Blood–brain barrier disruption and neuroinflammation in the hippocampus of a cardiac arrest porcine model: Single-cell RNA sequencing analysis

Abstract

Global brain ischemia and neurological deficit are consequences of cardiac arrest that lead to high mortality. Despite advancements in resuscitation science, our limited understanding of the cellular and molecular mechanisms underlying post-cardiac arrest brain injury have hindered the development of effective neuroprotective strategies. Previous studies primarily focused on neuronal death, potentially overlooking the contributions of non-neuronal cells and intercellular communication to the pathophysiology of cardiac arrest-induced brain injury. To address these gaps, we hypothesized that single-cell transcriptomic analysis could uncover previously unidentified cellular subpopulations, altered cell communication networks, and novel molecular mechanisms involved in post–cardiac arrest brain injury. In this study, we performed a single-cell transcriptomic analysis of the hippocampus from pigs with ventricular fibrillation-induced cardiac arrest at 6 and 24 hours following the return of spontaneous circulation, and from sham control pigs. Sequencing results revealed changes in the proportions of different cell types, suggesting post-arrest disruption in the blood–brain barrier and infiltration of neutrophils. These results were validated through western blotting, quantitative reverse transcription-polymerase chain reaction, and immunofluorescence staining. We also identified and validated a unique subcluster of activated microglia with high expression of S100A8, which increased over time following cardiac arrest. This subcluster simultaneously exhibited significant M1/M2 polarization and expressed key functional genes related to chemokines and interleukins. Additionally, we revealed the post-cardiac arrest dysfunction of oligodendrocytes and the differentiation of oligodendrocyte precursor cells into oligodendrocytes. Cell communication analysis identified enhanced post–cardiac arrest communication between neutrophils and microglia that was mediated by neutrophil-derived resistin, driving pro-inflammatory microglial polarization. Our findings provide a comprehensive single-cell map of the post-cardiac arrest hippocampus, offering potential novel targets for neuroprotection and repair following cardiac arrest.

Introduction

The high incidence and generally poor survival rates of out-of-hospital cardiac arrest (OHCA) represent a significant global public health challenge. It is estimated that the annual incidence of OHCA treated by emergency medical services ranges from 28 to 244 cases per 100,000 people across different countries, with discharge survival rates varying between 3.1% and 20.4% (Nishiyama et al., 2023). Post-resuscitation brain injury is a common and severe sequela of cardiac arrest (CA), potentially leading to cerebral edema with mass effect, impaired cerebral blood flow, brain herniation, and brain death (Sandroni et al., 2021). Despite successful initial resuscitation, up to 70% of patients admitted to a hospital die from the effects of brain injury following CA (Perkins et al., 2021). Approximately 10% of all deaths among hospital-admitted OHCA patients are specifically attributed to severe brain injuries (Sandroni et al., 2016).

Following CA, the pathophysiology of brain injury primarily involves two stages: the initial ischemic damage and the subsequent reperfusion injury (Sekhon et al., 2017; Sandroni et al., 2021). The disruption of cerebral blood flow immediately leads to the cessation of aerobic metabolism, followed by rapid depletion of the energy storage molecule ATP, which in turn triggers dysfunction in energy-dependent ion pumps. This results in a massive intracellular accumulation of sodium and water, ultimately causing cytotoxic edema. As resuscitation begins, partial restoration of cerebral blood flow occurs; however, this triggers a series of biochemical reactions that further contribute to brain injury. These include activation of Ca²⁺-dependent lysosomal enzymes following the initial damage-induced increase in intracellular Ca²⁺, resulting in excessive release of glutamate and subsequent further influx and accumulation of Ca²⁺, thereby exacerbating neuronal damage (Silver and Erecinska, 1992). Furthermore, reperfusion injury is also associated with a significant neuroinflammatory response involving glial activation, peripheral immune cell infiltration, and the release of pro-inflammatory mediators, such as cytokines and adhesion molecules. Innate immune cells, including microglia and macrophages, are rapidly activated post ischemia, releasing cytotoxic substances that trigger inflammation. This response can induce programmed cell death (e.g., pyroptosis and necroptosis), further amplifying inflammation and exacerbating secondary brain injury (Stoll et al., 1998; Zhang et al., 2023). Reperfusion injury exacerbates neuroinflammation by disrupting the blood–brain barrier (BBB), allowing peripheral immune cells (e.g., lymphocytes) to infiltrate the brain parenchyma during early reperfusion, compounding the ischemia-induced damage (Zhang et al., 2018; Hosseini et al., 2020).

Current therapies for post-CA brain injury have shown limited benefits (Papastylianou and Mentzelopoulos, 2012), while primarily narrowly focused research (e.g., neuronal death mechanisms) has overlooked the intricate cellular organizational levels and gene expression locales (Rosenthal et al., 2003; Niizuma et al., 2008). The changes in cellular population structures, temporal dynamics, and intercellular communication networks resulting from CA-induced global cerebral ischemia remain underexplored. A comprehensive brain cell atlas is essential to map the effects of CA with single-cell precision, identify cellular subpopulations, and investigate its impacts on cell–cell communication.

To address critical gaps in understanding the post-CA cerebral landscape, we employed single-cell RNA sequencing (scRNA-seq) of a porcine model of induced CA and resuscitation to map the cellular composition, population dynamics, and intercellular communication networks in the hippocampus, a region highly susceptible to ischemia–reperfusion injury (Petito et al., 1987; Calle et al., 1989; Fujioka et al., 2000; Haglund et al., 2019; van Putten et al., 2019; Davidson and Stevenson, 2024). Our findings help uncover mechanisms driving post-CA brain injury and provide insights to guide the development of novel therapeutic strategies to mitigate neuroinflammation and promote recovery.

Methods

Ethics statement

This study was approved by the Animal Care and Use Committee of Qilu Hospital, Shandong University (approval No. DWLL-2023-181, January 5, 2024). All experiments and tissue sampling procedures were designed and reported in accordance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines (Percie du Sert et al., 2020), and were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (8^th ed., National Research Council, 2011). Tissue sampling procedures strictly adhered to all applicable institutional and national animal care and welfare guidelines.

Ventricular fibrillation cardiac arrest and resuscitation pig model

We used 11 healthy male and female Landrace pigs (age: 14–16 weeks; body weight: 35–40 kg) purchased from the Xilingjiao Breeding Center in Jinan, China (License No. SCXK (Lu) 2020-0004). The pigs were of clean grade, housed under controlled conditions (20–25°C, 50%–70% humidity, 12-hour light/dark cycle), and had free access to food and water. The animals were randomly assigned to the following three groups using a computer-generated randomization schedule: sham control (Sham; n = 5), return of spontaneous circulation (ROSC) for 6 hours (ROSC6h; n = 3), and ROSC for 24 hours (ROSC24h; n = 3). Anesthesia was induced via ear vein injection of 2 mg/kg propofol (Libang Pharmaceutical Company, Xi’an, China). The animals were intubated with a 6.5-mm endotracheal tube connected to a volume-controlled ventilator (R419, RWD Life Science, Shenzhen, China), which was set to deliver a tidal volume of 10 mL/kg, an inspiration-to-expiration ratio of 1:2, and an inspired oxygen fraction of 21%. After initiating mechanical ventilation, anesthesia was maintained with 2% isoflurane (RWD Life Science). A fluid-filled catheter was advanced from the left femoral artery into the thoracic aorta to measure aortic pressure. Right atrial pressure was measured using a Swan-Ganz catheter (131F7, Edwards Life Sciences, Irvine, CA, USA), advanced from the left femoral vein into the pulmonary artery. Additionally, a pacing catheter (1300F85, Edwards Life Sciences) was advanced from the right internal jugular vein into the right ventricle to induce ventricular fibrillation (VF) using a programmable electrical stimulator with pacing electrodes (KST 2-CH, Kardiotek, Shanghai, China) set to an amplitude of 10,000 mV, a pulse width of 1 ms, and a 20-ms interval. VF was diagnosed based on characteristic electrocardiographic waveforms and a rapid drop in arterial blood pressure. Upon successful induction of VF, mechanical ventilation was stopped. After 6 minutes of untreated VF, cardiopulmonary resuscitation (CPR) was initiated using the LUCAS2 device to provide consistency of compressions at a depth of 5 cm and a target rate of 100 compressions per minute. The compression-to-ventilation ratio was 30:2 with 100% oxygen. Defibrillation was performed 3 minutes after the start of CPR using a BeneHeart D3 Defibrillator (Mindray, Mahwah, NJ, USA). If VF persisted, defibrillation was repeated every 2 minutes. If the initial two defibrillation attempts were unsuccessful, 1 mg of epinephrine was administered intravenously, followed by continuous CPR. Additional doses of epinephrine were given every 3 minutes if necessary until ROSC was achieved, as defined by a systolic blood pressure > 60 mmHg for more than 10 minutes. Failure to achieve ROSC within 30 minutes was considered a resuscitation failure. Upon achieving ROSC, the animals underwent a 2-hour intensive care period with mechanical ventilation set to pre-CA parameters. Standard post-CA intensive care included titration of oxygen concentration, tidal volume, and ventilation rate to maintain blood oxygen saturation at 90%–100%. Anesthesia was maintained with inhaled isoflurane (1.0%–2.5%). At 2 hours post ROSC, the vascular catheters were removed, and the animals were extubated and allowed to recover from anesthesia in observation cages, where they were monitored at the specified time points. Sham group animals underwent the same anesthesia duration without CA or CPR. At the specified post-CA time points, the animals were re-anesthetized, a craniotomy was performed, and hippocampal tissue was rapidly extracted (Figure 1A). The detailed methods for tissue dissociation and cell purification are shown in Additional file 1 ^{(149.4KB, pdf)} .

Figure 1.

Single-cell atlas of hippocampus in cardiac arrest and resuscitation pig model.

(A) Schematic representation of the experimental procedure to induce ventricular fibrillation by electrocution, followed by cardiopulmonary resuscitation, in pigs. Hippocampal tissue harvested from euthanized animals was dissociated and subjected to scRNA-seq. (B) Representative images of TUNEL staining and quantification of the percentage of TUNEL-positive cells (n = 3 pigs; 2–3 slices per pig). Values presented as means ± standard deviation. Statistical analysis was performed using one-way analysis of variance followed by Tukey’s post hoc test; ****P < 0.0001. (C) UMAP visualization of scRNA-seq data from the hippocampus, showing 26 distinct clusters. (D) Ten distinct cell types identified in the hippocampus. (E) Pie chart showing the proportions of different cell types identified across all samples. (F) Bar graphs of cell type proportions displayed by cell type (left) and by group (right). (G) Box plots showing the relative proportions of various cell types between the ROSC6h vs. Sham groups (left) and the ROSC24h vs. Sham groups (right), as calculated by the scCODA model. ROSC: Return of spontaneous circulation; scRNA-seq: single-cell RNA sequencing; TUNEL: TdT-mediated dUTP nick end labeling UMAP: Uniform Manifold Approximation and Projection.

BV2 cell culture

It is infeasible to isolate and culture primary microglia from pigs. Thus, we used BV2 mouse microglial cells, a well-established in vitro model (Henn et al., 2009) and a practical, reproducible, and species-compatible alternative for studying resistin-mediated signaling with mouse neutrophils. BV2 cells (iCell Biosciences, Shanghai, China; Cat# iCell-m011, RRID: CVCL_0182) were cultured in high-glucose DMEM (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. Cells were maintained in T75 flasks at 37°C in a humidified atmosphere containing 5% CO₂ when not in use for experiments.

Oxygen–glucose deprivation and reoxygenation

The culture medium was replaced with glucose-free DMEM (Thermo Fisher Scientific, Waltham, MA, USA; Cat# 11966025), which is modified to include L-glutamine, phenol red, and HEPES, but not glucose or sodium pyruvate. BV2 cells were placed in a 37°C incubator with a humidified atmosphere of 1% O₂, 5% CO₂, and 94% N₂ for 3 hours of oxygen deprivation. The cells were then transferred to normoxic conditions (21% O₂, 5% CO₂) in a standard cell culture incubator at 37°C and cultured in high-glucose DMEM supplemented with 10% fetal bovine serum to terminate oxygen–glucose deprivation and simulate reoxygenation.

Neutrophil treatment and neutrophil–microglia co-culture

Mouse peripheral blood neutrophils and their complete culture medium (Cat# CM-M150) were purchased from Pricella Biotechnology (Wuhan, China). This preformulated ready-to-use medium, containing a combination of basal medium, various growth factors, and antibiotics, does not require additional serum or additives. To activate neutrophils, lipopolysaccharide (LPS) (L2880, Sigma-Aldrich) was added to the neutrophil culture medium at a concentration of 100 ng/mL for 12 hours (Hou et al., 2019). The cells were then collected for phenotypic verification.

For the co-culture system, which allows for direct cell–cell communication between neutrophils and microglia, BV2 cells were cultured in 24-well plates and subjected to 3 hours of oxygen–glucose deprivation (OGD). Thereafter, activated neutrophils were added at a ~1:1 ratio and incubated under normoxic conditions. The neutrophils, as suspension cells, were then removed from the co-culture system by gently washing with medium, leaving behind microglia for subsequent experiments.

Immunofluorescence staining

For immunostaining, BV2 cells grown on a cover glass were fixed with 4% paraformaldehyde, while hippocampal tissue sections and cell slides were incubated with 0.1% Triton X-100 and blocked with 5% goat serum for 1 hour at room temperature. They were then incubated with the following primary antibodies at 4°C overnight: rabbit anti-ionized calcium binding adaptor molecule 1 (IBA1) (1:200 dilution; Abcam, Cambridge, UK, Cat# ab178846, RRID: AB_2636859); mouse anti-S100A8 (1:200; Proteintech, Wuhan, China, Cat# 66853-1-Ig, RRID: AB_2882193); rabbit anti-myelin basic protein (MBP) (1:50, Cell Signaling Technology, Danvers, MA, USA; Cat# 78896, RRID: AB_2799920); rabbit anti-myeloperoxidase (MPO) (1:100, Abcam, Cat# ab208670, RRID: AB_2864724); mouse anti-IBA1 (1:200, Santa Cruz Biotechnology, Santa Cruz, CA, USA, Cat# sc-32725, RRID: AB_667733); and rabbit anti-inducible nitric oxide synthase (iNOS) (1:200, Proteintech, Cat# 18985-1-AP, RRID: AB_2782960). After washing, the sections were incubated with the following appropriate secondary antibodies for 2 hours at 25°C: goat anti-mouse IgG (H+L), Alexa Fluor 488 (1:200, Abcam; Cat# ab150117, RRID: AB_2688012) and goat anti-rabbit IgG (H+L), Alexa Fluor 594 (1:200, Abcam; Cat# ab150080, RRID: AB_2650602). Finally, 4′,6-diamidino-2-phenylindole (DAPI) (Abcam; Cat# ab104139) was used to stain the nucleus. The images were observed under fluorescence or confocal microscopy (Olympus, Tokyo, Japan). Quantification of the S100A8⁺IBA1⁺/IBA1⁺ cell proportion and MBP fluorescence intensity was measured using ImageJ software version 1.8.0 (National Institutes of Health, Bethesda, MD, USA).

TdT-mediated dUTP nick end labeling staining

Hippocampal tissue sections from pigs were fixed in 4% paraformaldehyde, dehydrated, embedded in paraffin, and sectioned at 4 μm thickness using a microtome. The sections were deparaffinized through a series of graded ethanol and clearing solutions, rehydrated in distilled water, and then stained using a TdT-mediated dUTP nick end labeling (TUNEL) assay kit (Servicebio, Wuhan, China, Cat# G1507) in accordance with the manufacturer’s instructions. This procedure included proteinase K treatment for antigen retrieval, permeabilization of cell membranes, and incubation in the TUNEL reaction mixture followed by streptavidin-horseradish peroxidase (HRP) reaction solution (Servicebio, Cat# G1507-4). Visualization was achieved using 3,3′-diaminobenzidine staining (Servicebio, Cat# G1212). Hematoxylin (Servicebio, Cat# G1004) was used to counterstain nuclei. Finally, the samples were dehydrated, mounted, and observed under an optic microscope (Olympus). Quantification of the percentage of TUNEL-positive cells was measured using ImageJ (version 1.8.0).

Enzyme-linked immunosorbent assay

Resistin levels in culture supernatants collected from neutrophils, BV2 microglial cells, or the neutrophil–BV2 co-culture system were measured using an enzyme-linked immunosorbent assay (ELISA) kit (Boster, Wuhan, China, Cat# EK0582), following the manufacturer’s instructions. Culture supernatants were diluted 1:10 to fall within the detection range of the assay. Absorbance was read at 450 nm using a microplate reader (EON, BioTek Instruments, Winooski, VT, USA).

Western blotting

Protein samples extracted from hippocampal tissue were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Beyotime Biotechnology, Shanghai, China, Cat# P0015) and transferred to polyvinylidene fluoride membranes (Merck Millipore, Burlington, MA, USA, Cat# ISEQ00010). The membranes were blocked with 5% skim milk and incubated overnight at 4°C with the following primary antibodies: rabbit anti-MPO (1:1000, Abcam, Cat# ab208670, RRID: AB_2864724); rabbit anti-Claudin 5 (1:1000, Abcam, Cat# ab131259, RRID: AB_11157940); rabbit anti-MBP (1:50, Cell Signaling Technology, Cat# 78896, RRID: AB_2799920); and mouse anti-β-actin (1:1000, Proteintech, Cat# 66009-1-Ig, RRID: AB_2687938). The membranes were then incubated for 2 hours at room temperature with HRP-conjugated secondary antibodies, including goat anti-rabbit IgG (H+L) (1:5000, Proteintech; Cat# SA00001-2, RRID: AB_2722564) and goat anti-mouse IgG (H+L) (1:5000, Proteintech, Cat# SA00001-1, RRID: AB_2722565). Finally, the membranes were scanned and detected using an enhanced chemiluminescence assay kit (Merck Millipore, Cat# P90718). The gray values of the target protein bands were normalized to the internal loading control, β-actin, and band densities were quantified using ImageJ software (version 1.8.0).

Quantitative reverse transcription-polymerase chain reaction

Hippocampal tissue was mechanically homogenized, and total RNA was extracted using TRIzol Reagent (Sigma, Cat# T9424). Reverse transcription into cDNA was performed using the Reverse Transcription System Kit (Vazyme, Nanjing, China, Cat# R223-01). The synthesized cDNA was then amplified using a standard quantitative reverse transcription-polymerase chain reaction (qRT-PCR) protocol, employing ChamQ Universal SYBR qPCR Master Mix (Vazyme, Q711-02). Amplification products were normalized against β-actin mRNA, which served as an internal control in the same reaction. Data were analyzed using the 2^–∆∆Ct method, and a fold change (FC) in mRNA expression was quantified relative to the β-actin control. The primers used for qRT-PCR (BioSune, Shanghai, China) are listed in Additional Table 1.

Additional Table 1.

Primers used for quantitative reverse transcription-polymerase chain reaction

Primer Name	Sequence (5' -> 3')
Sus-MBP-83F	AGCAGATTTAGCTGGGGCG
Sus-MBP-83R	TTGTGAGCCGGTTTGTAGTCA
Sus-PLP-80F	TTTGGAGCGGGTGTGTCATT
Sus-PLP-80R	CGGTCAGGGCATAGGTGATG
Sus-ACTB-200F	GATATTGCTGCGCTCGTGGTC
Sus-ACTB-200R	GGGTACTTGAGGGTCAGGATG
Sus-PDGFRA-161F	GATTAAGCCGGTCCCAACCT
Sus-PDGFRA-161R	TGACTTGGTACAGGCTCCCA
Sus-CNP-163F	CTGGTGGCCTACTTTGGGAA
Sus-CNP-163R	TCACAAAGAGCGCGGAGATG
Sus-GALC-181F	CAACGCCGGGTGAAAGTCA
Sus-GALC-181R	GAAGTCGGGAGGTTGCCC
Sus-NG2-78F	TGTGTCCGTGTCTTTCGAGG
Sus-NG2-78R	TCGGAGACCCAGAGACCTT

scRNA-seq and analysis

scRNA-seq was performed with the Chromium Single Cell 3′ v3.1 kit (10x Genomics, Pleasanton, CA, USA) and sequenced on an Illumina NovaSeq X Plus platform (San Diego, CA, USA). Data were processed using Cell Ranger (v7.1.0) (10x Genomics) with alignment to the Sus scrofa genome (Sscrofa11.1). Quality control, normalization, clustering, and Uniform Manifold Approximation and Projection (UMAP) visualization were performed in the R package Seurat (v4.1.1), created by Robert Gentleman and Ross Ihaka. Low-quality cells were excluded based on transcript counts, number of detected genes, and mitochondrial gene percentages (means ± 2 standard deviation [SD] thresholds), as shown in detail in Additional file 1 ^{(149.4KB, pdf)} .

Cell-type identification

Cell identities were annotated using the cell type-specific markers in the Pig Single-Cell Transcriptome Atlas and Cell Taxonomy database (https://dreamapp.biomed.au.dk/pigatlas/). The detailed methods are shown in Additional file 1 ^{(149.4KB, pdf)} .

Compositional and differential analysis

Cell type composition changes were analyzed using scCODA, a Bayesian model, applying a false discovery rate (FDR) threshold of 0.4. Differentially expressed genes (DEGs) were identified using Seurat’s FindMarkers function (|log2 FC| > 0.25, Q ≤ 0.05). Gene Ontology (GO) enrichment analysis was performed using Goatools (Python package, Fujian Agriculture and Forestry University, Fuzhou, China) (Bonferroni-corrected P ≤ 0.05). The detailed methods are shown in Additional file 1 ^{(149.4KB, pdf)} .

Pseudotime and cell–cell interaction analysis

Trajectory analysis was conducted using Monocle 3 (Trapnell Lab, University of Washington, Seattle, WA, USA). Cell–cell interactions were evaluated using CellChat (Jin & Nie Labs, University of California, Irvine, CA, USA), focusing on secreted signaling, extracellular matrix (ECM)–receptor interactions, and cell–cell contact. The detailed methods are shown in Additional file 1 ^{(149.4KB, pdf)} .

Statistical analysis

No statistical methods were used to predetermine sample sizes, which were similar to those reported in previous publications (Zheng et al., 2022; Ma et al., 2023; Zhang et al., 2024). All data are presented as means ± SD. Cell type proportions across the Sham, ROSC6h, and ROSC24h groups were analyzed using two methods: 1) two-Proportion Z-Test to compare the relative proportions of specific cell types between groups, with a significance threshold of P < 0.05; and 2) scCODA, a Bayesian model for compositional analysis, to address biases in raw proportion comparisons and identify significant changes in cell type abundance (FDR threshold: 0.4). For comparisons involving more than two groups, we used one-way analysis of variance followed by Tukey’s post hoc test. A P value < 0.05 was considered statistically significant. All analyses were performed using GraphPad Prism 10.0 software.

Results

Increased apoptosis in the hippocampus following cardiac arrest

We assessed the CA-induced pathological changes in the pig hippocampus using TUNEL staining (Figure 1B). The results showed significantly more TUNEL-positive cells in both the ROSC6h and ROSC24h groups (both P < 0.0001 vs. Sham group), with a significantly greater increase in ROSC24h than in ROSC6h (P < 0.0001), suggesting that CA induces progressive damage to the hippocampus.

Major brain cell types in the hippocampus

Utilizing the 10x Genomics platform to perform high-throughput scRNA-seq, we examined the transcriptional profiles of hippocampal cells in the Sham and CA model groups. After processing, we generated 90,176 single-cell gene expression profiles: 39,665 from the Sham group, 27,997 from the ROSC6h group, and 22,514 from the ROSC24h group (Additional Table 2). We employed UMAP to visualize all sample data in two dimensions for cellular heterogeneity analysis, identifying 26 clusters (Figure 1C; Becht et al., 2018). Leveraging cell type-specific markers from a pig single-cell atlases (https://dreamapp.biomed.au.dk/pigatlas/), as well as classical markers of known cell types from the Cell Taxonomy database (https://ngdc.cncb.ac.cn/celltaxonomy/), we annotated the single-cell data, identifying 10 distinct cell types (Additional Table 3; Wang et al., 2022; Jiang et al., 2023). These included four glial classes (microglia, oligodendrocytes, oligodendrocyte progenitor cells (OPCs), and astrocytes), two leukocyte classes (neutrophils and T cells), three vascular classes (endothelial cells, smooth muscle cells and fibroblasts), and neurons (Figure 1D). Notably, among the identified cell types, microglia and oligodendrocytes were the most prevalent, accounting for 49.47% and 38.53% of the total cells, respectively (Figure 1E).

Additional Table 2.

Cell counts of different cell types in sham and post-cardiac arrest groups at two time points

Group	Cell type	Cell num	Group num	Celltype num	Ratio group (%)	Ratio Celltype (%)
Sham	Fibroblasts	47	39665	63	43.99	0.07
ROSC6h	Endothelial cells	679	27997	3782	31.05	4.19
Sham	Oligodendrocytes_progenitor_cells	271	39665	509	43.99	0.56
ROSC24h	Fibroblasts	10	22514	63	24.97	0.07
ROSC24h	Smooth muscle cells	52	22514	1229	24.97	1.36
ROSC24h	Neutrophils	505	22514	957	24.97	1.06
ROSC6h	Fibroblasts	6	27997	63	31.05	0.07
ROSC6h	Neurons	127	27997	402	31.05	0.45
ROSC6h	Oligodendrocyte cells	10983	27997	34743	31.05	38.53
ROSC6h	Astrocytes	437	27997	2417	31.05	2.68
Sham	Astrocytes	1736	39665	2417	43.99	2.68
ROSC24h	Neurons	112	22514	402	24.97	0.45
ROSC6h	Neutrophils	331	27997	957	31.05	1.06
ROSC24h	T cells	594	22514	1468	24.97	1.63
Sham	T cells	239	39665	1468	43.99	1.63
ROSC24h	Astrocytes	244	22514	2417	24.97	2.68
ROSC6h	Oligodendrocytes_progenitor_cells	129	27997	509	31.05	0.56
ROSC24h	Endothelial cells	205	22514	3782	24.97	4.19
Sham	Oligodendrocyte cells	14254	39665	34743	43.99	38.53
Sham	Endothelial cells	2898	39665	3782	43.99	4.19
Sham	Neurons	163	39665	402	43.99	0.45
Sham	Neutrophils	121	39665	957	43.99	1.06
ROSC24h	Oligodendrocyte cells	9506	22514	34743	24.97	38.53
ROSC6h	T cells	635	27997	1468	31.05	1.63
Sham	Microglias	19066	39665	44606	43.99	49.47
ROSC24h	Oligodendrocytes_progenitor_cells	109	22514	509	24.97	0.56
ROSC6h	Smooth muscle cells	307	27997	1229	31.05	1.36
Sham	Smooth muscle cells	870	39665	1229	43.99	1.36
ROSC24h	Microglias	11177	22514	44606	24.97	49.47
ROSC6h	Microglias	14363	27997	44606	31.05	49.47

Cell_num: Number of cells of this type in the sample.

Group_num: Total number of cells in the specific group.

Celltype_num: Total number of cells of this cell type across all samples.

Ratio_group(%): Percentage of cells from this group out of the total number of cells in the sample.

Ratio_Celltype(%): Percentage of this specific cell type out of the total number of cells of this cell type across all samples.

Additional Table 3.

List of marker genes used for determining cell-type identity

Astrocytes	Neurons	Microglia	Oligodendrocytes	Oligodendrocyte_precursor_cells	T_cells	B_cells	Endothelial_cells	Fibroblasts	Epithelial_cells	Macrophage_M onocyte	NK_cells	Neutrophils
AQP4	CAMK2A	TREM2	MBP	PDGFRA	CD3D	CD79A	PECAM1	PDGFRA	EPCAM	CD68	XCL1	S100A8
GFAP	CBLN2	C1QA	MOG	GPR37L1	CD3E	CD79B	VWF	DCN	CDH1	MS4A7	XCL2	S100A9
S100B	LDB2	TMEM119	MAG	CDO1	CD3G	MZB1	CLDN5	COL1A1	CD24	CD163	KLRB1	CSF3R
ALDOC	NEUROD2	FCRLS	PLP1	CSPG4	CD4	MS4A1	ADGRL4	COL3A1	KRT4	CD14	KLRC1	CCR1
GPC5	NEUROD6	P2RY12	ST18	EPN2	CD8A	MKI67	CDH5		KRT8	FCGR3A	KLRD1	CCR2
SLC1A2	SLC17A6	CTSL	CLDN11	LHFPL3		NEIL1	RNASE1		KRT18	MRC1	TRDC
SLC1A3	SLC17A7	APOE	CNP39	PCDH15			PCDH17		ESR1		NKG7
SOX9	SATB2	IBA1	OLIG1	C1QL1			FABP4		PGR
ALDH1L1	CAMK1	CD45	OLIG2	SOX10
ATP1B2	ATP1A1	CX3CR1	ZDHHC9
	SDK2	CD74	GALC
	ETL4	C3	KLK6
	DCX	LRMDA	OPALIN
	ELAVL4	DOCK8	SEC14L5
	MAP1B	CSF1R	UGT8A
	NEUROD1	ADRB2	SOX8
	RTN1	GPR183	GJB1
	SCG5	SALL1	MOBP
	SNAP25	CCL4	OMG
	TAGLN3		ASPA

Dynamic cellular changes following cardiac arrest implicate blood–brain barrier disruption and leukocyte infiltration

The distribution and relative abundance of the different cell categories across the ROSC6h, ROSC24h, and Sham groups are described in Figure 1F, G and Additional Tables 2–4. The two-proportion Z-test showed that, compared with the Sham group, the proportions of astrocytes, endothelial cells, smooth muscle cells, fibroblasts, oligodendrocytes, and OPCs were significantly reduced in both the ROSC6h and ROSC24h groups (all P < 0.0001). By contrast, the proportions of microglia, neutrophils, and T cells were significantly higher in the ROSC6h and ROSC24h groups than in the Sham group (all P < 0.0001). Notably, the ROSC24h group showed lower proportions of astrocytes (P < 0.0001), endothelial cells (P < 0.0001), microglia (P < 0.001), neutrophils (P < 0.0001), oligodendrocytes (P < 0.0001), smooth muscle cells (P < 0.0001), and T cells (P < 0.01) compared with the ROSC6h group. No significant differences in neuronal proportions were observed among the three groups (Additional Table 4).

Additional Table 4.

Comparison of cell type proportions using Two-Proportion Z-Test

Cell type	Group1	Group2	Cell_num1	Group_num1	Proportion1	Cell_num2	Group_num2	Proportion2	Z-stat	P-value	Higher Proportion Group
Fibroblasts	Sham	ROSC6h	47	39665	0.001185	6	27997	0.000214	4.444662	< 0.0001	Sham
Fibroblasts	Sham	ROSC24h	47	39665	0.001185	10	22514	0.000444	2.933369	< 0.01	Sham
Fibroblasts	ROSC6h	ROSC24h	6	27997	0.000214	10	22514	0.000444	-1.44296	0.149	ROSC24h
Endothelial ce lls	Sham	ROSC6h	2898	39665	0.073062	679	27997	0.024253	27.94453	< 0.0001	Sham
Endothelial ce lls	Sham	ROSC24h	2898	39665	0.073062	205	22514	0.009105	35.19977	< 0.0001	Sham
Endothelial ce lls	ROSC6h	ROSC24h	679	27997	0.024253	205	22514	0.009105	12.9039	< 0.0001	ROSC6h
Oligodendrocy tes_progenitor cells	Sham	ROSC6h	271	39665	0.006832	129	27997	0.004608	3.717626	< 0.001	Sham
Oligodendrocy tes_progenitor cells	Sham	ROSC24h	271	39665	0.006832	109	22514	0.004841	3.061219	< 0.01	Sham
Oligodendrocy tes_progenitor cells	ROSC6h	ROSC24h	129	27997	0.004608	109	22514	0.004841	-0.38138	0.7029	ROSC24h
Smooth muscl e cells	Sham	ROSC6h	870	39665	0.021934	307	27997	0.010965	10.74777	< 0.0001	Sham
Smooth_muscl e_cells	Sham	ROSC24h	870	39665	0.021934	52	22514	0.00231	19.45793	< 0.0001	Sham
Smooth_muscl e_cells	ROSC6h	ROSC24h	307	27997	0.010965	52	22514	0.00231	11.5104	< 0.0001	ROSC6h
Neutrophils	ROSC6h	ROSC24h	331	27997	0.011823	505	22514	0.02243	-9.28812	< 0.0001	ROSC24h
Neurons	Sham	ROSC6h	163	39665	0.004109	127	27997	0.004536	-0.83695	0.4026	ROSC6h
Neurons	Sham	ROSC24h	163	39665	0.004109	112	22514	0.004975	-1.5627	0.1181	ROSC24h
Neurons	ROSC6h	ROSC24h	127	27997	0.004536	112	22514	0.004975	-0.71378	0.4754	ROSC24h
Oligodendrocy te_cells	Sham	ROSC6h	14254	39665	0.35936	10983	27997	0.392292	-8.7242	< 0.0001	ROSC6h
Oligodendrocy te_cells	Sham	ROSC24h	14254	39665	0.35936	9506	22514	0.422226	-15.5051	< 0.0001	ROSC24h
Oligodendrocy te_cells	ROSC6h	ROSC24h	10983	27997	0.392292	9506	22514	0.422226	-6.81023	< 0.0001	ROSC24h
Astrocytes	Sham	ROSC6h	1736	39665	0.043767	437	27997	0.015609	20.46046	< 0.0001	Sham
Astrocytes	Sham	ROSC24h	1736	39665	0.043767	244	22514	0.010838	22.47508	< 0.0001	Sham
Astrocytes	ROSC6h	ROSC24h	437	27997	0.015609	244	22514	0.010838	4.621425	< 0.0001	ROSC6h
T cells	Sham	ROSC6h	239	39665	0.006025	635	27997	0.022681	-18.8967	< 0.0001	ROSC6h
T_cells	Sham	ROSC24h	239	39665	0.006025	594	22514	0.026384	-21.2214	< 0.0001	ROSC24h
T_cells	ROSC6h	ROSC24h	635	27997	0.022681	594	22514	0.026384	-2.68448	< 0.01	ROSC24h
Microglias	Sham	ROSC6h	19066	39665	0.480676	14363	27997	0.513019	-8.28774	< 0.0001	ROSC6h
Microglias	Sham	ROSC24h	19066	39665	0.480676	11177	22514	0.496447	-3.78145	< 0.001	ROSC24h
Microglias	ROSC6h	ROSC24h	14363	27997	0.513019	11177	22514	0.496447	3.702859	< 0.001	ROSC6h

Cell_num: The number of cells within the specific cell type being compared.

Group_num: The total number of cells within the respective group.

Proportion: The proportion of the specific cell type relative to the total cell count in the group.

Z-stat: The Z-statistic for the comparison between groups.

P-value: The P-value for statistical significance.

Higher Proportion Group: The group with the higher proportion of the specific cell type.

To minimize potential biases inherent in raw proportion comparisons, we further applied the scCODA model for analysis (Buttner et al., 2021), which revealed a higher proportion of T cells in the ROSC6h group than in the Sham group. Additionally, the ROSC24h group had lower proportions of astrocytes, endothelial cells, and smooth muscle cells, and a higher proportion of neutrophils, compared with the Sham group (Figure 1G and Additional Table 5).

Additional Table 5.

Comparison of cell type proportions using the scCODA model

Sham vs. ROSC6h (reference group:Sham)
Covariate	Cell Type	Final Parametei HDI 3%		HDI 97% SD		Inclusion proba	Expected Samp	log2-fold change
Condition[T.Sham]	Astrocyte	0	-0.348	1.17	0.325	0.508266667	277.4663282	0.177229888
Condition[T.Sham]	Endothelial cel	0	-0.251	1.062	0.29	0.5084	374.9150946	0.177229888
Condition[T.Sham]	Fibroblast	0	-0.482	1.703	0.516	0.5674	32.44997219	0.177229888
Condition[T.Sham]	Microglia	0	-1.076	0.16	0.318	0.555466667	4859.542545	0.177229888
Condition[T.Sham]	Neuron	0	0	0	0	0	91.44151485	0.177229888
Condition[T.Sham]	Neutrophil	0	-1.299	0.413	0.383	0.550666667	109.9141477	0.177229888
Condition[T.Sham]	Oligodendrocyt	0	-1.09	0.147	0.336	0.582533333	2296.505049	-0.360254638
Condition[T.Sham]	Oligodendrocyt	0	-0.71	0.785	0.263	0.453266667	107.414973	0.177229888
Condition[T.Sham]	Smooth muscle	0	-0.549	0.794	0.255	0.518266667	199.078573	0.177229888
Condition[T.Sham]	T_cell	-0.489151703	-1.584	0.296	0.491	0.635933333	109.0218023	-0.528466848
Sham vs. ROSC24h (reference group:Sham)
Covariate	Cell Type	Final ParameterHDI 3%		HDI 97% SD		Inclusion proba	Expected Samp	log2-fold change
Condition[T.Sham]	Astrocyte	0.351232623	-0.306	1.355	0.382	0.566333333	338.2388686	0.446676438
Condition[T.Sham]	Endothelial cel	0.56591235	-0.21	1.483	0.485	0.650533333	433.3007607	0.756393815
Condition[T.Sham]	Fibroblast	0	-0.533	1.207	0.33	0.514733333	50.47709831	-0.060045126
Condition[T.Sham]	Microglia	0	-0.458	0.392	0.135	0.377866667	3650.136104	-0.060045126
Condition[T.Sham]	Neuron	0	-0.744	0.734	0.26	0.478866667	93.36529801	-0.060045126
Condition[T.Sham]	Neutrophil	-0.325715455	-1.279	0.481	0.4	0.5886	94.6138886	-0.529953198
Condition[T.Sham]	Oligodendrocyt	0	-0.637	0.286	0.184	0.469466667	2663.828718	-0.060045126
Condition[T.Sham]	Oligodendrocyt	0	0	0	0	0	112.4512433	-0.060045126
Condition[T.Sham]	Smooth muscle	0.409778997	-0.302	1.441	0.437	0.6164	195.8401434	0.531141001
Condition[T.Sham]	T_cell	0	-1.096	0.506	0.306	0.5074	140.1228776	-0.060045126
ROSC6h vs. ROSC24h (reference group:ROSC6h)
Covariate	Cell Type	Final ParameterHDI 3%		HDI 97% SD		Inclusion proba	Expected Samp	log2-fold change
Condition[T.ROSC6h]	Astrocyte	0	-0.516	0.621	0.206	0.519533333	157.8380032	0
Condition[T.ROSC6h]	Endothelial cel	0	-0.366	0.865	0.253	0.521	170.4716847	0
Condition[T.ROSC6h]	Fibroblast	0	-0.79	0.677	0.281	0.513866667	17.00605081	0
Condition[T.ROSC6h]	Microglia	0	-0.189	0.447	0.133	0.4618	4241.062943	0
Condition[T.ROSC6h]	Neuron	0	-0.68	0.603	0.228	0.523733333	72.57127706	0
Condition[T.ROSC6h]	Neutrophil	0	-0.58	0.558	0.206	0.523533333	176.3673809	0
Condition[T.ROSC6h]	Oligodendrocyt	0	-0.345	0.332	0.119	0.4446	3211.743486	0
Condition[T.ROSC6h]	Oligodendrocyt	0	0	0	0	0	80.60568744	0
Condition[T.ROSC6h]	Smooth muscle	0	-0.464	0.91	0.274	0.559066667	91.97982052	0
Condition[T.ROSC6h]	T_cell	0	-0.294	0.895	0.219	0.501266667	198.8536667	0

The table presents the analysis of intergroup differences in cell type proportions.

The "Final Parameter" column serves as an indicator of differences in cell type proportions.

A value of 0 denotes no significant difference in the proportion of the corresponding cell type between groups. Positive values indicate an increased proportion of the cell type relative to the reference group.

Negative values reflect a decreased proportion.

The High-Density Interval (HDI) columns (3% and 97%) represent the credible interval for the Final Parameter.

The BBB is a multicellular system that acts as a selectively permeable membrane between the intravascular space and the brain parenchyma. Structurally, it comprises vascular endothelial cells, pericytes, astrocyte end-feet, and the ECM at the capillary basement membrane (Abbott et al., 2010). Taken together, the decreased proportions of astrocytes and endothelial cells may reflect BBB disruption, while the increased abundance of neutrophils and T cells could indicate infiltration of peripheral immune cells into the brain parenchyma through the compromised BBB following CA.

Microglia characteristics following cardiac arrest

Next, we examined CA-induced alterations in microglia, which constituted the majority of cells in our hippocampal samples. Recent studies have confirmed that microglia in the brain are extensively activated following CA (Chang et al., 2020; Ousta et al., 2022). First, we sought to identify potential microglia marker genes (Figure 2A), which confirmed the expression of classical markers (CSF1R, P2RY12, and CX3CR1), alongside those previously used to identify microglia in swine (e.g., CALCR, RNF128, and NAV3) (Wang et al., 2022). To further understand these changes in gene expression, we conducted DEG analyses among the Sham and CA model groups (Figure 2B and Additional Table 6 ^{(283.2KB, pdf)} ). The top DEGs in the hippocampal microglia included ENSSSCG00000033089, ARHGAP15, and GBP1 at the ROSC6h time point, and S100A9, ENSSSCG00000018063, and S100A12 at the ROSC24h time point. The DEGs were then characterized using GO analysis to observe potential functional changes in microglia cells at these post-CA time points (Figure 2C). Specific pathways observed in the ROSC6h group included the regulation of leukocyte differentiation and activation, transmembrane receptor signaling, large ribosomal subunits, lymphocyte proliferation, actin polymerization or depolymerization, cytokine secretion, B-cell proliferation, proteolysis involved in cellular protein catabolic process, and effects of the transforming growth factor (TGF)-β receptor signaling pathway. In the ROSC24h group, distinct pathways included cytosolic small ribosomal subunit, negative regulation of RNA catabolic process, positive regulation of mononuclear cell proliferation, regulation of tumor necrosis factor superfamily cytokine production, regulation of signal transduction by p53 class mediator, antigen processing and presentation, and cytoplasmic side of endoplasmic reticulum membrane. Shared pathways included regulation of leukocyte cell–cell adhesion and proliferation, rRNA binding, polysomal ribosome, and cytoplasmic side of membrane.

Figure 2.

Identification of characteristic markers of microglia.

(A) UMAP plots showing the expression of key microglial markers. (B) DEGs in microglia from the three groups: ROSC6h vs. Sham, ROSC24h vs. Sham, and ROSC24h vs. ROSC6h. (C) GO pathway enrichment analysis of DEGs in microglia, showing significant pathways affected at two time points following cardiac arrest. (D) UMAP visualization of the seven subclusters of microglia. (E) Dot plot of key marker genes across microglia subclusters. DEGs: differentially expressed genes; GO: Gene Ontology; ROSC: return of spontaneous circulation.

Microglia exhibit distinct heterogeneity rather than M1/M2 polarization following cardiac arrest

To further explore the diversity of microglia following CA, we re-clustered 44606 microglia from all groups into seven subclusters, designated MG-0 to MG-6 (Figure 2D). We utilized a dot plot to illustrate the expression of marker genes in each subcluster (Figure 2E). Microglial polarization into M1 or M2 phenotypes post stroke is well documented (Lan et al., 2017; You et al., 2024a), but whether this occurs after global cerebral ischemia-reperfusion following CA remains unclear. Initially, we sought to examine the expression of three sets of classical polarization markers corresponding to the M1, M2a, and M2b/c phenotypes. However, our data revealed that only a few of these were highly expressed in some subclusters, with co-expression of genes from different polarization groups within the same subcluster (Figure 3A; Lan et al., 2017). The representative genes of microglial polarization are listed in Additional Table 7. This suggested that the simple binary classification into M1 and M2 may not be adequate. Therefore, we sought to elucidate the post-ROSC heterogeneity of microglia by exploring the phenotypic scenarios of each subcluster through examination of the expression of functional genes that regulate chemokines, interleukins, and anti-inflammatory and neurotrophic molecules (e.g., neurotrophins, growth factors, and neuropeptides) (Figure 3B and Additional Table 8). GO enrichment analysis of DEGs for each subpopulation and scoring of gene sets representing classical microglial functions led to the identification of an intriguing subpopulation (Additional Figure 1 ^{(37.1MB, tif)} ).

Figure 3.

Characterization of microglia subclusters after cardiac arrest.

(A) Dot plots and violin plots of marker genes representing the M1 (upper), M2a (middle), and M2b/c (lower) phenotypes of microglial polarization across the identified microglia subclusters. (B) Violin plots of functional genes related to major microglial functions, showcasing their expression patterns across the microglia subclusters. (C) Stacked bar graphs depicting the proportions of each microglia subcluster within the Sham, ROSC6h, and ROSC24h groups (left), and the proportions of each group within individual microglia subclusters (right). (D) Violin plots displaying the top 10 marker genes in the MG-3 subcluster. (E) Double immunofluorescence staining of S100A8⁺IBA1⁺ cells (arrows) at each post–cardiac arrest time point (n = 3 pigs; 2–3 slices per pig). Representative images show samples stained with antibodies against IBA1 (green, representing microglia) and S100A8 (red, MG-3 marker), and DAPI (blue, nuclei). The bar graph depicts quantification of the percentage of S100A8⁺IBA1⁺ cells relative to IBA1⁺ cells (2–3 slices per pig). Values presented as means ± standard deviation. Statistical differences determined by one-way analysis of variance followed by Tukey’s post hoc test; *P < 0.05, ***P < 0.001, ****P < 0.0001. (F) Gene Ontology enrichment analysis of the MG-3 subcluster. DAPI: 4′,6-Diamidino-2-phenylindole; GO: Gene Ontology; IBA1: ionized calcium binding adaptor molecule 1; MG-3: microglia subcluster 3; ROSC: return of spontaneous circulation.

Additional Table 7.

List of classic polarization markers of microglia

Ml
ENSSSCG00000060860
ENSSSCG00000032963
ENSSSCG00000020970
ENSSSCG00000017755
ENSSSCG00000017705
ENSSSCG00000017044
ENSSSCG00000011877
ENSSSCG00000009856
ENSSSCG00000008087
ENSSSCG00000001472
ENSSSCG00000001459
ENSSSCG00000001456
ENSSSCG00000001457
ENSSSCG00000001470
ENSSSCG00000001472
ENSSSCG00000006350
ENSSSCG00000008088
ENSSSCG00000016254
ENSSSCG00000032764
ENSSSCG00000033667
ENSSSCG00000034182
ENSSSCG00000036288
ENSSSCG00000036618
ENSSSCG00000036704
ENSSSCG00000039214
ENSSSCG00000040808
ENSSSCG00000053490
ENSSSCG00000055451
ENSSSCG00000060276
ENSSSCG00000061450
ENSSSCG00000061569

Additional Table 8.

List of genes related to classical functions of microglia

Neurotrophic Factor

ENSSSCG00000013333 ENSSSCG00000009134 ENSSSCG00000022126 ENSSSCG00000010224 ENSSSCG00000009630 ENSSSCG00000014336 ENSSSCG00000060151 ENSSSCG00000003872 ENSSSCG00000013857 ENSSSCG00000000600 ENSSSCG00000024954 ENSSSCG00000036213 ENSSSCG00000061049 ENSSSCG00000012870 ENSSSCG00000058425 ENSSSCG00000014047 ENSSSCG00000010698 ENSSSCG00000030827 ENSSSCG00000015815 ENSSSCG00000036512 ENSSSCG00000001624 ENSSSCG00000000493 ENSSSCG00000031244 ENSSSCG00000009565 ENSSSCG00000010375 ENSSSCG00000000669 ENSSSCG00000023261 ENSSSCG00000034943 ENSSSCG00000028931 ENSSSCG00000014288 ENSSSCG00000038727 ENSSSCG00000008035 ENSSSCG00000006902 ENSSSCG00000010345 ENSSSCG00000014054 ENSSSCG00000017206 ENSSSCG00000017495 ENSSSCG00000014362 ENSSSCG00000006468 ENSSSCG00000015403 ENSSSCG00000033082 ENSSSCG00000000857 ENSSSCG00000030560 ENSSSCG00000035293 ENSSSCG00000036695 ENSSSCG00000011795 ENSSSCG00000023128 ENSSSCG00000004044 ENSSSCG00000017472 ENSSSCG00000035392 ENSSSCG00000058216 ENSSSCG00000016728 ENSSSCG00000034818 ENSSSCG00000024555 ENSSSCG00000016619 ENSSSCG00000034898 ENSSSCG00000015045 ENSSSCG00000029656 ENSSSCG00000025085 ENSSSCG00000016295 ENSSSCG00000032417 ENSSSCG00000017548 ENSSSCG00000055255 ENSSSCG00000039399 ENSSSCG00000012942 ENSSSCG00000053344 ENSSSCG00000016664 ENSSSCG00000016718 ENSSSCG00000015839 ENSSSCG00000010344 ENSSSCG00000026211 ENSSSCG00000011102 ENSSSCG00000026383 ENSSSCG00000033413 ENSSSCG00000026180 ENSSSCG00000006464 ENSSSCG00000010959 ENSSSCG00000005106 ENSSSCG00000000932 ENSSSCG00000038278 ENSSSCG00000008625 ENSSSCG00000037207 ENSSSCG00000056699 ENSSSCG00000004279 ENSSSCG00000021184 ENSSSCG00000032115 ENSSSCG00000007541 ENSSSCG00000033327 ENSSSCG00000024960 ENSSSCG00000014994 ENSSSCG00000008841 ENSSSCG00000022741 ENSSSCG00000006988 ENSSSCG00000050616 ENSSSCG00000000602 ENSSSCG00000025661 ENSSSCG00000016732 ENSSSCG00000008326 ENSSSCG00000003017 ENSSSCG00000035952 ENSSSCG00000010816 ENSSSCG00000002385 ENSSSCG00000014316 ENSSSCG00000011226 ENSSSCG00000005382 ENSSSCG00000006911 ENSSSCG00000040598 ENSSSCG00000024636 ENSSSCG00000038864 ENSSSCG00000001695 ENSSSCG00000054506 ENSSSCG00000012135 ENSSSCG00000015770 ENSSSCG00000039053

A distinct microglial subcluster is associated with neuroinflammation following cardiac arrest

One subcluster of microglia, MG-3, was significantly increased in the ROSC6h group compared with that in the Sham group, showing a further increase in the ROSC24h group (Figure 3C). The top marker genes for MG-3 included S100A6, S100A8, S100A9, S100A12, TIMP117, and ARG1 (Figure 3D), which were confirmed by immunofluorescence staining analysis (Figure 3E). Additionally, MG-3 exhibited significant gene expression of M1/M2 polarization markers, chemokines, and interleukins, suggesting that it represents activated microglia. GO pathway enrichment for MG-3 revealed key processes in energy metabolism, including respiratory chain and ATPase complexes, alongside immune regulation pathways such as tumor necrosis factor production and neutrophil chemotaxis, while highlighting the role of p53 in DNA damage response. Other notable areas involves reactive oxygen species regulation, viral response control, and leukocyte apoptosis. These pathways illustrate the diverse roles of MG-3 in immune response and cellular maintenance (Figure 3F).

Oligodendrocyte characteristics indicate post–cardiac arrest disruption in myelin integrity

Oligodendrocytes, crucial for axonal integrity and myelination in the central nervous system (CNS), are particularly vulnerable to ischemic damage, which can lead to demyelination, axonal instability, and long-term neurological dysfunction (Huang et al., 2023). Thus, we sought to explore functional changes in hippocampal oligodendrocytes at the single-cell level following CA. We identified expression of the classical oligodendrocyte markers MBP, MOG, and OLIG1, as well as previously identified swine markers, including PLP, TMEFF2, and TF (Figure 4A). Next, an analysis of differential gene expression in hippocampal oligodendrocytes among the Sham and CA model groups revealed the following notable DEGs: SGCZ, FKBP5, MOBP, ISG15, ISG12(A), and IFI6 in the ROSC6h group; and ENSSSCG00000057738, H3-3A, ND2, RIMS2, ISG12(A), and IFI6 in the ROSC24h group (Figure 4B and Additional Table 6 ^{(283.2KB, pdf)} ). This analysis further revealed decreased expression levels of oligodendrocyte and myelin integrity markers, such as MOBP, at both 6 and 24 hours post CA, with MBP and MOG showing reductions at 24 hours. Western blot analysis of hippocampal tissues demonstrated that, compared with the Sham group, MBP levels were significantly decreased at 6 hours post CA (P < 0.001), showing a further decline at 24 hours post CA (P < 0.01; Figure 4C). Immunofluorescence staining validated that the reduction in hippocampal MBP expression was sustained at 6 and 24 hours post CA (Figure 4D). Pathway analysis revealed both shared and unique biological processes at 6 and 24 hours. Common pathways included myelin sheath formation, cell body maintenance, and chaperone-mediated protein folding. At 6 hours, there were exclusive pathways related to phosphorus metabolism, synaptic functions, and cellular response to organic substances, among others focused on cell motility, projection, and morphogenesis. By 24 hours, pathways shifted to include protein complex assembly regulation, protein localization, and various aspects of protein stability, along with unique cellular rebuilding processes such as endocytosis and cytokinesis regulation (Figure 4E).

Figure 4.

Characteristics of hippocampal oligodendrocytes and OPC following cardiac arrest.

(A) UMAP plots showing the expression of key oligodendrocyte markers. (B) DEGs in oligodendrocytes: ROSC6h vs. Sham, ROSC24h vs. Sham, and ROSC24h vs. ROSC6h. (C) Immunoblotting analysis and quantification of MBP in hippocampal tissue (n = 3). (D) Representative images and quantification of immunofluorescence staining of MBP in the hippocampus (n = 3 pigs; 2–3 slices per pig). (E) GO pathway enrichment analysis of DEGs, showing significant pathways affected in oligodendrocytes at different time points following cardiac arrest. (F) UMAP visualization of the seven oligodendrocyte subclusters. (G) DEGs across oligodendrocyte subclusters. (H) GO enrichment analysis of the oligodendrocyte subclusters. (I) UMAP plots showing the expression of key OPC markers. (J) GO pathway enrichment analysis of DEGs in OPCs. (K) Pseudotime analysis (left) of the oligodendrocyte lineage, showing the trajectory of differentiation from OPCs to mature oligodendrocytes. The color gradient represents pseudotime progression, from early (purple) to late (yellow) stages of differentiation. UMAP visualization (right) of the differentiation progression from OPCs (yellow) to oligodendrocytes (pink). (L) Quantitative reverse transcription polymerase chain reaction analysis of oligodendrocyte lineage markers at different maturation stages. Statistical differences determined by one-way analysis of variance followed by Tukey’s post hoc test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. DEGs: Differentially expressed genes; GO: Gene Ontology; OPC: oligodendrocyte precursor cell; qRT-PCR: quantitative reverse transcription-polymerase chain reaction; ROSC: return of spontaneous circulation;. UMAP: Uniform Manifold Approximation and Projection.

To further investigate the post-CA diversity of oligodendrocytes, we re-clustered 34743 oligodendrocytes from the three groups into six subclusters, designated OLG-0 to OLG-5 (Figure 4F). A volcano plot illustrating the top DEGs for each subcluster showed that the genes with the greatest changes in OLG-0 were PALM2AKAP2, EGR1, and MX1. GO enrichment analysis of OLG-0 showed the involvement of key processes, including cell movement, adhesion, and metabolic pathways involving lipids and steroids, highlighting its roles in immune function activities involving antigen presentation via MHC class I. In OLG-1, the top DEGs were HMGCS1, GRIN2B, and FCHSD2, with pathway enrichment analysis highlighting cell movement regulation, actin cytoskeleton organization, and cell adhesion as key processes. OLG-1 was also enriched for molecular functions involved in GTPase signaling and biosynthesis of lipids and steroids, essential for cellular metabolism and structural integrity. In OLG-2, the top DEGs were ISG15, ISG12(A), and IFI6, with involvement in pathways related to immune responses, including antigen processing and presentation, particularly via MHC class I. Key pathways also involved responses to external stimuli and biotic threats, regulation of viral genome replication, and defense mechanisms against viruses. In OLG-3, the top DEGs were FOS, EGR1, and JUNB, with pathway enrichment related to oligodendrocyte responses to various stimuli, including oxygen-containing compounds, extracellular factors, organic substances, and cytokines. These pathways involved numerous binding activities, including sequence-specific DNA binding by RNA polymerase II, double-stranded DNA binding, and transcription factor binding, all within cis-regulatory regions. Additionally, there was involvement in the regulation of cell death processes, including both positive and negative regulation of apoptosis and programmed cell death. Other notable functions included ubiquitin-like protein ligase and transferase activities. In OLG-4, the top DEGs were CST3, CALCR, and C1QB, with GO enrichment results highlighting processes involving actin polymerization, myeloid and lymphocyte differentiation, and immune response via MHC protein complexes. It also showed regulation of leukocyte activities and involvement of endoplasmic reticulum components. In OLG-5, the top DEGs were ISG15, STAT1, and SLC6A20, with pathway enrichment analysis highlighting its involvement in metabolic regulation, immune responses, and stress reactions. Key processes included the regulation of nitrogen and protein metabolism, antigen processing and presentation via MHC class I, and defense responses to various stimuli, particularly viruses. These findings suggested a multifaceted role for oligodendrocytes in neuroimmune interactions and environmental response within the CNS (Figure 4G and H).

Maturation of oligodendrocyte progenitor cells

OPCs remain undifferentiated under both physiological and pathological conditions, and are widely distributed throughout the CNS in adults (Dimou et al., 2008; Psachoulia et al., 2009). In the brains of both rodents and humans, a certain proportion of newly formed or pre-existing OPCs differentiate into mature oligodendrocytes, replacing those that have degenerated under physiological or pathological conditions (Winkler et al., 2018). Here, we observed the expression classical OPC markers such as PDGFRA and C1QL1, as well as those previously identified in swine, including TNR, SHISA9, and GPR17 (Figure 4I). Because we only detected a total of 509 OPCs, subgroup analysis was not conducted. GO analysis revealed changes in the myelin sheath and regulation of nervous system development pathways at 24 hours post–ROSC, while only the protein folding pathway was altered at 6 hours post–ROSC (Figure 4J). Furthermore, pseudotime analysis confirmed that the OPCs could develop into mature oligodendrocytes (Figure 4K). To further investigate this post-CA differentiation, we used qRT-PCR analyze the mRNA expression levels of marker genes at three maturation stages: OPCs, immature oligodendrocytes, and mature oligodendrocytes (Armstrong, 1998). Compared with the Sham group, the mRNA levels of OPC markers NG2 and PDGFRA were significantly reduced, immature oligodendrocyte markers CNP and GALC were elevated, and mature oligodendrocyte markers MBP and PLP were decreased in the CA model groups (Figure 4L). These results suggested an impaired myelination process accompanied by a compensatory acceleration in the differentiation and maturation of OPCs in response to CA-induced injury.

Characteristics of infiltrating leukocytes following cardiac arrest

Following brain injury, inflammation-mediated BBB damage triggers the recruitment of peripheral immune cells, which directly interact with resident brain cells or release immune mediators that shape the immune niche and influence recovery (Cunningham et al., 2022). We observed the expression of classical T-cell markers CD3D and CD3E, and previously identified swine markers, including GNLY, CTSW, and SKAP1 (Figure 5A). Differential gene expression analysis of T cells revealed S100A1, HSPA6, and APOE as key DEGs in the ROSC6h group and ENSSSCG00000018063, ND2, and ND3 in the ROSC24h group (Figure 5B and Additional Table 6 ^{(283.2KB, pdf)} ).

Figure 5.

Single-cell RNA sequencing analysis of T cells, neutrophils, and endothelial cells in the hippocampus following cardiac arrest.

(A, D, I) UMAP plots showing the expression of key markers of T-cells (A), neutrophils (D), and endothelial cells (I). (B, G, J) Volcano plots illustrating DEGs in T cells (B), neutrophils (G), and endothelial cells (J) among the three groups. (C, H, L) GO pathway enrichment analysis of DEGs in T cells (C), neutrophils (H), and endothelial cells (L), showing significant pathways affected at two time points following cardiac arrest. (E) Immunoblotting analysis and quantification of MPO in hippocampal tissue (n = 3). (F) Representative images and quantification of immunofluorescence staining of MPO⁺ cells (arrows) in hippocampal sections stained with anti-MPO antibodies (red, neutrophil marker) and DAPI (blue, nuclei). Scale bars: 50 µm. The bar graph shows quantification of the number of MPO⁺ cells/mm² (2–3 slices per pig). (K) Immunoblotting analysis and quantification of Claudin-5 in hippocampal tissue (n = 3). Statistical significance determined by one-way analysis of variance, followed by Tukey’s post hoc test; *P < 0.05, ****P < 0.0001. DAPI: 4′,6-Diamidino-2-phenylindole; DEGs: differentially expressed genes; GO: Geno Ontology; MPO: myeloperoxidase; ns: not significant; ROSC: return of spontaneous circulation; UMAP: Uniform Manifold Approximation and Projection.

At 6 hours post–CA, T cells were found to be involved in pathways related to leukocyte activation, protein polymerization, cell death regulation, translation, and actin dynamics. Key processes included regulation of lymphocyte activation, cell proliferation, protein complex assembly, unfolded protein binding, and response to cytokines. By 24 hours, the focus shifted to pathways regulating mononuclear cell proliferation, immune response, leukocyte and T-cell chemotaxis, and proliferation, along with RNA polymerase II-specific transcription factor binding. Mitochondrial electron transport, cellular respiration, and antigen receptor-mediated signaling were also prominent. Common to both time points was the generation of precursor metabolites and energy (Figure 5C).

A previous study has observed T-cell infiltration in the brain parenchyma following CA (Smida et al., 2021). In contrast, we detected neutrophil infiltration, a novel finding not previously reported, as evidenced by expression of the classical T-cell marker CSF3R (Figure 5D). Western blot analysis of hippocampal tissue showed a significant increase in the neutrophil marker protein MPO in the ROSC6h group (P < 0.05), with a further increase in the ROSC24h group (P < 0.05), compared with that in the Sham group (Figure 5E). Additionally, immunofluorescence staining of hippocampal tissue confirmed neutrophil infiltration into the brain parenchyma of CA model pigs that increased as ROSC progressed (Figure 5F). These findings provided clear evidence of neutrophil infiltration following CA. Next, we conducted a differential gene expression analysis of hippocampal neutrophils among the Sham and CA model groups, revealing CXCL10, RETN, and S100A9 as key DEGs in the ROSC6h group and S100A9, S100A12, and TIMP1 in the ROSC24h group (Figure 5G and Additional Table 6 ^{(283.2KB, pdf)} ).

At 6 and 24 hours post CA, neutrophils shared common pathways, including regulation of cell adhesion, interspecies interaction between organisms, response to external biotic stimulus, and regulation of immune system processes. Unique pathways that were expressed at 6 hours involved the regulation of nitrogen compound metabolism, mononuclear cell proliferation, various immune responses, and T cell-related processes. At 24 hours, the unique pathways were associated with stress response regulation, signal transduction, nitric oxide biosynthesis, cell death, apoptotic processes, and innate immune responses, highlighting a shift towards stress and death mechanisms at the later time point (Figure 5H).

Endothelial dysfunction following cardiac arrest

We identified 3782 endothelial cells, which are crucial to the vascular lineage of the BBB and regulation of cerebral blood flow, originating from the inner walls of blood vessels. We observed expression of the classical endothelial cell markers CLDN5, PECAM1, and VWF, alongside swine-specific endothelial cell markers, including CYYR1 and ADGRF5 (Figure 5I). Differential gene expression analysis of endothelial cells among the Sham and CA model groups identified CXCL10, CXCL9, and GBP1 as key DEGs in the ROSC6h group and ENSSSCG00000032591, ENSSSCG00000057738, and MT1A in the ROSC24h group (Figure 5J and Additional Table 6 ^{(283.2KB, pdf)} ). Additionally, at both time points following CA, there were reduced expression levels of tight junction protein 1 and occludin in endothelial cells. Additionally, the ROSC6h group showed decreased expression of claudin 5 in endothelial cells, indicating significant dysfunction in cerebral vascular endothelial cells and a notable increase in paracellular permeability following CA. Western blot analysis of hippocampal tissue confirmed the significant reduction in claudin 5 expression in the CA groups (P < 0.05) compared with that in the Sham group (Figure 5K). At 6 hours post–CA, endothelial cells exhibited pathway activity related to the regulation of cellular protein catabolic processes, proteolysis, cell adhesion, and endothelial cell migration, with a notable response to interferon-gamma and alpha, and regulation of signal transduction by p53. Unique pathways included cytoplasmic ribonucleoprotein granules, aerobic respiration, and proteasome core complex involvement. By 24 hours, the focus shifted towards biosynthetic and metabolic processes, specifically ribonucleoside triphosphate and purine nucleoside triphosphate metabolism, cellular component biogenesis, and regulation of vasculature development. Endothelial cells also exhibited increased nucleoside-triphosphatase regulator activity, integrin binding, and small GTPase-mediated signal transduction, alongside the presence of ribosomal subunits and components of the endoplasmic reticulum membrane (Figure 5L).

Enhanced intercellular communication drives neuroinflammatory responses following cardiac arrest

The post-CA neuroinflammatory response is dependent on complex interactions between resident brain cells and bone marrow-derived immune cells that occur through molecular adhesion on adjacent cell surfaces or the ECM, as well as the release of chemical signaling molecules. Next, we conducted cell communication analysis, using ligand–receptor interactions to infer intercellular communication within the hippocampus and to identify CA-related changes. To simulate the potential early intercellular regulatory relationships following CA, this analysis included all cells. The results indicated interactions among various cell types (Figure 6A). For instance, microglia were observed to regulate neutrophil recruitment by secreting chemokines (e.g., C–C motif chemokine ligand 8 [CCL5] and [CCL23]) of the CCL pathway, while endothelial cells modulated the adhesion of microglia and T cells by secreting ICAM1 and ICAM2 of the ICAM pathway (Additional Table 9 ^{(520.3KB, pdf)} ). The chord diagram shown in Figure 6B illustrates the 30 most significant ligand–receptor pairs.

Figure 6.

Enhanced intercellular communication drives neuroinflammatory responses following cardiac arrest.

(A) Heatmap illustrating the numbers of inferred interactions between different cell types based on LR pairs. (B) Chord diagram of the top 30 LR pairs between cell types. (C, D) Network diagrams of the changes in the number of intercellular communications between the Sham and ROSC6h groups (C) and the Sham and ROSC24h groups (D). Red represents an increase, and blue represents a decrease, in the number of interactions. (E) Visualization of enhanced LR pairs involved in microglia–neutrophil communication after cardiac arrest in the ROSC6h and ROSC24h groups vs. the Sham group. Each dot represents an LR pair, with its position indicating the group. Dot size corresponds to the P value (larger dots indicate lower P values), while the color represents the interaction probability, with the gradient ranging from purple (low probability) to yellow (high probability). (F) Secretion of resistin into the supernatant of the neutrophil–microglia co-culture system over time. (G) Comparison of resistin secretion, with and without OGD/R and the presence of neutrophils. (H) Immunofluorescence analysis and quantification of iNOS expression in microglia. The left panel shows representative images of iNOS (red), IBA1 (green, microglia marker), and DAPI (blue, nuclei), with merged images in the bottom row. The bar graph shows quantification of relative iNOS fluorescence intensity. Statistical significance determined by one-way analysis of variance followed by Tukey’s post hoc test; ****P < 0.0001. Scale bars: 20 μm. IBA1: Ionized calcium binding adaptor molecule 1; iNOS: inducible nitric oxide synthase; LR: ligand receptor; ns: not significant; OGD/R: oxygen-glucose deprivation/reperfusion; ROSC: return of spontaneous circulation.

In the ROSC6h group, we observed an increase in microglia–neutrophil communication, along with enhanced interactions between neutrophils and other cells, namely endothelial cells, astrocytes, T cells, smooth muscle cells, OPCs, and oligodendrocytes, compared with the Sham group (Figure 6C and Additional Table 10 ^{(366.5KB, pdf)} ). At 24 hours post–CA, there were further increases in communication between neutrophils and endothelial cells, astrocytes, smooth muscle cells, and oligodendrocytes (Figure 6D and Additional Table 10 ^{(366.5KB, pdf)} ). Microglia–neutrophil communication plays a critical role in neuroinflammation, leading to increased recruitment and activity of neutrophils in the brain parenchyma while simultaneously activating microglia (Kim et al., 2020). Our cell communication analysis further indicated that, post CA, neutrophils likely interact with microglia by releasing resistin, a ligand that binds to microglial receptors. Compared with the Sham group, this interaction was significantly stronger in both the ROSC6h and ROSC24h groups (Figure 6E and Additional Table 10 ^{(366.5KB, pdf)} ). To verify the communication between neutrophils and microglia, we simulated post-CA interactions in an in vitro model comprising co-culture of LPS-activated neutrophils and OGD and reoxygenation (OGD/R)-treated BV2 microglial cells. First, neutrophils were activated with LPS and co-cultured with microglia under normoxic conditions. Using ELISA, we detected a time-dependent increase in resistin concentration in the co-culture supernatant that peaked at 9 hours, which was used to determine the optimal co-culture duration (Figure 6F). When the microglia underwent OGD followed by co-culture with LPS-activated neutrophils during the reoxygenation phase, they did not independently secrete resistin into the supernatant. However, resistin levels in the supernatant of the co-culture system were significantly higher than those in the OGD/R-treated microglia alone (P < 0.0001) or the control group, which consisted of microglia maintained under normoxic conditions without OGD/R treatment or neutrophil co-culture (P < 0.0001; Figure 6G). To further assess the polarization status of the microglia, we performed immunofluorescence staining to evaluate the levels of inducible nitric oxide synthase (iNOS). OGD/R-treated microglia co-cultured with neutrophils exhibited significantly higher iNOS levels than OGD/R-treated microglia cultured alone (P < 0.0001), indicating a shift toward a pro-inflammatory phenotype (Figure 6H). These results confirmed that neutrophils communicate with microglia via resistin, driving microglial polarization and promoting neuroinflammation.

Discussion

The BBB, composed of tightly connected endothelial cells, pericytes, astrocytes, and microglia, regulates the transport of plasma components into the brain parenchyma (Daneman and Prat, 2015; Alahmari, 2021). Hypoxia can increase the permeability of tight junctions in the BBB, leading to its disruption (Takahashi and Macdonald, 2004; Hayman et al., 2018). Interestingly, the protection provided by the BBB varies across different regions of the brain, with the hippocampus potentially being less protected than others (Wilhelm et al., 2016; Davidson and Stevenson, 2024). Here, we observed significantly reduced abundance of astrocytes and endothelial cells following CA, similar to the changes observed in a scRNA-seq study of ischemic stroke (Zheng et al., 2022). At the RNA level, we found decreased expression of the tight junction protein 1 (TJP1) and occludin (OCLN) genes in endothelial cells at both 6 hours and 24 hours post–ROSC, with claudin 5 (CLDN5) levels also declining at the 6-hour time point. We confirmed this reduction in claudin 5 at the protein level using western blotting, suggesting substantial endothelial dysfunction and increased paracellular permeability. These findings align with recent reports in patients with CA, in whom BBB permeability gradually increased within the first 24 hours after ROSC, despite post-resuscitation care (You et al., 2024b). Similarly, our findings are consistent with animal studies on post-CA BBB disruption (Li et al., 2017, 2018, 2023; Chen et al., 2022).

Disruption of the BBB after cerebral ischemia–reperfusion allows peripheral immune cells to infiltrate the brain, exacerbating BBB damage by releasing inflammatory factors and oxidative stress products. Preclinical studies across various animal models suggest that, after CA, T cells rapidly accumulate in the brain and produce pro-inflammatory cytokines, contributing to brain injury (Zhang et al., 2018; Smida et al., 2021; Cunningham et al., 2022). However, there are multiple types of infiltrating immune cells in the brain. Here, we made the novel discovery that neutrophils from the circulation were also recruited to the brain following resuscitation from CA. This increase in the proportion of neutrophils within post-CA hippocampal tissue observed in our scRNA-seq data was confirmed through western blot analysis and immunofluorescence staining. GO analysis of post-CA neutrophils revealed pathways enriched in cell adhesion, regulation of nitrogen compound metabolism, monocyte and lymphocyte proliferation, T-cell activation and cytokine-mediated signaling. Similarly, GO analysis of post-CA T cells showed pathways enriched in leukocyte and lymphocyte activation, migration, and proliferation, suggesting a synergistic effect of peripheral immune cells on brain injury. In addition to interventions targeting T cells, strategies to reduce neutrophil infiltration may represent a novel therapeutic approach to mitigate brain injury following CA.

Previous research on CA indicates that the hippocampal microglia, representing the primary resident immune cells of the brain, undergo extensive activation, marked by increased IBA1 expression and morphological changes, which correlate with severe neurological damage (Wang et al., 2021; Ousta et al., 2022; Dou et al., 2023; Stommel et al., 2023; Wu et al., 2023). Studies also demonstrated that activated microglia polarize into M1/M2 phenotypes after CA (Chen et al., 2022; Wang et al., 2023). Researchers have used the M1/M2 paradigm, borrowed from macrophage research, to classify the activation states of microglia and distinguish between their various functional states. The M1 phenotype represents a pro-inflammatory state, characterized by high levels of pro-inflammatory factors, such as tumor necrosis factor-α, interleukin (IL)-1β, chemokines, and nitric oxide. By contrast, the M2 phenotype represents an anti-inflammatory or tissue repair state, associated with high levels of anti-inflammatory factors, such as IL-10, TGF-β, Arg1, and Ym1. This M1/M2 framework has been instrumental in understanding microglial roles in various neurological conditions. M1 microglia are thought to contribute to neuroinflammation and neuronal damage, while M2 microglia are believed to aid in tissue repair and inflammation resolution (Hu et al., 2015). However, recent research suggests that microglia exhibit a wide range of activation states in different pathological contexts, which do not always fit neatly into the M1/M2 categories (Ransohoff, 2016; Guo et al., 2022; Ma et al., 2023). Therefore, we used scRNA-seq to gain a comprehensive understanding of microglial functions in post-CA states. Our data indicated that only a few classical markers are significantly expressed in microglial subclusters. For example, while the predominant subcluster, MG-0, did not express any polarization marker genes, other subclusters expressed both M1 and M2 marker genes. This prompted us to focus on the specific subcluster functions instead of microglial polarization, leading to the identification of a unique subpopulation, MG-3, which exhibited high gene expression of M1/M2 polarization markers, chemokines, and interleukins. Using immunofluorescence staining, we confirmed that the MG-3 subcluster exists, with increased abundance over time following CA, highlighting its potential as a therapeutic target. Notably, four of the MG-3 marker genes, S100A6, S100A8, S100A9, and S100A12, encode calcium-binding proteins of the S100 family. In the CNS, S100A8/A9 is predominantly expressed in microglia, exerting its biological effects through interactions with receptors (e.g. RAGE and TLR4). Importantly, the involvement of S100A8/A9 in the pathology of various CNS disorders is well documented (Shepherd et al., 2006; Ziegler et al., 2009; Wang et al., 2024). S100A8 is highly expressed in the brains of mice following focal cerebral ischemia–reperfusion injury (Sun et al., 2013). In vitro experiments have also demonstrated high expression of S100A8 in microglia exposed to OGD/R. Additionally, silencing S100A8 can effectively alleviate OGD/R-induced microglial damage by inhibiting inflammation and oxidative stress, thereby reducing apoptosis (Hu and Lin, 2021). Significant upregulation of S100A8 and S100A9 has been observed in microglia from a tauopathy mouse model and an accelerated aging mouse model. A unique microglial subset characterized by high expression of S100A8/A9 was identified, showing partial similarity with the disease-associated microglial phenotype. Enrichment of this subset in the hippocampus and its association with cognitive impairment were validated in brain samples from human patients with Alzheimer’s disease and tauopathy (Gruel et al., 2024). Upregulation of microglial S100A8/A9 was also observed in a sepsis mouse model, where it plays a critical role in microglial activation (Denstaedt et al., 2018).

While the pathophysiology of neuronal death in ischemic brain diseases has been extensively studied, the ischemic injury to glial cells has received relatively little attention. Our study is among the first to focus on oligodendrocytes and OPCs following CA. Oligodendrocytes are the only CNS cells capable of forming myelin, a crucial lipid layer surrounding neurons that is essential for communication and neuronal health (Nave et al., 2023). Although these cells are particularly vulnerable to oxygen and nutrient deprivation, leading to inevitable myelin breakdown, this aspect of the pathophysiological mechanisms of brain injury following CA has often been overlooked. Studies have demonstrated that ischemic damage to oligodendrocytes contributes to neuronal injury and brain dysfunction (Jia et al., 2019; Cheng et al., 2024). OPCs, which can differentiate into new oligodendrocytes, are widely distributed throughout the adult CNS (Hughes et al., 2013; Simons et al., 2024). Our DEG analysis showed that the expression levels of three oligodendrocyte and myelin integrity markers, MOBP, MBP, and MOG, decreased following CA (Yamamoto et al., 1999; Montague et al., 2006; Santos et al., 2018). Between-group comparisons of oligodendrocyte marker expression using immunofluorescence, western blotting, and qRT-PCR all showed reductions in MBP in the CA model groups. Additionally, pseudotime analysis provided insights into the evolution and progression of OPCs into oligodendrocytes. Beyond demonstrating that the myelination process is impaired at the RNA level post CA, we also identified a compensatory acceleration in oligodendrocyte lineage progression in response to the injury. This phenomenon was observed in prior studies using transient middle cerebral artery occlusion animal models (Tanaka et al., 2001, 2003; Miyamoto et al., 2010). Currently, neuroprotective strategies for post-CA treatment include antioxidation, calcium and ionic control, lipid membrane regulation, and mitochondrial protection to support energy production. However, such drugs have shown limited efficacy in treating neurological damage caused by CA (Choudhary et al., 2021, 2023). Oligodendrocyte generation is a critical reparative process during recovery from brain ischemia, thus promoting differentiation of OPCs into oligodendrocytes may represent an attractive strategy to enhance endogenous myelin regeneration (Dewar et al., 2003; Zhang et al., 2013). Further exploration of the roles of oligodendrocytes and OPCs in post-CA repair mechanisms, and the development of interventions to enhance or support these natural processes, may be crucial in treating demyelination following CA.

The neuroinflammatory response following CA is driven by complex interactions between resident brain cells and peripheral immune cells that infiltrate the brain through a disrupted BBB (Zhang et al., 2023). Here, we inferred potential intercellular communication within the hippocampus, identifying significant CA-related changes and further validating the post-CA increase in microglia–neutrophil communication. Activated neutrophils infiltrating the brain parenchyma were found to communicate with microglia via resistin, promoting microglial polarization and exacerbating neuroinflammation. Previous studies have demonstrated that resistin is highly expressed in neutrophils and is upregulated upon stimulation, with both membrane-bound and granule-stored forms being released (Li et al., 2021). Additionally, resistin has been shown to induce increased IL6 mRNA expression in microglial cell lines (Abgrall et al., 2022). The crosstalk between neutrophils and microglia is a well-documented phenomenon in neuroinflammation. We confirmed this interaction in the context of CA and identified resistin as a key mediator, paving the way for further mechanistic investigations (Kim et al., 2020).

In single-cell transcriptomics, isolating individual cells from living tissues is particularly challenging for morphologically complex cells, such as those in the CNS. Furthermore, this separation exposes the cells to specific environmental and chemical conditions, including harsh dissociation steps, which may introduce biases in downstream data analysis. Moreover, although millions of cells can easily be isolated from a single pig hippocampus, the 10x Genomics kit used for scRNA-seq has a processing capacity of only ~10,000 cells and, with a pipeline diameter of 50 μm, it captures cells no larger than 40 μm, making it more suitable for small-sized cell types (e.g., immune cells) than larger cell types (e.g., neurons), which often cannot pass through the sequencing chip. This limitation led to a significant bias against neurons, most of which were filtered out during sequencing, resulting in a low proportion of neurons in the total cell count and no significant between-group differences. Therefore, the UMAPs generated here represent the proportion of processed cells, rather than the absolute cell count. Notably, in post-CA hippocampal tissue, microglia and oligodendrocytes became the predominant cell types, while other cell types were relatively scarce, as indicated by their lower percentages in the UMAP analysis of the total cell population. Nevertheless, the proportions of sequenced cell groups in our study are consistent with previous studies. In a scRNA-seq study of the mouse cortex following ischemic stroke, microglia and oligodendrocytes constituted the majority of the total cell population, with microglia representing the largest proportion (Guo et al., 2021).

In this study, we utilized domestic pigs as a CA model for several reasons. First, the pig brain has a clear regional structure and multiple gyri, making it more similar to the human brain than commonly used small laboratory animals that lack gyri (e.g., rodents). Second, domestic pigs are widely bred, making them relatively easy to obtain and more ethically acceptable than primates (Lind et al., 2007; Gieling et al., 2011). Third, from a cardiovascular perspective, pigs and humans have remarkable similarities in resting heart rate, blood pressure, and serum biochemical markers, providing a crucial physiological basis for studying diseases such as CA. Additionally, the large thoracic cavity of pigs allows for effective chest compressions and defibrillation, closely resembling clinical scenarios. Finally, the pig brain is the largest among commonly used laboratory animals, providing ample tissue samples for detailed biochemical and histological analyses of specific brain subregions (Cherry et al., 2015).

In summary, our study investigated the cellular composition and transcriptomic changes in the hippocampus following CA in a porcine model, revealing BBB disruption, leukocyte infiltration, and enhanced neutrophil–microglia communication mediated by neutrophil-derived resistin, thereby driving pro-inflammatory microglial polarization. We identified a unique subcluster of activated microglia with high S100A8 expression and uncovered evidence of remyelination involving oligodendrocyte myelin production and OPC differentiation. These findings highlight the complexity of the hippocampal environment post CA and suggest neutrophil–microglia communication and S100A8-expressing microglia as therapeutic targets to mitigate neuroinflammation and promote recovery. Supporting oligodendrocyte and OPC functioning may further enhance post-CA regeneration and neurological repair.

Our conclusions are based on a limited number of biological replicates (three samples at each time point), necessitating further research with a larger sample size to validate the generalizability of these results. Additionally, we only examined gene expression changes within the hippocampus at 6 and 24 hours post–CA, which represent relatively early changes. Importantly, the effects of CA and ischemia–reperfusion on the brain are long lasting, with neurocognitive declines (e.g., memory deficits and depression) that potentially persist for up to 4 years (Roine et al., 1993; Buanes et al., 2015). Consequently, future studies should consider longer observation periods to fully understand the long-term impacts of these changes.

Additional files:

Additional Figure 1 ^{(37.1MB, tif)} : GO enrichment analysis of the microglia subclusters.

Additional Figure 1

GO enrichment analysis of the microglia subclusters.

Additional Table 1: Primers used for quantitative reverse transcription-polymerase chain reaction.

Additional Table 2: Cell counts of different cell types in sham and post-cardiac arrest groups at two time points.

Additional Table 3: List of marker genes used for determining cell-type identity.

Additional Table 4: Comparison of cell type proportions using Two-Proportion Z-Test.

Additional Table 5: Comparison of cell type proportions using the scCODA model.

Additional Table 6 ^{(283.2KB, pdf)} : Differentially expressed genes between Sham and two time points post-ROSC for each cell type.

Additional Table 6

Differentially expressed genes between Sham and two time points post-ROSC for each cell type

Additional Table 7: List of classic polarization markers of microglia.

Additional Table 8: List of genes related to classical functions of microglia.

Additional Table 9 ^{(520.3KB, pdf)} : Number and weight of ligand-receptor interactions among different cell types.

Additional Table 9

Number and weight of ligand-receptor interactions among different cell types

Additional Table 10 ^{(366.5KB, pdf)} : Enhanced ligand-receptor pairs after cardiac arrest.

Additional Table 10

Enhanced ligand-receptor pairs after cardiac arrest

Additional file 1 ^{(149.4KB, pdf)} : Supplementary information on methods.

Funding Statement

Funding: This work was supported by the National Science Foundation of China, Nos. 82325031 (to FX), 82030059 (to YC), 82102290 (to YG), U23A20485 (to YC); Noncommunicable Chronic Diseases-National Science and Technology Major Project, No. 2023ZD0505504 (to FX), 2023ZD0505500 (to YC); and the Key R&D Program of Shandong Province, No. 2022ZLGX03 (to YC).