Analysis of ferroptosis-related genes in cerebral ischemic stroke via immune infiltration and single-cell RNA-sequencing

Abstract

Ischemic stroke (IS) represents a harmful neurological disorder with limited treatment options. Ferroptosis accounts for the iron-dependent, nonapoptotic cell death pattern, which shows the feature of fatal lipid ROS accumulation. Nonetheless, ferroptosis-related biomarkers for identifying IS early are currently lacking. The present study focused on investigating the possible ferroptosis-related biomarkers for IS and analyzing their effects on immune infiltration. Altogether five hub differentially expressed ferroptosis-related genes (DEFRGs) were identified from the relevant databases. Additionally, single-cell RNA-sequencing (seq) analysis was conducted for the comprehensive mapping of cell populations based on the IS database. These five hub DEFRGs were analyzed using gene set enrichment analysis, miRNA prediction, and single-cell RNA-seq analysis. A transient middle cerebral artery occlusion mouse model was constructed. We also adopted bioinformatics methods combined with western blot, changes to mitochondria, hematoxylin & eosin staining, Nissl staining, ROS fluorescence staining, immunohistochemistry, and quantitative real-time polymerase chain reaction (qRT-PCR) to show the involvement of ferroptosis in IS progression. The results revealed that nuclear factor erythroid-derived 2-like 2 (Nfe2l2) was the potential candidate biomarker for IS diagnosis, and ferroptosis may be suppressed via the Nfe2l2/HO-1 pathway. Thus, drug targeting Nfe2l2 can shed novel lights on IS treatment.

Introduction

Globally, stroke is a major factor leading to mortality and is frequently responsible for acquired physical disability [1], with acute ischemic stroke (AIS) accounting for a significant proportion of the cases [2].Thrombolysis, using recombinant tissue plasminogen activator and endovascular thrombectomy, is recognized as the reperfusion strategy for AIS; however, it is only beneficial for a few patients because of the time window [3]. The IS pathology involves excitatory amino acid toxicity, inflammatory reactions, free radical injury, neuron death, and mitochondrial energy metabolic disorder [4,5,6]. As a new regulatory cell death pattern, ferroptosis involves iron-dependent lipid reactive oxygen species (ROS) aggregation, distinct from other cell death types, including autophagy, pyroptosis, necrosis, or apoptosis [7, 8].Ferroptosis is known to participate in different nervous system disorders, like Alzheimer’s disease, Parkinson’s disease, intracerebral hemorrhage, and traumatic brain injury [9,10,11,12].Although reports exist on the involvement of ferroptosis in IS [13], they are inadequate to explain the regulatory mechanism of IS.

In this study, the IS transcription microarray dataset was acquired based on Gene Expression Omnibus (GEO) database. Later, we chosen differentially expressed genes (DEGs) and intersected them with ferroptosis-related genes for obtaining IS-related differentially expressed ferroptosis-related genes (DEFRGs). We used DEFRGs to establish the protein–protein interaction (PPI) network. Finally, the top five hub DEFRGs, including Rela, Jun, Myc, Stat3, and nuclear factor erythroid-derived 2-like 2 (Nfe2l2), were acquired, and their expression levels were analyzed. Nfe2l2 is an extensively recognized transcription factor that is important for the endogenous antioxidant stress system, and it regulates the antioxidant response element (ARE) genes [14].Nfe2l2 is associated with critical genes and lipid metabolism during ferroptosis [15].Meanwhile, Nfe2l2 is related to mitogen-activated protein kinase (MAPK), tumor necrosis factor (TNF), nuclear factor kappa B (NF-κB), and heme oxygenase (HO)-1 pathways [16,17,18], and it can mitigate cerebral ischemic injury. Activation of the Nfe2l2/HO-1 pathway is beneficial for inhibiting ROS and proinflammatory cytokine levels and remarkably reduces oxidative injury to the ischemic brain tissue [19, 20]. Single-cell RNA-sequencing (seq) analysis is an excellent high-throughput sequencing technique [21], which helps to analyze molecular and cellular alterations in cerebral ischemia/reperfusion (I/R) injury.

This study downloaded GSE154396 database, a single-cell RNA-seq analysis on cerebral hematopoietic cells in brain ischemic model mice, for the analysis. Then, the five hub DEFRGs were used for analyzing the relationship with immune cells along with their functions, gene set enrichment analysis (GSEA), miRNA prediction, and single-cell RNA-seq analysis. The transient focal cerebral ischemic mouse model was also validated. The expression levels of Nfe2l2, together with their underlying mechanisms in ferroptosis, were explored using bioinformatics analyses, western blot, changes to mitochondria, hematoxylin & eosin (H&E) staining, Nissl staining, ROS fluorescence analysis, quantitative real-time polymerase chain reaction (qRT-PCR), and immunohistochemical analysis. In addition, this present study focused on conducting high-throughput sequencing on IS samples and applying bioinformatics analyses to explore the candidate targets and identifying the pathogenic mechanisms underlying IS. Figure 1 outlines our study flowchart. We used the Nfe2l2 activator sulforaphane (SFN) for exploring Nfe2l2’s role during cerebral I/R injury.

Fig. 1

Study flowchart and data processing procedure. DEGs, differentially expressed genes; DEFRGs, differentially expressed ferroptosis-related genes

Materials and methods

Datasets and data preprocessing

We extracted the transcriptome dataset based on GEO (http://www.ncbi.nlm.nih.gov/geo). GSE30655 dataset, including GPL1261 annotation platform, contained seven IS and three control groups. Additionally, GSE154396 dataset was also obtained to conduct single-cell RNA-seq analysis. DEGs in IS versus control groups were identified using Limma package upon p < 0.05 &|logFC| > 0.585. Totally 619 ferroptosis-related genes were acquired using the analysis of the ferroptosis gene set obtained in GeneCards database (https://www.genecards.org/).

Functional annotation of DEFRGs

DEFRGs functions were annotated with “ClusterProfiler” package in R for the comprehensive exploration of the functional relationships. In addition, related functional categories were assessed in accordance with Gene Ontology (GO) as well as the Kyoto Encyclopedia of Genes and Genomes (KEGG) with significance levels of p- and q-values < 0.05.

PPI network analysis

Protein interactions corresponding to DEFRGs were constructed through STRING (http://string-db.org) online database with a confidence score > 0.4. Subsequently, we built a PPI network by importing the results of STRING into Cytoscape v3.9.0 software and visualizing it. The nodes in this network represented genes, whereas the edges indicated between-gene links. We also utilized molecular complex detection (MCODE) for investigating modules with highest significance from the PPI network [22]. Furthermore, according to the similarity of the GO terms used in functional annotation, the proteins were ranked by functional similarity to the corresponding interactive proteins. Herein, between-protein functional similarity was determined as the geometric average of GO terms similarity categorized as molecular function (MF), biological process (BP), and cellular component (CC). Besides, the strength of the relationships of proteins with corresponding interactive proteins was determined by considering the functions and pathways.

Immune infiltration analyses

The CIBERSORT algorithm has been extensively used to evaluate the immune cell types within the microenvironment. In line with the support vector regression principle, we conducted a deconvolution analysis of the immune cell expression matrix [23]. It covers 547 biomarkers that distinguish 25 mouse immune cell phenotypes, such as T cells, B cells, myeloid cells, and plasma cells. We utilized the CIBERSORT algorithm for analyzing sample data to infer the relative abundances of 25 immune infiltrating cell types. The immune cell abundances and gene levels were analyzed through Pearson’s correlation coefficient.

GSEA

Preset gene sets were utilized in GSEA analysis for ranking genes based on the corresponding differential expression within the two samples and later inspecting if those predefined gene sets were enriched on the top or bottom of the list. In this work, we analyzed differences in pathways in low- versus high-expression groups through GSEA and explored the molecular mechanism related to the hub genes in both the groups, separately under the replacement number and type of 1000 and phenotype.

Single-cell RNA-seq

First, “Seurat” package was employed to process the data [24], whereas the position relationship among clusters was evaluated with t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm. The cluster was annotated via the celldex package to cells related to IS development. Finally, marker genes were obtained from single-cell expression data for every cell subtype using the logfc.threshold parameter FindAllMarkers of 1. Specific marker genes were selected from genes satisfying p < 0.05 and|avg log2FC| >1 for every subtype.

Animal experiments

We obtained SPF-grade C57BL/6 male mice (20–25 g) in Liaoning Changsheng Biotechnology and strictly treated them following Laboratory Animal Care guidelines of Fourth Affiliated Hospital of Harbin Medical University, as recommended by the ARRIVE guidelines (IACUC2022136). All animals were allowed to take water and food freely and raised at 23–25 °C, 50–60% humidity, and 12-/12-h light/dark cycle conditions.

Establishment of transient middle cerebral artery occlusion model (tMCAO)

Mouse anesthesia was accomplished via isoflurane (2–3% in oxygen) administration; thereafter, mice were put in their supine position, and their rectal temperature was maintained under 37 °C with the homoeothermic blanket intraoperatively. When an incision was made at the cervical midline of each mouse, the common carotid artery, external carotid artery (ECA), and internal carotid artery (ICA) were separated successively. Thereafter, we inserted one Doccol suture (diameter, 0.21 mm) via this incision on the ECA to about 9 ± 1 mm at the bifurcation of ICA and ECA for occluding the middle cerebral artery (MCA) origin. After feeling slightest resistance, the thread was terminated. This suture was removed 1 h later, followed by a 24 h recovery period. The mice in the sham group received an identical procedure with the exception of inserting an intraluminal filament. One technician performed every surgery. After the dissolution of SFN (HY-13755, MCE) in dimethylsulfoxide (DMSO, < 2%), the mixture was given into mice of the SFN + tMCAO and V + tMCAO groups through intraperitoneal injection with 5 mg/kg SFN and vehicle (DMSO < 2%), respectively, in the process of reperfusion [25].Subsequently, mice were randomized as following groups: (1) Sham; (2) tMCAO; (3) SFN + tMCAO; and (4) V + tMCAO group. Ischemic injury severity was evaluated by measuring the extent of neurological deficit. 24 h postischemia, blinded researchers rated neurological deficits as 0–5, which represented no neurological deficits, forelimb weakness and torso turning to ipsilateral side through tail holding, incapability of extending contralateral forelimb, circling onto contralateral side, falling to left side, and absence of spontaneous activity or low consciousness level or death, respectively.

Western blotting analysis

The right brain tissues were subjected to homogenization using RIPA lysis buffer (Beyotime, China) containing protease/phosphatase inhibitors, followed by 15 min of centrifugation at 12,000× g. To analyze protein levels, we adopted the BCA protein assay kit (Beyotime, China). Thereafter, separation of 30 µg protein aliquots was completed using 10% or 12.5% SDS-PAGE gels, prior to electrotransfer on polyvinylidene difluoride membranes. The membranes then received 1 h of blocking with 5% defatted milk, prior to overnight primary antibody incubation under 4 °C, such as anti-ferritin heavy chain (FTH1, 1:1000, ab183781, Abcam); anti-HO-1 (1:2000, 10701–1-AP, Proteintech), anti-TFR1 (1:1000, 16375–1-AP, Proteintech), anti-COX2 (1:1000, 12375-1-AP, Proteintech), anti-GAPDH (1:50000, 60004–1-Ig, Proteintech), anti-TNF-α (1:1000, 60291–1-Ig, Proteintech), anti-IL-17 (1:1000, 10663–1-AP, Proteintech). Membranes were later rinsed using tris-buffered saline that contained Tween 20, and later incubated for 1 h using horseradish peroxidase-labeled secondary antibodies. Additionally, protein bands were visualized with the enhanced chemiluminescence detection kit. Signals were quantified relative to GAPDH with ImageJ.

H&E staining

Brain tissue underwent 24 h of fixation using 4% paraformaldehyde, followed by dehydration with an ethanol gradient, clearing with xylene, and paraffin embedding. Finally, 4-µm sections were prepared for H&E staining and resin sealing. Cortical histology was analyzed with an optical microscope.

Nissl staining

Following deparaffinage and rehydration, sections were stained with toluidine blue solution for a 40-min period under 50–60 °C, and later rinsed by distilled water and dehydrated with an ethanol gradient. Thereafter, each slide was immersed twice in dimethylbenzene (10 min each) prior to covering with neutral balsam coverslips. Finally, using an optical microscope, the tissue sections were examined and photographed.

Transmission electron microscopy (TEM)

TEM (Japan Electron microscopy Hitachi, H-7650) was utilized for examining the ultrastructure within peri-infarct region 24 h postischemia. Briefly, after 4% paraformaldehyde perfusion of the mice tissues, the tissue was prepared in 1 mm³ cubes and subjected to treatment using 2.5% glutaraldehyde as well as 1% osmic acid. Following hydration, the embedded tissue was sliced into sections for lead citrate and uranyl acetate staining. Random fields of view were obtained for every sample to calculate the ferroptotic mitochondrial number.

ROS production

The ROS contents were identified using the fluorescent probe-2,7-dichlorofluorescein diacetate (DCFH-DA). After dilution with DMSO and loading buffer solution, the probe at 10 µmol/L was obtained. Thereafter, the tissues were subject to 3 h of incubation using DCFH-DA under 37 °C, followed by analyzing the fluorescence intensity using a fluorescence microscope.

Immunohistochemical analysis

After paraffin embedding, tissues were dewaxed, rehydrated, and incubated with 0.3% hydrogen peroxide. After 4 min of heating in citric acid solution in a pressure cooker to retrieve antigen, the tissue sections were immersed in 5% bovine serum albumin for 30 min, followed by primary antibody incubation (TNF-α [1:50, 60291–1-Ig, Proteintech], IL-17 [1:100, 13082–1-AP, Proteintech]), and later corresponding secondary antibody (PV-6002; Zhongshan Biotechnology) incubation under 37 °C for 30 min. Diaminobenzidine (DAB, ZLI-9018; Zhongshan Biotechnology) was utilized for section development, followed by observation under the microscope.

Immunofluorescence analysis

After fixation in 4% paraformaldehyde, brain sections were subjected to 20 min of permeabilization using 0.2% Triton X-100 and later overnight primary antibody Iba1 (Abcam, 1:500) incubation under 4 °C. Next, the sections were subject to incubation with the corresponding secondary antibody (BA1105, 1:100, BOSTER) for 3 h in the dark at ambient temperature. 4′,6-diamidino-2-phenylindole (DAPI) (Abcam, ab104139) was later added for nuclear staining, whereas the fluorescence microscope (Nikon, Y-TV55, JAPAN) was adopted for acquiring images, and ImageJ was adopted for analyzing the immunofluorescence area (peri-ischemic area on ischemic hemisphere). The experiment was double-blinded, and diverse group members read and analyzed the images.

qRT-PCR

By applying Trizol reagent (Invitrogen, Thermo Fisher Scientific), total RNAs were isolated from brain tissue for preparation of cDNA with reverse transcriptase ((TIANGEN, China) according to corresponding instructions. SYBR green kit (TIANGEN, China) was then adopted for qRT-PCR based on the PCR system (BioRad, Singapore). Primer sequences include: HO-1, AAGCCGAGAATGCTGAGTTCA (forward), GCCGTGTAGATATGGTACAAGGA (reverse), and GAPDH, AGGTCGGTGTGAACGGATTTG (forward), TGTAGACCATGTAGTTGAGGTCA (reverse). At last, we used 2^–∆∆Ct method for analyzing gene expression.

Statistical analysis

R language (version 3.6) and GraphPad Prism software 8.0 were employed in statistical analyses. Results were represented by means ± SEM. To explore between-group differences, we used one-way ANOVA and Tukey’s post-hoc test. p < 0.05 was considered to be of statistical significance.

Results

DEGs identification

We recruited a total of ten samples (seven IS and three control samples) in the present study. DEGs were screened at p < 0.05 &|log2FC| >0.585, and 1493 DEGs (804 upregulated and 689 downregulated genes) were directly identified. (Figure 2A and B). Then, 66 DEFRGs were obtained following interaction with ferroptosis-related genes for further analysis (Fig. 2C).

Fig. 2

Differential expression analysis. (A, B) Volcano and heatmaps for DEGs in GSE30655. (C) Venn diagram showing DEGs intersecting with ferroptosis-related genes. Bar plot showing GO (D) as well as KEGG analysis (E) on DEFRGs. (F) PPI network for DEFRGs and (G) distinct gene module obtained from the PPI network. (H) Functional similarity analysis. Horizontal and vertical axes represent similarity score and genes, respectively. GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; PPI, protein–protein interaction

Functional enrichment analysis of DEFRGs and five hub DEFRGs

We further analyzed the pathway of the 66 DEFRGs. Based on GO functional annotation, genes were primarily associated with response to oxidative stress, cellular response to drug, response to reactive oxygen species, or other pathways (Fig. 2D). KEGG enrichment results showed that genes were mostly associated with ferroptosis, TNF, IL − 17, and MAPK signaling pathways (Fig. 2E). In addition, we constructed a protein interaction network to understand the interactions between DEFRGs (Fig. 2F). To further identify hub DEFRGs, we used the MCODE algorithm to screen the 66 DEFRGs and observed that the cluster with the greatest score had 13 nodes and 70 connections (Fig. 2G). Subsequently, among the 13 genes, we screened the more critical genes using the GO semantic similarity method. The ranking of the 13 genes is shown in Fig. 2H, and the five hub DEFRGs identified, namely Rela, Jun, Myc, Stat3, and Nfe2l2 were selected for subsequent analysis.

Immune infiltration analyses

The immune microenvironment mostly comprises immune cells, immune-related fibroblasts, inflammatory factors, growth factors, extracellular matrix, as well as specific physicochemical characteristics. It has a significant impact on the diagnosis, treatment, and prediction of IS prognosis. By exploring the relationship of the five hub DEFRGs with immune infiltration of the IS dataset, we investigated the underlying mechanism of the hub DEFRGs on cerebral ischemia progression. Figure 3A displayed immune cell proportions in every sample. Several prominent correlations were observed among immune infiltration levels (Fig. 3B). Moreover, relative to the normal group, the IS group had markedly increased proportions of active and immature dendritic cells (DC), M2 macrophages, and monocytes (Fig. 3C). The analysis of the relationship of the five hub DEFRGs (Rela, Jun, Myc, Stat3, and Nfe2l2) with immune cells showed that DEFRGs were closely related to immune cells (Fig. 3D-H). The correlation between the five hub DEFRGs with diverse immune factors, such as chemokines, immune stimulators, immune inhibitors, major histocompatibility complex (MHC), and receptors, were obtained from TISIDB database. The hub DEFRGs exhibited a close relationship with the immune factors (Fig. 3I-M) and thus are tightly associated with immune cell infiltration levels, which are crucial for the immune microenvironment. Noncoding RNA networks of the five hub DEFRGs were predicted via the Targetscan database. There were 70 miRNAs, and 73 mRNA–miRNA relation pairs were predicted. A noncoding RNA network of hub DEFRGs was successfully constructed (Fig. 4A). We obtained the regulatory genes of cerebral I/R through the GeneCards database, and the differential expression of the 20 most significant genes obtained from this dataset were examined. Results showed that numerous genes, such as Col4a1, Cst3, Eng, F2, Icam1, and Il6, demonstrated variation in expression levels (Fig. 4B). Subsequently, we conducted a correlation analysis on the five hub DEFRGs and IS regulatory genes. Nfe2l2 was significantly negatively related to Cst3 (r = − 0.903), and Stat3 showed a significant relationship to Icam1 (r = 0.985) (Fig. 4C).

Fig. 3

Infiltration of immune cells of the IS and control groups. (A) Bar plot showing percentages of 25 immune cells of the IS and control groups. (B) Correlation between 25 immune cells. (C) Comparison of 25 immune cell subtypes between the IS and control groups. (D–H) Correlations of immune cells with Jun, Myc, Nfe2l2, Rela, and Stat3. Dots of a larger size represent greater p-values. (I–M) Correlation analysis between the five hub DEFRGs (Jun, Myc, Nfe2l2, Rela, and Stat3) and chemokines, immune inhibitors, immune stimulators, MHC, and receptors. *p < 0.05; **p < 0.01; ***p < 0.001; N.S., nonsignificant

Fig. 4

Correlation of IS regulatory genes with hub DEFRGs. (A) miRNA‒mRNA network diagram of hub DEFRGs. (B) Differential expression of IS regulatory genes. (C) Correlation of IS regulatory genes with hub DEFRGs showing that Nfe2l2 was significantly negatively related to Cst3 and Stat3 was significantly positively related to Icam1. *p < 0.05; **p < 0.01; ***p < 0.001; N.S., nonsignificant

GSEA

Next, the pathways related to the five hub DEFRGs were investigated to explore their underlying molecular mechanisms that are critical to IS progression. Rela was correlated with the positive regulation of phosphatidylinositol 3-kinase activity, tricarboxylic acid cycle, peptidyl-tyrosine phosphorylation, spindle microtubule formation, nonmembrane spanning protein tyrosine kinase activity, and potassium channel activity. Jun was enriched in DNA damage checkpoint, positive regulation of protein binding, erythrocyte differentiation, potassium channel activity, hydrolase activity acting on acid anhydrides in phosphorus-containing, and Smad binding. Myc was enriched in dorsal-ventral axis specification, lipid glycosylation, regulation of translational initiation, ribosome biogenesis, rRNA processing, and protein complex scaffolds. Stat3 was enriched in the negative regulation of MAPK signaling cascade, protein k48-linked ubiquitination, ubiquitin-dependent protein catabolic process, stereocilium, peroxidase activity, and ubiquitin-specific protease activity. Nfe2l2 was enriched in ferrous iron transport, negative regulation of the MAPK signaling cascade, protein deubiquitination, histone methyltransferase complex, intermediate filament cytoskeleton, and bHLH transcription factor binding (Fig. 5A and E).

Fig. 5

Gene set enrichment analysis (GSEA) results. GSEA investigation of Rela, Jun, Myc, Stat3, and Nfe2l2 (A–E)

Single-cell RNA-seq

The cells were clustered via the t-SNE algorithm, and R package SingleR was used for annotations of various subtypes in the seven cell categories: microglia, granulocytes, T cells, B cells, monocytes, macrophages, or NK cells (Fig. 6A). The expression levels of the five 5 hub DEFRGs in the seven types of cells were shown in Fig. 6B and C.

Fig. 6

Transcriptomic atlas of IS summarized using single-cell RNA-sequencing analysis on mouse brain. (A) t-SNE plot showing single cell clustering; different colors indicate different cell types. (B) t-SNE plots showing expression levels of the five hub DEFRGs, including Nfe2l2, Stat3, Myc, Jun, and Rela. (C) Correlation analysis of the five hub DEFRGs (Nfe2l2, Stat3, Myc, Jun, and Rela) with microglia, NK cells, B cells, T cells, macrophages, monocytes, and granulocytes

Activation of Nfe2l2 inhibited ferroptosis injury post-tMCAO in mice

Several key factors, such as TFR1, COX2 and FTH1, have been recognized as crucial proteins for regulating ferroptosis [26, 27]. Thus, these protein levels were determined following I/R to verify ferroptosis as well as the impact on I/R. Our previous studies showed that Nfe2l2 had the lowest expression at 24 h post-tMCAO; therefore, the brain tissue was collected at this time. It was indicated that, in relative to the sham group, FTH1 expression level was obviously downregulated (p < 0.0001), COX2 and TFR1 were simultaneously significantly upregulated (p < 0.01) 24 h postischemia in the tMCAO group (Fig. 7A and B). Based on the above results, I/R induces ferroptosis, suggesting the possible involvement of ferroptosis in the pathology of cerebral I/R. When Nfe2l2 was activated by SFN, the level of FTH1 was upregulated (p < 0.0001), whereas that of COX2 and TFR1 was downregulated (p < 0.01 and p < 0.05, separately) (Fig. 7A and B). Nfe2l2 activation promoted typical alterations of the mitochondrial morphology in ferroptosis compared with 24 h post-tMCAO group (Fig. 7C). Then, mouse brain sections were subjected to immunofluorescence analysis (Fig. 7D). At 24-h post-tMCAO, fluorescence intensity of ROS was enhanced within the peri-ischemic regions (p < 0.05). Following SFN treatment, the ROS fluorescence intensity decreased (p < 0.01) (Fig. 7E). Consequently, Nfe2l2 activation post-tMCAO decreased the ROS levels, suggesting that Nfe2l2 participates in regulating post-I/R oxidative stress.

Fig. 7

Nfe2l2 activation inhibited ferroptosis injury post tMCAO in mice. (A, B) TFR1, COX2 and FTH1 protein levels were determined through western blotting (n = 4). (C) Ferroptosis-related mitochondrial morphology was analyzed through transmission electron microscopy. White arrows indicate neuronal outer mitochondrial membrane rupture and mitochondrial cristae disappearance or reduction. Scale bar = 2 μm/1 µm. (D, E) DAPI (blue)/ROS (green) immunofluorescence images and statistical analyses on mouse brain tissue sections post-tMCAO. n = 3. Scale bar = 100 μm. The findings indicate means ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 and N.S. nonsignificant versus sham group

Activation of Nfe2l2 prevented I/R-mediated cerebral dysfunction of tMCAO/R mice

Next, we investigated whether Nfe2l2 activation conferred neuroprotective efficacy to the MCAO/R mice. Compared with 24 h tMCAO group, SFN + tMCAO group showed improved neurological deficits of the mouse brain (p < 0.05) (Fig. 8A). H&E staining revealed peri-infarct tissue injury post-tMCAO, with swollen penumbra area, neuropil vacuolation, and glial cell hyperplasia. Typically, peripheral neurons showed the features of nuclear shrinkage with darker staining. Nonetheless, based on the H&E image analysis, the number of degenerated neurons that showed deep nuclear staining decreased within the penumbra of SFN + 24 h-tMCAO group (Fig. 8B). Simultaneously, Nissl body disappearance was observed within the neurons of 24 h-tMCAO relative to SFN + 24 h-tMCAO groups (Fig. 8C). Thus, activation of Nfe2l2 prevented against brain I/R injury. Nfe2l2 was also highly expressed in microglia, as shown in Fig. 6B. Staining of brain tissue for Iba1 showed that Iba1-positive cell number within the peri-infarct area was dramatically elevated in 24 h-tMCAO and V + 24 h-tMCAO groups, whereas it was reduced in SFN + 24 h-tMCAO group (Fig. 8D).

Fig. 8

Nfe2l2 activation prevents I/R-mediated cerebral dysfunction in MCAO/R mice. (A) Mouse neurological deficit scores recorded 24 h post-tMCAO (n = 4). (B) Hematoxylin and eosin staining analysis of mouse brain tissues post-tMCAO following treatment with the Nfe2l2-specific activator. (*) indicates vacuolated neuropil within brain tissue. White arrow indicates glial cells and black arrow suggests neuron nuclei showing shrinkage and dark staining (→), (C) Nissl staining of mouse brain tissue post-tMCAO following treatment with the Nfe2l2-specific activator. Black and white arrows indicate neurons with and without Nissl bodies, respectively. (D) DAPI (blue)/Iba1 (green) immunofluorescence images showing mouse brain tissue sections post-tMCAO. n = 4, scale bar = 100 μm. The data are represented by mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 and N.S., nonsignificant versus sham group

SFN exerted anti-inflammatory effect via Nfe2l2 activation

Ferroptosis occurs together with proinflammatory factor production [28]. In Fig. 2E, we observed that Nfe2l2 was enriched in the IL-17 and TNF signaling pathways. To analyze the role of SFN in inflammatory responses in I/R, we measured TNF and IL-17 levels using western blotting. TNF-α and IL-17 levels markedly elevated within brain tissue of 24 h-tMCAO group in relative to sham group (p < 0.001 and p < 0.0001, separately), and these impacts were blocked by SFN (p < 0.01 and p < 0.0001, separately) (Fig. 9A and B). Immunohistochemical staining with anti-TNF-α and anti-IL-17 antibodies showed that SFN treatment decreased the positive staining for TNF-α and IL-17 relative to that of 24 h-tMCAO and V + 24 h-tMCAO groups. Thus, immunohistochemical staining analysis demonstrated a similar tendency to western blotting (Fig. 9C and D).

Fig. 9

SFN exerts an anti-inflammatory effect via Nfe2l2 activation. (A, B) Western blot analysis was used for determining the TNF-α and IL-17 protein expression levels (n = 4). (C, D) Immunohistochemical staining for TNF-α (C) and IL-17 (D) of sham, 24 h-tMCAO, SFN + 24 h tMCAO, and V + 24 h tMCAO groups. Data stand for means ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 and N.S. nonsignificant versus sham group. Scale bar = 100 μm

SFN may inhibit ferroptosis injury post-tMCAO in mice through the Nfe2l2/HO-1 pathway

Nfe2l2/HO-1 pathway activation inhibits ferroptosis to protect the brain against I/R injury [29, 30].In the PPI network, Nfe2l2 was correlated with HO-1. Western blotting analysis revealed that HO-1 level was notably lowered in the tMCAO group compared with sham group (p < 0.001). Nfe2l2 activation upregulated HO-1 (p < 0.0001) (Fig. 10A and B). Additionally, compared with sham group, 24 h-tMCAO group had reduced HO-1 mRNA transcript level (p < 0.05); however, Nfe2l2 activation upregulated HO-1 transcript level (p < 0.001) (Fig. 10C). Thus, our results suggest that I/R-induced ferroptosis could be inhibited through the Nfe2l2/HO-1 pathway.

Fig. 10

SFN inhibited ferroptosis via Nfe2l2/HO-1 pathway. (A, B) Western blot assay was conducted to analyze HO-1 expression (n = 4). (C) qRT-PCR was used for analyzing HO-1 expression levels (n = 4). Data represent means ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 and N.S. nonsignificant versus sham group

Discussion

IS is a commonly seen disorder that endangers human health and life. Cerebral ischemia can induce various complex and hazardous events, such as inflammation, excitotoxicity, or cell death. Oxidative stress has a critical effect on the pathogenic mechanism and therapeutic targets of cerebral I/R. Additionally, ferroptosis may induce an imbalance of oxidative stress and cause oxidative damage in vivo. Single-cell RNA-seq allows us to comprehensively explore IS at a single-cell transcriptome level, uncover the different phenotypes of brain cells, and investigate the possible critical immune cell clusters in IS. We explored the functions of ferroptosis-related genes during IS and identified potential efficient therapeutic targets based on bioinformatics analyses and experimental verification.

We identified five hub DEFRGs (Rela, Jun, Myc, Stat3, and Nfe2l2). Nfe2l2, a transcription factor, is related to various biological events, including inflammation, oxidative stress, and cellular homeostasis regulation. The role of Nfe2l2 in cerebral I/R remains unknown. Functional and pathway enrichment analyses demonstrated that Nfe2l2 was mainly associated with pathways, such as ferroptosis, TNF-α, IL-17, and MAPK, and inflammatory bowel disease. TNF-α and IL-17, which may participate in stroke-associated neuronal inflammatory injury stage. Selective targeting of IL-17 and TNF-α signaling pathways might provide a novel therapeutic target for treating stroke [31,32,33].In this study, activation of Nfe2l2 decreased IL-17 and TNF-α levels. This observation was corroborated by immunohistochemical analysis within the penumbra tissue. Nfe2l2 can suppress proinflammatory gene transcription and inhibit inflammation by reducing ROS production [34].SFN, the NRF2 activator, can suppress inflammation by downregulating NF-κB-mediated inflammatory factors and proinflammatory cytokines, including iNOS, TNF-α, and COX2 [35].

Ferroptosis is related to IS genesis and progression. Nonetheless, studies on ferroptosis-related genes, their possible regulatory roles, and immune infiltration landscapes is lacking, which is crucial for treating patients with IS. Thus, we utilized CIBERSORT in analyzing immune infiltration within IS. Combining the immune infiltration difference analysis results and the relevance of ferroptosis in infiltration, we found that DCs, M2 macrophages, and monocytes were significantly higher in the IS group. The brain DC counts elevated at 24 and 72 h postischemia, which aggregated on an infarct border close to the invasive T cells. The number of peripheral DCs that entered the brain apparently increased at 72 h post-I/R, which were mainly located at the center of an ischemic infarct. Further research is warranted to address the possible role of ischemia-activated brain DC in stimulating cytokines or chemokines expression in T cells [36]. Tetrahydroxy stilbene glucoside ameliorates the phenotypes of IS [37], which is related to the higher infiltrating M2 anti-inflammation macrophage proportion in the brain post-stroke and provides remedy for IS. Studies have found IL-1β, IL-4, and TNF-α mRNA levels increased, whereas IL-10 mRNA expression levels decreased at 24 and 48 h post-IS within total monocytes of patients compared with those of controls. Disability and stroke severity may be the possible factors triggering CD163 + expression within the circulating CD16 + monocytes [38].

We downloaded the GSE154396 database by high throughput sequencing analysis of the MCAO mice and used bioinformatics analyses for exploring candidate targets and the pathogenic mechanism of IS. The distribution of the five hub DEFRGs is observed in the microglia, granulocytes, B cells, T cells, macrophages, monocytes, and NK cells. Nfe2l2 is mainly distributed in the microglia. GSK-3β phosphorylation-mediated activation of Nfe2l2 is associated with sphingosylphosphorylcholine-related resistance to cerebral ischemia through the transformation of microglia/macrophages to anti-inflammatory phenotype following cerebral I/R [39]. Sinomenine treatment also promoted Nfe2l2dependent microglia, M1/M2 polarization, and inhibited NF-κB inhibitor alpha phosphorylation as well as NF-κB nuclear translocation [40].In an experimental stroke model, nonmitogenic fibroblast growth factor 1 promoted functional recovery through regulation of the microglia/macrophage-induced neuroinflammation by the Nfe2l2 and NF-κB pathways [41].

GSEA of Rela was rich in peptidyl-tyrosine phosphorylation, tricarboxylic acid cycle, and positive regulation of phosphatidylinositol 3-kinase activity. p50/Rela dimer is related to the pathogenic mechanism of injury postischemia, and activation of c-Rel-containing dimers enhances ischemia tolerance of neurons [42].Rela acetylation within Lys310 contributes to dictating the NF-κB-mediated proapoptotic responses, which is an appropriate target for dissecting the pathology through neuroprotective NF-κB activation in the case of brain ischemia [43]. GSEA of Jun was rich in DNA damage checkpoint, erythrocyte differentiation, and positive regulation of protein binding. c-Jun and NF-κB are upregulated, which aggravate cerebral I/R induced damage [44].MiR-139 upregulation exerts neuroprotection on oxygen-glucose deprivation/re-oxygenation (OGD/R)-mediated neurological injury through the negative regulation of c-Jun/NLRP3 inflammasome pathway [45]. GSEA of Myc was rich in dorsal ventral axis specification, lipid glycosylation, regulation of translational initiation, and ribosome biogenesis. C-myc can prevent IS in mice by upregulating miR-200b-5p-targeted SIRT1 [46]. Multimodal rehabilitation program contributes to promoting the recovery of motor functions in rats following IS by increasing the GAP-43 and SYN levels in the arteriae cerebri anterior zone while increasing HSP70 and C-Myc levels within the brain tissues [47].GSEA of Stat3 was rich in negative regulation of MAPK signaling cascade, protein k48-linked ubiquitination, and ubiquitin-dependent protein catabolic process. Rux shows neuroprotection on CIRI through mitigating neuroinflammation by inhibiting NLRP3 inflammasome via JAK2/STAT3 pathway inactivation [48]. Resveratrol exerts its neuroprotection effect on cerebral I/R injury through activation of JAK2/STAT3 and PI3K/AKT/mTOR pathways [49]. GSEA of Nfe2l2 was rich in ferrous iron transport, negative regulation of the MAPK signaling cascade, and protein deubiquitination. In an aforementioned KEGG analysis, we discovered that Nfe2l2 was rich in the MAPK signaling pathway, which is probably critical for cerebral I/R. Kaempferol protects against OGD/R-mediated ferroptosis partially through activation of Nfe2l2/SLC7A11/GPX4 pathway [50].Diosmetin suppresses oxidative stress to mitigate cerebral I/R injury by SIRT1/Nfe2l2 pathway [51]. Upregulation of Nfe2l2 and MAPK phosphorylation can relieve oxidative pressure [52]. Lupeol protects from cerebral I/R by inhibiting the oxidative and inflammatory factor levels via activation of Nfe2l2 while inhibiting p38-MAPK in rats [16]. Overexpression of GIP attenuates ferroptosis by regulating Nfe2l2 through the MAPK signaling pathway [53]. The MAPK signaling pathway is classified into three families, including extracellular signal regulated kinases (ERK), c-JNK, and p38-kinases, and they have critical and cell-type specific functions in ferroptosis [19].

Cerebral I/R is alleviated through suppressing inflammation, oxidative stress, inhibiting ferroptosis, and maintaining mitochondrial homeostasis. Ferroptosis is related to inflammatory response, and targeting ferroptosis may prevent and treat inflammatory disorders [54]. Nfe2l2 regulates the factors involved in regulating iron homeostasis. Ferritin is a target of the Nfe2l2-ARE pathway [55, 56], and its expression is related to Nfe2l2 levels [57]. COX2 is a ferroptosis marker gene, whose upregulation is a symbol of ferroptosis [58].Nfe2l2 can modulate FPN1-dependent iron efflux and promote the expression of FPN1 and TFR1 [59, 60]. In the future, we can use siRNA or CRISPRi to knock down Nfe2l2 and overexpress it via CRISPRa or a plasmid in cell models to observe the effects on ferroptosis and inflammation.

Nonetheless, Nfe2l2 upregulation is found to impact iron metabolism by increasing the HO-1 expression, inducing iron deposition while regulating the ferroptosis pathway, which may be related to the interactions between Nfe2l2 and other key transcriptional factors. As the stabilized transcription factor, Nfe2l2 has antioxidative effect and is important for regulating ferroptosis [15, 61].Nfe2l2/HO-1 pathway significantly contributes to alleviating inflammation, ferroptosis, and lipid peroxidation within numerous organs and during different pathological processes [62, 63].

Conclusion

In summary, using bioinformatics analyses combined with mouse model verification, we identified five hub DEFRGs together with the signaling pathways related to IS, which could become potential diagnostic biomarkers and therapeutic targets for IS. Experimental verification showed that ferroptosis does occur during IS. The Nfe2l2/HO-1 pathway might make a vital impact on inhibiting ferroptosis and inflammation. We established immune infiltration and single-cell RNA-seq related to Nfe2l2 and identified the potential drug SFN for clinically treating IS. Our results offer novel insights on the pathogenic mechanism and treatment of IS and the effect of Nfe2l2/HO-1 pathway on regulating ferroptosis during IS.