Mapping Neutrophil Fate and Function in Ischemic Stroke: A Single-cell Roadmap for Translational Insights

doi:10.21203/rs.3.rs-5300274/v1

Download PDF

Research Article

Mapping Neutrophil Fate and Function in Ischemic Stroke: A Single-cell Roadmap for Translational Insights

https://doi.org/10.21203/rs.3.rs-5300274/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Ischemic stroke (IS) accounts for 71% of all strokes, whose diagnosis and prognosis require further exploration. Neutrophil extracellular traps (NETs) are produced by neutrophils, and there is already evidence that NETs play a role in IS, but further studies about crosstalk between immune cells, pathways and NETs are still needed.

Materials and Methods

To assess the expression of neutrophil extracellular traps (NETs), we utilized single sample Gene Set Enrichment Analysis. Stroke-associated NETs genes (SN genes) were identified through differential expression analysis combined with Weighted Correlation Network Analysis. Based on these SN genes, we developed a sophisticated diagnostic model incorporating machine learning techniques. Furthermore, we constructed a single-cell atlas of neutrophil transitions in post-stroke mice. Validation of our findings was conducted both in vitro and in vivo. In vitro, we employed oxygen-glucose deprivation (OGD) experiments to simulate ischemic conditions, facilitating the assessment of NETs formation and monitoring alterations in SN genes expression within neutrophils. In vivo, validation involved tracking changes in peripheral blood levels of these genes in a mouse model of transient middle cerebral artery occlusion (tMCAO) post-cerebral ischemia.

Results

A detailed single-cell landscape depicting the dynamic transitions of neutrophils within the cerebral microenvironment post-stroke has been elaborately constructed.NETs displayed significant differential expression between IS and control groups in peripheral blood, correlating strongly with the activities of neutrophils and macrophages.. Pathways pertinent to IS and NETs were delineated. A diagnostic model incorporating two SN genes was developed, demonstrating an AUC greater than 0.98, effectively pinpointing the hyperacute phase of IS. Additionally, the ceRNA networks concerning IS and NETs were mapped out. In vitro validation with oxygen-glucose deprivation (OGD) experiments revealed marked changes in NET formation and SN genes expression in neutrophils, corroborating our computational predictions. In vivo validation using a mouse transient middle cerebral artery occlusion (tMCAO) model confirmed significant changes in peripheral blood levels of F12 and PLXDC2 after cerebral ischemia, proving the excellent predictive value of these markers for IS.

Conclusion

This study elucidates the complex roles and dynamic changes of neutrophils within the cerebral microenvironment of mice from 3 hours to 3 days following stroke onset. We have identified key genes, immune cells, signaling pathways, and ceRNA networks implicated in the formation of NETs in IS. Our study constructed a robust diagnostic model capable of detecting the hyperacute phase of IS, with an AUC value greater than 0.98. The inclusion of experimental validation for the SN genes F12 and PLXDC2 not only corroborates our model's predictive accuracy but also underscores its potential utility in clinical settings. These findings offer promising avenues for improving early diagnosis and potentially guiding therapeutic strategies in IS.

neutrophil extracellular traps

ischemic stroke

gene signature

WGCNA

ceRNA

PPI

diagnosis model

Stroke remains a leading cause of mortality worldwide, with ischemic stroke (IS) accounting for 71% of all cases and resulting in approximately 5.5 million deaths annually¹. Neuroinflammation plays a pivotal role in the pathogenesis of IS², with the immune response being initiated immediately upon vascular occlusion³. This immune cascade persists for several weeks⁴, involving a range of peripheral immune cells including neutrophils, T cells, B cells, and monocytes/macrophages³. These cells not only participate actively during an IS event but also secrete various cytokines that exacerbate the condition⁵. Neutrophils, the first blood-derived immune cells to infiltrate the ischemic brain, are critical in the progression of post-ischemic inflammation⁶. They impair tissue perfusion and disrupt the blood-brain barrier (BBB) ⁷, facilitating the entry of other peripheral leukocytes ⁴. This disruption has a significant impact on the prognosis of IS patients. Given their central role in disease progression, there is an urgent need for further research into the mechanisms by which neutrophils influence IS pathophysiology.

Neutrophil Extracellular Traps (NETs) are formed by neutrophils extruding DNA, histones, and granular proteins in response to microbial invasion, inflammatory stimuli, or activated platelets ⁸. NETs are hypothesized to initiate coagulation through intrinsic pathways and to act as scaffolds for platelet aggregation, thereby promoting thrombosis. Additionally, evidence suggests that NETs contribute to increased resistance to thrombolysis ⁸. A crucial link between NET formation and ischemic stroke is the identification of abundant NETs in thrombus samples from IS patients⁹. While the contribution of NETs to thrombosis is documented, the mechanisms by which NETs enhance coagulation and thrombosis in IS remain poorly understood¹⁰. Furthermore, the regulatory interactions between NETs and other immune cells, which are critical for understanding the prognosis and treatment of IS patients, are still largely unexplored.

Furthermore, in clinical practice, the treatment window for ischemic stroke patients remains critically narrow. Intravenous thrombolysis is recommended within 4.5 hours and endovascular thrombectomy within 6 hours of stroke onset¹¹. However, delays in diagnosing stroke often cause patients to miss these optimal treatment windows. Diagnostic challenges are compounded by the limited accessibility of MRI and the suboptimal sensitivity of CT in detecting hyperacute IS¹². Consequently, there is an urgent need for a rapid and reliable screening method that can promptly identify stroke patients at the earliest stages, ensuring accurate and timely intervention.

To bridge the existing knowledge gap, our study integrates single-cell profiling with bulk RNA-sequencing to map the complex landscape and dynamic trajectory of neutrophils within the cerebral microenvironment of mice following stroke onset. Our comprehensive bioinformatic analysis has identified key traditional pathways and immune cells associated with the involvement of Neutrophil Extracellular Traps (NETs) in stroke. Additionally, we introduced 'SNSig,' a robust diagnostic tool designed for the rapid detection of ischemic stroke in its hyperacute phase. The formulation of competing endogenous RNA (ceRNA) networks offered novel insights and identified targets for elucidating NETs' function in IS. Subsequent experimental validation of this model has confirmed its effectiveness, providing a robust foundation for its application in clinical diagnostics and enhancing our understanding of NETs in the pathophysiology of stroke.

The overall workflow of our study in shown in Fig. 1.

2.1 Preparation of Data

In this study, the 10x Genomics single-cell RNA sequencing (scRNA-seq) data and the bulk discovery cohort were obtained from stroke patients in Gene Expression Omnibus (GEO, available at https://www.ncbi.nlm.nih.gov/geo/), specific information is listed in Table 1. In our study, the control group refers to people who were non-stroke neurologically healthy, while the IS group includes who were diagnosed with IS based on MRI results and diagnosis by professional pathologists. All series were background corrected and normalized before downloading the validation cohort from the GEO database.

Table 1

Summary of dataset information
Disease	Model	Reperfusion	Strain of Mice	Position	Time (post-MCAO)	Number of cases used	GSE number or fastq download	References
IS	tMCAO	Yes	B6.129	brain	3d	4	GSE247102	PMID: 38509618
IS	tMCAO	Yes	C57BL/6	brain	3h, 12h, 3d	4	PRJNA912889 and PRJNA912890	PMID: 38082331
IS	tMCAO	Yes	C57BL/6	brain	1d,2d	4	GSE197731	PMID: 35688114
IS	tMCAO	Yes	C57BL/6	brain	3d	2	GSE210986	PMID: 37183260
IS	pMCAO	No	C57BL/6	brain	3d	2	ST GSA accession number CRA006677 and scRNA-seq GSA accession number CRA006645	PMID: 38324639
IS	tMCAO	Yes	C57BL/6	brain	1d	6	GSE174574	PMID: 34496660

2.2 Quality control of scRNA‑seq data

Seurat version 4.1.0 was used to further analyses. We performed quality control of the scRNA‑seq data, the following filter standards were set: (1) cells were excluded with detected genes (nFeature) < 200, percentage of mitochondrial gene expression > 30%, percentage of ribosomal gene expression < 40%; (2) genes with expression in fewer than three cells were excluded.

2.3 Data integration, dimensional reduction, visualization, and cell type annotation

The integrated dataset was log-normalized and scaled using Seurat workflow. Harmony version 0.1.0 was used to reduce the batch effects. Principal component analysis (PCA) and uniform manifold approximation and projection (UMAP) were consecutively employed for dimensional reduction and visualization. FindMarker function using Wilcoxon ranksum test in Seurat was used to identify the differentially expressed genes (DEGs) in one cluster compared with other clusters.

2.4 Cell scoring and functional annotation

We used the algorithm AddModuleScore of the R package ‘Seurat’ to evaluate NETs expression and maturity in all cells. Gene set enrichment analysis (GSEA) was used to find the enriched hallmark pathways in a cell-specific or condition-specific manner, based on the DEGs between treatment conditions (e.g., MCAO versus sham), respectively.

2.5 Trajectory analyses and Cell communication inference

The Slingshot algorithm was used to define a computational estimated pseudotime trajectory in alveolar epithelial cells ¹³. CellChat version 1.1.3 was utilized to infer the intercellular communication pathways between each cell type¹⁴.

2.6 Assessment of Degree of Immune Cell Infiltration

To ensure the accuracy of immune infiltration analysis, we employed two different algorithms to assess the extent of immune cell infiltration in the discovery cohort, that is, xCell and CIBERSORT.

The xCell online tool (https://xcell.ucsf.edu/) was used to perform enrichment analysis of infiltrating immune cell level, Immune Score, Stroma Score, and Microenvironment score based on gene expression data. Finally, we screened 22 common immune cells from all the results and showed them.

The CIBERSORT algorithm (https://cibersort.stanford.edu/index.php) was used to evaluate the relative abundances of 22 immune cells in the discovery cohort¹⁵. Standard annotation files were used to organize gene expression profiles. Standard annotation files were used to organize gene expression profiles. Conversion of mRNA data to levels of immune cell infiltration was achieved using the R package "CIBERSORT". When performing data processing, some cell types with low immune cell infiltration were removed and a total of 19 common immune cell types were retained.

2.7 Identification of NETs Score

In order to obtain accurate and efficient NETs markers, we referred to relevant published articles¹⁶ and obtained 69 genes as biomarkers to assess NETs expression. The extent of NETs in the discovery cohort was assessed using the ssGSEA algorithm.

2.8 Assessment of Hallmark Gene Set and Gene Set Enrichment Analysis (GSEA)

To further understand how NETs function in stroke, we downloaded hallmark gene sets from MSigDB (www.gsea-msigdb.org/gsea/index.jsp) and quantified the degree of activation of the hallmark pathway using the ssGSEA algorithm for all samples in the discovery cohort. To ensure the reliability of the conclusions, we simultaneously performed GSEA analysis using hallmark gene sets.

2.9 Identification of Stroke Related Differential Expression Genes (DEGs)

According to clinical information, all samples were divided into 2 groups, that is, Control group and Stroke group. Using FDR < 0.05 and | log2 (fold change) | > 1 as thresholds, the LIMMA algorithm was applied to identify DEGs between different clusters.

2.10 Construction of Gene Co-Expression Network

Weighted correlation network analysis (WGCNA) is a bioinformatics analytical method that is used frequently to explore effectively the relationships between genes and phenotypes¹⁷. The WGCNA tool in R was used to build a weighted co-expression network for the discovery cohort expressing data before a subset of genes with absolute deviations greater than 25% from the median were selected for further investigation. PickSoftThreshold was used to select and verify an optimum soft threshold, which is 5 in the study. In order to find modules based on the topological overlap, the matrix data were transformed into an adjacency matrix, and then clustered. Clustering dendrograms were generated after the computation of module eigengene (ME) and the merging of related modules in the tree based on ME. Using phenotypic data and modules, the importance of genes and clinical data was assessed, and the relationship between models and modules was examined. Modules with correlations > 0.8 and P < 0.05 were considered as modules highly associated with NETs phenotypes.

2.11 Functional Enrichment Analysis

R packages “clusterProfiler” and “enrichplot” were used to perform GO assays, KEGG assays. Gene sets with FDR < 0.05 were considered highly enriched in GO and KEGG assays.

2.12 Protein-Protein Interaction (PPI) Network Building

Currently, the Search Tool for the Retrieval of Interaction Genes (STRING) database is the highest species coverage and the largest interaction. A protein-protein interaction network was constructed based on STRING. In this study, the PPI network of DEGs was constructed based on the minimum interaction (> 0.4). The PPI network was uploaded to the Cytoscape software (version 3.9.1) for visualization. In PPI, we used MCODE to select the genes in Degree top 50 and intersect with SN genes, and finally obtained six genes as mRNAs for constructing ceRNA networks.

2.13 Identification of LncRNAs Co-Expressed with mRNAs

Differential expression lncRNAs (DElncRNAs) were identified from DEGs in the discovery cohort. The spearman algorithm was used to identify the association of DElncRNAs with six mRNAs. P < 0.05 and cor > 0.3 were taken as thresholds to obtain co-expressed DElncRNAs for each mRNA.

2.14 Construction of CeRNA

The mRNA required for ceRNA construction has been obtained in previous analyses. Prediction of interactions between miRNAs and mRNAs was based on miRDB (http://mirdb.org/), TarBase(https://dianalab.ece.uth.gr/html/diana/web/index.php?r=tarbasev8), TargetScan (https://www.targetscan.org/vert_80/), and miRwalk (http://mirwalk.umm.uni-heidelberg.de/), taking miRNAs emerging in at least two databases and intersecting them with differential expression miRNAs obtained through dataset GSE110993 to obtain miRNAs used to construct ceRNA networks. The construction of interactions between lncRNAs and miRNAs was based on LncBase (https://diana.e-ce.uth.gr/lncbasev3/home), and the results of mRNA-lncRNA co-expression analysis were synthesized to construct corresponding complete ceRNA networks for each of the six mRNAs. Visualization was performed by Cytoscape (version 3.9.1).

2.15 Isolation of Neutrophils from Mice Peripheral Blood

Peripheral blood was collected via cardiac puncture from anesthetized C57BL/6 mice to ensure minimal stress and discomfort, in compliance with the ethical guidelines set by the Hunan Province Administration Committee of Affairs Concerning Experimental Animals, China. All the experimental protocols were approved by the Animal Ethics Committee of Central South University. Blood was drawn directly from the left ventricle using a heparin-precoated syringe to prevent coagulation. The collected blood was diluted 1:1 with isotonic saline (0.9% NaCl) and mixed with an equal volume of erythrocyte sedimentation solution from the Peripheral Blood Neutrophil Isolation Kit (Solarbio, P9201). The mixture was allowed to settle at room temperature for 30–40 minutes for erythrocyte sedimentation. The leukocyte-enriched supernatant was then carefully collected. For sample volumes less than 5 mL, a density gradient was established in centrifuge tubes using 4 mL of Solution A and 2 mL of Solution C from the kit. For larger samples, a 2:1 ratio of Solution A to Solution C was adjusted according to the post-sedimentation sample volume. The samples were centrifuged at 600-1000g for 25–30 minutes at room temperature to optimize neutrophil integrity and yield. Post-centrifugation, the neutrophil layer was transferred to a new centrifuge tube and washed twice with the wash solution provided in the kit, first at 250g for 10 minutes and then for 5 minutes after discarding the supernatant. The final pellet was resuspended for further experiments.

2.16 Flow Cytometric Identification of Neutrophils

Neutrophils were isolated from mouse peripheral blood and subjected to flow cytometry for phenotypic characterization. After centrifugation at 1500 rpm for 5 minutes at 4°C, the supernatant was removed, and the cells were stained with PE-conjugated anti-CD11b antibody (Abcam, ab25175) and FITC-conjugated Ly-6G monoclonal antibody (CST, 88876) at a 1:200 dilution in 100 µL of PBS. This staining occurred in the dark for 30 minutes to prevent fluorophore degradation. Cells were then washed with 500 µL of ice-cold PBS and centrifuged at 1500 rpm for 5 minutes. The resuspended cell pellet in 200 µL of ice-cold PBS was analyzed via flow cytometry, with FlowJo software determining the percentage of positive cells and their mean fluorescence intensity.

2.17 Primary culture of Neutrophils and oxygen-glucose deprivation (OGD) treatment

Following isolation, neutrophils were resuspended and cultured in RPMI-1640 medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Hyclone) and 1% penicillin-streptomycin (Hyclone). The cells were incubated at 37°C in a humidified atmosphere containing 5% CO2 for approximately 4 hours to allow for stabilization. For OGD treatment, the neutrophils were washed and then transitioned to glucose-free RPMI-1640 medium (Thermo Fisher Scientific), and subjected to OGD in a hypoxic incubator (NuAire) set to an atmosphere of 0.5% O2, 5% CO2, and 94.5% N2. Exposure durations were set at 0, 3, 5, and 24 hours, respectively. Cells were collected after each time point for subsequent analysis.

2.18 Enzyme-linked immunosorbent assay (ELISA)

Cells were harvested and washed three times with pre-chilled PBS, and lysed using non-denaturing lysis buffer to ensure complete lysis. After lysis, cell debris was pelleted by centrifugation at 10000g for 10 minutes at 4°C, and the clear supernatant containing soluble proteins was collected. These lysates were then used for the quantitative analysis of MPO-DNA complexes using an ELISA kit (hnybio, HS1194-Mu), following the manufacturer’s protocol.

2.19 Establishment of Transient Focal Cerebral Ischemia Model

This research employed the Middle Cerebral Artery Occlusion (MCAO) procedure to induce transient focal cerebral ischemia in male C57BL6 mice aged between 10 to 12 weeks, with body weights in the range of 25 to 28 grams. Four mice were used in each experimental and control (sham-operated) group. Animals were housed under controlled temperature conditions and subjected to a 12-hour light/dark cycle, with unrestricted access to water and food. A 7-day acclimatization period was observed prior to experimentation. Animal care and usage complied with the guidelines set by the Hunan Province Administration Committee of Affairs Concerning Experimental Animals in China. The Animal Ethics Committee granted approval for all experimental procedures. Anesthesia was administered using a blend of isoflurane, oxygen, and nitrous oxide. The procedure began with a longitudinal neck incision to reveal the common carotid artery. The artery's branches were electrocoagulated in preparation for the occlusion. A nylon suture was introduced through the external carotid artery, navigated into the internal carotid artery, and extended towards the middle cerebral artery to block blood flow. A Laser Doppler Flowmeter was utilized to confirm a substantial decrease in regional cerebral blood flow (CBF), aiming for a reduction to below 25% of pre-occlusion levels. The blockage was maintained for 60 minutes before removal to initiate reperfusion, followed by observation periods of 3h, 5h, and 24h. For control animals, sham operations mirrored all steps except the arterial blockage. Mice failing to meet the CBF reduction criteria or not surviving the surgery were excluded from the study. The work has been reported in line with the ARRIVE criteria¹⁸.

2.20 Peripheral Blood Collection

Peripheral blood samples were collected from the mice at predetermined time points (3h, 5h, and 24h post-reperfusion) to monitor the temporal expression of targeted genes. The blood collection was performed via cardiac puncture under deep anesthesia to minimize distress. The collected blood was immediately processed for RNA extraction to ensure the integrity of the RNA.

2.21 Quantitative Real-Time PCR

RNA was isolated using Trizol reagent (Invitrogen) from harvested neutrophils for in vitro validation and from the peripheral blood of both experimental and control (sham-operated) mice for in vivo validation, respectively.. The RNA thus isolated was then reverse transcribed into cDNA employing the iScript cDNA Synthesis Kit. Quantitative real-time PCR (qRT-PCR) was conducted using iTaq Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) on a Bio-Rad CFX Connect Thermocycler, in adherence to standardized protocols. Specific primer pairs were designed for F12, PLXDC2, and GAPDH genes to facilitate their amplification during the qRT-PCR process. GAPDH served as the housekeeping gene, providing an internal reference to normalize the relative mRNA expression levels across different samples. To assure the accuracy and reproducibility of the results, each qRT-PCR assay was performed in triplicate using independent samples. The primer sequences used for the amplification of target genes were as follows: For F12, the Forward Primer sequence is 5'- GCAGCAAACACAACCCGTGC − 3', and the Reverse Primer sequence is 5'- CTCGTGGAAGAACTTGAGAA − 3'. For PLXDC2, the Forward Primer sequence is 5'- GTTGTGGCCCTTGTGTGTCC − 3', and the Reverse Primer sequence is 5'- CCACTGTCCACCCAGTCCTG − 3'. For GAPDH, serving as the internal control, the Forward Primer sequence is 5'- TGGCATTGTGGAAGGGCTCA − 3', and the Reverse Primer sequence is 5'- ATGCAGGGATGATGTTCTGG − 3'.

2.22 Statistical Analysis

Statistical analyses were performed using Prism software version 10.2.3 and R version 4.2.1. For normally distributed data (e.g., qPCR and ELISA results), differences between groups were evaluated using one-way ANOVA followed by Tukey’s post hoc multiple comparisons test. Non-parametric data were analyzed using the Wilcoxon rank-sum test for comparisons between two groups, and the Kolmogorov-Smirnov test for comparisons among multiple groups. Correlation analyses were conducted using Spearman's rank correlation coefficient. The ssGSEA algorithm from the 'GSVA' R package was employed to calculate NETs scores. Statistical significance was set at *P < 0.05, **P < 0.01, ***P < 0.001, and **P < 0.0001.

3.1 Preliminary Analysis of MCAO Mice Brain Single Cell Profiles

We downloaded mouse single-cell RNA sequencing (scRNA-seq) data from the GEO, comprising brain tissue samples collected at 3 hours, 12 hours, 1 day, 2 days, and 3 days post-stroke, with detailed information provided in Supplementary Table 1. Following quality control, a total of 213,033 cells were progressed to downstream analyses. To mitigate batch effects between different cohorts, Harmony was employed. Eventually, 37 cell clusters were identified (Fig. 2a). Using published marker genes, we excluded clusters expressing multiple lineage markers, deemed as doublets, and re-annotated the remaining cells into 14 cell clusters (Fig. 2b), with their characteristic marker genes depicted by a dot plot (Fig. 2c). We examined the average proportion of neutrophils across samples from varying time points post-stroke and observed an expansion of neutrophils in all stroke groups compared to the sham group (Fig. 2d). At 1 day, 2 days, and 3 days post-stroke, neutrophil proportions initially rose then fell, with significantly more neutrophils present on the ipsilateral side of the stroke lesion compared to the contralateral side (Fig. 2d). Intriguingly, neutrophil abundance peaked at the 2 days before sharply declining. At 3 days, although the ipsilateral side retained more neutrophils than the first day, the neutrophil count on the contralateral side reached a nadir, even lower than that in the sham group (Fig. 2d). Assessing neutrophil extracellular trap (NETs) expression across all cells revealed, as expected, the highest expression in neutrophils, followed by monocytes and macrophages (Figs. 2e and 2f).

3.2 Analysis of NETs in Neutrophils of MCAO Mice Brain

Following reannotation, a total of 8,079 neutrophils were advanced for further downstream analysis. After reclustering, 8 neutrophil clusters were obtained (Fig. 3a). Cluster 0 and 2 exhibited overly similar marker gene profiles (Fig. 3b), leading us to annotate them as a single cluster during the reannotation process (Fig. 3c). Finally, 7 neutrophil clusters were identified (Fig. 3c), with their characteristic marker genes visualized via a dot plot (Fig. 3d). Assessment of NETs expression across all neutrophils revealed Mmp8 + Neu as the most active in NETs expression, closely followed by Ifit3 + Neu (Fig. 3e). Additionally, we evaluated the maturity of neutrophils, a crucial feature of neutrophils, and found a strong correlation between NETs expression and maturity (Fig. 3f). The proportions of neutrophil subclusters in different locations over time are depicted (Figs. 3g and 3h), showing that Mmp8 + Neu dominate during the acute phase of stroke. Consistent with this finding, functional enrichment analysis of Mmp8 + Neu indicates its function is mainly concentrated in lymphocyte chemotaxis and migration (Fig.S1a), enabling its rapid recruitment to the brain from the periphery and marking it as the predominant cell type during the acute stage. The functions of C1qc + Neu are related to antigen presentation and cell migration (Fig.S1b), suggesting they also originate from the peripheral blood but infiltrate the brain less swiftly than Mmp8 + Neu. Their numbers increase at 1 day but subsequently decrease (Fig. 3g), with a higher proportion observed on the ipsilateral side of the stroke lesion compared to the contralateral or sham group (Fig. 3h), implying a preference for entering the stroke-affected hemisphere. Mbnl2 + Neu emerge later, starting from 1 day, with their proportion gradually increasing over time. Their functions center around mRNA metabolism and cellular adhesion (Fig.S1c). Pseudotime analysis, initiating from Cd34 + neutrophils, reconstructed differentiation trajectories (Fig. 3i). Among these trajectories, three paths consistently progressed from Cd34 + Neu to C1qc + Neu and then to Mbnl2 + Neu, which suggests a potential transition between Mbnl2 + Neu and C1qc + Neu. Considering the substantial amplification of C1qc + Neu observed at 1 day post-stroke followed by a subsequent decrease at 2 days in proportion, coupled with the increasing presence of Mbnl2 + Neu starting from day 1, which continues to rise over time, there is grounds to speculate that Mbnl2 + Neu may derive from C1qc + Neu. Ifit3 + Neu, with stable proportions, primarily respond to viral infections (Fig.S1d), while Camp + Neu, Fcnb + Neu, and Cd34 + Neu, abundant in sham groups, are associated with NADP metabolism, chromosome segregation, and oxidative phosphorylation (Figures S1e, S1f, S1g).

Expression of NETs varied over time and location, peaking at 2 days post-stroke before decreasing below sham levels by day 3 (Fig. 3j). Similarly, NETs expression on the stroke-lesioned side was lower than the contralateral side only on day 3 (Fig. 3k). Reflecting this decline, NETs expression in Mmp8 + Neu, the highest NETs-expressing cells, also dropped below sham levels by day 3 (Fig. 3l), hinting at exhaustion of NETs expression by the third day post-stroke.

3.3 Dominant Cells for Cell Communication Change with Time after Stroke

To investigate temporal variations in cellular communication following stroke, we used the sham group as a control and compared it with experimental groups at 3h, 12h, 1day, 2d, and 3d post-event. We initially quantified overall communication numbers and strength across these time points for both control and experimental groups (Figures S2a-e), along with interactions between neutrophil clusters and other cell clusters (Figures S2f-j), and the differential interactions of neutrophil clusters with other clusters between experimental and control groups (Figures S2k-o). By dividing the communication frequency/intensity of the experimental groups by that of the control group, we represented the resultant ratios in line graphs (Fig. 4a), positing this ratio as a succinct indicator of cellular communication change. We observed congruent trends in frequency and intensity, with peaks at 12h and 2d and a trough at 3d (Fig. 4a). Notably, the ratio fell below 1 at 3d, implying reduced cellular communication in the stroke group compared to the sham group at this stage, potentially indicating exhaustion of communication processes. Subsequently, we depicted 2D maps of cellular communications at different time points for both control and experimental groups, highlighting the intensity of signals received and transmitted by various cell subsets (Figs. 4b-f). Unlike the consistent dominance of Astro in the sham group, leadership in communication networks shifted among cell types in stroke groups over time. Specifically, at 3h, the dominant communicators were Ifit3 + Neu, Mmp8 + Neu, and C1qc + Neu; at 12h, they were Cd34 + Neu, Ifit3 + Neu, C1qc + Neu, and Mmp8 + Neu; at 1d, Mbnl2 + Neu and Mmp8 + Neu took precedence; at 2d, Mmp8 + Neu, Ifit3 + Neu, C1qc + Neu, and Mbnl2 + Neu dominated; yet, by 3d, communication patterns resembled those of the sham group, with Astro regaining prominence (Figs. 4b-f). This suggests a final surge of neutrophil communication at 2 d before an aberrant pattern gradually normalized by 3 d. This decline, resembling a depletion pattern, aligns with prior observations of cellular proportion shifts, NETs scores, and overall communication dynamics. A heatmap further detailed the specific changes in intercellular communications and signaling pathways involved (Figure S3).

3.4Immune Microenvironment and NETs Expression in Whole Blood of IS Patients

Expression matrices (GSE58294) from whole blood of IS patients were downloaded from the GEO as a discovery cohort, including a total of 92 samples from the control group, within 3 hours after stroke attack, 5 hours after stroke attack, and 24 hours after stroke onset. Immune cell infiltration and immune microenvironment score were assessed in each group by xCell (Fig. 5a, b and Supplementary Table S1). The results showed that compared with the control group, the infiltration of B cells and CD4 + T cells was significantly decreased, while the infiltration of monocytes/macrophages, neutrophils, and NKT cells were significantly increased in the stroke group. To validate the finding, CIBERSORT was selected to re-evaluate these samples for immune cell infiltration (Fig. 5c and Supplementary Table S2). After some unimportant cells or cells that were poorly infiltrated were removed, both CIBERSORT and xCell yielded similar results. We found that the degree of neutrophil infiltration in whole blood was particularly altered in stroke patients. Given the recent increasing emphasis on the relationship between neutrophils and NETs and their role in stroke, we obtained a signature gene set of NETs from a published study¹⁶ and assessed NETs using the ssGSEA algorithm for 92 samples in the discovery cohort (Supplementary Table S3). The boxplot showed that the expression of NETs was significantly increased in the stroke group (Fig. 5d). ROC curves showed that NETs could distinguish the control group from the stroke group well regardless of the duration of the stroke (Fig. 5e-g), suggesting that NETs may play an important role in the development and progression of stroke.

3.5 Exploration of the Role of NETs in IS Patients

To understand how NETs affect stroke patients, we selected data from 69 stroke patients in the discovery cohort and divided them into two groups by the median expression of NETs expression (4.067074), which means 34 samples in the Low NETs group and 35 samples in the High NETs group. By comparing the immune cell infiltration between the two groups, we found that the degree of infiltration of monocytes/macrophages and neutrophils increased significantly in the High NETs group compared with the Low NETs group, and the degree of infiltration of other immune cells decreased to varying degrees (Fig. 6a, b). We performed correlation analysis between immune cell infiltration and NETs expression in the control and stroke groups, respectively, and the heatmap showed that NETs were most associated with neutrophils in the stroke group, followed by macrophages and CD4 + T cells (Fig. 6c). Interestingly, we found that the association of NETs with some immune cells changed because of different groups. For example, Th2 cells, Tregs, DCs, and macrophages, which were significantly negatively correlated with NETs in the stroke group, were not significantly correlated with NETs expression in the control group, which seems to imply changes in a subtle cell-to-cell link related to NETs before and after IS attack, which is very worthy of further study. To further identify pathways that are associated with NETs, we selected the HALLMARK pathway gene set from public websites (MSigDB) and quantified the activation degrees of all pathways by ssGSEA algorithm. The heatmap shows all pathways associated with NETs (P < 0.05) in the control or stroke groups (Fig. 6d). By GSEA analysis, we found the hallmark pathways that are most likely to be related to NETs' role in stroke (Fig. 6e, f and Supplementary Table S4).

3.6 Identification of Stroke-NETs Related Genes (SN genes)

In order to search for SN genes, differential gene expression analysis was first performed between the control and stroke groups, and a total of 156 differentially expressed genes (DEGs) were obtained using log2FC > 1 and adjusted P < 0.05 as thresholds (Fig. 7a and Supplementary Table S5). To identify genes related to NETs, WGCNA was performed. First, we selected 5000 genes based on the top 25% median absolute deviation, and in the next step, we chose the soft-threshold 5 (based on the topology on the scale-free criterion with R² = 0.9) in order to get a scale-free network (Fig. 7b). Finally, 15 modules were obtained through average hierarchical clustering and dynamic tree clipping (Fig. 7c). To obtain important modules associated with NETs, we evaluated the association of modules and NETs by Pearson’s correlation analysis. Two modules with |correlation coefficient| > 0.8 and P < 0.05 were selected as gene modules strongly associated with NETs, namely the brown module and the turquoise module, which contained 663 and 1341 genes (Fig. 7d). Significant existed correlation in the MM and gene significance modules (GS) of the brown and turquoise modules. Key genes were screened out with geneTraitSignificance > 0.6 and geneModuleMembership > 0.8, and a total of 295 key genes were found in the two modules (Fig. 7e, f). To ensure that we found the correct gene modules, we performed GO enrichment analysis of the brown module that is strongly positively correlated with NETs and found that it was related to the function of neutrophils (Fig. 7g), illustrating their strong potential relationship with NETs. Finally, we intersected 156 DEGs and 295 module genes to obtain 20 SN genes (Supplementary Table S6).

3.7 Establishment of Hyperacute Stroke Diagnosis Model Based on SN Genes

With the help of medical imaging, the diagnosis of IS has been greatly promoted, but the diagnostic time required for imaging greatly hinders the efficiency of correct diagnosis for patients, especially for patients with hyperacute stroke. For this reason, we hope to construct a simple and robust diagnostic model to help clinicians quickly identify stroke patients, especially those in the hyperacute phase. To ensure the robustness of the model and prevent overfitting, we randomly divided the discovery cohort into the training cohort and test cohort according to a ratio of 4:6, that is, 38 samples in the training cohort and 54 samples in the test cohort (Supplementary Table S7 and S8). Based on R package “caret”, we selected two SN genes to construct a model for identifying control and hyperacute stroke patients (stroke 3h) in the training cohort. Using the Linear regression function, the diagnostic model “SNSig” is as followed: Sig.Score=-5.8467 + 0.2172*expression of F12 + 0.5016*expression of PLXDC2 (Table 2). The boxplot shows the expression of two genes used to construct SNSig and Sig.score in each group of the training cohort (Fig. 8a). ROC curves showed that SNSig had an extremely strong diagnostic ability for control and hyperacute stroke patients (Fig. 8b). In order to verify the holistic diagnostic ability of SNSig, we sought to test the diagnostic power of SNSig in the control and groups with longer time to stroke attack (stroke 5h and stroke 24h). ROC curves showed that SNSig had an incredible diagnostic ability not only for patients in the hyperacute phase, but also for patients with slightly longer time to stroke attack. We checked gene expression and Sig.score in the test cohort and verified the diagnostic power of SNSig (Fig. 8c). ROC curves indicate its strong robustness (Fig. 8d). Next, to maintain SNSig integrity, we validated the diagnostic power of SNSig in the original discovery cohort (Fig. 8e, f). To further validate the diagnostic effect of SNSig, we downloaded an external independent cohort as the validation cohort (GSE16561). The boxplot shows that SNSig genes and Sig.score were significantly different in both the control and stroke groups in the validation cohort (Fig. 8g). The ROC curve shows that in the validation cohort, SNSig remained extremely diagnostic ability (Fig. 8h). Besides, the expression of SNSig genes and their robustness were not affected by other common clinical indicators of patients, such as gender and age (Fig. 8i, j). In conclusion, in the discovery cohort, the AUC values of our proposed SNSig are all above 0.98. In the external independent cohort, the overall AUC value of SNSig was also above 0.98 and did not decline below 0.96 because of common clinical factors. SNSig can not only diagnose stroke patients in the hyperacute phase, but also its robustness exceeds most of the currently proposed diagnostic models. So far, we propose a brand-new and extremely robust diagnostic model for hyperacute stroke, which will play an important role in clinical diagnosis and further scientific research.

Table 2

The calculated coefficient of the 2 SN genes
Gene	Coefficient
F12	0.2172
PLXDC2	0.5016

3.8 Neutrophil Identification and Response to Oxygen-Glucose Deprivation

Flow cytometric analysis successfully identified neutrophils in the peripheral blood, with a high positive rate of 96.77% for CD11b + Ly6G + cells, indicating a robust isolation and detection methodology (Fig. 9a). Quantitative PCR revealed a dynamic response of neutrophils to oxygen-glucose deprivation (OGD). F12 expression was elevated at 3 hours post-OGD (p < 0.01) and reached its peak at 5 hours (p < 0.0001), demonstrating a rapid increase in response to hypoxic conditions. This expression remained higher than baseline at 24 hours (Fig. 9b). Similarly, PLXDC2 showed an immediate increase at 3 hours (p < 0.01) with a peak expression at 5 hours post-OGD (p < 0.0001). Although a slight reduction was observed at 24 hours, the expression was still significantly above the control levels (Fig. 9c). To further validate neutrophil activation, we quantified the MPO-DNA complex, a marker of neutrophil extracellular trap (NET) formation, via an ELISA. There was a pronounced increase in MPO-DNA levels at 3 hours (p < 0.0001), which further escalated at 5 hours (p < 0.0001) and maintained a significantly elevated concentration at 24 hours (p < 0.0001) post-OGD (Fig. 9d). This pattern reflects robust neutrophil activation and sustained NETosis under ischemic-like conditions, reinforcing the potential of F12 and PLXDC2 as biomarkers for early ischemic detection and monitoring of neutrophil activity.

3.9 Validation of Predictive SN Genes and Their Dynamic Expression Post-Ischemia

In line with our bioinformatic predictions, experimental qPCR analysis confirmed significant upregulation of F12 and PLXDC2 in the peripheral blood following tMCAO at 3h, 5h, and 24h time points, as compared to controls. F12 expression showed a substantial increase at 3h, which remained elevated at 5h and 24h post-MCAO. PLXDC2 expression demonstrated a pronounced elevation at 3h and continued to rise at 5h, before presenting a slight decrease at 24h still maintaining higher levels than control (Fig. 9e,f). These temporal expression patterns highlight the genes' responsiveness to ischemic events and reinforce their potential as reliable markers for the hyperacute phase of IS.

3.10 Identification of Key SN Genes and Construction of CeRNA Network

To identify key SN genes and provide further guidance for clinical and scientific research, we first constructed PPI using 156 DEGs in the STRING database, and then used Cytoscape to remove genes without interaction and visualize them (Fig. 10a). The genes in Degree top 50 were identified as stroke core genes using the MCODE (Fig. 10b). Fifty stroke core genes and 20 SN genes were intersected to obtain six SN critical genes. The heatmap reflects the correlation of SN critical genes and Sig.score with immune score and more than 20 immune cells infiltration (Fig. 10c, d), implying that monocytes/macrophages, NKT cells, neutrophils, and CD4 + T cells may be involved in the progression of critical genes affecting the disease. To construct the ceRNA network, based on DEGs, co-expression analysis of mRNA and lncRNA was performed to search for differentially expressed lncRNAs with co-expression relationship with six key genes as alternative lncRNA (Fig. 10e). We downloaded and selected miRNAs that may interact with six key genes from four public websites (TargetScanHuman, miRDB, miRWalk, TarBase) and intersected them with stroke-related differential expressed miRNAs obtained from an external independent cohort GSE110993 to get alternative miRNAs for constructing ceRNA networks. A ceRNA network was constructed by downloading lncRNAs that may interact with alternative miRNAs on the LncBase website and combining alternative lncRNA (Fig. 10f), and the Sankey diagram further demonstrated this interaction network (Fig. 10g). Our study provides seminal suggestions for exploring these critical SN genes.

This study elucidates the role of neutrophil extracellular traps (NETs) in the occurrence, progression, diagnosis, and prognosis of stroke, leveraging existing scRNA-seq and bulk RNA-seq datasets. Due to the lack of post-stroke human brain data, we analyzed scRNA-seq data from the brains of mice subjected to middle cerebral artery occlusion (MCAO) at five critical time points—3 hours, 12 hours, 1 day, 2 days, and 3 days post-stroke—to observe NETs function during the transition from the hyperacute to the acute phase. Simultaneously, we utilized bulk RNA-seq data from human peripheral blood at three time points post-stroke (3 hours, 5 hours, and 1 day) to integrate with our mouse brain data, providing a comprehensive view of NETs evolution and their impact on stroke progression and prognosis. This integration also supports the development of a diagnostic model for rapidly identifying hyperacute stroke, potentially enhancing the efficacy of thrombolysis and mechanical thrombectomy interventions.

Ischemic stroke initiates with thrombus formation, obstructing cerebral blood supply. Traditionally linked to aberrant platelet activation and dysregulated blood coagulation, the past decade has broadened our understanding to include the significant role of peripheral immune cells, such as monocytes and neutrophils, in thrombus development¹⁹, aligning with our assessments of immune cell infiltration in the whole blood of post-stroke patients. Notably in 2017, study demonstrated substantial neutrophil and NETs accumulation in stroke-related thrombi, with elevated plasma NETs levels²⁰, consistent with our evaluation of NET expression in whole blood. This suggests a profound connection between NETs and thrombus formation. Recent studies have elucidated various mechanisms by which NETs promote thrombosis, such as their DNA serving as a scaffold for F12 to activate the coagulation cascade, and their histones recruiting and activating platelets via TLR2 and TLR4²¹. Current research on NETs in thrombosis has largely focused on the effects of individual components rather than the holistic role of NET structure itself. Utilizing a bioinformatics approach, we conducted a comprehensive evaluation of NETs, analyzing their interactions with other cells and pathways to uncover previously neglected targets. Our findings link NET expression in stroke patients to critical biological processes including hypoxia, glycolysis, inflammatory responses, and the JAK/STAT pathway. This aligns with existing literature indicating the JAK/STAT pathway's role in neutrophil activation and NET release²², the dual role of inflammation in stimulating and being exacerbated by NETs, and the contribution of reduced glycolysis to decreased NET production²³. Notably, while hypoxia has been implicated in NET formation in contexts such as gastric cancer²⁴, its specific relationship to NETs in ischemic stroke remains unexplored. Considering the frequent co-occurrence of hypoxia and inflammation in stroke, we hypothesize that hypoxia significantly contributes to NET formation, presenting an overlooked yet critical area of study in ischemic stroke's pathophysiology.

We propose that the effects of NETs following the onset of IS can be broadly categorized into effects within the brain and those in the periphery. Peripherally, NETs primarily contribute to thrombus stability. Evidence indicates that thrombi in patients who experience stroke onset beyond 4.5 hours contain significantly increased NET levels²⁵. Intriguingly, older thrombi not only show denser NETs accumulations but also display enhanced resistance to dissolution²⁶. Although the specific mechanisms underlying thrombus reinforcement by NETs remain largely unexplored, our data reveal a marked increase in NETs expression in the stroke group compared to controls. Notably, the concentration of NETs peaks at five hours post-onset and remains at a relatively high level throughout the day. This sustained elevation likely plays a critical role in maintaining thrombus stability. We propose that these persistently elevated NET levels are essential for thrombus reinforcement, representing a potential target for therapeutic intervention to mitigate the progression of ischemic damage.

Moreover, current research predominantly focuses on the interaction between platelets and neutrophils, while neglecting to discuss the relationships between neutrophils and other immune cells in the periphery. Although certain studies suggest that in mice, monocytes, NETs and platelets may collaborate to facilitate thrombus formation²⁷, such research is inherently limited by the species and thus remains superficial. According to our findings, we observed no correlation between NETs and M0 macrophages or M2 macrophages in the control group of human whole blood. Conversely, in the stroke group, a significant positive correlation emerged among these cells. These results imply that macrophages may become activated following thrombus formation and subsequently engage in interactions with NETs. Such interactions could potentially contribute to maintaining high levels of NETs or reinforcing thrombi.

Currently, many researchers believe that the primary roles of NETs in the brain involve breaching the blood-brain barrier (BBB), exacerbating neuronal death, hindering vascular remodeling, and promoting inflammation²¹. Owing to the brief lifespan of neutrophils, their limited absolute numbers in the brain, and the intricate architecture of the brain, the spatiotemporal distribution of neutrophils in the brain following stroke remains poorly understood, particularly regarding temporal dynamics. Some studies suggest that neutrophils begin to appear in the brain as early as 12 hours following tMCAO, peaking within 1–2 days²⁸. Other studies contend that due to neutrophil accumulation in the peri-infarct cortex, the greatest accumulation of neutrophils in the cortex and striatum is observable on day 3 post-cerebral ischemia²³. Conversely, in certain transient MCAO models, only a minority of ischemic mice exhibit substantial neutrophil infiltration within the first day, while generally the number of neutrophils increases on days 2 and 4²⁹. Our findings generally support the first notion, indicating that neutrophils infiltrate the brain as early as 12 hours post-insult, reaching a maximum by the second day and declining by the third day. In essence, we have elucidated the temporal trajectory of neutrophils in the brain following MCAO from a single-cell perspective.

To date, analyses of neutrophils and NETs in post-stroke brains have predominantly been limited to tissue sections, lacking integration with single-cell RNA sequencing to thoroughly explore neutrophil clusters and their specific roles. Our study fills this gap in the field, providing a detailed landscape of neutrophil dynamics in MCAO mice from 3 hours to 3 days. Initially, we identified seven neutrophil clusters. The Mmp8 + neutrophil cluster are extremely sensitive to chemokines, and they first respond to chemokines into the brain. Given its high expression of Mmp8, we speculate that it may disrupt the BBB through MMPs related pathways³⁰, resulting in sustained entry of peripheral immune cells into the brain and mediating deleterious effects during the acute phase of IS.

The C1qc + neutrophil cluster, responsive secondarily to chemokines, predominantly migrates towards the stroke-affected hemisphere and may undergo a transformation into Mbnl2 + neutrophils. We propose a novel hypothesis: this transformation could involve C1qc + neutrophils releasing mitochondrial DNA to form NETs²¹, subsequently transitioning into Mbnl2 + neutrophils. Analysis of mitochondrial gene expression across different clusters revealed that Mbnl2 + neutrophils exhibit notably lower mitochondrial levels compared to the high levels in C1qc + neutrophils (Figure S1h), supporting our hypothesis. Functionally, enrichment analysis suggests that these transformations may be linked to processes such as cytoplasmic translation and protein folding, potentially enhancing NETs production. Intriguingly, at 1 day post-stroke, Mbnl2 + neutrophils show a significant increase in communication frequency and intensity with C1qc + neutrophils relative to the sham group (Figure S3g), suggesting a tight association and a possible feedback mechanism that promotes this phenotypic shift. While past research has predominantly focused on lytic NETs formation, our findings emphasize the significance of C1qc + neutrophils, potentially opening a new perspective on NETs in stroke research, particularly regarding mitochondrial NETs formation, an underexplored area.

At 2d, the expression of NETs reaches its peak, whereas by 3d, there is a comprehensive decline in both neutrophil count and NET expression: The NET expression in the stroke group even falls below that of the sham group, with the ipsilateral stroke side exhibiting lower NET expression than the contralateral side. The Mmp8 + neutrophils, which contribute most to NET production, show a significant decrease in their NET expression at 3d. Neutrophils, which dominate cellular communication, revert to levels comparable to those in the sham group at 3d. We speculate that this may signify functional exhaustion of neutrophils within the brain.

Notably, by 3d, our cellular communication analyses reveal a significant increase in the number and strength of communication between macrophages/microglia and Mbnl2 + neutrophils. Thus, we propose that the neutrophil exhaustion observed at 3d could be related to the phagocytic clearance of NETs and neutrophils by these scavenger cells²³.

Consequently, based on the analysis of neutrophil proportions and pseudotime ordering, we propose the following timeline for the differential neutrophil clusters: 1. Mmp8 + neutrophils are the first responders to chemokines, rapidly infiltrating the brain where they produce NETs and participate in the disruption of the BBB. 2. C1qc + neutrophils, as the second wave of chemoattractant-responsive cells, preferentially migrate to the stroke-affected hemisphere and undergo conversion within the hemisphere into Mbnl2 + neutrophils. This transformation may occur through a mechanism involving mitochondrial NETs formation. 3. The transformed Mbnl2 + neutrophils accumulate within the brain tissue and are gradually cleared by macrophages and microglia. Our work provides a clearer understanding of neutrophil dynamics in the brain, furnishing more precise information and references for the identification of therapeutic targets.

Presently, the US Food and Drug Administration (FDA) endorses only two therapeutic strategies for IS: pharmacological thrombolysis with tissue plasminogen activator (tPA) and mechanical thrombectomy ³¹. The time window for tPA is limited to 4.5 hours, and delays in acquiring medical imaging results often jeopardize timely treatment for many patients³². Thus, developing a rapid, simple and robust method for initial screening of suspected patients is crucial for their treatment and prognosis. Analysis of clinical samples has revealed that virtually all thrombi in stroke patients contain NETs, with plasma levels of NETs also being elevated²⁰. Consequently, we propose that NETs represent a highly promising peripheral blood biomarker for stroke. Through differential analysis and WGCNA, we identified key genes associated with NETs in the peripheral blood of stroke patients. Leveraging machine learning techniques, we established a straightforward and robust linear model capable of identifying patients in the hyperacute stage of stroke. This model achieved the AUC value exceeding 0.98 in both training and validation cohort, with minimal variation influenced by gender or age. These findings underscore the remarkable robustness and vast potential of our model.F12, that is Factor XII (FXII), is the zymogen of serine protease factor XIIa (FXIIa)³³. It has been reported that F12 ^–/– mice have reduced thrombosis, but normal hemostasis, and several substances supporting FXII autoactivation have been identified, which contain NETs³³. In addition, the mechanism by which it interacts with NETs to generate thrombi has been addressed earlier. Plexin domain containing 2 (PLXDC2), also known as TEM7R, is a member of the tumor endothelial marker (TEM) family³⁴. PLXDC2 has been identified as a regulator of several brain activities, including the coordinated control of proliferation and cell fate specification along and across the neuroaxis³⁵, which has been reported in a few pieces of literature that it can be coordinated with the remaining nine genes to diagnose IS ³⁶.

Building on our primary findings, the subsequent validation through targeted in vitro and in vivo experiments provides crucial biological insights that not only support our computational predictions but also deepen our understanding of their clinical implications. The observed consistency across different experimental settings underscores the reliability of our data and presents a compelling case for the translational potential of our findings. The in vitro validation using OGD to simulate ischemic conditions significantly reinforces the biological relevance and mechanistic understanding of NETs formation and SN gene expression changes in response to stroke-like conditions. Notably, the alterations in NET formation and subsequent gene expression changes observed in neutrophils provide compelling evidence supporting our bioinformatics predictions and underline the sensitivity of these cells to ischemic conditions. These observations are pivotal as they not only validate the robustness of our computational analyses but also provide a mechanistic link to potential therapeutic targets. Furthermore, our in vivo experiments employing the tMCAO model demonstrate significant changes in peripheral blood levels of key SN genes post-cerebral ischemia. This highlights the translational potential of these biomarkers for early detection and monitoring of stroke progression in a clinical setting. The consistency between our in vitro and in vivo findings not only strengthens the case for these SN genes as reliable biomarkers but also advances our understanding of the cellular dynamics at play during the hyperacute phase of ischemic stroke.

The key to the treatment and prognosis of IS lies in the reperfusion of the brain; however, current thrombolytic agents are constrained by a narrow therapeutic time window, inconsistent efficacy, and potential side effects. Evidence suggests that NETs within thrombi increase over time following stroke onset, rendering clots more rigid and resistant to lysis²⁶. This observation partly elucidates why the therapeutic window for thrombolysis is so brief and its efficacy unpredictable. Besides dissolving existing NETs, inhibiting further NETs formation is equally vital. We postulate that peripheral NETs induction might be linked to the Jak/Stat signaling pathway, implying that blocking this pathway could be pivotal. Furthermore, we speculate that peripheral blood macrophages may contribute to sustained NETs generation, presenting another potential target. To facilitate further investigation, we constructed a ceRNA network focusing on peripheral SN genes. Within the brain, we have characterized distinct neutrophil clusters and inferred their roles, suggesting that Mmp8 + neutrophils could be a target during the acute phase of stroke, potentially pivotal in alleviating BBB damage. Meanwhile, C1qc + neutrophils are postulated as a primary source of NETs; targeting their clearance and inhibition may reduce NETs production, thereby mitigating cerebral tissue destruction.

Our study has several limitations. Firstly, although we utilized single-cell technologies to map the temporal and functional landscape of neutrophils, our findings are constrained by the use of murine models, which may not fully replicate the intracranial environment of human patients. Secondly, the dataset we used lacks sufficient time points, and the sample size for constructing diagnostic models was limited. Lastly, while our experimental validation addressed key aspects, further functional and predictive testing are required to fully substantiate our hypotheses.

Despite these limitations, the pioneering findings of our study have opened numerous avenues for further research and practical applications. The elucidation of neutrophils' roles and their dynamic transitions within the context of stroke sets the stage for more targeted investigations into their mechanistic contributions, both in animal models and potentially in human clinical settings. These insights are pivotal for refining our models and deepening our understanding of stroke pathophysiology. Moving forward, we aim to expand our dataset and refine our methodologies to address the identified gaps, thereby enhancing the reliability and applicability of our findings to broader clinical scenarios.

In conclusion, we have successfully mapped a single-cell atlas elucidating the complex dynamics of neutrophils in mice post-stroke. Our analysis reveals that Mmp8 + neutrophils, prevalent during the hyperacute phase, potentially contribute to NETs production and disrupt the BBB. Additionally, C1qc + neutrophils are identified as key players in NETs generation, potentially undergoing transformation into Mbnl2 + neutrophils through processes associated with mitochondrial NETs formation. Furthermore, we have developed a straightforward yet robust diagnostic model that demonstrates exceptional accuracy, with an AUC value exceeding 0.98, in detecting the hyperacute phase of ischemic stroke. Complementing our computational findings, both in vitro and in vivo experiments have validated the predictive accuracy of our model. By integrating single-cell profiling with bulk RNA-sequencing and experimental validation, our comprehensive approach not only identifies novel therapeutic targets and early biomarkers but also advances the development of precision medicine strategies aimed at mitigating stroke damage and establishing targeted therapeutic interventions.

Funding

Declaration: This work was supported by grants from the National Natural Science Foundation of China (No.82471361), Natural Science Foundations for Excellent Young Scholars of Hunan Province (No.2021JJ20095), the Key Research and Development Program of Hunan Province (No. 2020SK2063), Research Project on Education and Teaching Innovation of Central South University (No. 2021jy145), the Natural Science Foundations of Hunan Province (No. 2020JJ4134).

Conflicts of interest:

none.

Author Contribution

M.Z.: conceptualization, methodology, writing—review & editing, supervision, funding acquisition, resources. J.Z.: data curation, methodology, formal analysis, writing—original draft, validation. Z.C.: data curation, formal analysis, visulization. Y.P., B.X: writing—review & editing, visualization.

Acknowledgments:

Not applicable.

Data Availability

Publicly available datasets were analyzed in this study. These data are available in the GEO, https://www.ncbi.nlm.nih.gov/geo/.

GBD 2016 Lifetime Risk of Stroke Collaborators, Feigin VL, Nguyen G, et al. Global, Regional, and Country-Specific Lifetime Risks of Stroke, 1990 and 2016. N Engl J Med. 2018;379(25):2429–37. 10.1056/NEJMoa1804492.
Jayaraj RL, Azimullah S, Beiram R, Jalal FY, Rosenberg GA. Neuroinflammation: friend and foe for ischemic stroke. J Neuroinflammation. 2019;16(1):142. 10.1186/s12974-019-1516-2.
Anrather J, Iadecola C. Inflammation and Stroke: An Overview. Neurotherapeutics. 2016;13(4):661–70. 10.1007/s13311-016-0483-x.
Levard D, Buendia I, Lanquetin A, Glavan M, Vivien D, Rubio M. Filling the gaps on stroke research: Focus on inflammation and immunity. Brain Behav Immun. 2021;91:649–67. 10.1016/j.bbi.2020.09.025.
Maida CD, Norrito RL, Daidone M, Tuttolomondo A, Pinto A. Neuroinflammatory Mechanisms in Ischemic Stroke: Focus on Cardioembolic Stroke, Background, and Therapeutic Approaches. Int J Mol Sci. 2020;21(18):6454. 10.3390/ijms21186454.
Iadecola C, Buckwalter MS, Anrather J. Immune responses to stroke: mechanisms, modulation, and therapeutic potential. J Clin Invest. 2022;130(6):2777–88. 10.1172/JCI135530.
Garcia-Bonilla L, Moore JM, Racchumi G, et al. Inducible nitric oxide synthase in neutrophils and endothelium contributes to ischemic brain injury in mice. J Immunol Baltim Md 1950. 2014;193(5):2531–7. 10.4049/jimmunol.1400918.
Martinod K, Wagner DD. Thrombosis: tangled up in NETs. Blood. 2014;123(18):2768–76. 10.1182/blood-2013-10-463646.
Essig F, Kollikowski AM, Pham M, et al. Immunohistological Analysis of Neutrophils and Neutrophil Extracellular Traps in Human Thrombemboli Causing Acute Ischemic Stroke. Int J Mol Sci. 2020;21(19):7387. 10.3390/ijms21197387.
Guo Y, Zeng H, Gao C. The Role of Neutrophil Extracellular Traps in Central Nervous System Diseases and Prospects for Clinical Application. Oxid Med Cell Longev. 2021;2021:9931742. 10.1155/2021/9931742.
Goyal M, Menon BK, van Zwam WH, et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet Lond Engl. 2016;387(10029):1723–31. 10.1016/S0140-6736(16)00163-X.
Bal S, Bhatia R, Menon BK, et al. Time dependence of reliability of noncontrast computed tomography in comparison to computed tomography angiography source image in acute ischemic stroke. Int J Stroke Off J Int Stroke Soc. 2015;10(1):55–60. 10.1111/j.1747-4949.2012.00859.x.
Street K, Risso D, Fletcher RB, et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics. 2018;19(1):477. 10.1186/s12864-018-4772-0.
Jin S, Guerrero-Juarez CF, Zhang L, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12(1):1088. 10.1038/s41467-021-21246-9.
Gentles AJ, Newman AM, Liu CL, et al. The prognostic landscape of genes and infiltrating immune cells across human cancers. Nat Med. 2015;21(8):938–45. 10.1038/nm.3909.
Zhang Y, Guo L, Dai Q, et al. A signature for pan-cancer prognosis based on neutrophil extracellular traps. J Immunother Cancer. 2022;10(6):e004210. 10.1136/jitc-2021-004210.
Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559. 10.1186/1471-2105-9-559.
Kilkenny C, Browne WJ, Cuthill IC, Emerson M, Altman DG. Improving Bioscience Research Reporting: The ARRIVE Guidelines for Reporting Animal Research. PLoS Biol. 2010;8(6):e1000412. 10.1371/journal.pbio.1000412.
Stark K, Massberg S. Interplay between inflammation and thrombosis in cardiovascular pathology. Nat Rev Cardiol. 2021;18(9):666–82. 10.1038/s41569-021-00552-1.
Laridan E, Denorme F, Desender L, et al. Neutrophil extracellular traps in ischemic stroke thrombi. Ann Neurol. 2017;82(2):223–32. 10.1002/ana.24993.
Luo H, Guo H, Zhou Y, Fang R, Zhang W, Mei Z. Neutrophil Extracellular Traps in Cerebral Ischemia/Reperfusion Injury: Friend and Foe. Curr Neuropharmacol. 2023;21(10):2079–96. 10.2174/1570159X21666230308090351.
Zhao Z, Pan Z, Zhang S, et al. Neutrophil extracellular traps: A novel target for the treatment of stroke. Pharmacol Ther. 2023;241:108328. 10.1016/j.pharmthera.2022.108328.
Li C, Xing Y, Zhang Y, Hua Y, Hu J, Bai Y. Neutrophil Extracellular Traps Exacerbate Ischemic Brain Damage. Mol Neurobiol. 2022;59(1):643–56. 10.1007/s12035-021-02635-z.
Li J, Xia Y, Sun B, et al. Neutrophil extracellular traps induced by the hypoxic microenvironment in gastric cancer augment tumour growth. Cell Commun Signal CCS. 2023;21(1):86. 10.1186/s12964-023-01112-5.
Datsi A, Piotrowski L, Markou M, et al. Stroke-derived neutrophils demonstrate higher formation potential and impaired resolution of CD66b + driven neutrophil extracellular traps. BMC Neurol. 2022;22:186. 10.1186/s12883-022-02707-0.
Liaptsi E, Merkouris E, Polatidou E, et al. Targeting Neutrophil Extracellular Traps for Stroke Prognosis: A Promising Path. Neurol Int. 2023;15(4):1212. 10.3390/neurolint15040076.
Zhang S, Cao Y, Du J, et al. Neutrophil extracellular traps contribute to tissue plasminogen activator resistance in acute ischemic stroke. FASEB J Off Publ Fed Am Soc Exp Biol. 2021;35(9):e21835. 10.1096/fj.202100471RR.
Cai W, Liu S, Hu M, et al. Functional Dynamics of Neutrophils After Ischemic Stroke. Transl Stroke Res. 2020;11(1):108–21. 10.1007/s12975-019-00694-y.
Otxoa-de-Amezaga A, Gallizioli M, Pedragosa J, et al. Location of Neutrophils in Different Compartments of the Damaged Mouse Brain After Severe Ischemia/Reperfusion. Stroke. 2019;50(6):1548–57. 10.1161/STROKEAHA.118.023837.
Yang C, Hawkins KE, Doré S, Candelario-Jalil E. Neuroinflammatory mechanisms of blood-brain barrier damage in ischemic stroke. Am J Physiol - Cell Physiol. 2019;316(2):C135–53. 10.1152/ajpcell.00136.2018.
Powers WJ, Rabinstein AA, Ackerson T, et al. Guidelines for the Early Management of Patients With Acute Ischemic Stroke: 2019 Update to the 2018 Guidelines for the Early Management of Acute Ischemic Stroke: A Guideline for Healthcare Professionals From the American Heart Association/American Stroke Association. Stroke. 2019;50(12):e344–418. 10.1161/STR.0000000000000211.
Hankey GJ, Stroke. Lancet Lond Engl. 2017;389(10069):641–54. 10.1016/S0140-6736(16)30962-X.
Stavrou EX, Fang C, Bane KL, et al. Factor XII and uPAR upregulate neutrophil functions to influence wound healing. J Clin Invest. 2018;128(3):944–59. 10.1172/JCI92880.
Guan Y, Du Y, Wang G, et al. Overexpression of PLXDC2 in Stromal Cell-Associated M2 Macrophages Is Related to EMT and the Progression of Gastric Cancer. Front Cell Dev Biol. 2021;9:673295. 10.3389/fcell.2021.673295.
Miller-Delaney SFC, Lieberam I, Murphy P, Mitchell KJ. Plxdc2 Is a Mitogen for Neural Progenitors. PLoS ONE. 2011;6(1):e14565. 10.1371/journal.pone.0014565.
O’Connell GC, Petrone AB, Treadway MB, et al. Machine-learning approach identifies a pattern of gene expression in peripheral blood that can accurately detect ischaemic stroke. NPJ Genomic Med. 2016;1:16038. 10.1038/npjgenmed.2016.38.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Mapping Neutrophil Fate and Function in Ischemic Stroke: A Single-cell Roadmap for Translational Insights

Status:

Version 1

Abstract

Background

Materials and Methods

Results

Conclusion

Figures

1. Introduction

2. Materials and Methods

2.1 Preparation of Data

2.2 Quality control of scRNA‑seq data

2.3 Data integration, dimensional reduction, visualization, and cell type annotation

2.4 Cell scoring and functional annotation

2.5 Trajectory analyses and Cell communication inference

2.6 Assessment of Degree of Immune Cell Infiltration

2.7 Identification of NETs Score

2.8 Assessment of Hallmark Gene Set and Gene Set Enrichment Analysis (GSEA)

2.9 Identification of Stroke Related Differential Expression Genes (DEGs)

2.10 Construction of Gene Co-Expression Network

2.11 Functional Enrichment Analysis

2.12 Protein-Protein Interaction (PPI) Network Building

2.13 Identification of LncRNAs Co-Expressed with mRNAs

2.14 Construction of CeRNA

2.15 Isolation of Neutrophils from Mice Peripheral Blood

2.16 Flow Cytometric Identification of Neutrophils

2.17 Primary culture of Neutrophils and oxygen-glucose deprivation (OGD) treatment

2.18 Enzyme-linked immunosorbent assay (ELISA)

2.19 Establishment of Transient Focal Cerebral Ischemia Model

2.20 Peripheral Blood Collection

2.21 Quantitative Real-Time PCR

2.22 Statistical Analysis

3. Result

3.1 Preliminary Analysis of MCAO Mice Brain Single Cell Profiles

3.2 Analysis of NETs in Neutrophils of MCAO Mice Brain

3.3 Dominant Cells for Cell Communication Change with Time after Stroke

3.4Immune Microenvironment and NETs Expression in Whole Blood of IS Patients

3.5 Exploration of the Role of NETs in IS Patients

3.6 Identification of Stroke-NETs Related Genes (SN genes)

3.7 Establishment of Hyperacute Stroke Diagnosis Model Based on SN Genes

3.8 Neutrophil Identification and Response to Oxygen-Glucose Deprivation

3.9 Validation of Predictive SN Genes and Their Dynamic Expression Post-Ischemia

3.10 Identification of Key SN Genes and Construction of CeRNA Network

4. Discussion

5. Conclusions

Declarations

Funding

Author Contribution

Acknowledgments:

Data Availability

References

Additional Declarations

Supplementary Files

Status:

Version 1