonatal mice that received adult monocytes transplantation [61], suggesting that microglial loss in neonates and in adults creates a niche that allows circulating CD93+ monocytes to colonize the brain. Functional analyses have shown that CD93 downregulation in primary microglia significantly impairs microglial phagocytosis of synaptosomes [73], and Cd93-deficient (Cd93⁻/⁻) mice develop exacerbated neuroinflammation in both the brain and spinal cord during experimental autoimmune encephalomyelitis (EAE) [74]. However, CD93 is also highly expressed by endothelial cells, and CD93 deficiency may lead to blood–brain barrier disruption and exacerbate inflammation [73]. Therefore, the specific role of CD93 in microglia after stroke requires further investigation. Similar to the previous study, iMM progressively lost monocytic marker genes and acquired transcriptomic characteristics of tissue-resident macrophages [44]. We observed a time-dependent shift in the transcriptional profile of iMM following stroke: metabolic activation dominated on D7, while interferon-stimulated gene expression became prominent on D14, indicating a dynamic transition in their functional states. Furthermore, due to the overlapping gene expression profile in microglia vs. iMM in the injured brain, distinguishing these two cell populations has presented a challenge for the field and led to conflicting results in various neurological conditions [20–25]. We have identified that TMEM119CreER/CreER-Ai14 mouse line is capable of pre-labeling almost 100% of microglia in the brain, which offers a unique advantage for the precise identification and characterization of microglia, enabling the distinction of resident microglial populations from infiltrating monocytes with high fidelity. Our data revealed that a small subset of markedly activated microglia (tdTomato⁺ cells in cluster iMM4) was automatically annotated with iMM in the initial cluster (Cluster 3), exhibiting a highly similar gene expression profile with iMM. tdTomato+ cells in this cluster expressed the canonical lineage marker for microglia Sall1, while other cells (tdTomato−) in cluster 3 are Sall1−. Therefore, we termed this cluster of tdTomato+ cells iMM-like microglia. When comparing this iMM-like microglia cluster (tdTomato+, activated microglia that upregulate some iMM markers) vs. true iMM (tdTomato− cells) in cluster 3, we observe very distinct marker genes and distinct enriched gene functions: iMM-like microglia are enriched for gliogenesis and crosstalk with leukocytes, while iMM is enriched for metabolic remodeling. Cell–cell communication analysis further demonstrated that these iMM-like microglia exhibited distinct interaction patterns with other immune cells and EA compared with iMM. This exciting data suggests that, despite sharing certain gene expression similarities, activated microglia and iMM exhibit distinct transcriptomic signatures and functional pathways, indicating that they may perform divergent biological roles and could be differentially targeted to modulate stroke recovery and outcome. Based on DEGs analysis between tdTomato⁺ iMM-like microglia and tdTomato⁻ true iMM, we identified Gm21188, a specific gene maker of activated iMM. We compared the expression pattern of this novel activated iMM markers across several independent brain scRNA-seq datasets and found striking consistency in the unique enrichment of these markers in either microglia or iMM in all datasets examined [44, 64, 65]. A recent study also reported that Gm21188 was highly expressed in monocyte-derived BAM in BAM-ablated mice compared with classical BAM, suggesting that Gm21188 could serve as a potential marker for immune cells originating from monocytes [61]. Gm21188 remains a poorly characterized gene and was initially annotated as annexin A2 receptor 1 (Anxa2r1) in mice. However, a recent study named this gene RESIST (regulated stimulator of interferon via stabilization of transcript) and proposes RESIST has a novel role of stabilizing interferon-β (Ifnb1) mRNA [75]. Therefore, the specific role of Gm21188 in iMM after stroke warrants further investigation. As a promising future direction, genetic lineage tracing using Gm21188 knockin reporter or creER driver could potentially enable the labeling and tracking of iMM in the CNS under conditions where iMM infiltrations are implicated. Experimental and recent clinical evidence highlights the protective role of systemic IGF-1 in ischemic stroke. Elevated serum IGF-1 levels are associated with reduced stroke risk [76], and IGF-1 delivered through various routes (e.g., cerebroventricular, intravenous, and intranasal) has been shown to attenuate neuronal death and enhance recovery after ischemic injury [77, 78]. A recent clinical trial testing the IGF-1–based agonist SCP776 delivered during the acute stage of stroke systemically demonstrated early signs of neurological improvement in patients with large-vessel occlusion stroke treated within 24 h [68]. However, how chronically elevated IGF-1 in the CNS affects later strategies of stroke pathology is not known and needs further investigation to direct clinical strategies. Our scRNA-seq analysis shows both activated microglia and iMM upregulate Igf1 at D7 and D14 post stroke; therefore, we investigated the effects of abolishing this upregulation during the later stages of stroke. Our findings in Cx3cr1ERT2 Igf1 post-stroke iKO mice revealed enhanced sensorimotor function recovery post-MCAo, suggesting that inhibition of this signaling pathway during the subacute/chronic stage might be beneficial despite the beneficial effects of boosting this signaling pathway during the acute stage of stroke. Conversely, Igf1 iKO after stroke also leads to further exacerbated anxiety-like behavior in stroke mice, further emphasizing the complex role of this signaling pathway in stroke pathology. These apparently divergent effects may reflect distinct circuit- and region-specific roles of IGF-1 signaling in the post-stroke brain. In particular, recovery of sensorimotor function is primarily associated with cortical and striatal plasticity and network reorganization [79, 80], where reduced IGF-1 signaling in microglia/iMM may facilitate functional reorganization of motor circuits. In contrast, anxiety-like behaviors are more closely associated with limbic circuits, including the hippocampus and amygdala, which are highly sensitive to stress-related neuroplasticity and immune modulation. IGF-1 signaling has been shown to support hippocampal trophic function and synaptic plasticity [81, 82], and may therefore contribute to the regulation of affective behaviors. Therefore, the opposing behavioral phenotypes may reflect differential effects of IGF-1 signaling across neural circuits rather than a uniform systemic action. IGF-1–mediated neurogenesis has been well documented in both the developing and injured brain [83–87]. A recent study further revealed a specific role for microglia in IGF-1–mediated neurogenesis within the human medial ganglionic eminence (hMGE), whereas microglia appeared to play no such role in neurogenesis in the mouse brain [70]. Consistently, we observed no significant difference in neurogenesis between stroke and Sham groups in Cx3cr1ERT2 Igf1 iKO mice, suggesting that microglial IGF-1 may not play a major role in regulating mouse hippocampal adult neurogenesis. Instead, Cx3cr1ERT2 Igf1 iKO stroke group exhibited reduced numbers of CD8⁺ T cells within the infarct core at day 30 post-stroke, suggesting that microglial and iMM IGF-1 signaling might play an important role in mediating chemoattraction or survival of CD8+T cells in stroke brain. Our LIANA analysis suggests that Igf1⁺ myeloid cells promote CD8⁺ T-cell infiltration not through a single pathway, but via coordinated enhancement of chemotaxis, antigen presentation, adhesion, and extracellular matrix interactions. Interestingly, enforced IGF-1 signaling has been shown to restrict leukemia stem cell activity in both mouse and human models of T-cell acute lymphoblastic leukemia (T-ALL) [88]. Therefore, it is possible that Igf1 deficiency in microglia may also similarly constrain T-cell expansion after stroke, potentially by modulating local immune activation or the metabolic microenvironment. Our findings may have important implications for the development of cell type– and time-specific therapeutic strategies for human stroke. The distinct temporal dynamics and functional divergence between iMM and resident microglia suggest that these populations represent differential therapeutic targets across disease stages. Specifically, the rapid accumulation and subsequent disappearance of iMM indicate that modulation of peripheral monocyte recruitment and activation may be most effective during the acute phase of stroke. In contrast, the sustained presence and dynamic transcriptional states of microglia during the subacute and chronic phases suggest that targeting resident microglial signaling pathways may be more relevant for long-term tissue remodeling and functional recovery. In this context, our findings on IGF-1 signaling further support a stage-dependent framework, in which transient IGF-1 activity in the acute phase may contribute to neuroprotection, whereas prolonged or chronic IGF-1 signaling in myeloid cells may differentially regulate immune–neuronal interactions, including T-cell infiltration and behavioral outcomes. Together, these data support a temporally staged immunomodulatory strategy, in which therapeutic interventions are tailored to the evolving cellular landscape of the post-stroke brain. In summary, our data emphasize the importance of careful characterization of cell type and temporally specific characterization of the signaling pathway after stroke to better guide treatment strategies. Additionally, TMEM119CreER-Ai14 Reporter line could be a useful tool for the field to clearly delineate the cellular dynamic and transcriptomic profile of resident microglia vs. iMM in a variety of CNS pathological conditions. Conclusions In summary, Tmem119CreER-Ai14 tdTomato reporter mouse line is a useful tool for the field to clearly delineate the cellular dynamics and transcriptomic profiles of resident microglia vs. iMM in a variety of CNS pathological conditions, which could have broad implications beyond stroke studies. Moreover, we have characterized novel subtypes and markers for activated microglia and iMM, as well as a possible novel role for CNS myeloid cell-derived IGF-1 in driving T-cell infiltration and/or survival in the post-stroke brain. Supplementary Information Supplementary Material 1. Additional file 1: Fig. S1-S7. Supplementary Material 1. Additional file 1: Fig. S1-S7. Supplementary Material 2. Additional file 2: Table S1. Quantification of microglia (tdTomato⁺/IBA1⁺) and iMM cells (tdTomato⁻/IBA1⁺) at days 3, 7, 14, and 30 following stroke. Table S2. Differential gene expression (DEG) results from scRNA-seq data of CD45⁺ immune cell populations isolated from the brains of Tmem119CreER-Ai14 mice. Table S3. DEG results from scRNA-seq analysis of tdTomato⁺ microglia subclusters. Table S4. DEG results from scRNA-seq analysis of iMM subclusters. Table S5. DEG results of marker genes distinguishing tdTomato⁺ and tdTomato⁻ cells. Table S6. Quantification of CD8⁺ and CD4⁺ T cells at days 3, 7, 14, and 30 following stroke. Supplementary Material 2. Additional file 2: Table S1. Quantification of microglia (tdTomato⁺/IBA1⁺) and iMM cells (tdTomato⁻/IBA1⁺) at days 3, 7, 14, and 30 following stroke. Table S2. Differential gene expression (DEG) results from scRNA-seq data of CD45⁺ immune cell populations isolated from the brains of Tmem119CreER-Ai14 mice. Table S3. DEG results from scRNA-seq analysis of tdTomato⁺ microglia subclusters. Table S4. DEG results from scRNA-seq analysis of iMM subclusters. Table S5. DEG results of marker genes distinguishing tdTomato⁺ and tdTomato⁻ cells. Table S6. Quantification of CD8⁺ and CD4⁺ T cells at days 3, 7, 14, and 30 following stroke. Supplementary Material 3. Additional file 3: Source of Materials. A complete list of all antibodies, reagents, and commercial sources used in this study, including manufacturers and catalog numbers. Supplementary Material 3. Additional file 3: Source of Materials. A complete list of all antibodies, reagents, and commercial sources used in this study, including manufacturers and catalog numbers.