C1qc Mediates Blood–Brain Barrier Disruption in a Mouse Model of Intracerebral Haemorrhage Through Microglia‐Derived Migrasomes

ABSTRACT

Extracellular vesicles (EVs) are essential for intercellular communication and significantly influence neuroimmune responses after intracerebral haemorrhage (ICH). This study elucidates the immune mechanisms underlying secondary brain injury following ICH by identifying microglia‐derived migrasomes. By integrating immune cell clustering analysis from single‐cell RNA sequencing (scRNA‐seq) with Mendelian randomization, we identified five key genes associated with ICH. Further experiments revealed that C1qc accumulates around hematomas, worsening blood‒brain barrier (BBB) disruption through migrasome‐mediated pathways. Our results showed that microglial migrasomes transport C1qc to induce pyroptosis in vascular endothelial cells, thereby compromising BBB integrity. Genetic ablation of C1qc substantially attenuated this BBB damaging effect, C1qc‐deficient migrasomes exhibited protective properties. Our findings, for the first time, reveal the mechanistic role of C1qc and migrasomes in ICH pathogenesis and establish microglia‐derived migrasomes as promising therapeutic targets, providing novel avenues to improve neurologic outcomes after ICH.

Microglia‐derived migrasomes induce vascular endothelial pyroptosis through C1qc delivery, uncovering a novel mechanism of blood–brain barrier (BBB) disruption after intracerebral haemorrhage (ICH) and proposing a novel therapeutic strategy targeting extracellular vesicles (EVs). Intracellularly, the delivered C1qc triggers the Nlrp3/Caspase‐1/GSDMD pyroptotic cascade. This signalling results in endothelial pyroptosis mediated by GSDMD‐N, ultimately causing the disruption of tight junctions and BBB breakdown.

1. Introduction

Intracerebral haemorrhage (ICH) is a severe type of stroke characterized by high rates of morbidity and mortality. Although ICH accounts for only 20% to 25% of strokes, its age‐standardized death rate is almost equal to that of ischemic stroke, which accounts for 45% of all strokes (Collaborators 2024). Compounding this clinical challenge, efficacious therapeutic strategies that can substantially improve ICH outcomes remain notably absent (Guo et al. 2024; Ronaldson and Davis 2020). Increasing evidence indicates that complex interactions among oedema, inflammation, iron‐induced injury, oxidative stress, and excessive immune activity contribute to secondary brain damage following ICH (Xiang et al. 2025). Central to these processes is the disruption of the blood‒brain barrier (BBB), a critical event that amplifies secondary injury and substantially worsens clinical outcomes (Betterton et al. 2021; Profaci et al. 2020).

Immune responses constitute a pivotal determinant of secondary brain injury following ICH, with dysregulated immune activity exacerbating BBB breakdown and perpetuating neuroinflammation. Among the immune cells engaged in post‐ICH responses, microglia—the brain's resident innate immune cells—are among the first responders to haemorrhagic injury and exert multifaceted effects on BBB integrity (Ronaldson and Davis 2020; Xue and Yong 2020).

Migrasomes are defined as a newly discovered category of extracellular vesicles (EVs) generated during cell migration, with immune cells serving as principal producers (Ma et al. 2015). Studies have demonstrated that neutrophils release migrasomes to eliminate damaged mitochondria, thereby supporting cell survival (Jiao et al. 2021; Zheng et al. 2023). Furthermore, macrophage‐derived migrasomes encapsulate complement molecules that have a pivotal role in advancing cerebral amyloid angiopathy (CAA) (Hu et al. 2023). Although macrophages are established migrasome producers (Schmidt‐Pogoda et al. 2018), the physiological functions and pathological mechanisms of migrasomes derived from microglia, the CNS resident macrophages, remain unexplored.

Pyroptosis is a highly inflammatory form of programmed cell death that plays a critical role in the secondary brain injury following ICH. The canonical pyroptosis process is primarily orchestrated by the NLRP3‒Caspase‐1‒GSDMD pathway (Shi et al. 2015; Swanson et al. 2019). Upon sensing damage‐associated molecular patterns (DAMPs), the NLRP3 inflammasome assembles and activates Caspase‐1. Active Caspase‐1 subsequently cleaves Gasdermin D (GSDMD) to release its N‐terminal domain (GSDMD‐N), which oligomerizes and forms pores in the plasma membrane, ultimately leading to cell swelling, membrane rupture, and the massive release of pro‐inflammatory cytokines. However, whether this canonical pathway is directly triggered by microglial‐derived migrasomes in brain endothelial cells remains largely unknown.

Mendelian randomization (MR) leverages single‐nucleotide polymorphisms (SNPs) as instrumental variables to draw causal connections between exposures and outcomes (Plomin and Haworth 2010; Tin and Köttgen 2021; Smith 2010). Integrating single‐cell RNA sequencing (scRNA‐seq) with MR analysis of genetic and plasma metabolomic data offers a powerful approach for identifying disease‐relevant molecular targets. However, studies employing such integrated strategies to perform immune‐focused single‐cell sequencing in ICH models and identify key factors associated with post‐ICH brain injury remain absent.

In the present study, we integrated scRNA‐seq with MR analyses to identify C1qc as a novel susceptibility gene implicated in ICH. Using ICH mouse models and cellular models, we demonstrate that microglia generate migrasomes that carry the complement component C1qc. These microglia‐derived migrasomes attach to cerebrovascular endothelial cells, triggering pyroptosis and disrupting the BBB. Targeting C1qc attenuates BBB damage following ICH, thereby offering potential therapeutic strategies and prognostic biomarkers for ICH management.

2. Materials and Methods

2.1. Ethics

The animal research proposal bearing the identifier 2021‒KY‒0156‒002 was approved by the First Affiliated Hospital of Zhengzhou University's Ethics Committee. Zhengzhou University and its Institutional Review Board granted ethical approval for the human study (Approval No. 021‒KY‒0156), which adhered to the Declaration of Helsinki guidelines. Informed consent was acquired from all participants or their legal representatives.

2.2. Patient Selection

Human brain samples, comprising five ICH cases and five normal controls, were obtained from the Neurosurgery Department at the First Affiliated Hospital of Zhengzhou University. Perihaematomal brain tissues were collected during minimally invasive surgery from five ICH patients whose basal ganglia hematomas exceeded 40 mL. Control samples were obtained from nonfunctional regions during surgical resection for deep‐seated meningiomas. The control tissue resided in normal, nonfunctional brain parenchyma and did not originate from the tumour margin or adjacent areas. It was generated as a necessary byproduct of the specific surgical corridor created to access a deep‐seated meningioma. The demographic details of the cohort are outlined in Table S1.

2.3. Animals

Adult male C57BL/6 J mice (8–10 weeks, 20–25 g) were obtained from SPF Biotechnology Co., Ltd. (Beijing), and C1qc^flox/+ mice were obtained from Cyagen Company (Suzhou, China). Under specific pathogen‐free (SPF) conditions, the animals were maintained at 23°C–25°C with 45%–55% humidity. They followed a regular day‒night pattern with 12 hours of light and 12 hours of darkness, and had unrestricted access to food and water. In the migrasome transfer experiments, microglia‐derived migrasomes were injected into the striatum of C57/BL6 mice, which were subsequently euthanized on day 2. Every aspect of these experiments adhered strictly to both national regulations and the ARRIVE guidelines, while group assignments for mice were determined using an online randomization tool.

2.3.1. ICH Mouse Model

After administering isoflurane anaesthesia (70% N₂O and 30% O₂; 4% induction, 2% maintenance) to male C57BL/6 mice, the subjects were carefully placed in a stereotaxic frame. Subsequently, a 1–1.5 cm surgical incision was made along the sagittal suture. A micropipette crafted from glass and containing paraffin wax was linked to a microinfusion pump, which allowed for the delivery of collagenase VII‐S (Sigma, St. Louis, MO) into the left striatum via a dual‐arm brain stereolocator positioned (77001S, RWD, Shenzhen, China) at 0.3 mm anterior to the bregma, 2.3 mm lateral to the midline, and 3.6 mm beneath the skull's surface. A microinfusion pump (R‐480; RWD; Shenzhen, China) delivered 0.4 µL of collagenase VII‐S (concentration: 0.15 U/µL; total dose: 0.06 U) was injected into the striatum at a rate of 0.1 µL/min. Throughout the surgical procedure, we utilized a body temperature monitoring system (RWD; Shenzhen, China) to keep core temperature stable at 37.0 ± 0.5°C. The needle was left in place for a full ten minutes before being slowly and carefully removed. To close the surgical site, we applied bone wax to seal the burr hole and then sutured the incision. All animals were kept under close watch until they regained full consciousness following the anaesthesia. For the control group, the surgical protocol was identical except they did not receive the collagenase treatment. Any mice that failed to survive the first 24‐h postoperative period were not included in our final analysis. Our primary evaluation was conducted three days after the procedure.

2.3.2. Tissue Collection

Mice were chosen for this study according to their neurological deficit scores. A qualifying mouse from each group was anaesthetized, and its brain was collected for preservation or experimentation. We collected sections of the brain tissue containing the striatal haemorrhage area and tissue samples from a region of 1–2 mm surrounding the hematoma from the ICH group. For the sham group, we collected normal striatal tissue from the corresponding plane of the sections to ensure consistency in the sampling areas across all groups.

2.4. scRNA‐seq Analysis of Mouse Tissues and Screening of CD45‐Positive Immune Cells

The tissue was freshly harvested from the mice with a scalpel. After rinsing the cells a couple of times with sterile, ice‐cold RPMI 1640 medium fortified with 0.04% BSA, the tissue was diced into roughly 0.5 mm³ chunks using surgical tools. These pieces were then transferred to a brand‐new enzyme cocktail and left to marinate in a 37°C incubator for half to an hour, giving the mixture a good shake every 5 min. Cell suspensions were twice filtered through a BD 40 µm sieve at 4°C, then centrifuged at 300 × g for 5 min. The sediment was brought back into suspension using the medium, then combined with an identical portion of red blood cell lysis solution (MACS, 130–094–183). Following a 10‐minute chill session at 4°C, the mixture was spun down at 300 × g for 5 min, after which the liquid portion was carefully poured off. The precipitates were subsequently washed with medium and centrifuged at 300 × g for 5 min. The single‐cell suspension was stained with CD45 antibodies and fluorescein diacetate for FACS sorting (FDA; Cat# F1303, SH800S Cell Sorter), resulting in the isolation of live CD45+ immune cells. The sedimented cells were reintroduced into 100 µl of medium, then quantified and evaluated for viability via Luna Cell Counter analysis. Samples whose activity was greater than 85% were subjected to scRNA‐seq library preparation. The single‐cell suspension was adjusted to 700–1200 cells/µl for oil‐in‐water preparation and cDNA library amplification using the 10X Genomics Chromium Next GEM Single‐Cell 5ʹ Reagent Kit (v2.0.1000165) as per the manufacturer's guidelines. A DNA library was created with the Chromium single‐cell 3'/5' kit (1000020) and sequenced on the Illumina Nova 6000 in PE150 mode.

2.5. scRNA‐seq Data Analysis

2.5.1. Quality Control

The expression profiles underwent analysis using the Seurat package, where we implemented filters based on each cell's total UMIs, the gene count, the respective percentages of mitochondrial and ribosomal gene activity. The ratio of cells displaying mitochondrial and ribosomal genes reveals their proportion in the grand scheme of total gene expression. Cells with heightened levels of these genes usually exhibit lower RNA levels, suggesting they're on the brink of apoptosis. To maintain accuracy, we applied the median absolute deviation (MAD) method for quality control. Any value more than three MADs away from the median is generally considered an outlier and is therefore excluded. We used DoubletFinder (version 2.0.4) to eliminate any potential doublet cells in each sample, effectively wrapping up the cell quality assurance process.

2.5.2. Data Standardization

We applied the LogNormalize method to scale total cell expression to 10,000 using a coefficient s0, then applied logarithmic transformation to normalize the data. To assess cell cycle progression, we utilized CellCycleScoring to derive respective phase scores, and leveraged FindVariableFeatures to pinpoint genes exhibiting high variability. The ScaleData function was used to combat the inconsistencies in gene expression levels, which stemmed from disparities in the ratios of mitochondrial and ribosomal gene expressions and the fluctuations of the cell cycle. RunPCA was employed for linear dimensionality reduction and principal component selection of the expression matrix. We leveraged harmony to address batch effects, and then applied RunUMAP for non‐linear dimensionality reduction through the unified manifold approximation and projection technique (UMAP).

2.5.3. Analysis of Ligand‐Receptor Interactions Using CellChat

CellChat is a quantitative instrument designed to deduce and dissect intercellular communication networks from single‐cell data. It employs network analysis and pattern detection to pinpoint key cellular signals both incoming and outgoing, as well as their functional coordination. In this study, standardized single‐cell expression profiles were used as primary dataset, and cellular subtypes from scRNA‐seq were used as cellular information. We examined the interactions between cells and utilized the connection strengths (weights) and frequency (counts) to measure how closely the cells interacted, which enabled us to track the functional dynamics and impact of each cellular subset within the pathological context.

2.5.4. GO and KEGG Functional Analyses

Genes with varied expression were meticulously annotated using the Metascape database (find it at www.Metascape.org) to delve deep into their functional connections. By performing Gene Ontology (GO) pathway analysis on these specific genes, we were able to map out their participation in key biological processes and molecular functions. Statistical significance was defined as a minimum overlap of at least 3 and a p value of 0.01 or less.

2.6. Mendelian Randomization Analysis

The MR data collected included (1) self‐test scRNA‐seq data; (2) exposure expression quantitative trait loci (eQTL) data from the eQTLGen consortium database; and (3) outcome data from outcome‐related GWAS participants, primarily of European ancestry. The outcome summary data were obtained from the EBI database (ebi–a–GCST90018870). Our study analysed 1,935 ICH patients and 471,578 controls. SNPs achieving locus‐wide significance (P < 1e–8) per gene were selected as candidate instrumental variables (IVs). Linkage disequilibrium (LD) between SNPs was assessed, retaining only variants with LD at R2< 0.001 within a 10,000 kb window and p2< 5e‐8. Sequentially, the process incorporates Inverse Variance Weighting (IVW), MR Egger, Weighted Median method and the Weighted Mode. IVW is a technique that employs a meta‐analysis approach alongside the Wald estimate for each SNP. The central tenet of the Egger approach posits that the strength of instrumental variables is unrelated to direct effects—this assumption is referred to as INSIDE. Weighted Median method can accurately gauge causality in up to 50% of scenarios where instrumental variables are not valid. The Weighted Mode, on the other hand, boasts enhanced detection of causal effects, less bias, and a reduced Type I error rate compared to the MR‐Egger approach. This amalgamation of statistical techniques serves to assess the dependability of causal links. Ultimately, the identified causal associations are validated and dissected using the leave‐one‐out verification process.

2.7. Cell Culture and In Vitro Model

BV2 microglial and bEnd.3 cell lines were sourced from the American Type Culture Collection. They were nurtured in DMEM supplemented with 10% fetal bovine serum, and were also treated with 1% penicillin and streptomycin. These cell cultures were kept under incubation at 37°C with a 5% CO₂ environment to ensure optimal growth conditions.

2.7.1. Hemin Exposure

Hemin (100 µM; Sigma, USA) was incorporated into the culture medium of BV2 cells. Migrasomes were extracted after 12 h.

2.7.2. SiRNA

For knockdown experiments, BV2 cells were transfected with siRNAs (GenePharma, Shanghai, China) using the Neon NxT Electroporation System (Thermo Fisher Scientific, USA) at 1100 V, 10 ms, and 4 pulses, following the manufacturer's guidelines. Si‒C1qc cell samples were harvested for RT‒qPCR or Western blot analysis after 24 hours of culture.

siRNA #1 sense GCCUCUACUACUUCGUCUATT

and antisense: UAGACGAAGUAGUAGAGGCTT;

siRNA #2 sense GCCGAUACAAACAGAAGCATT

and antisense UGCUUCUGUUUGUAUCGGCTT;

siRNA #3 sense GACCACAUGUUCAACAGCATT

and antisense UGCUGUUGAACAUGUGGUCTT;

siNC sense UUCUUCGAACGUGUCACGUTT

and antisense ACGUGACACGUUCGGAGAATT.

2.7.3. Drug Administration

Ten milligrams of the inflammasome inhibitor MCC950 (Nlrp3 inhibitor)(Selleck Chemicals, Houston, TX, USA) was dissolved in 1 mL of 1× phosphate‐buffered saline (PBS). MCC950 was diluted in cell culture medium to reach a 100 µM concentration. Following a 60‐minute incubation with diluted MCC950, microglia‐derived migrasomes were added into bEnd.3 cells.

Pyroptosis Inducer—Nigericin (Nlrp3 activator) (HY‐127019; MCE Med‐Chem Express, USA). Nigericin was diluted in cell culture medium to reach a 20 µM concentration. Following a 2 h incubation with diluted Nigericin, subsequent detection was performed.

2.8. C1qc Knockdown in Brain Tissue

2.8.1. Plasmid Construction

An adeno‐associated virus vector (AAV9–C1qc–RNA) containing a short hairpin RNA targeting the C1qc gene sequence (GCCGATACAAACAGAAGCACC) was synthesized and cloned, and inserted into the GV478 vector at the BsmBI site, after which the recombinant vector was confirmed by sequencing. This vector was obtained from Shanghai Genechem Co., Ltd.

2.8.2. Adeno‐Associated Virus (AAV) Production

293T cells received viral vector transfection via Lipofectamine 2000 (Invitrogen; Thermo Fisher Scientific). Adeno‐associated viruses were harvested at 72 h after transfection, through iodixanol gradient ultracentrifugation AAV9 was purified. AAV virus titres were assessed via a PCR‐based quantitative analysis method. The AAVs in this study were prepared in a 0.001% Pluronic F‐68 solution (Poloxamer 188; Caisson Laboratories, Smithfield, UT, USA).

AAV#1: GCCGATACAAACAGAAGCACC

AAV#2: GGCCTCTACTACTTCGTCTAC

AAV#3: GGTGTGGCTATCAGTCAATGA

AAV con: CGCTGAGTACTTCGAAATGTC

2.8.3. AAV Inhibition of C1qc Expression in Microglia

Male C1qc^flox/flox mice, aged 8–9 weeks, were anaesthetized and positioned in a stereotaxic frame. The microglia‐specific Cre viral vector employed in this study was designated GV884: Iba1–MCS–3flag–FT2A–miR–9.T×4–miR129–2–3p. T–SV40pA and was of serotype 11, which was procured from Shanghai Genechem Co., Ltd. AAV11–Iba1–Cre (AAV–Iba1‒Cre) and AAV11–Iba1–Empty (AAV‐con) were injected into the left striatum hemisphere of C1qc^flox/flox mice using a microdispenser. The injection coordinates matched the coordinates of the basal ganglia as previously described. As the pipette retracted, it was halted at 0.5 mm intervals to facilitate virus deposition. At each pause, the viral solution was administered gradually into the tissue samples, with 150 nL dispensed over a period of 5 min at every stop. Two weeks post‐virus injection, the mice were utilized in subsequent experiments.

2.9. Neurobehavioral Tests

A treatment‐blinded experimenter evaluated the mice's neurological function before and three days post‐ICH. The Neurological Deficit Scores evaluate bodily symmetry, gait, climbing abilities, circular movements, forelimb symmetry and compulsive circular behaviour. Forelimb use asymmetry assessment measured the percentage of limb movements by the unimpaired forelimb (represented by I), the contralateral forelimb (represented by C), and both forelimbs (represented by B). The asymmetry score = (I − C)/(I + C + B). During the corner turn experiment, the mice were directed into a 30° corner, and their turning direction was recorded over 20 trials, with the results shown as the percentage of left turns. Rotarod Test: During the testing phase, mice were placed on the rotating rod, and the speed was linearly accelerated from 4 to 40 rpm over a period of 5 min. The trial ended when the mouse fell off the rod or when the maximum time of 300 s was reached. If a mouse clung to the rod and completed a full passive rotation without walking, it was also considered a fall. The latency to fall was recorded, and the average time of 3 trials was calculated for statistical analysis. Foot‐fault Test: Mice were placed on an elevated wire grid apparatus 30 cm × 30 cm, with 1 cm × 1 cm and allowed to move freely for 5 min. A video camera was placed below the grid to record their movements. A foot‐fault was defined as a misstep where the mouse's paw slipped through the grid opening instead of accurately grasping the wire. The total number of steps and the number of foot faults for the limbs were counted. The severity of the motor deficit was expressed as the percentage of foot faults, calculated using the following formula: (Number of foot faults/Total number of steps) × 100%. To eliminate investigator bias, all behavioural tests and data analyses were conducted in a double‐blinded manner. Briefly, mice were randomized into different groups and assigned encoded identification numbers. Two investigators, who were strictly unaware of the allocation and treatment history, performed all behavioural testing and scoring.

2.10. Haemorrhage Volume and Brain Water Content

A validated method was applied to assess intracranial hematoma magnitude and brain swelling. The hematoma area determined from mouse brain sections was multiplied by the sectional thickness to determine the volume per slice; hematoma volume was then obtained by summing the volumes from all the slices. Brain water content = (wet weight − dry weight)/wet weight × 100%.

2.11. Enzyme‐Linked Immunosorbent Assay (ELISA)

ELISA kits were utilized as per the manufacturer's guidelines. Serum samples from human and mouse blood were isolated by 20‐min centrifugation at 4°C, followed by supernatant analysis. C1qc levels were determined using ELISA kits (cat. #RX106505H, RX2DW249666, Ruixin, China).

2.12. Isolation of Migrasomes

Cells were incubated in fibronectin‐treated plates with microglial medium. After treatment, the cells and migrasomes underwent trypsin digestion for subsequent collection. Following centrifugation at 1000 × g (10 min) and 4000 × g (20 min) at 4°C, cell and debris removal was achieved. Migrasomes were collected at 20000 × g for 30 min, PBS washed, and re‐pelleted twice under identical conditions. Blood samples were collected into anticoagulant tubes. The samples were centrifuged at 1000 × g for 10 min, followed by centrifugation at 4000 × g for 20 min to remove large debris. Subsequently, the samples were centrifuged at 20000 × g for 30 min to collect the crude migrasome pellet. The crude migrasome pellet was washed with PBS and then centrifuged again at 20000 × g for 30 min; this washing step was repeated twice. Subsequently, the resulting pellet was resuspended in fresh, Hemin‐free complete medium to enable further analysis or in vitro/in vivo application. Migrasome protein content was measured using a BCA assay to determine dosing. A Litesizer 500 was used to analyse the particle diameter and zeta potential.

2.13. Ultrasonic Disruption of the Migrasome Structure

Ultrasound at 600 W with a 3‐s on/off cycle for 10 min disrupted the intrinsic structure of the migrasomes. Despite this disruption, the protein and RNA contents of the migrasomes remained intact throughout the ultrasonic procedure. These preserved protein and RNA samples were subsequently utilized for further experimental analyses.

2.14. Real‐Time qPCR

Samples were collected from perihaematomal regions from the ipsilateral hemispheres. We utilized RNA extraction kits (Vazyme, China) to extract the mRNA, which was subsequently reverse transcribed into cDNA. The resulting cDNA was then analysed through RT‒qPCR using SYBR Green mix on a Real‐Time PCR System (RWD, CFXDUET, China). Data expression was normalized using GAPDH, with primer details provided in Table S2.

2.15. Western Blotting

The mice received isoflurane anaesthesia followed by cold PBS transcardial perfusion. Perihaematomal regions from the ipsilateral hemispheres were collected. Tissue lysates and migrasome samples were prepared by homogenization in RIPA buffer. Protein concentration was measured via BCA analysis. Protein samples were separated using SDS‐PAGE, then transferred onto PVDF membranes. Molecular weight markers: Multicolor Prestained Protein Ladder (10‒250 kDa, Epizyme Biotech, China) for Figure 4G and Figure 5L; The PageRuler Prestained Protein Ladder (10‒180 kDa, Thermo Fisher Scientific, USA) for other blots shown. To prevent cross‐reactivity, the membranes were blocked with a 5% nonfat milk solution. After blocking, they were subjected to an overnight incubation at 4°C with the primary antibodies. Subsequently, the membranes were exposed to secondary antibodies for a duration of 1.5 h at room temperature. Immunoblot detection utilized an ECL kit, with band intensity quantified by ImageJ and normalized to GAPDH levels. Table S3 displays antibody details.

FIGURE 4.

C1qc adheres to brain blood vessels by microglia‐derived migrasomes and contributes to BBB disruption after ICH. (A) Microglia‐derived migrasomes were detected in brain tissues from ICH model mice, and the brain tissues of sham and ICH model mice were subjected to TEM. Images of representative migrasomes detected in the two groups: red indicates the blood vessel lumen, yellow indicates microglia, and green indicates migrasomes. Microglia were identified based on standard ultrastructural criteria, including an electron‐dense cytoplasm, irregular nuclei with marginated heterochromatin, and the presence of lysosomes/phagosomes. (B) Representative immunofluorescence co‐staining images and quantification analyses of TSPAN4 (a migrasome marker, red) and C1qc (white) in the brains of Sham and ICH groups. The quantitative data demonstrate the total number of migrasomes per filed and the fraction of C1qc‐positive migrasomes (C1qc + TSPAN4 + ratio) (n = 5). Data shown represent mean ± SD with significance levels; ****p < 0.0001. (C) Staining and fluorescence colocalization analysis of the vascular endothelial cell marker CD31 and the migrasome markers TSPAN4 and C1qc. Sham: Colocalization coefficient (CD31 + TSPAN4) = 0.108. ICH: colocalization coefficient (CD31 + TSPAN4) = 0.7243; colocalization coefficient (C1qc + TSPAN4) = 0.7071. (D) Staining and fluorescence colocalization analysis of the microglial marker IBA1 and the migrasome markers TSPAN4 and C1qc. Sham: colocalization coefficient (IBA1 + TSPAN4) = 0.09. ICH: colocalization coefficient (IBA1 + TSPAN4) = 0.496; colocalization coefficient (C1qc + TSPAN4) = 0.8047. White box across the merged images in Figure 4C,D indicate the specific regions utilized for the spatial fluorescence intensity profiling. Co‐localization analysis was performed using ImageJ and Prism software. Spatial fluorescence intensity profiles were generated by plotting the grey values along manually drawn lines across the regions of interest (white box). (E) Electron microscopy image of migrasomes extracted from in vitro vehicle or hemin‐treated BV2 cells (Samples a and b). (F) Particle size detection (left) and potentiometric detection (right) of the migrasome samples shown in Figure 4E. (G) The expression of the migrasome markers TSPAN9, TSPAN4, ROCK, and C1qc in the purified migrasomes was assessed by Western blot. (H) Schematic of the in vitro model: Stimulation of BV2 cells with hemin, extraction of BV2‐derived migrasomes, and addition to bEND.3 cells. (I) Staining for the tight junction protein Claudin‐5 in endothelial cells after 12 h of treatment with or without microglial migrasomes (n = 3). DAPI was used to label nuclei. Quantification of the fluorescence intensity of Claudin‐5. Data shown represent mean ± SD with significance levels; ns = no significant difference; *p < 0.05. (J) The cytotoxicity assay, with a verified concentration (40 µg/mL) using the CCK‐8 assay (n = 3). Data shown represent mean ± SD with significance levels; ns = no significant difference; ***p < 0.001. (K)Annexin V/PI dual‐staining flow cytometry was used to quantitatively assess cell death. The statistical graph displays the percentage of Annexin V⁺/PI⁺ cells after various treatments (n = 3). Data shown represent mean ± SD with significance levels; ns = no significant difference; ***p < 0.001. (L) Endothelial cells were stained for the migrasome marker TSPAN4. Phalloidin staining was used to visualize the overall morphology of endothelial cells. The occurrence of migrasome‐attached endothelial cells and the number of migrasomes per cell were assessed (n = 5). The cluster diameters of control (vehicle + microglial migrasomes) and treated (hemin + microglial migrasomes) endothelial cells were also quantified. The data are presented as the mean ± SD with the indicated levels of significance; ***p < 0.001. (M) Endothelial cells were PI and calcein‐AM stained after treatment, as shown in Figure 4H (n = 3). The percentage of injured/dead cells (calcein–AM⁻ PI⁺) was quantified relative to the total number of cells per field. Intact cells are shown in grey and injured/dead cells are shown in red. The data are presented as the mean ± SD, with significance levels indicated. (N) Two days after receiving an injection of BV2 cell‐derived migrasomes (0.5 mg/kg) into the brain striatum, healthy wild‐type C57/BL6 mice were perfused with PBS followed by 4% paraformaldehyde solution. (O) Coronal brain sections were immunostained for CD31 (red) and Claudin‐5 (green) to assess tight junction integrity (n = 5). The data are presented as the mean ± SD with the indicated levels of significance; ***p < 0.001.

FIGURE 5.

C1qc regulates microglial migrasomes involved in BBB injury in ICH model mice with microglia‐specific C1qc knockout. (A) Images of brain histopathology in C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con and statistical analysis of the hematoma volume (n = 6). The data are presented as the mean ± SD; ***p < 0.001; ****p < 0.0001. (B) Statistical analysis of the brain water content in C1qc^flox/flox model mice injected with AAV‒Iba1–Cre or AAV–con (n = 6). The data are presented as the mean ± SD; ***p < 0.001. (C) Statistical analysis of the neurological deficit scores of C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con (n = 6). The data are presented as the mean ± SD; ***p < 0.001; ****p < 0.0001. (D) Percentage of left turns in the corner test for C1qc^flox/flox model mice injected with AAV‒Iba1–Cre or AAV–con (n = 6). The data are presented as the mean ± SD; **p < 0.01; ***p < 0.001. (E) Percentage of forelimb use asymmetry in C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con (n = 6). The data are presented as the mean ± SD. **p < 0.01; ***p < 0.001. (F) Images of brain histopathology and Evans blue extravasation in C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con after a tail vein injection of Evans blue. (G) Spectrophotometric readings at 620 nm were measured to assess Evans blue leakage in the samples shown in Figure 5G (n = 6). The data are presented as the mean ± SD; ****p < 0.0001. (H) IgG staining in C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con. DAPI staining was employed for nuclear identification. (I) Measurement of IgG Fluorescence Intensity. The data are presented as the mean ± SD; ns, no significant difference; *p < 0.05; **p < 0.01; ***p < 0.0001. (J) Electron microscopy visuals depicting tight junction proteins within the perihaematomal zone of ICH mouse model. The area marked with the red triangle is magnified. (K) Immunofluorescence staining for Claudin‐5, Occludin or ZO‐1, and the endothelial cell marker CD31 in the perihaematomal area of C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con (n = 5). Claudin‐5, Occludin, or ZO‐1 fluorescence intensity was quantified relative to CD31 area. The data are presented as the mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (L) WB assessment of Claudin‐5, Occludin, and ZO‐1 protein levels shown in Figure 5A (n = 5).(M) Densitometric quantification of protein levels; the data are presented as the mean SD; **p < 0.01, ***p < 0.001, ****p < 0.0001. (N) Schematic diagram of the in vitro cell model. (O) Claudin‐5 staining of endothelial cells that were treated as described in Figure 5N (n = 3). Quantification of the fluorescence intensity of Claudin‐5. The data are presented as the mean ± SD. *p < 0.05; **p < 0.01.

2.16. Immunofluorescence Staining

Following transcardial perfusion with ice‐cold PBS, the mouse brains were harvested. These samples were kept in a fixative solution of 4% paraformaldehyde at 4 degrees Celsius overnight, before being progressively dehydrated using 15% and 30% sucrose solutions. Brain samples were incorporated into OCT, frozen, and cut into 10 µm coronal slices with Leica CM 1950 cryostat. Following three 10‐minute washes with PBST, both the brain sections and the cell samples were incubated with 5% fetal bovine serum solution at standard room temp for a full hour. The cell samples were also fixed with 4% paraformaldehyde and then treated with primary antibodies. The subsequent day, the brain tissue sections were subjected to a 1‐hour incubation at ambient temperature in the presence of fluorescent secondary antibodies. Tissue sections were affixed using an antifade mounting medium containing DAPI, followed by microscopic examination with a Leica DMi8 fluorescence microscope (Germany). Multiplex immunofluorescent staining employed kit reagents (Aifang biological, China) for targets detection according to manufacturer protocols. Table S3 displays antibody details.

2.17. Cell Viability Analysis

Endothelial viability was assessed with the LIVE/DEAD viability analysis, using Pl–Calcein–AM–labeled kit (Solarbio, China). The cells were either photographed under a Leica DMi8 fluorescence microscope (Germany).

Cell Counting Kit‐8 (Catalogue No.B34304, Selleckchem, USA) according to the manufacturer's instructions. Briefly, bEnd.3 cells were seeded into 96‐well plates in 100 µL of complete culture medium and allowed to attach overnight. Following the Hemin‐M treatments, 90 µL culture medium and 10 µL of the CCK‐8 solution were added to each well. The plates were then incubated at 37°C in a 5% CO₂ humidified incubator for 2 h protected from light. The absorbance was subsequently measured at 450 nm using a microplate reader (BioTek, USA). Cell viability was calculated as the percentage of the absorbance of the treated cells relative (OD = 450 nm) to that of the control cells.

Cell apoptosis was quantitatively evaluated using the Annexin V‐FITC/Propidium Iodide (PI) Detection Kit (Catalogue No.40302ES50, Yeasen Biotechnology, Shanghai) following the standard protocol. After the treatments, the cells were harvested using EDTA‐free trypsin to prevent false‐positive Annexin V staining, washed twice with ice‐cold PBS, and gently resuspended in 100 µL of Binding Buffer at a concentration of approximately 2 × 10⁵ cells. Next, 5 µL of Annexin V‒FITC and 10 µL of PI were added to the cell suspension. Incubated for 15 min at room temperature in the dark. The samples were immediately analysed using a flow cytometer (BD Biosciences). The data were analysed using FlowJo software (version 10.8.1, Treestar, USA).

2.18. Transmission Electron Microscopy (TEM)

Brain tissues were rapidly dissected within 1–3 min and trimmed into small blocks (<1 mm³). These tissue blocks were immediately immersed in TEM fixative and stored at 4°C for subsequent processing and transportation. Ultrastructure was examined using a Hitachi HT7800/HT7700 transmission electron microscope after the preparation of sections using standard procedures. The samples were negatively stained with 3% phosphotungstic acid prior to TEM observation to visualize the membranes of the isolated migrasomes. In TEM images, microglia were identified based on established ultrastructural morphological criteria. Specifically, microglia were distinguished by their irregular or elongated nuclei with prominent electron‐dense heterochromatin characteristically marginated along the inner nuclear envelope. The cytoplasm of microglia appeared more electron‐dense (darker) relative to adjacent astrocytic processes and neuropil. Additionally, in the context of ICH pathology, these cells frequently exhibited activated features, including the presence of numerous electron‐dense lysosomes and phagosomes.

2.19. Evans Blue Extravasation

A 2% Evans blue solution was prepared by dissolving the dye (Sigma‒Aldrich, USA) in normal saline. Following intravenous injection of 100 µl of Evans blue per mouse, the dye was circulated for 4 hours. The mouse underwent isoflurane anaesthesia before transcardial perfusion with chilled PBS. Following separation of the brains into ipsilateral and contralateral hemispheres, the tissues were homogenized in 50% trichloroacetic acid and centrifuged. The supernatant was diluted 1:4 with ethanol, and EB fluorescence was measured at 620 nm excitation using a plate reader (Thermo Fisher Scientific, USA).

2.20. HRP Extravasation

HRP type II (Sigma‒Aldrich, USA) was intravenously administered at 0.5 mg/g body weight in 0.2 mL of PBS through the tail vein. The brains were collected after 30 min of circulation without transcardial perfusion. They were then cut into 3‐mm‐thick coronal blocks and fixed by immersion in 4% paraformaldehyde. Brain blocks were cryoprotected using sucrose and sectioned coronally at a thickness of 30 µm. To evaluate HRP extravasation, sections received 10‐minute DAB solution (Solarbio, China) incubation at ambient temperature.

2.21. Statistical Analysis

A solid MR analysis stands on three fundamental assumptions: first, the correlation criterion demands that genetic instruments share a tight bond with the exposure while steering clear of direct connections to the outcomes; second, the independence principle insists these instruments remain untethered from any confounding variables that might muddy the waters; and third, the exclusivity provision requires that instruments affect outcomes exclusively by way of their impact on the exposure variable. Gene pleiotropy comes into play when instrumental variables impact results through different routes of influence. The investigation was carried out using R software (V4.3.0.) All data have been presented as mean values accompanied by standard errors of the mean (SEMs).

In this study, we employed either a Student's t‐test or the Mann‒Whitney test for comparing groups. When it came to evaluating differences across three or more groups, we relied on one‐way ANOVA, which was then followed by Tukey's post hoc test to determine statistical significance. We considered a finding statistically significant if it fell below the threshold of p < 0.05.

3. Results

3.1. scRNA‐seq and MR Analysis Identify 5 Key Genes Associated With ICH

To examine immune cell changes after ICH, we collected perihaematomal tissue from mice three days post‐ICH and from sham controls. CD45‐positive cells were sorted by flow cytometry and then subjected to single‐cell RNA sequencing (Figure 1A). Cells with fewer than 500 genes were excluded. Doublets were identified and removed using DoubletFinder, resulting in 9,561 cells. Figure S1A,B show the filtered data via violin and scatter plots. The dataset was normalized, cell cycle scores calculated, and 2,000 hypervariable genes identified (Figure S1C). PCA and UMAP were used to standardize and analyse the data (Figure S1D). UMAP plot of integrated single‐cell data after Harmony batch correction (Figure S1E). The U‐map of CD45 (+) cells in ICH model differs between the sham group, prompting further cluster analysis of these cells (Figure 1B). Ten distinct cellular clusters were defined, based on distinct molecular characteristics (Figure 1C,E). Figure 1D presents the clusters in the brains of sham (left map) or ICH model mice (right map).

FIGURE 1.

scRNA‐seq analysis of the brains of ICH model mice. (A) The workflow of the ICH model, sample collection, cell screening, single‐cell sequencing, and data analysis is illustrated. (B) U‐map of all CD45(+) cells in the brains of the sham and ICH model mice. (C) Cluster analysis of CD45(+) cells in the brains of sham and ICH model mice derived from single‐cell sequencing by Seurat. (D) U‐map of CD45(+) cells in the brains of sham (left map) or ICH model mice (right map). (E) Annotation of CD45(+) cell populations identified using single‐cell sequencing. (F) Cellular interaction networks between the various cell types. (G) The CellChat network's interaction frequency across cell types appears in descending sequence from left to right. (H) The scatterplot visualizes the Mendelian Randomization analysis results for key genes, with various colours representing distinct statistical approaches, while the line slopes illustrate the causal impact attributed to each methodology.(I) Forest plot of the leave‐one‐out analysis of SNPs in key genes. The red line shows the results of the combined analysis.

Utilizing the CellChat software, we analysed ligand‐receptor interactions by integrating single‐cell samples from both sham and ICH model groups, revealing complex interactions among cellular subtypes (Figure 1F,G). Statistical analysis indicated that microglia had higher interaction probabilities with other cell types. We identified the top five genes with significant differential expression for each cell type (Figure S2), using criteria of avg_log2FC > 2 and p_val_adj < 0.05, resulting in 192 unique genes for further analysis. Metascape pathway analysis of these marker genes revealed significant enrichment for pathways related to leukocyte activation, immune response enhancement, and regulation of cellular activation (Figure S3).

To identify the pivotal genes impacting cerebral haemorrhage, we employed the 192 genes from the preceding stage and then retrieved the outcome ID ebi‒a‒GCST90018870 through the summary statistics of 473,513 cases (Controls: 471,578; Cases: 1,935) related to cerebral haemorrhage. We applied the extract_instruments and extract_outcome_data functions in sequence to clarify the causal relationships between genes and outcomes. Using MR analysis (MR Egger, Inverse variance weighted, Weighted median, Weighted mode), we subsequently determined the causal relationships for five gene pairs corresponding to positive eQTLs (IVW p value < 0.05) (Figure 1H). Heterogeneity Assessment and Pleiotropy Assessment results are shown in Table S4. The genes identified were Blk, C1qc, Cytip, F13a1, and Sparc. The genes Blk, Cytip, F13a1, and Sparc were associated with a decreased risk of cerebral haemorrhage, whereas the C1qc gene was associated with an increased risk of ICH. A sensitivity analysis on the five gene pairs showed that removing any single SNP hardly affected the overall margin of error (Figure 1I). The genes Blk, C1qc, Cytip, F13a1, and Sparc were deemed essential for further study.

3.2. The Risk Gene C1qc Shows Increased Expression After ICH

The Seurat R package's Dotplot and FeaturePlot functionalities were applied to evaluate the expression of key genes across separate cell units (Figure 2A,B). Figure 2C shows a mouse brain coronal section to indicate the exact imaging area (the perihaematomal region, marked with red box). Additionally, we validated the expression of key genes in a mouse model of ICH. RT‐qPCR analysis indicated increased mRNA expression of Cytip, C1qc, BLK, and Sparc, and decreased F13a1 expression in the brains of ICH model mice (Figure 2D). Protein expression levels of C1qc and F13a1 were elevated in the ICH model (Figure 2E). C1qc, a gene significantly associated with increased disease risk, showed elevated expression at mRNA and protein levels. Additionally, substantial accumulation of C1qc was observed near the hematoma site in the brain tissue sections three days post‐modelling (Figure 2F). The pronounced propensity for C1qc to accumulate around ICH lesions indicates the potential direct involvement of C1qc in brain damage associated with ICH. To further validate the relationship between C1qc and ICH, we examined brain tissue and serum from ICH patients. The findings indicated significantly elevated levels of C1qc in both brain tissue adjacent to the haemorrhage site (Figure 2H,I) and serum (Figure 2J) in patients with ICH, compared with control groups. Similarly, the serum levels of C1qc were also elevated in mice with ICH (Figure 2G). To definitively determine whether the elevated circulating C1qc originated from secreted migrasomes, we isolated migrasomes from the serum of control and ICH groups. Western blot analysis of these circulating vesicles confirmed the presence of migrasome markers and C1qc. Importantly, the levels of migrasome‐associated C1qc were significantly enriched in the serum of the ICH group compared to controls (Figure S4).

FIGURE 2.

Validation, localization, and expression of key genes. (A) Key gene expression visualization from scRNA‐seq cell data. (B) Bubble visualization of key gene expression across cellular populations based on scRNA‐seq data. In this representation, blue indicates low expression, and red indicates high expression. (C)Schematic diagram of a mouse brain coronal section. The red‐outlined perihaematomal region marks the sampling and imaging area used throughout this study. (D) qPCR measured brain tissue gene expression in sham and intracerebral haemorrhage murine models (n = 5); the data are presented as the mean ± SD; *p < 0.05; **p < 0.01; ****p < 0.0001. (E) Immunoblots showing the expression of key proteins in the brain tissues of sham‐treated and ICH model mice; densitometric quantitation of proteins was performed (n = 3); the data are presented as the mean ± SD; ns, not significantly different; *p < 0.05. (F) Expression and localization of the proteins encoded by key genes in the brain tissues of ICH model mice; DAPI was used to label the nuclei. (G) Serum C1qc levels in ICH model mice (n = 5) are shown as the mean ± SD; ****p < 0.0001. (H) Expression of C1qc in the brain tissue of ICH patients. (I) Densitometric quantification of the proteins in Figure 3G (n = 5); the data are presented as the mean ± SD; **p < 0.01. (J) Serum C1qc levels in ICH patients (n = 5); data are presented as the mean ± SD; ***p < 0.001. (K) Immunofluorescence staining for the neuronal markers NeuN and C1qc in the brains of sham or ICH model mice (n = 6). DAPI was used to label nuclei—quantification of C1qc fluorescence intensity. The data are presented as the means ± SD with the indicated levels of significance; ns, no significant difference. (L) Immunofluorescence staining for the astrocyte markers GPAP and C1qc in the brains of sham or ICH model mice (n = 6). DAPI was used to label the nuclei—quantification of C1qc fluorescence intensity. The data are presented as the means ± SD with the indicated levels of significance, no significant difference. (M) Iba1 and C1qc immunofluorescent staining in sham and ICH model mice brains (n = 6). Quantitative analysis of C1qc fluorescence signal density after ICH in (K‐M), C1qc fluorescence signal co‐localized with respective cellular marker was quantified and normalized by the cellular marker area. DAPI was used to label nuclei—quantification of the fluorescence intensity of C1qc. Data are shown as the means ± SD with significance levels; ***p < 0.001.

To determine the precise localization of increased C1qc expression following ICH, we performed immunofluorescence staining. The results indicated a distinct microglial localization of C1qc in the ICH group (Figure 2M), corroborating our single‐cell sequencing data (Figure 2B). Since Iba1 cannot distinguish between microglia and macrophages, we additionally employed TMEM119, a well‐established microglia‐specific marker, to examine its co‐localization with C1qc. The results showed that C1qc predominantly co‐localized with microglia‐specific marker TMEM119 (Figure S5). In contrast, no significant colocalization was detected with astrocyte markers (GFAP) or neuronal markers (NeuN) (Figure 2K,L). Transcriptional profiling identified three distinct microglial subsets (Microglia 1, 2, and 3) alongside infiltrating macrophages, characterized by specific marker genes (Figure 1E). Microglia 1 was defined by homeostatic markers (P2ry12, Tmem119, Plxdc2), representing the resting state. In contrast, Microglia 2 exhibited a strong disease‐associated activation signature marked by the upregulation of Spp1, Mif, and Pgk1. Microglia 3 highly expresses Clu, Ecrg4, Mt3, and Ttr, potentially representing a distinct reactive phenotype or boundary‐associated macrophage subset. Notably, while C1qc maintained an expression level in homeostatic Microglia 1 and infiltrating macrophages—consistent with its canonical roles—its expression was profoundly upregulated following ICH, reaching a high level of enrichment specifically within the highly activated Microglia 2 population. This preferential enrichment within the pathogenic cluster identifies C1qc as a potentially pivotal mediator of ICH‐induced brain injury. We confirmed this functional significance through further experimental validation, highlighting the transition of C1qc from a homeostatic protein to a key driver of neuroinflammation.

Furthermore, we utilized the AUCell tool to quantitatively evaluate the involvement of genes in immunometabolism‐related pathways at the single‐cell level, and the results indicated C1qc have higher activity in coagulation, inflammatory response and complement pathways (Figure S6). A correlation analysis was then performed to examine the coexpression network between key genes and disease‐related genes. This analysis revealed a significant co‐expression between C1qc and disease‐associated genes (Figure S7A–E).

3.3. C1qc Knockdown Mitigates Brain Injury and BBB Disruption in ICH Mouse Model

To investigate the role of C1qc in brain injury induced by ICH, we used AAV‐shC1qc to interfere with its expression in brain tissue (Figure 3A). The efficiency of AAV‐mediated knockdown of C1qc expression in brain tissue is shown in Figure S8. ICH can cause brain injury and cerebral oedema, a marked reduction in lesion size and a decrease in brain water content were detected after AAV‐shC1qc injected (Figure 3B,C). C1qc knockdown significantly improved ICH‐induced alterations in forelimb use asymmetry, Corner turn Test, neurological deficit scores (Figure 3D–F). Knockdown of C1qc also improved the outcomes of the Rotarod Test and the Foot‐fault Test following ICH (Figure S9). Evans blue (EB) extravasation assays and subsequent staining of brain tissues from ICH mouse model revealed substantial BBB damage. AAV‐shC1qc significantly decreased EB leakage and preserved BBB integrity in the perihaematomal area of mice compared to AAV‐Ctrl (Figure 3G,H). Similarly, IgG staining revealed increased BBB permeability in ICH mice, which was significantly mitigated following AAV‐shC1qc administration (Figure 3I). Horseradish peroxidase (HRP) extravasation in the perihaematomal area was significantly exacerbated post‐ICH but was alleviated in mice treated with AAV‐shC1qc (Figure 3J). These findings indicate that C1qc knockdown alleviates ICH‐induced brain injury, demonstrated by decreased cerebral oedema, improved functional recovery in behavioural tests, and notably reduced BBB disruption.

FIGURE 3.

The knockdown of C1qc alleviates brain injury and mitigates BBB disruption in ICH mouse model. (A)AAV‐mediated knockdown of C1qc and subsequent molecular biology assays. (B) Images of brain histopathology in sham mice and ICH model mice treated with AAV–shC1qc or AAV–Ctrl alongside the statistical evaluation of haemorrhage size. The data are presented as the mean ± SD with the indicated levels of significance (n = 6); ***p < 0.001; ****p < 0.0001. (C) Statistical analysis of the brain water content in sham mice and ICH model mice treated with AAV–shC1qc or AAV–Ctrl (n = 6). The data are presented as the mean ± SD with the indicated levels of significance; ***p < 0.001. (D) Percentage of left turns in the corner test for sham mice and ICH model mice treated with AAV‐shC1qc or AAV‐Ctrl (n = 6). The data are presented as the mean ± SD with the indicated levels of significance; **p < 0.01; ***p < 0.001. (E) Statistical analysis of the neurological deficit scores of sham mice and ICH model mice treated with AAV–shC1qc or AAV–Ctrl (n = 6). The data are presented as the mean ± SD with the indicated levels of significance; **p < 0.01; ***p < 0.001. (F) Statistical analysis of the results of the forelimb use asymmetry test for sham mice and ICH model mice treated with AAV–shC1qc or AAV–Ctrl (n = 6). The data are presented as the means ± SDs with the indicated levels of significance; *p < 0.05; ***p < 0.001. (G) Images of brain histopathology and Evans blue extravasation in sham mice and ICH model mice treated with AAV–shC1qc or AAV–Ctrl after a tail vein injection of Evans blue. (H) Spectrophotometric measurements at 620 nm were performed to assess dye leakage in the samples shown in Figure 3H (n = 6). The data are presented as the mean ± SD with the indicated levels of significance; ***p < 0.001. (I) IgG staining of the brains of ICH model mice and sham mice (n = 6). DAPI was used to label nuclei. The findings from the quantified analysis of the fluorescence intensity of IgG are presented as the mean ± SD with the indicated levels of significance; ns, no significant difference. ***p < 0.001. (J) Images of dye extravasation, the extravasation band width and positive area ratio after the injection of HRP dye into the tail vein of mice treated in the same manner as in Figure 3A. N indicates normal, E indicates extravasation, and H indicates hematoma (n = 6). The findings from the quantified analysis of the HRP extravasation band width and the quantitative analysis of the HRP extravasation‐positive area ratio are presented as the mean ± SD; *p < 0.05, **p < 0.01, and ***p < 0.001.

3.4. C1qc‐Positive Microglial Migrasomes Contribute to BBB Disruption After ICH

BBB disruption represents a critical pathological change in the ICH (Jia et al. 2024; Chen et al. 2024). Ultrastructure alterations around blood vessels in the brain after ICH were investigated with TEM. Our findings showed that the presence of EVs were deposited around the blood vessels adjacent to the hematoma. The observed vesicles range from 0.5 to 3.0 µm in diameter and contain varying numbers of smaller vesicles, aligning with the established morphological characteristics of migrasomes as previously documented in the literature (Hu et al. 2023) (Figure 4A). Quantitative analysis revealed that ICH induced both a robust increase in total migrasome production and a significantly enhanced fraction of C1qc‐positive migrasomes compared to the Sham group (Figure 4B). IF staining revealed more pronounced perivascular deposition of C1qc and migrasomes in ICH mice, with C1qc clearly colocalizing with these migrasomes (Figure 4C). Using IF staining, we visualized the spatial relationships among microglia (Iba1+), migrasomes (TSPAN4+), and C1qc, revealing C1qc within the migrasomes around microglia and its colocalization with migrasomal markers (Figure 4D). Statistical quantification of the Pearson correlation coefficient (PCC) demonstrate a high correlation of C1qc and TSPAN4, supporting our co‐localization conclusions (Figure S10). It should be noted that although C1qc is basally expressed in homeostatic microglia, our imaging parameters were optimized for the exceptionally high fluorescence peak in the ICH group, rendering the inherently weak basal signal in the sham group visually obscure. We treated mice with PLX5622 to eliminate microglia, resulting in the disappearance of migrasomes in the ICH‐induced brain (Figure S11A). To elucidate whether C1qc regulates migrasome biogenesis or serves as a pathogenic cargo, we examined migrasome formation using the specific marker TSPAN4. Notably, knockdown of C1qc (ICH + AAV‐C1qc) failed to attenuate the ICH‐induced increase in total TSPAN4+ migrasomes, despite successfully eliminating C1qc expression within them (Figure S11B). These data demonstrate that microglial migrasome biogenesis is triggered by ICH independently of C1qc, and that C1qc functions as a pathogenic cargo packaged into these vesicles to mediate downstream endothelial injury.

To observe the effects of microglia‐derived migrasomes on endothelial cells, migrasomes were obtained by culturing microglia (BV2) in vitro, both in the presence and absence of hemin. Firstly, TEM confirmed that the purified vesicles had a migrasome structure (Figure 4E), with diameter uniformity and structural stability (Figure 4F) (Li et al. 2022). Western blot subsequently verified migrasome markers and C1qc expression in the isolated vesicles (Figure 4G). These migrasomes were subsequently introduced into cerebrovascular endothelial cells (bEnd.3) (Figure 4H). Moreover, unlike those from vehicle‐treated microglia (Vehicle‐M), migrasomes from hemin‐treated microglia (Hemin‐M) significantly disrupted endothelial cell tight junctions (Figure 4I), Western blot results are consistent with our immunofluorescence observations (Figure S12), indicating that migrasomes play a critical role in affecting the BBB during ICH. Hemin‐M exhibited clear dose‐dependent toxicity, cell viability markedly decreased when the concentration reached 40 µg/mL (Figure S13). The cytotoxicity assay, with a verified concentration (40 µg/mL) is shown in Figure 4J. We utilized Annexin V/PI dual‐staining flow cytometry to quantitatively assess cell death. Compared to the Vehicle‐M group, bEnd.3 cells treated with Hemin‐M (40 µg/mL) exhibited a higher proportion of Annexin V/PI positive cells (Figure 4K). Compared with those derived from unstimulated microglia, migrasomes derived from hemin‐treated microglia exhibited stronger adherence to endothelial cells (Figure 4L). Endothelial injury caused by Hemin‐M was assessed through PI and calcein‐AM staining (Figure 4M). In addition, we introduced microglia‐derived migrasomesthe migrasomes generated by microglia into the striatum of mice through stereotaxic injection to assess their effects on the BBB (Figure 4N). Our in vitro findings were corroborated in vivo, as the administration of hemin‐induced migrasomes led to BBB damage. The brain microvessels of recipient mice exhibited reduced Claudin‐5 expression (Figure 4O). Collectively, the results of this study suggest that ICH‐induced microglia‐derived migrasomes, containing C1qc, adhere to and disrupt cerebrovascular endothelial cells. This represents a novel mechanism for BBB disruption following haemorrhagic stroke.

3.5. Microglia‐Specific Knockout of C1qc Improved Behavioural Outcomes in Mice Following ICH and Preserved BBB Integrity

To ensure the proper functioning of the complement system within brain tissue, we specifically generated microglia‐specific C1qc knockout mice and established an ICH model. AAV–Iba1–Cre or AAV–con was injected into C1qc^flox/flox mice, which were bred from C1qc^flox/+ heterozygous mice (Figure S14), to create microglia‐specific C1qc knockout mice and control mice. The specificity and infection efficiency of the AAV–Iba1‒Cre virus are shown in Figure S15. After the knockout of C1qc in microglia, there was a significant reduction in lesion size and a decrease in brain water content (Figure 5A,B). Furthermore, microglia‐specific knockout of C1qc improved behavioural outcomes, including forelimb use asymmetry, corner turn test, and neurological deficit scores (Figure 5C–E), the results from the Rotarod test and Foot‐fault test were also improved (Figure S16). To investigate the protective mechanism, BBB integrity was assessed. EB extravasation assays revealed significant EB leakage in the ICH mice. However, EB leakage was significantly decreased in the perihaematomal area of AAV–Iba1–Cre mice, compared to AAV‐con mice (Figure 5F,G). Similarly, IgG staining revealed improved BBB permeability in the ICH mice following AAV–Iba1–Cre administration (Figure 5H,I). TEM images demonstrated that tight junctions were disrupted in the perihaematomal area after ICH. In ICH model mice, the integrity of the structural integrity of BBB was notably preserved following the knockout of C1qc from microglia (Figure 5J).

Subsequently, knocking out microglial C1qc was observed to affect endothelial tight junction proteins. In ICH model mice, the fluorescence intensities of Claudin‐5, Occludin, and ZO‐1 were notably lower than in the sham group. However, the fluorescence intensities of these proteins significantly increased in the AAV–Iba1–Cre group compared to the AAV‐con group (Figure 5K). Western blot analysis confirmed significantly reduced expression levels of Claudin‐5, Occludin, and ZO‐1 in the brains of ICH model mice compared to sham mice. However, these protein levels were restored following C1qc knockdown in microglia (Figure 5L,M).

Experiments in vitro were performed to further validate that the effect of C1qc on endothelial cells is mediated by migrasomes. We first employed siNC/siRNA to generate microglial migrasomes with/without C1qc expression in BV2 cells. The efficacy of siRNA‐mediated C1qc knockdown in BV2 cells and reduced C1qc expression in the migrasomes is demonstrated in Figures S17 and S18. Subsequently, we employed bEnd.3 as an in vitro model to investigate the role of C1qc‐positive microglial migrasomes in BBB disruption (Figure 5N). Immunofluorescence staining demonstrated migrasomes from hemin‐treated microglia (Hemin‐M) significantly decreased fluorescence intensity of Claudin‐5 in the bEnd.3 cells. Compared with those derived from siNC and hemin treatment (siNC‒Hemin–M), migrasomes derived from siC1qc and hemin‐treated microglia (siC1qc–Hemin–M) resulted in an increase of Claudin‐5 in the cerebrovascular endothelial cells (Figure 5O).

The results of in vivo and in vitro indicate that microglial migrasomes significantly contributes to BBB disruption via C1qc. The genetic deletion of C1qc specifically in microglia preserved BBB integrity, reduced brain oedema, and enhanced functional outcomes, thereby underscoring the significant role of C1qc‐positive microglial migrasomes in the pathogenesis of ICH.

3.6. C1qc‐Positive Microglial Migrasomes Exert Detrimental Effects on Endothelial Cells Through the Nlrp3–Caspase‐1–GSDMD Pathway

To further clarify the mechanism of C1qc damaging to endothelial cells, we first assessed Nlrp3 and pyroptosis‐related proteins in C1qc^flox/flox mice post‐ICH. Western blot revealed elevated levels of C1qc, cleaved caspase‐1, GSDMD‐N and Nlrp3 in ICH mice brain compared to the sham group. However, these markers were notably lower in ICH mice subjected to C1qc gene knockout (Figure 6A). Furthermore, double immunofluorescence staining confirmed that C1qc, cleaved caspase‐1, and GSDMD‐N colocalize with the endothelial marker CD31. Expression levels increased after ICH and decreased following C1qc knockdown in microglia (Figure 6B,C). These findings reveal that microglial C1qc triggers Nlrp3‐mediated pyroptosis in endothelial cells post‐ICH and inhibiting microglial C1qc mitigates this process in vivo.

FIGURE 6.

C1qc mediates the effects of migrasomes on the BBB via the Nlrp3–Caspase 1–GSDMD pathway. (A) Immunoblots revealed C1qc, Nlrp3, cleaved caspase‐1, and GSDMD‐N protein expression levels in C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con (n = 5). Densitometric quantification of protein levels; the data are presented as the mean SD; **p < 0.01, ***p < 0.001, ****p < 0.0001. (B) Immunofluorescence staining for GSDMD–N and the endothelial cell marker CD31 in the perihaematomal area of C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con (n = 5). Normalized quantitative measurement of GSDMD‐N fluorescence intensity relative to CD31 area. The data are presented as the mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (C) Immunofluorescence staining for cleaved caspase‐1 and the endothelial cell marker CD31 in the perihaematomal area of C1qc^flox/flox model mice injected with AAV–Iba1–Cre or AAV–con (n = 5). Fluorescence signal density from cleaved caspase‐1, normalized against the CD31 area, was analysed quantitatively. The data are expressed as ± SD; ***p < 0.001, ****p < 0.0001. (D) GSDMD‐N staining of endothelial cells that were treated as described in Figure 5N (n = 3). Quantification of the fluorescence intensity of GSDMD‐N of endothelial cells in the vehicle + microglia‐secreting migrasome group, the hemin + microglia‐secreting migrasome group, and the hemin + siC1qc microglia‐secreting migrasome group were also quantified. Data are presented as the mean ± SD with the indicated significance, *p < 0.05, **p < 0.01, ***p < 0.001. (E) Cleaved caspase‐1 staining of endothelial cells that were treated as described previously in Figure 5N (n = 3). Quantification of the fluorescence intensity of Cleaved caspase‐1 of endothelial cells in the vehicle + microglial migrasome groups, hemin + microglial migrasome group, and hemin + siC1qc microglial migrasome group were also quantified. The data are presented as the means ± SDs; *p < 0.05, **p < 0.01. (F) Immunoblots showing the levels of the Nlrp3 and GAPDH proteins in endothelial cells that were treated as described in Figure 5N (n = 6). Densitometric quantification of protein levels. The data are presented as the mean ± SD; ***p < 0.001. (G) Claudin‐5 staining of endothelial cells that were treated with MCC950 and microglial migrasomes without C1qc (n = 5). Quantification of the fluorescence intensity of Claudin‐5. The data are presented as the mean ± SD; ns, no significant difference; **p < 0.01, ***p < 0.001. (H) GSDMD‐N immunofluorescence in endothelial cells treated with migrasomes, C1qc‐deficient migrasomes, or C1qc‐deficient migrasomes plus Nigericin. Quantification of the fluorescence intensity of GSDMD‐N. The data are presented as the mean ± SD; **p < 0.01, ***p < 0.001. (I) Claudin‐5 immunofluorescence in endothelial cells treated with migrasomes, C1qc‐deficient migrasomes, or C1qc‐deficient migrasomes plus Nigericin. Quantification of the fluorescence intensity of Claudin‐5. The data are presented as the mean ± SD; *p < 0.05, ***p < 0.001. (J) Immunoblots showing the levels of the GSDMD‐N, Claudin‐5 and GAPDH proteins in endothelial cells that were treated with migrasomes at different time points (n = 6). Densitometric quantification of protein levels. The data are presented as the mean ± SD; ns = no significant difference, **p < 0.01, ***p < 0.001. (K) Cell viability dramatically decreased endothelial cells that were treated with migrasomes at different time points (n = 3). Data are shown as the means ± SD with significance levels; ns = no significant difference, *p < 0.05, ***p < 0.001, ****p < 0.0001.

To demonstrate that C1qc triggers endothelial pyroptosis through migrasomes, experiments in vitro were carried out. Immunofluorescence staining revealed that migrasomes from Hemin‐M significantly increased fluorescence intensities of GSDMD‐N and cleaved caspase‐1 in the bEnd.3 cells. Treatment with C1qc‐deficient migrasomes resulted in a reduction of GSAMD‐N and cleaved caspase‐1 fluorescence intensities in the vascular endothelial cells (Figure 6D,E). Additionally, in vitro experiments revealed that exposure to migrasomes released from hemin‐treated microglia led to upregulation of Nlrp3 in endothelial cells, which was reversed by treatment with C1qc‐deficient migrasomes (Figure 6F).

To verify pyroptosis is a key pathway in endothelial injury induced by C1qc‐positive microglial migrasomes we assessed the effect of the pyroptosis inhibitor MCC950 on the Claudin‐5 expression in vitro. Both MCC950 and siC1qc significantly prevented tight junction degradation and reduced GSDMD‐N expression compared to the Hemin‐M group. Interestingly, siC1qc‐Hemin‐M exhibited an even more robust restoration of tight junctions than the inhibitor MCC950 (Figure 6G). These findings indicate that targeting C1qc is an efficient strategy to protect the endothelial barrier, yet C1qc may also exert detrimental effects on endothelial cells via other pathways. To further substantiate this axis through a reverse strategy. We first evaluated the cytotoxicity of the inducer (Nigericin) on endothelial cells and selected an appropriate concentration (20 µM) based on the concentration gradient (Figure S19). Subsequently, by examining tight junction proteins and GSDMD‐N, we found that the application of Nigericin counteracted the protective effect conferred by siC1qc, leading to a re‐elevation of GSDMD‐N levels and the re‐disruption of tight junctions (Figure 6H,I). Taken together, these results demonstrate that Hemin‐M damages endothelial tight junctions via a C1qc‐dependent pyroptosis mechanism. This finding establishes endothelial pyroptosis as a critical mechanistic link between C1qc‐positive migrasomes and disruption of the BBB following ICH.

To determine the temporal sequence between endothelial pyroptosis and tight junction disruption, we conducted a time‐course analysis (0–24 h) of endothelial cells treated with migrasomes. Western blot analysis revealed a significant increase in the pyroptosis executor GSDMD‐N at 6 h post‐treatment (Figure 6J). Notably, the degradation of the tight junction protein Claudin‐5 was not evident until 12 h, which coincided temporally with a significant reduction in cell viability as measured by the CCK‐8 assay (Figure 6K). These data establish that the cleavage of GSDMD precedes the loss of tight junction proteins and overt cell death, indicating that tight junction disruption is a secondary consequence of the pyroptotic process rather than an independent initial event.

4. Discussion

ICH indicates a poor prognosis due to complex primary and secondary injury mechanisms (Gong et al. 2025). While hypertension and cerebrovascular malformations are major risk factors, genetic factors are increasingly acknowledged for their significant impact on individual susceptibility and outcomes (Li et al. 2022; Iona et al. 2024). This study introduces a novel framework that integrates two advanced technologies to overcome the limitations of traditional methods. We conducted scRNA‐seq on perihaematomal mouse brain tissue to obtain a detailed view of cellular transcription during the acute phase of ICH. Additionally, we employed MR analysis, which uses genetic variants to infer causality, thereby avoiding confounding biases and reverse causality. While scRNA‐seq identifies disease‐related cellular states and gene expression changes, MR establishes causal relationships from genomic data, their combination uniquely enhances our research. This approach effectively links observational cellular dynamics with strong genetic evidence, allowing us to identify causal mediators of the disease (Yu et al. 2024). Through this innovative approach, we identified five key genes—Blk, C1qc, Cytip, F13a1, and Sparc—linked to ICH susceptibility and outcomes. These genes are involved in immune responses, coagulation, and cytoskeletal integrity, highlighting their roles in ICH pathology. The study uncovers new genetic markers and highlights overlooked functional pathways involved in ICH pathogenesis, providing a comprehensive understanding of the disease.

C1qc is significantly upregulated at both mRNA and protein levels after ICH, particularly around haemorrhagic lesions. Although constitutively expressed in homeostatic microglia and infiltrating macrophages, C1qc undergoes a marked expression shift during the transition from homeostatic Microglia 1 to pathogenic Microglia 2. This preferential enrichment in the pathogenic cluster positions C1qc as a key cellular mediator of ICH‐induced brain injury. Also, the isolation of C1qc‐enriched migrasomes from peripheral serum demonstrates that brain‐derived migrasomes enter the circulation after ICH‐induced BBB disruption, supporting our mechanistic model and highlighting circulating migrasomal C1qc as a potential liquid biopsy biomarker for microglial activation and ICH severity. Mechanistically, C1qc sits at the nexus of canonical and novel inflammatory pathways. The established role of C1q is to initiate the classical complement cascade by binding to DAMPs and lysing RBCs (Zheng et al. 2022; Bally et al. 2021; Huang et al. 2025), which generates anaphylatoxins (C3a and C5a) and the membrane attack complex (C5b–9) (Jiang et al. 2022; Wang et al. 2023; Li et al. 2023). However, while this canonical pathway undoubtedly contributes to pathogenesis, our further experimental validation reveals a novel, complementary mechanism by which C1qc exacerbates ICH injury. Together, these data highlight the critical transition of C1qc—from a homeostatic physiological protein to a key pathological driver of neuroinflammation.

In mouse and cellular models, we observed that the activation of pyroptosis‐related proteins was reversed upon microglial C1qc deletion. C1qc‐deficient microglial migrasomes similarly failed to induce pyroptosis. Given the established role of vascular endothelial pyroptosis in BBB disruption after ICH (Xu et al. 2024; Feng et al. 2024), our findings demonstrate that microglial migrasomes exacerbate endothelial damage via C1qc‐mediated activation of the pyroptotic pathway, leading to severe BBB impairment. This expands the traditional view by showing that C1qc‐induced damage involves more than just the lytic effects of the MAC or anaphylatoxins. We propose that C1qc exacerbates neuroinflammation by inducing pyroptosis in perhaematomal endothelial cells, connecting complement activation to BBB disruption in ICH. Moreover, we found significantly higher C1qc expression in both animal models and blood samples from ICH patients, suggesting its potential as a biomarker. This expands our research from brain tissue to a systemic level, indicating that monitoring C1qc could aid in the early diagnosis, risk assessment, and prognosis of ICH.

Our in vitro data explicitly demonstrated that microglial migrasomes induced brain endothelial cell pyroptosis in an NLRP3‐Caspase 1‐GSDMD dependent manner. Literature has established that the activation of the NLRP3 inflammasome and subsequent GSDMD cleavage are central events in cell dysfunction during various cerebrovascular diseases (Ma et al. 2014). For instance, previous studies have shown that pharmacological inhibition of NLRP3 can significantly attenuate pyroptosis and preserve BBB integrity after ICH. MCC950 improved blood–brain barrier (BBB) integrity and diminished cell death (Ren et al. 2018). Consistent with these established concepts, our current study identifies microglial migrasomes—and specifically the C1qc protein—as a novel upstream trigger for this canonical pyroptotic cascade in endothelial cells.

Time‐course analysis provides critical mechanistic insights into how migrasomes compromise the blood–brain barrier. We found that the generation of GSDMD‐N precedes the degradation of Claudin‐5 and overt cell death. This temporal hierarchy suggests that tight junction disruption is secondary to pyroptosis execution. Mechanistically, the insertion of GSDMD pores into the endothelial membrane likely disrupts cellular osmotic gradients and ion homeostasis (Liu et al. 2016), which secondarily destabilizes the cytoskeleton‐anchored tight junction complexes before culminating in catastrophic cell lysis (Ruhl et al. 2018). Thus, targeting the early stages of migrasome‐induced pyroptosis could preserve tight junction architecture before irreversible structural collapse occurs.

This study further investigated its intercellular mechanism. Migrasomes are EVs produced during cell migration, primarily by immune cells, and contain lipids, proteins, and RNA (Yu et al. 2022). They can be directed to specific sites to stimulate adjacent cells, thereby enhancing intercellular communication and signal transduction (Li et al. 2025). Migrasomes originating from macrophages are crucial for cellular communication, development, vascular homeostasis, and substance transport (Zhang et al. 2025; Ye et al. 2025; Huang et al. 2025). Microglia are the primary immune cells in the nervous system. However, research on the role of microglia‐derived migrasomes in the central nervous system is somewhat limited. Our study revealed, for the first time, that migrasomes are closely associated with microglia, suggesting that microglia may participate in brain injury through migrasome‐mediated intercellular communication. Notably, migrasomes are not exclusive to the post‐ICH state; they are also present in normal brain tissue, albeit in smaller quantities. In the ICH condition, the substances within migrasomes vary from those in the normal state, potentially playing a crucial role in brain tissue or BBB damage. This suggests a shift in migrasome cargo from homeostatic to pathological components across physiological states, a critical mechanism controlling their role in disease progression.

Hu et al. demonstrated that macrophage‐derived migrasomes contribute to BBB impairment in CAA (Hu et al. 2023). While their work established a link between migrasomes and vascular injury in a chronic amyloid‐driven setting, our study advances this paradigm in acute brain injury. First, in ultra‐early ICH—an acute catastrophic event—brain‐resident microglia, not peripheral macrophages, are the primary initiators of migrasome formation in response to hematoma toxicity. Second, we uncovered a distinct molecular mechanism: C1qc as a specific pathogenic cargo enriched in microglial migrasomes, which directly orchestrates endothelial damage via the canonical NLRP3‐Caspase 1‐GSDMD pyroptotic cascade. Thus, our findings extend migrasome pathology from chronic neurodegeneration to acute haemorrhagic stroke and define a targetable C1qc‐pyroptosis axis in microglial‐endothelial crosstalk.

Inhibiting the entire complement system can increase the risk of infection (Garred et al. 2021). Compared to C1qc knockout mice, specifically targeting C1qc in microglia offers a safer treatment. We generated C1qc^flox/flox mice and established a microglia‐specific C1qc knockout model using Iba1–Cre. After ICH injury, wild‐type mice suffer severe BBB disruption with broken tight junctions between vascular endothelial cells, whereas microglia‐specific C1qc‐deficient animals exhibit significantly mitigated damage. In vitro experiments further confirmed that migrasomes derived from microglia with siRNA‐mediated reduction in C1qc did not cause significant tight junction damage. These findings indicate that microglial C1qc is transported via migrasomes, leading to vascular endothelial damage and BBB dysfunction. By selectively knocking out C1qc in microglia, we can assess its neuroprotective potential, supporting the development of precise immunotherapy.

The limitation of this study is the focus on acute‐phase behavioural assessments after ICH. While we confirmed acute sensorimotor deficits using objective tests, we did not evaluate cognitive functions because acute motor impairments could confound spatial learning tasks. Future studies extending into the chronic phase (>28 days) are needed to determine if targeting microglial migrasomes or the C1qc pathway provides long‐term cognitive protection. Furthermore, another limitation regarding our in vitro mechanistic experiments is the reliance on immortalized murine cell lines (BV2 and bEnd.3). While these cell lines provided a highly pure and reproducible system to isolate and validate the microglia‐endothelial crosstalk (specifically, effectively excluding the interference of peripheral macrophages), they may not fully capture the phenotypic complexity, metabolic state, and exact transcriptomic profiles of primary central nervous system cells. Future studies employing primary microglia‐endothelial co‐cultures or human‐derived brain organoids will be highly valuable to further substantiate the translational relevance of the migrasome‐C1qc axis in ICH pathology.

While we observed the accumulation of migrasomes and their harmful effects, the mechanisms of their formation, cargo sorting, and delivery to endothelial cells remain unclear.

Based on the current literature on EVs and our own observations, we hypothesize that the internalization of C1qc‐positive migrasomes by endothelial cells likely involves multiple pathways. Firstly, similar to other large EVs, the internalization of migrasomes by target cells is generally considered to be driven by macropinocytosis, phagocytosis, or membrane fusion (Mathieu et al. 2019), these processes are primarily mediated by surface structural proteins such as tetraspanins and integrins, rather than by specific internal cargo (Andreu et al. 2014). Secondly, although in our model C1qc primarily functions as a pathogenic effector molecule (cargo) mediating endothelial injury (pyroptosis), we cannot exclude the possibility that if a portion of C1qc is exposed on the migrasome membrane surface, it may facilitate targeted uptake. Interactions between migrasome‐associated C1qc and C1q receptors are highly likely to enhance the recognition and receptor‐mediated endocytosis of these specific pathogenic migrasomes.

An important mechanistic feature identified by our study is the precise role of C1qc in the migrasome‐dependent pathogenic cascade. We demonstrated that C1qc is not a structural driver of migrasome formation. Instead, ICH triggers a parallel response: it induces extensive migrasome biogenesis (the delivery vehicle) while simultaneously enriching them with C1qc (the pathogenic cargo). This highlights migrasomes as highly specialized pathogenic carriers in the post‐ICH microenvironment, where targeting specific cargos like C1qc‐rather than blindly blocking all EV secretion or deplete all the microglia‐might offer a safer, more precise therapeutic strategy.

In conclusion, our work reveals a previously unrecognized pathway through which microglial‐derived migrasomes function as crucial carriers of C1qc, establishing a direct connection between neuroinflammation and BBB disruption through endothelial pyroptosis (Song et al. 2022 ) (Figure 7). This paradigm not only enhances our understanding of ICH pathology but also presents novel opportunities for therapeutic intervention and biomarker discovery.

FIGURE 7.

Microglia‐derived migrasomal C1qc induces endothelial injury following intracerebral haemorrhage. After the occurrence of ICH, microglia secrete migrasomes enriched with C1qc. These migrasomes adhere to brain microvascular endothelial cells. Intracellularly, the delivered C1qc triggers the Nlrp3/Caspase‐1/GSDMD pyroptotic cascade. This signalling leads to GSDMD‐N‐mediated pore formation and endothelial pyroptosis, ultimately causing the disruption of tight junctions and blood–brain barrier breakdown.

Author Contributions

Dengpan Song: investigation, formal analysis, visualization. Menghan Deng: investigation, methodology, validation. Moxin Wu: investigation, resources. Fang Wang: methodology, validation. Mengyuan Li: conceptualization, methodology, data curation, formal analysis, validation, visualization, writing – original draft, investigation. Jian Wang: validation, writing – review and editing, supervision, software. Mengzhao Feng: data curation, validation. Zhihua Li: writing – original draft, visualization, project administration, conceptualization, methodology, formal analysis. Yating Wang: data curation, resources. Yanyan Sun: methodology, investigation. Xuemei Chen: formal analysis, validation. Junmin Wang: writing – review and editing, supervision, software. Tingting Zhai: investigation, methodology, formal analysis. Fuyou Guo: conceptualization, funding acquisition, project administration, supervision, resources, writing – review and editing. Xiaoping Yin: software, resources. Qiong Yang: validation, methodology. Chi‐Tai Yeh: writing – review and editing, project administration. Junyang Wang: methodology, data curation.

Funding

This work was supported by the National Natural Science Foundation of China under Grant 32471043, Leading Project of Henan Province Young Medical Research under Grant LJRC2023010, and Key Project of International Science and Technology Cooperation, Science and Technology Department of Henan Province under Grant 241111520600.

Ethics Statement

All procedures were conducted in compliance with the relevant guidelines and regulations. This study received approval from the Institutional Review Board of Zhengzhou University granted ethical approval for this study (Approval No.021–KY–0156) and adhered to the principles of Good Clinical Practice and the Declaration of Helsinki. Informed consent was acquired from all participants or their legal representatives.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Table: jev270377‐sup‐0001‐TableS1.xlsx

Supplementary Table: jev270377‐sup‐0002‐TableS2.docx

Supplementary Table: jev270377‐sup‐0003‐TableS3.docx

Supplementary Table: jev270377‐sup‐0004‐TableS4.xlsx

Supplementary Figure: jev270377‐sup‐0005‐FigureS1‐S19.docx

Li, M. , Li Z., Song D., et al. 2026. “C1qc Mediates Blood–Brain Barrier Disruption in a Mouse Model of Intracerebral Haemorrhage Through Microglia‐Derived Migrasomes.” Journal of Extracellular Vesicles 15, no. 10: e70377. 10.1002/jev2.70377

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.