Cerebral venous outflow disturbance exacerbates ischemic stroke via an endothelium-initiated inflammatory cascade: an integrated single-cell transcriptomic analysis

Abstract

Background

Unfavorable cerebral venous outflow (VO) critically worsens ischemic stroke (IS) prognosis, yet the underlying cellular mechanisms driving this aggravation remain obscure. This study aimed to explore the impact of VO disturbance on ischemic brain injury.

Methods

We superimposed the left internal jugular vein occlusion (LIJVO) onto a rat model of ischemia/reperfusion (I/R). Key pathological indices were assessed and the transcriptomic landscape of the injury using the integrated single-cell and single-nucleus RNA sequencing approach, followed by immunofluorescence validation.

Results

Superimposed LIJVO significantly exacerbated neurological deficits and infarction, precipitating malignant brain edema and blood-brain barrier breakdown (P < 0.05). The pathological expansion of pro-inflammatory Endothelial0 and Microglia1 subpopulations were identified. We traced these deteriorations to an endothelial-initiated inflammatory cascade: a specific Endothelial0 emerged to recruit leukocytes and trigger a feed-forward loop with hyper-reactive Microglia1. This crosstalk may drive a cytokine storm and tight junction collapse, ultimately leading to severe perivascular infiltration and accelerated neuronal injury.

Conclusion

Collectively, this study elucidates that unfavorable VO functions as a critical aggravator in IS via an endothelial-initiated inflammatory cascade. Targeting upstream endothelial activation represents a promising therapeutic strategy for the IS patients with compromised venous drainage.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12987-026-00842-1.

Background

Ischemic stroke (IS) remains a primary cause of global mortality and long-term disability [1, 2]. The abrupt cessation of cerebral perfusion triggers a devastating ischemic cascade, characterized by inflammation and oxidative stress, that compromises the blood-brain barrier (BBB) and precipitates neuronal death. The resulting sequelae place a staggering socioeconomic burden on patients and healthcare systems worldwide.

While traditional stroke research has predominantly focused on arterial inflow, the critical role of venous outflow (VO) in cerebral hemodynamics has historically been underappreciated. Physiological recovery following ischemia relies not only on restoring arterial supply but also on the integrity of microvascular circulation and efficient venous drainage [3, 4]. Emerging clinical evidence highlights that impaired cortical venous drainage significantly exacerbates the pathophysiology of stroke [3, 5, 6]. Specifically, VO obstruction or insufficiency, often manifesting as insufficient cortical venous filling, is strongly associated with the progression of malignant cerebral edema and hemorrhagic transformation [5–10]. This link is further substantiated by pharmacological findings; for instance, Argatroban has been reported to improve IS prognosis specifically by facilitating basal vein drainage [11].

While recanalization therapies like tissue plasminogen activator effectively restore arterial inflow, they cannot immediately correct downstream venous insufficiency caused by pre-existing stenosis, thrombosis, or compression from elevated intracranial pressure [3, 12]. In these refractory cases, a pathological “inflow-outflow mismatch” occurs where blood flow is restored to the brain, but drainage remains compromised. This sustained venous hypertension acts as a critical driver for malignant edema and BBB collapse [5, 9].

Despite these clinical observations serving VO as a potential prognostic indictor, the mechanisms underlying VO-mediated exacerbation remain obscure. Current assessments often overlook venous metrics, and research has been constrained by a lack of optimal experimental models that accurately replicate ischemia combined with venous obstruction. Consequently, how unfavorable VO reshapes the cellular microenvironment and exacerbates neurovascular unit (NVU) injury remains a critical knowledge gap.

To interrogate the cellular underpinnings, we superimposed left internal jugular vein occlusion (LIJVO) in the rat model of ischemia/reperfusion (I/R) to simulate outflow insufficiency following IS recanalization. We selected LIJVO because the internal jugular vein serves as the primary direct drainage route for the cerebral sinuses in rodents. Functionally, this model creates a state of high-resistance outflow and we employed an integrated single-cell and single-nucleus RNA sequencing (scRNA-seq/snRNA-seq) approach to map the transcriptomic consequences of this hemodynamic mismatch.

Unraveling the cellular heterogeneity of the central nervous system demands high-resolution genomic approaches. However, conventional single-cell RNA sequencing (scRNA-seq) often suffers from selective cellular depletion during enzymatic tissue dissociation, leading to the underrepresentation of fragile neurons and the artificial activation of stress genes [13–16]. In contrast, single-nucleus RNA sequencing (snRNA-seq) circumvents this limitation by profiling nuclei directly from frozen tissues, thereby ensuring the robust capture of vulnerable neuronal populations [17–20]. Consequently, the integration of these dual modalities offers an unbiased strategy to map the complete cellular atlas, which is critical for decoding the dynamic microenvironmental shifts during stroke progression.

Materials and methods

Transient cerebral I/R and internal jugular vein occlusion model

The LIJVO model was established by permanent ligation of the left internal jugular vein. Briefly, under deep anaesthesia induced by 1% pentobarbital sodium, a small midline cervical incision was made to expose the left internal jugular vein. The vein was carefully separated from the surrounding tissues and permanently ligated using a sterile silk suture. For animals subjected to the combined I/R with LIJVO model, the left internal jugular vein was ligated immediately before middle cerebral artery occlusion (MCAO) and remained occluded throughout the ischemia and reperfusion periods. Subsequently, transient MCAO was performed according to the procedure described below.

The left transient cerebral I/R was induced by middle cerebral artery (MCA) closure using the modified Zea Longa method [21]. Briefly, rats were anesthetized and subsequently positioned in a supine placement. By making a midline incision in the neck, the left common carotid artery (CCA), external carotid artery (ECA), and internal carotid artery (ICA) were exposed and separated from the surrounding tissue. Further, for the blockage of the MCA, ECA was introduced using a rounded tip nylon monofilament suture (Cat. No. A4-263650; Beijing Cinontech Co. Ltd., Beijing, China), which was subsequently inserted into the ICA. To achieve reperfusion, the nylon suture was withdrawn cautiously 2 h after the occlusion. The identical procedure was conducted on the sham rats, except for suture retention just inside the ICA.

The rats had a 12-hour fasting period before to the operation but were permitted unrestricted access to drink water. The rats were provided with food and water inside their original cages upon reviving from anesthetisation. Animals were anesthetized and euthanized with overdose of pentobarbital sodium at 1d, 3d or 7d post cerebral ischemia and tissues were meticulously harvested for experiments.

Efforts were made to minimize the number of animals and their suffering.

Animals and groups

Adult male Sprague–Dawley rats weighing 230–250 g were obtained from the Shanghai Slacom Experimental Animal Technology Co. Ltd (Shanghai, China). All related experimental animal protocols followed the “Guide for the Care and Use of Laboratory Animals”, which was approved and regulated by the Zhejiang Chinese Medical University’s Institutional Animal Care and Ethical Committee.

The rats were housed in a specific pathogen-free condition with 12-hour light/dark cycle at the Laboratory Animal Centre of Zhejiang Chinese Medical University (Hangzhou, China). The rats were maintained at the ambient temperature and humidity at 22 ± 2 °C and 45 ± 3%, respectively. All of the animals were given free access to clean water and adequate standard chow. Following a week of acclimatization, rats were randomly assigned to 4 groups: the sham operation group (Sham), I/R model group, LIJVO group, and I/R+LIJVO group.

Inclusion and exclusion criteria

Based on the Zea Longa methods [21], neurological deficit assessment was performed in rats at 2 h post- I/R injury. Rats demonstrating neurological scores between 1 and 3 points on this standardized scale were subsequently enrolled for further experimental procedures.

Rats meeting any of the following criteria were excluded from the study: (1) Premature death prior to scheduled experimental timepoints despite successful model establishment; (2) presence of evident cerebral haemorrhage symptoms or intracranial haemorrhage identified during tissue harvesting; (3) absence of cerebral infarction.

Neurobehavioral test and infarct volume estimation

A modified neurological severity score (mNSS) was performed to estimate motor, sensory and balance function on days 1, 3 and 7 after modeling. This score ranged from 0 to 18. Higher scores indicated the more severe neurological deficiencies. All mNSS tests were carried out by a researcher who was blinded to group assignment.

Rats were anesthetized and decapitated on post-operation 1d, 3d, and 7d. Brain tissues were isolated, sectioned into 2 mm thick coronal slices and stained with TTC 2% 2, 3, 5-triphenyltetrazolium chloride (TTC, Solarbio, China). The olfactory bulb, cerebellum, and brainstem were exposed, and the isolated brain tissues were rapidly frozen at -20°C for 20 min. Brain tissues were sectioned into 2 mm thick coronal slices from the frontal lobe and subsequently incubated with 2% 2, 3, 5-triphenyltetrazolium chloride (TTC, Solarbio, China) for 20 min at 37 °C. The total infarct volume was calculated by summing the infarction measurements from each slice by ImageJ (NIH, Bethesda, MD, USA). The infarct volume was defined as V%=volume of the infarct area/volume of the total area ×100%.

Assessment of cerebral perfusion by laser speckle contrast imaging

Cerebral blood flow (CBF) was evaluated using laser speckle contrast imaging (LSCI; moorFLPI-2, Moor Instruments, UK) at 24 h after reperfusion. Briefly, rats were anesthetized with isoflurane and placed in a stereotaxic frame to minimize motion artifacts during imaging. Following a midline scalp incision, the skull was carefully exposed while maintaining its integrity. Perfusion images were acquired under identical imaging conditions using a 100mW laser source.

For quantitative analysis, a region of interest (ROI) encompassing the entire exposed cerebral surface was defined, and the mean perfusion flux (Flux Mean) was calculated using the manufacturer’s analysis software. To minimize inter-animal variability and facilitate comparison among experimental groups, CBF values were normalized to the mean value of the Sham group and expressed as a percentage of Sham. The normalized CBF values were subsequently used for statistical analysis.

Histopathology

The rats underwent deep anaesthesia and transcardial perfusion with saline, followed by 4% paraformaldehyde on days 1 and 7. Next, the brains were removed and fixed in 4% paraformaldehyde. The brain tissues were prepared for OCT embedding and sectioned into 10 μm coronal slices for staining. The morphological changes of the brain and the neuronal Nissl bodies were evaluated by staining with Haematoxylin and eosin (HE) and Nissl (Solarbio, China), followed by observation using light microscopy (Nikon, Japan).

TUNEL staining assessment

Cell apoptosis in the ischemic brain was observed through Terminal deoxynucleotide transferase dUTP nick-end (TUNEL) staining (Beyotime, China). TUNEL staining was performed in accordance with the manufacturer’s instructions. 10-µm thick sections were incubated for 15 min in proteinase K working buffer at 37 °C, followed by PBS rinsing of the tissues and staining for 60 min at room temperature with the green fluorescein-labeled dUTP solution per the manufacturer’s protocol, and then stained with DAPI for nuclear labelling. The TUNEL-positive cells were detected by a fluorescence microscope (Nikon, Japan). Image post-analysis was performed using Image J software.

Brain water content assessment

The brain was promptly excised adhering to deep anaesthesia with 1% pentobarbital sodium at 24 h after the operation. The olfactory bulb, cerebellum and lower brainstem were discarded, and the ipsilateral hemisphere was weighed prior to and after drying at 65 °C for 72 h. Brain edema was evaluated as brain water content, as follows: (wet weight- dry weight)/wet weight×100%.

BBB permeability detection

On postoperative day 1, the rats received an intravenous administration of 2% Evans blue (EB, 2 mL/kg, Aladdin) via the tail vein to detect BBB permeability. Post-injection 2 h, rats underwent transcardial perfusion with 0.9% saline, followed by the removal of brain tissue. The ipsilateral cortices were homogenized in N, N-dimethylformamide (10 µL/g wet weight tissue, Aladdin) and incubated at 60 °C for 24 h. After grinding and extraction, the mixture was centrifuged for 30 min at 12,000 rpm. The supernatants were collected, and the concentration of EB was quantified by standard curve, with absorbance measured at 620, 500 and 740 nm. The formula utilized for permeability evaluation is expressed as EB value= A₆₂₀-(A₅₀₀+A₇₄₀)/2.

scRNA-seq and snRNA-seq library preparation and sequencing

The left hemispheric brain tissues from the Sham, I/R, and I/R+LIJVO groups were rapidly collected. Each biological sample was equally divided into two portions for the preparation of scRNA-seq and snRNA-seq suspensions, respectively. Each group contained three independent biological samples obtained from the infarct core and peri-infarct regions of three individual rats.

To maximize the recovery of both immune and parenchymal cell populations, single-cell and single-nucleus suspensions were prepared separately and subsequently pooled prior to library construction. The pooled cell/nucleus preparations from each biological sample were processed as a single library using the Chromium Single Cell 3′ platform (10× Genomics, Pleasanton, CA, USA). Consequently, one unified library was generated for each biological sample.

For scRNA-seq preparation, freshly isolated brain tissues were preserved in MACS Tissue Storage Solution (Miltenyi Biotec, Germany) and mechanically dissociated followed by enzymatic digestion. Cell suspensions were filtered through cell strainers, counted, and subjected to quality assessment before encapsulation into gel beads-in-emulsion (GEMs) using the Chromium platform. For snRNA-seq preparation, nuclei were isolated from brain tissues through mechanical and enzymatic dissociation, followed by purification and quality control. The purified nuclei were then pooled with the corresponding single-cell suspension before library construction. Library preparation and high-throughput sequencing were performed by Oebiotech (Shanghai, China) according to the manufacturer’s protocols.

scRNA-seq/snRNA-seq data analysis

Raw sequencing data were processed using Cell Ranger software (version 7.0.1, 10× Genomics). Following barcode demultiplexing and alignment to the reference genome using STAR, gene expression matrices were generated for each biological sample.

Because single-cell and single-nucleus preparations were pooled before library construction, independent scRNA-seq and snRNA-seq datasets were not generated. Instead, a single unified expression matrix was obtained for each biological sample. In total, three expression matrices corresponding to the three biological samples were generated and imported into Seurat (version 5.1.0) using the Read10X function.

The expression matrices from all samples were merged into a single Seurat object. Metadata including sample identity and experimental group information were assigned to each cell/nucleus. Quality control filtering was performed using the following criteria: nFeature_RNA > 100, nFeature_RNA < 6000, nCount_RNA > 500, nCount_RNA < 15,000, and mitochondrial gene percentage < 3%.

To reduce technical variation among biological samples, the RNA assay was split into sample-specific layers and normalized using NormalizeData. Highly variable genes were identified using FindVariableFeatures, followed by data scaling with ScaleData and dimensionality reduction using principal component analysis (RunPCA). Batch-effect correction and data integration were subsequently performed using Seurat v5’s CCA-based integration workflow (IntegrateLayers with CCAIntegration). CCA-based integration was selected because it effectively identifies shared biological structures across samples while reducing technical variation. The resulting integrated embeddings (integrated.cca) were used for downstream clustering and analysis. Finally, the RNA layers were merged using JoinLayers.

Cell-cluster and annotation

The dimensionality of the gene expression matrix was diminished by selecting the top 2000 variable features using the FindVariableFeatures function, which employs a method such as variance stabilizing transformation by default. Subsequently, principal component analysis was performed on these 2000 characteristics to identify the most informative principal components. Based on the PCA results, we constructed a shared nearest neighbour graph using the FindNeighbors function. Then, the distinct cell clusters were delineated employing the FindClusters function.

For visualization, the RunTSNE function was leveraged for nonlinear dimensionality reduction, facilitating the examination of the distribution patterns among distinct cell clusters in a two-dimensional space. Cluster marker genes were identified using the FindMarkers function. Thorough cluster annotation was accomplished by a multi-faceted methodology: (1) automated annotation by tools like SingleR; (2) online annotation via databases such as CellMarker; (3) annotation based on a collection of classical markers derived from literature review and domain knowledge.

Cell-cell communication analysis

Intercellular communication analysis was conducted utilizing CellChat (version 1.16.0) R package, spanning all cell types with an emphasis on endothelial, microglia, and neuron subtypes. We constructed a CellChat object from normalized scRNA-seq data and adhered to the specified protocol. Key steps encompassed identifying overexpressed ligand-receptor pairs and interactions, calculating interaction probabilities with computeCommunProb, assessing signaling pathway activities with computeCommunProbPathway and aggregateNet, and visualizing outcomes with netVisual_bubble. This analysis revealed intricate communication patterns among cell types.

Gene ontology (GO) enrichment analysis

To elucidate the biological processes associated with our gene set of interest, enrichment analysis was conducted with the enrich GO function from the clusterProfiler (version 4.6.2) R package, concentrating on the “Biological Process” ontology. The analysis focused exclusively on differentially expressed genes (DEGs) rigorously filtered based on the criteria: average_log2_Fold_Change > 1 and P_value < 0.05.

Immunofluorescence staining

The immunofluorescence staining was conducted on 10-µm frozen sections. After acclimatization to room temperature for 30 min, the sections were permeabilized in 0.5% Triton X for 5 min, treated with antigen retrieval. Then they were blocked with 5% bovine serum albumin for 1 h and incubated overnight at 4 °C with primary antibodies (shown in Supplementary Material). Fluorochrome-conjugated secondary antibodies specific to the species were applied the next day and a DAPI-containing anti-fluorescence quenching agent was used for sealing. The sections were imaged using a fluorescence microscope.

Detection of the proinflammatory cytokines

The ischemic brain hemispheres were homogenized in pre-cooled PBS 1 day after the operation and centrifuged at 3000 rpm at 4°C. The supernatant was collected, and the concentrations of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumour necrosis factor α (TNF-α) were quantified using ELISA kits according to the manufacturer’s instructions (Liankebio, China; Nanjing Jiancheng Bioengineering Institute, China). Absorbance was measured at 450 nm using a microplate reader (SpectraMax M5, MD, China).

Western blot

The isolated brain tissues were lysed with RIPA Buffer containing a phosphatase inhibitor and the protein content was evaluated using BCA assay. Following normalization and thermal denaturation, protein samples were separated by SDS-PAGE gel and then transferred into the PVDF membrane. The membrane underwent blocking with skimmed milk to prevent nonspecific banding and was subsequently incubated overnight with primary antibodies at 4 °C (ZO-1, 1:5000, Proteintech, 21773-1-Ap; Occludin, 1:5000, Proteintech, 27260-1-Ap). On the next day, the membranes were treated with appropriate secondary antibodies at ambient room temperature. The protein bands were visualized and quantified via LI-COR Odyssey DLX.

Statistical analysis

Results were presented as mean ± standard deviation (SD). IBM SPSS statistical 25 software (SPSS, USA) was used to perform data analysis, whereas the one-way analysis of variance (ANOVA) was employed to analyze multiple groups, followed by the LSD test. The threshold for statistical significance was determined to be P < 0.05.

Results

LIJVO exacerbated cerebral I/R injury and neurological deficits

To define the impact of VO disturbance on IS progression, we systematically evaluated survival rates, neurological function, and infarct volumes (Fig. 1A). Survival analysis demonstrated that LIJVO alone did not compromise viability compared to the Sham group (P > 0.05). In contrast, I/R introduced significant mortality, which was further exacerbated in I/R+LIJVO group, resulting in the lowest survival rate among all cohorts (P < 0.05, Fig. 1B).

Fig. 1.

LIJVO exacerbated cerebral I/R injury and neurological deficits. (A) The schematic overview of the animal experimental design. Created in https://BioRender.com. (B) The survival rate of the Sham, LIJVO, I/R, I/R+LIJVO groups from the 1 day to 7d post-operation. (C) The mNSS was performed to evaluate the neurological deficit of each group at 1d, 3d and 7d. (D-E) The presentative images of TTC staining at different time points, and the quantified analysis of the 1d, 3d and 7d. Data were presented as mean ± SD, n = 20 per group for survival rate, n=10 per group for mNSS and n=5 for infarct volume. (F-G) Representative perfusion images of regional cerebral blood flow in the cranial window for each group at 24h post-operation and the quantified analysis (n=4).TTC, 2, 3, 5-triphenyltetrazolium chloride; mNSS, modified neurological severity score; LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion; ns, no significant difference. *, P<0.05; **, P<0.01; ***, P<0.001.

Neurobehavioral assessment via mNSS confirmed this trend. While rats in the Sham and LIJVO groups remained neurologically intact, those subjected to I/R injury exhibited substantial deficits. Crucially, I/R+LIJVO significantly compounded these functional impairments, resulting in consistently higher deficit scores compared to the I/R group across all post-operative time points (P < 0.05, Fig. 1C). This functional deterioration was mirrored by morphological damage. As visualized by TTC staining, Sham and LIJVO groups showed no detectable tissue damage. However, I/R+LIJVO group produced significantly larger infarct volumes than I/R at every assessed interval (P < 0.05, Fig. 1D-E). These data revealed that LIJVO potentiated neurobehavioral deficits and increased infarct volume following I/R injury. Additional LSCI experiments demonstrated a tendency toward reduced cerebral perfusion in the LIJVO group compared with the Sham group. Although CBF was markedly reduced in both I/R and I/R+LIJVO groups (P < 0.05, Fig. 1G-H), no statistically significant difference was observed between these groups at 24 h after reperfusion (P > 0.05, Fig. 1G-H).

LIJVO aggravated severe histopathological deterioration and BBB breakdown following I/R

To determine whether VO aggravates I/R injury, we examined histological staining severity and barrier integrity. Histopathological evaluation revealed a clear hierarchy of damage. While the Sham and LIJVO groups retained intact cortical architecture, the I/R group exhibited characteristic ischemic signatures, including nuclear pyknosis and cellular atrophy. Crucially, the I/R+LIJVO significantly accelerated this degeneration, progressing from early cellular disruption to severe nuclear membrane breakdown and vacuolization by day 7. This finding was substantiated by Nissl staining, which assessed neuronal viability. Consistent with HE results, neuronal structures in the I/R+LIJVO group displayed marked shrinkage and severe depletion of Nissl bodies, indicating that VO disturbance severely compromised neuronal structural integrity beyond I/R (Fig. 2A-B).

Fig. 2.

LIJVO aggravated severe histopathological deterioration and BBB breakdown following I/R. (A-B) Representative images of HE, Nissl and TUNEL staining within the brain ischemic regions from Sham, LIJVO, I/R, I/R+LIJVO groups at 1d (A) and 7d (B) post-ischemic. Scale bar = 20 μm. (C) Quantification analysis of apoptosis proportion measured by TUNEL staining at 1d and 7d. (D) Quantification analysis of ipsilateral brain tissue’s water content. (E) Representative images of EB extravasation at 1d after operation in the four groups. (F) Quantified analysis of EB extravasation. Data were presented as mean ± SD, n = 3 per group for TUNEL staining and n = 5 per group for water content and Evans blue leakage. EB, Evans blue; LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion. *, P < 0.05; **, P < 0.01; ***, P < 0.001

Like the morphological damage, TUNEL assays confirmed extensive cell death in the injured regions (Fig. 2A-B). Negligible apoptosis was observed in the Sham or LIJVO groups; however, I/R injury triggered a surge in apoptotic cells. Notably, the I/R+LIJVO group amplified this apoptotic cascade, exhibiting a significantly higher number of TUNEL-positive cells compared to the I/R group at both Day 1 and Day 7 post-operation (P < 0.05, Fig. 2C).

We next investigated the integrity of the NVU. Quantitative analysis of brain water content indicated that I/R+LIJVO significantly exacerbated fluid accumulation compared to the I/R group, leading to profound cerebral edema (P < 0.05, Fig. 2D). This edema was linked to BBB failure, as visualized by EB extravasation. While the barrier remained impermeable in Sham and LIJVO groups, the I/R+LIJVO group displayed catastrophic leakage, with dye extravasation significantly exceeding that of the I/R group (P < 0.05, Fig. 2E-F).

Above the results established that impaired VO acted as a potent aggravating factor, potentiating cerebral I/R injury through intensifying cellular apoptosis, compromising BBB integrity and worsening malignant cerebral edema.

Single-cell transcription unveiled a distinct cellular remodeling driven by I/R+LIJVO

To comprehensively map the cellular dynamics underlying the I/R+LIJVO injury, we profiled the transcriptional landscape of the brain using single-cell transcriptome (Fig. 3A). Rigorous quality control yielded 36,789 high-quality transcriptomes, which were projected into a global atlas via t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction. Unsupervised clustering resolved these cells into 36 distinct clusters, annotating 11 major lineages based on canonical marker expression (Fig. 3B-D, Figure S1A). These included resident central nervous system populations (Astrocytes, Microglia, Oligodendrocytes, Neurons, Endothelial cells) and infiltrating immune subsets (Macrophages, B cells, T cells, Granulocytes).

Fig. 3.

Single-cell transcription unveiled a distinct cellular remodeling driven by I/R+LIJVO. (A) Scheme flow of single-cell RNA-sequencing and single-nuclei RNA-sequencing. (B) The t-SNE plot visualizes 36,789 cells split into 36 cell clusters in sham, I/R and I/R+LIJVO groups of rat brains at 1d post-ischemic stroke. (C) The cell is clustered into 11 cell types based on core genes by cell type annotation. Colors depict cell-type clusters by t-SNE. (D) Violin plots showing logarithmic-scale expression of marker genes for each cell type identified in (C). (E) The t-SNE plot in the three samples. (F) Bar plot showing the fraction of the 11 cell-type subsets in each group. LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion. t-SNE, t-distributed stochastic neighbor embedding

Cellular proportions analysis revealed that ischemic injury dramatically reshaped the cellular composition compared to the Sham group (Fig. 3E). Consistent with the previous publications [20, 22], the ischemic brain was characterized by a robust infiltration of peripheral immune cells, specifically granulocytes and macrophages, induced by the BBB disruption. Most critically, I/R+LIJVO imposed a unique cellular signature distinct from I/R. While I/R injury triggered broad inflammatory recruitment, the I/R+LIJVO specifically drove an unexpected expansion of the endothelial cell compartment, concomitant with a pronounced reduction in the microglial population (Fig. 3F, Figure S1B). This divergence suggested that VO disturbance did not merely intensify the severity of IS injury but altered the cellular composition.

LIJVO promoted the accumulation of a specific pro-inflammatory endothelial subcluster post-I/R

Given the overarching expansion of the endothelial compartment observed, we performed sub-clustering to dissect its internal heterogeneity. The endothelial population was deconvoluted into 7 distinct transcriptomic phenotypes (Endothelial0–Endothelial6), defined by signature markers such as Adamts9 (Endothelial0) and Notch3 (Endothelial2) (Fig. 4A, Figure S2A).

Fig. 4.

LIJVO promoted the accumulation of a specific pro-inflammatory endothelial subcluster post-I/R. (A) The endothelial cells were identified as 7 subtypes. (B) Cell count of each population. (C) GO enrichment of endothelial0. (D) Volcano plots of differential expression genes in endothelial0 between the I/R and I/R+LJVO groups. (E) The relative expression of Egr1 in endothelial0 among three groups. (F) Representative images of immunofluorescence co-staining of Egr1 (red) and Claudin (green) in the ischemic brain. 400X, Scale bar = 20 μm. LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion

To determine the functional consequences of this shift, we interrogated the biological pathways enriched in Endothelial0. Unlike other subsets, Endothelial0 was characterized by a distinct immunogenic signature. Gene ontology (GO) analysis revealed a strong enrichment for “granulocyte migration” and “leukocyte cell-cell adhesion,” positioning Endothelial0 as a sticky, pro-inflammatory mediator capable of facilitating immune cell infiltration into the tissues (Fig. 4B-C, Figure S2B-D, Table S1).

To explore the molecular signature underlying this pathological state, we analyzed the DEGs. We identified Early Growth Response 1 (Egr1), an immediate-early gene often induced by fluid shear stress and ischemia, as one of the most upregulated transcription factors specific to Endothelial0. Egr1 expression levels were significantly elevated in the I/R+LIJVO group compared to the I/R group (Fig. 4D-E). This transcriptional elevation was spatially corroborated by IF staining, which confirmed an accumulation of Egr1 protein in the microvasculature of the I/R+LIJVO injury (Fig. 4F). Collectively, these data identified Endothelial0 as a specialized, LIJVO-sensitive endothelial state marked by pro-inflammatory features and high Egr1 expression.

I/R+LIJVO exacerbated neuroinflammation by driving the expansion of a specific NF-κB^high microglia subpopulation

To map the immunological landscape of the ischemic injury, we performed second-level clustering on the microglia, the primary immune defenders of the brain. The microglia were resolved into 6 distinct phenotypes (Microglia0–Microglia5), distinguishing homeostatic states from disease-associated subsets based on transcriptomic signatures, such as P2ry13 (Microglia0) and Cxcl2 (Microglia1) (Fig. 5A, Figure S3A). Functional phenotyping revealed a clear dichotomy. Microglia0 expressed high levels of homeostatic checkpoint genes, maintaining surveillance functions. In contrast, Microglia1 exhibited a robust chemokine-enriched profile, specifically associated with inflammatory chemotaxis and leukocyte migration. (Fig. 5B, Table S2).

Fig. 5.

I/R+LIJVO exacerbated neuroinflammation by driving the expansion of a specific NF-κB^high microglia subpopulation. (A) The microglia cells were identified as 6 subtypes. (B) GO enrichment analysis of microglia subtypes. (C) Fraction of each cell subtype population relative to all cells in three groups. (D) GO enrichment pathway of differential expression genes in microglia1 between I/R and I/R+LJVO groups. (E) Volcano plots of differential expression genes in microglia1 between I/R and I/R+LJVO groups. (F-G) Representative images and quantitative analysis of immunofluorescence co-staining of NF-κB (P65) (red) and Tmem119 (green) in the ischemic brain (n = 3). Scale bar = 20 μm. LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion. *, P < 0.05; **, P < 0.01

Further analysis uncovered a dramatic “state shift” following I/R+LIJVO. The homeostatic Microglia0, dominant in the Sham group, was severely depleted after ischemia. Conversely, the inflammatory Microglia1 and proliferative Microglia2 subsets emerged rapidly. Crucially, I/R+LIJVO specifically prioritized the expansion of the pro-inflammatory Microglia1 subpopulation, which occupied the largest proportion compared to I/R, suggesting that VO disruption served as a potent amplifier of the specific inflammatory response (Fig. 5C, Figure S3B).

To characterize the molecular signature of this I/R+LIJVO-expanded subset, we analyzed pathway enrichment. GO analysis revealed that Microglia1 in the I/R+LIJVO group displayed a hyper-activated state compared to the I/R group, with significant enrichment in “positive regulation of leukocyte cell-cell adhesion” and “NIK/NF-kappaB signaling”. This pathological state was marked by the transcriptional upregulation of key inflammatory mediators, including NF-κB and Il1b, corroborated by IF staining (Fig. 5D-E, Figure S3C-D). Quantitative IF analysis demonstrated a significant increase in the number of NF-κB⁺/TMEM119⁺ microglia in the I/R+LIJVO group compared with the I/R group (Fig. 5F-G, P < 0.05), indicating enhanced microglial inflammatory activation following LIJVO. These data identify Microglia1 as a highly reactive, I/R+LIJVO-associated subset, whose expansion underscored the severe inflammatory response characterized by NF-κB pathway activation.

I/R+LIJVO promoted the accumulation of a specific Cxcl3⁺ neuronal subpopulation that drove neuron-immune recruitment

To delineate the specific neuronal alterations underpinning the exacerbated neurological deficits, we performed high-resolution sub-clustering on the neuron. This analysis resolved 9 distinct neuronal phenotypes (Neuron0–Neuron8) (Fig. 6A-B). Proportion analysis revealed a dramatic remodeling of the neuronal landscape. While the Sham group was dominated by homeostatic populations (e.g. Neuron0 and Neuron1), ischemic injury triggered the emergence of disease-associated subsets. Most strikingly, Neuron3 was identified as a distinct pro-inflammatory subpopulation, uniquely characterized by the high expression of chemokines Cxcl3 and Cxcl2. In terms of abundance, the proportion of Neuron3 was not merely increased by I/R but was significantly further amplified in the I/R+LIJVO group (Fig. 6C-D, Table S3).

Fig. 6.

I/R+LIJVO promoted the accumulation of a specific Cxcl3⁺ neuronal subpopulation that drove neuron-immune recruitment. (A) The neuronal cells were identified as 9 subtypes. (B) Expression of marker genes for the subtypes. (C) GO enrichment analysis of neuron subtypes. (D) Fraction of each cell subtype population. (E). GO enrichment pathway of differential expression genes in neuron3 between I/R and I/R+LJVO groups. LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion

We next investigated the functional implications of this Neuron3 expansion using GO enrichment analysis. Rather than simply responding to stress, Neuron 3 appeared to function as an active neuro-immune hub. It showed significant enrichment in pathways governing “neutrophil chemotaxis” and “myeloid leukocyte migration” (Fig. 6C-D, Table S3). Crucially, I/R+LIJVO upregulated TNF production and leukocyte regulation pathways specifically within this subset (Fig. 6E). Collectively, these data suggested that I/R+LIJVO aggravates injury by prioritizing the survival or differentiation of Neuron 3, a disease-associated subset, that may actively orchestrate the infiltration of peripheral immune cells into the ischemic brain.

I/R+LIJVO established a pathogenic Endothelial-Microglia crosstalk

To decipher the intercellular communication driving inflammation, we utilized CellChat to map the interaction network among the three identified disease-associated subpopulations: pro-inflammatory Endothelial0, Microglia1, and Neuron3 (Fig. 7A).

Fig. 7.

I/R+LIJVO established a pathogenic Endothelial-Microglia crosstalk. (A) Cell-cell communication among Microglia1, Endothelial1 and Neuron3. (B) Relative information flow of the altered cell communication. (C-D) Dot plot showing the potential interacting ligand-receptor pairs sending from endothelial0 to microglia1 and neuron3. The size of the dots represents the probability of the ligand-receptor pair and the color indicates the communication probability. (E) The relative gene expression of the key ligand-receptor pairs in the cell subtypes. (F) The interaction network of three cell subtypes by MDK-Lrp1 ligand-receptor pairs in the microenvironment. The color of the line represents the signal emitted from the cell with the corresponding color. LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion

Further analysis focused on the specific signaling pathways that were enriched in these altered cellular interactions. Post I/R, the Spp1 and Ccl pathways exhibited the most significant enhancement in communication strength, while VEGF pathways showed a marked decline (Fig. 7B). Comparative pathway analysis revealed a hierarchical activation of signaling networks. While I/R broadly activated the Spp1 and Ccl pathways, I/+LIJVO triggered a distinct inflammatory signature. Specifically, the MK (Midkine), VISFATIN, and COMPLEMENT signaling pathways were uniquely predicted to be active in the I/R+LIJVO group. In contrast, VEGF signaling was markedly diminished, suggesting a shift from reparative angiogenesis to pathological inflammation (Fig. 7B).

We further dissected this network by analyzing 48 ligand-receptor (L-R) pairs, identifying Endothelial0 as a primary signaling hub. In the I/R+LIJVO, Endothelial0 significantly upregulated the expression of inflammatory ligands, including Mdk (MK signaling), Nampt (VISFATIN signaling), and C4a (Complement signaling) (Fig. 7C-D).

Crucially, the receptivity of target cells to these signals diverged sharply. Microglia1 showed a significant upregulation of receptors such as Cd44 (receptor for MDK) and Ccr5, indicating a primed state ready to receive endothelial inflammatory signals. Conversely, Neuron3 exhibited a downregulation of receptors like Lrp1 (an MDK receptor), Cd44, and C3ar1 (Fig. 7E, Figure S4A-E). This implied that Endothelial0 was more predisposed to connect with Microglia1 rather than Neuron3. The MDK-Lrp1 interaction, which was almost inactive in the sham and I/R groups, showed enhanced and more complex crosstalk among subtypes in the I/R+LIJVO group, especially between endothelial and microglia (Fig. 7F).

This differential receptor profile uncovered a critical cellular crosstalk: in the presence of I/R+LIJVO, Endothelial0 is predisposed to establish a high-affinity, pro-inflammatory loop with Microglia1, rather than sustaining communication with Neuron3. This pro-inflammatory Endothelial-to-Microglial signaling axis likely propagated the neuroinflammation and injury aggravation observed in the I/R+LIJVO model.

LIJVO exacerbated the I/R-induced inflammatory environment disorders

To further validate the inflammatory cascades predicted by scRNA-seq, we examined the cytokine and vascular architecture. ELISA analysis confirmed a specific pro-inflammatory profile: brain concentrations of IL-1β and TNF-α were significantly elevated in the I/R+LIJVO group compared to I/R (P < 0.05), although IL-6 levels remained comparable (P > 0.05, Fig. 8A).

Fig. 8.

LIJVO exacerbated the I/R-induced inflammatory environment disorders. (A) The levels of IL-1β, TNF-α and IL-6 in the brain. (n = 5)(B) Representative immunostaining and quantitative analysis of ICAM and VCAM in the peri-infarct within sections from sham, I/R and I/R+LIJVO groups (n = 3). Scale bar = 20 μm. (C) the representative images and Western blotting analysis of the expression of the ZO-1 and Occludin (n = 3). (D-E) Representative images and quantitative analysis of immunofluorescence co-staining of MPO (Red) and CD11b (Green) in the peri-infarct brain (n = 3). Scale bar = 100 μm. (F-G) Representative images and quantitative analysis of immunofluorescence co-staining of NeuN (Red) and Iba-1 (Green) in the peri-infarct brain (n = 3). Scale bar = 100 μm. LIJVO, left internal jugular vein occlusion; I/R, ischemia/reperfusion;IL-1β,interleukin-1β;IL-6, interleukin-6; TNF-α, tumour necrosis factor α; ICAM-1, intercellular adhesion molecule 1; VCAM-1, vascular cell adhesion molecule 1. Data were presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001

This cytokine surge coincided with a pathological activation of the vascular. IF staining revealed a marked upregulation of intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1) in the I/R+LIJVO group. The overexpression of these adhesion molecules serves as a critical mechanism recruiting circulating leukocytes to the luminal surface of the endothelium (P < 0.05, Fig. 8B).

We next determined whether this endothelial activation compromised barrier integrity. As expected, the I/R+LIJVO group exhibited severe vascular leakage, quantified by a profound reduction in the tight junction proteins ZO-1 and Occludin (P < 0.05, Fig. 8C). Consequently, this dysfunctional barrier facilitated the massive influx of peripheral immune cells. Staining for CD11b (general myeloid marker) and MPO (neutrophil marker) demonstrated distinct perivascular clustering and infiltration of leukocytes, which was drastically intensified by I/R+LIJVO (Fig. 8D-E, P < 0.05).

Finally, we assessed the impact of this dysregulated microenvironment on neurons. The combination of barrier loss and leukocyte infiltration created a highly toxic microenvironment, characterized by exacerbated microglial hyperactivation and significant neuronal injury in the I/R+LIJVO group compared to the I/R group (Fig. 8F-G, P < 0.05). These results demonstrated that I/R+LIJVO resulted in more severe inflammation and environmental disorders, including enhanced leukocyte infiltration, microglial activation and neuronal injury.

Collectively, these results substantiated our single-cell findings, confirming that I/R+LIJVO created an inflammatory cycle: elevated cytokines triggered endothelial adhesion, leading to BBB disassembly, which in turn permitted aggressive leukocyte infiltration and resultant neuronal degeneration.

Discussion

Unfavorable cerebral venous outflow acts as an aggravator in IS, yet the mechanism linking it to worsened prognosis has remained obscure. By simulating this clinical scenario using the I/R+LIJVO model, our study confirmed that VO obstruction functioned as a critical pathogenic catalyst, transforming manageable ischemia into a malignant phenotype characterized by unremitting cerebral edema and exacerbated neurological deficits.

Beyond describing these phenotypic outcomes, our analysis proposed the specific cellular mechanism behind this aggravation. We found a paradigm of endothelial-initiated neuroinflammation: cerebral VO obstruction appeared to act as a stressor that drove the emergence of a specific pro-inflammatory endothelial subpopulation (Endothelia0). Marked by mechanosensitive signatures (e.g., Egr1), this endothelial subset serves as the primary transducer, converting hemodynamic stress into a biological distress signal [23, 24].

This specific endothelial subtype subsequently resulted in a pathogenic crosstalk network. Specifically, endothelial0 recruited and amplified a pro-inflammatory axis involving reactive microglia (Microglia1) and chemokine-secreting neurons (Neuron3). The findings suggested that the inflammatory endothelial cells, rather than neurons or glia alone, may as the strategic key point in IS with unfavorable VO, highlighting endothelial immunomodulation as a vital and therapeutic frontier.

Current clinical research increasingly recognizes that venous of cerebral circulation is a vital, yet overlooked, determinant of stroke trajectory. Pivotal studies by Faizy et al. have established that a favorable VO profile is not merely a bystander but an active prerequisite for sustaining collateral perfusion and minimizing tissue swelling post-thrombectomy [6, 12, 25]. This clinical evidence posits that compromised venous drainage acts as a bottleneck, impeding blood transit even when arterial patency is restored.

Translating this concept into animal models requires precise anatomical targeting. While previous study reported that bilateral external jugular vein occlusion exacerbates ischemia injury [26], the internal jugular vein represents the principal venous outflow pathway for the intracranial dural sinuses and plays a more direct role in regulating cerebral venous drainage [27]. Although rodent venous anatomy differs from that of humans and includes extensive collateral drainage pathways, isolated LIJVO did not induce overt brain injury in our study, suggesting that alternative venous drainage routes may partially maintain cerebral venous homeostasis under physiological conditions.

Importantly, laser speckle contrast imaging revealed only a modest reduction in cerebral perfusion following LIJVO, while no significant difference in cerebral blood flow was observed between the I/R and I/R+LIJVO groups at 24 h after reperfusion. These findings suggest that the exacerbation of ischemic injury induced by LIJVO may not be solely attributable to further reductions in bulk cerebral perfusion. Instead, impaired venous outflow may promote endothelial dysfunction, blood–brain barrier disruption, and neuroinflammatory responses, consistent with our transcriptomic analyses and histological observations, including endothelial Egr1 activation, enhanced endothelial–microglial communication, increased inflammatory cytokine expression, and loss of tight-junction proteins.

Therefore, we propose that LIJVO creates an inflow–outflow imbalance that increases the vulnerability of the NVU to ischemic injury, thereby aggravating stroke pathology even in the absence of marked additional reductions in global cerebral perfusion.

A striking paradox emerged from our analysis: while I/R injury led to the expected loss of endothelial cells, I/R+LIJVO triggered a substantial, albeit pathological, expansion of the endothelial compartment. This phenomenon aligns with the “venous hypertension hypothesis,” where mechanical congestion induces vascular proliferation but fails to maintain barrier maturity [28]. Our sub-clustering analysis resolved this expansion to the specific enrichment of Endothelial0, a disease-associated subpopulation virtually absent in the Sham group.

Deep exploration of the Endothelial0 transcriptomic signature found that Egr1 may as a pivotal molecular determinant. Egr1 serves as an immediate-early regulator sensitive to both hypoxia and hemodynamic shear stress, where two conditions exacerbated by venous occlusion [23, 29]. We postulate that the I/R+LIJVO sustained over-activates Egr1, which upregulation has profound functional consequences. Literature corroborates that Egr1 acts as a transcriptional gateway for inflammatory cascades, directly driving the expression of downstream effectors such as TNF-α, IL-6, and adhesion molecules (ICAM-1/CD44) [30, 31]. Consistent with a study in meningitic E. coli [32], where Egr1 governs BBB initiation of neuroinflammation, our data suggested that during I/R+LIJVO, Egr1 transforms the endothelium from a passive barrier into an active inflammatory mediator.

Consequently, the accumulation of the endothelial0 population provides a cellular explanation for the severe BBB disruption and leukocyte infiltration observed in our model. Rather than serving a reparative role, this endothelial subset represents a dysregulated, sticky phenotype that propagates the inflammatory injury.

Beyond the endothelial barrier, the impact of VO obstruction cascaded rapidly into the brain’s immune environment. While endothelial dysfunction initiates the breach, our data suggest that microglia serve as the crucial amplifiers of this injury. Specifically, the I/R+LIJVO environment selectively prioritized the expansion of the inflammation-associated Microglia1 subpopulation, transforming the local landscape into a hostile, pro-inflammatory milieu.

Transcriptional profiling of Microglia1 revealed a distinct hyper-reactive signature, anchored by the marked upregulation of Nfkb1 and Il1b. Nuclear Factor Kappa B (NF-κB) functions as the central regulator for inflammatory storms [33]. In I/R+LIJVO, the activation of this pathway, potentially fueled by the same signals (such as Egr1) that triggered the endothelium [34, 35], propels the sustained secretion of cytotoxic cytokines like IL-6, IL-1β, and TNF-α [36, 37].

The consequence of this pro-inflammatory microglial explosion is evident in the neuron. Coinciding with the expansion of NF-κB-driven microglia, we observed a specific susceptibility in Neuron3. While most neurons possess some resilience, this specific subset appears to lack the defensive buffer against inflammatory barrage induced by Microglia1. This Glial-Neuronal Mismatch, where immune hyperactivity directly targets specific vulnerable neuronal clusters, likely underpins the irreversible neurological deficits observed in I/R+LIJVO.

Ultimately, our interaction analysis suggested a remodel of the NVU interactome under I/R+LIJVO. The most defining feature of I/R+LIJVO was the establishment of a hyper-active Endothelial-Microglial axis. We observed a two-fold surge in the interaction frequency between Endothelial0 and Microglia1 compared to I/R. This quantitative jump implies that VO disruption forces the endothelium to shift from a blood-brain barrier guardian into a broadcaster of inflammatory signals.

This axis appears to impact the Midkine (MDK) signaling pathway. While physiological MDK often supports neuronal growth [38], the I/R+LIJVO injury repurposes it for pathology. Endothelial0 emerged as the dominant MDK source, targeting Microglia1 (and to a lesser extent, Neuron3) via the LRP1 receptor. Consistent with evidence linking endothelial MDK to leukocyte trafficking [39, 40], this Endothelial0-MDK to Microglia1-LRP1 loop may function as a key driver of the adhesive inflammation we observed.

This interaction constructs a self-perpetuating cycle of injury. By saturating the microenvironment with MDK, cytokines (Tnfsf12) and chemokines (Ccl3), the Endothelial0-Microglia1 interaction amplifies neurotoxic aggression. Neuron3 serves as the specific victim in vicious cycle. Critically, this neuronal subset loses homeostatic trophic support from the endothelium while being simultaneously targeted by inflammatory attack. Deprived of protection and besieged by signaling stress, Neuron3 succumbs to dysfunction, offering a cellular explanation for the severe deficits seen in IS with unfavorable VO.

Study limitation

While this study offers a novel cellular landscape of LIJVO-mediated exacerbation in stroke, we acknowledge several limitations that warrant future investigation. First, our interaction analysis prioritized the “Endothelial-Microglia-Neuron” triad due to their profound transcriptomic remodeling. However, the NVU functions as a holistic entity. We recognize that other resident populations, particularly astrocytes and pericytes, likely contribute to this pathogenic network, yet their specific roles under I/R+LIJVO were not the primary focus of this analysis.

Second, the inherent nature of single-cell dissociation leads to the loss of spatial architecture. Although we identified pathogenic subsets like Endothelial0, we cannot map their precise distribution relative to the ischemic core versus the penumbra. Future integration with spatial transcriptomics is essential to visualize these inflammatory environments in situ and correlate cell location with regional hemodynamic stress.

Third, our study represents a snapshot of the acute phase. Stroke pathology is highly dynamic, whether the pro-inflammatory “Endothelial0-Microglia1” axis persists into the chronic phase or hinders long-term repair remains to be defined. Further studies tracking cellular dynamics across multiple time points are necessary to fully elucidate the temporal evolution of this I/R+LIJVO-driven pathology.

Finally, while our bioinformatic analyses predict robust ligand-receptor signaling, such as MDK-LRP1, these interactions are inferred from transcriptomic data. To definitively establish causality, functional validation using conditional knockout models or specific pathway inhibitors is required. We aim to bridge this gap in future work to identify precise therapeutic targets for clinical intervention.

Conclusion

In summary, this study elucidates that unfavorable VO functions as a critical aggravator in IS. Specifically, we identified that VO disturbance induced a pro-inflammatory endothelial environment that served as the igniter for microglial hyperactivation and subsequent neuronal degeneration. Therapeutic strategies aiming to intercept this upstream endothelial activation may offer superior clinical efficacy for IS patients with unfavorable cerebral venous drainage. This work provides the cellular blueprint necessary to translate these venous-centric insights into precision medicine.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

IS

Ischemic stroke

VO

Venous outflow

scRNA-seq

Single-cell RNA sequencing

snRNA-seq

Single-nucleus RNA sequencing

I/R

Ischemia/reperfusion

LIJVO

Left internal jugular vein occlusion

BBB

Blood-Brain barrier

EB

Evans blue

DEGs

Differentially expressed genes

GO

Gene ontology

IL-1β

Interleukin-1β

IL-6

Interleukin-6

TNF-α

Tumour necrosis factorα

GO

Gene ontology

Egr1

Early growth response 1

NF-κB

Nuclear Factor Kappa B Subunit 1

Author contributions

SY Q, Conceptualization, Investigation, Methodology, Formal analysis, Visualization, Writing-original draft; Y Z, Software, Formal analysis, Visualization, Writing-original draft.; GQ C, Methodology; MZ L, Validation; LX W, Validation; WB G, Validation; X S, Supervision; J L, Supervision; QH F, Validation; SS C, Writing-original draft, Supervision, Funding acquisition; XH F, Resources, Supervision, Funding acquisition; GQ Z, Conceptualization, Resources, Supervision, Funding acquisition.

Funding

This work was supported by grants from the National Natural Science Foundation of China (81973657, 82304883) and the Major Subject of the Zhejiang Provincial Natural Science Foundation of China (D23H270001).

Data availability

All relevant data are available from the authors upon request. The raw data files of RNA-seq have been deposited in the Gene Expression Omnibus (GEO) with the following association number: GSE281830. Supplementary data to this article can be found online.

Declarations

Ethics approval and consent to participate

Under the approval of No. 20240429-23, all related experimental animal protocols followed the “Guide for the Care and Use of Laboratory Animals”, which was approved and regulated by the Zhejiang Chinese Medical University’s Institutional Animal Care and Ethical Committee.

Consent for publication

Each author has seen and approved the contents of the submitted manuscript.

Competing interests

The authors declare no competing interests.