Introduction
Ischemic stroke (IS) accounts for 75–80% of all strokes and is the leading cause of disability and death worldwide [
1,
2]. Approximately 45% of IS cases are caused by blood clots in small or large arteries. Cardioembolic and lacunar strokes account for 14–30% and 15–25% of IS cases, respectively [
3]. The different stages of the cascade reaction have been extensively studied. However, traditional treatment methods, such as antithrombotic therapy, neuroprotective drugs, or surgical implementation, are substantially limited due to poor safety or therapeutic efficacy [
4,
5]. Therefore, uncovering the underlying molecular mechanisms and exploring innovative therapeutic targets for IS have always been a top priority.
Pyroptosis, a type of programmed cell death, is accompanied by the release of a large number of inflammatory factors [
6]. During IS, inflammasomes, such as abstract in melanoma 2 (AIM2), NLR-family CARD-containing protein 4 (NLRC4), and PYD domain-containing protein 3 (NLRP3), are activated in cells [
7,
8]. This process further activates caspase-1. Caspase-1 activates the cleavage of the N-terminal sequence of Gasdermin-D (GSDMD), allowing it to bind to the membrane and generate membrane pores. Similarly, caspase-1 can simultaneously cleave pro-IL-1β and pro-IL-18 into biologically active forms, producing mature proinflammatory cytokines, which are released into the extracellular environment. Proinflammatory cytokines are toxic to neural cells, leading to pyroptosis [
6]. Recent studies have shown that caspase-8 can also participate in pyroptosis by cleaving GSDMD [
9,
10]. A combination of several antiplatelet drugs has been shown to attenuate inflammasome-mediated pyroptosis by inhibiting the NF-κB/NLRP3 pathway in IS [
6,
11]. However, there are few studies of pyroptosis-related biomarkers in IS.
In this study, based on bioinformatics and experimental verification methods, the analysis of differentially expressed genes (DEGs) between healthy individuals and IS patients was used to reveal the mechanisms related to pyroptosis, identify new potential therapeutic targets, and investigate pyroptosis in IS. The foundation was laid for the development of conditioning treatment regimens.
Materials and methods
Processing of data and computational analysis
The Gene Expression Omnibus (GEO) database (
http://www.ncbi.nlm.nih.gov/geo) was used to obtain gene expression profiles of IS, including GSE58294, GSE22255, GSE66724, GSE16561, and GSE37587. The details of the dataset are presented in Additional file
1: Table S1. Among them, GSE58294, GSE22255, and GSE66724 are all GPL570 platforms integrated into one dataset. GSE16561 and GSE37587 are GPL6883 platforms integrated into one dataset. The corresponding dataset is represented by the platform number. For multiple probes corresponding to a gene, the average expression value was taken as the gene expression value. ComBat was utilized to eliminate batch-to-batch variation in the "sva" R package. Distribution patterns of IS and control samples (before and after batch correction using ComBat) were observed by principal component analysis (PCA).
Screening of pyroptosis-related DEGs
The two datasets were analyzed for differences using the limma package. The filter condition was |logFC|> log
2(1.5) and p < 0.05. The filtered DEGs were used for random forest analysis to obtain the important common genes of the two datasets. The pyroptosis-associated gene set, including 44 genes, is shown in Additional file
2: Table S2. After gene set variation analysis (GSVA), pyroptosis scores were obtained. Correlation analysis was performed using the obtained important common genes and pyroptosis scores in the two datasets. The genes with R > 0.2 and p < 0.05 were included in the next analysis, and the intersection of the two data sets was taken as potential pyroptosis-related genes in IS.
Animals
A total of 48 adult male Sprague–Dawley rats (250–280 g) were purchased from the Hunan Slack Jingda Experimental Animal Co., Ltd. All rats underwent a 12 h light–dark cycle in a room with constant humidity (40–70%) and temperature (22–26 °C). Rats were allowed to eat and drink freely before the experiment and fasted for 12 h before surgery. Animals were randomly divided into 2 groups, including the sham operation group (Sham, n = 8) and the middle cerebral artery occlusion/reperfusion (MCAO/R) model group (n = 8). The MCAO/R model was constructed as previously described [
11]. Rats were anesthetized by intraperitoneal injection of sodium pentobarbital (35 mg/kg). Rats were fixed in the supine position. Nylon (0.26 mm in diameter, 40 mm in length) was used to suture the internal carotid artery. The indwelling sutures were maintained for 2 h, and the filaments were removed to allow reperfusion for 24 h. During MCAO/R, the body temperature of the rats was maintained at 37 ± 0.5 °C. In the sham group, the same surgical induction was performed, but the artery was not ligated.
As previously described [
12,
13], 3 days before the MCAO/R intervention, rat brains were targeted for slow injection of 10 µl of 4 × 10
9 TU/ml sh-NC and sh-ANXA3 lentiviral suspensions further to explore the effects of ANXA3 on the animal model. The injection process was carried out at 0.2 μl/min. Following the injection, the rats were subjected to MCAO/R treatment. The rats in the model and sham-operated groups were treated as described above.
All experiments were performed in accordance with institutional guidelines. Rats were sacrificed by intraperitoneal injection of 150 mg/kg sodium pentobarbital. Brain tissue and peripheral blood were obtained for subsequent experimental detection. This study was approved by the Animal Ethical and Welfare Committee of Hunan University of Traditional Chinese Medicine (LL2021091501).
Neurological defect scoring
The neurological function of rats was scored according to Bederson's method [
14]. The rat tail was suspended. Forelimb flexion was observed. A score of 0 points indicated that the activity of the rat was normal; 1 point indicated that when the rat was lifted vertically, the left forelimb could not be fully extended; 2 points indicated that the lateral thrust resistance of the rat was reduced; 3 points indicated unilateral rotation when the rat walked freely; and 4 points indicated that the rat was unable to walk due to flaccid paralysis. A higher score indicates worse neurological function.
Infarct size measurement
Brain tissue was cut into 5 consecutive coronal brain sections. Afterward, brain slices were incubated with 1% 2,3,5 triphenyltetrazolium chloride (TTC, Sangon Biotech Co., Ltd.) for 15 min at 37 °C in the dark [
15]. With reference to previous work [
16,
17], the formula for calculating the percentage of infarcted area was calculated as follows: infarct rate (%) = (volume of the unlesioned hemisphere—volume of the lesioned hemisphere that is not infarcted)/volume of the unlesioned hemisphere × 100%.
Immunofluorescence (IF) staining
Tissue sections were grilled as previously described [
18] and then deparaffinized to water. After heat antigen repair and routine pre-treatments, sections were incubated overnight at 4 °C with primary antibody (caspase1, 22915–1-AP, 1:100, Proteintech, USA). Then, the sections were incubated with a secondary antibody (Goat Anti-Rabbit IgG(H + L), SA00013-2, Proteintech, USA) for 90 min at 37 °C. After nucleation and sealing, the sections were placed under a fluorescence microscope (XD-202, Ningbo Jiangnan Instrument Factory, China) for observation. Image J (ImageJ, National Institutes of Health, USA) was used to analyze the fluorescence intensity.
Cell treatment
A rat pheochromocytoma cell line (PC12, CL-0412, Pricella, China) was cultured in Dulbecco's modified Eagle's medium (D5796, DMEM, Gibco, USA) containing 10% fetal bovine serum (FBS, 10,099,141, Gibco, USA). Cells were cultured at 37 °C in a 5% CO2 environment.
Cells were first divided into control, oxygen–glucose deprivation/reoxygenation (OGD/R), and OGD/R + lipopolysaccharide (LPS) groups. The cells in the OGD/R group were placed in sugar-free DMEM under hypoxia (95% N2, 5% CO
2) for 2 h and then reoxygenated for 2 h. The cells in the OGD/R + LPS group were placed in sugar-free DMEM containing LPS (1 μg/mL, 82857-67-8, Sigma, USA) under hypoxia (95% N2, 5% CO
2) for 2 h and then reoxygenated for 2 h [
11].
Cells were divided into the control, OGD/R + LPS, OGD/R + LPS + silencing of negative control (si-NC, Invitrogen, USA), and OGD/R + LPS + silencing of Annexin A3 (si-ANXA3) groups. Cells in the OGD/R + LPS group were treated as above. Cells in the OGD/R + LPS + si-NC and OGD/R + LPS + si-ANXA3 groups were treated as follows. After cells were transfected with si-NC and si-ANXA3 with Lipofectamine 2000 (2028090, Invitrogen, USA), they were treated with OGD/R and LPS.
Cell counting kit-8 (CCK-8)
The cell viability was assessed using CCK-8 solution (10%) following the manufacturer’s instructions (AWC0114a, Abiowell, China). After incubating the cells at 37 °C with 5% CO2 for 4 h, the absorbance (OD) value at 450 nm was measured using an enzyme marker (MB-530, HEALES, China).
Quantitative real-time PCR (qRT‒PCR)
The above total RNA from peripheral blood, tissue, or cells was extracted and quantitatively analyzed. One microgram of RNA was used to synthesize cDNA. An UltraSYBR Mixture kit (CW2601, CWBIO, China) was utilized as a template for PCR. The primers are as follows. ANXA3, forward primer: TCAGCTCTCTGAGCCTTAGGT; reverse primer: CTCGTGGGGTGACCATTTCG; ANKRD22, forward primer: CAAAGCAGAATGAGGCTCTCG; reverse primer: GTAGCCGTAGCAGTCGGTAG; ADM, forward primer: TTGGACTTTGCGGGTTTTGC, reverse primer: GCTCCGATACCCTGCTGAAA; GAPDH, forward primer: ACAGCAACAGGGTGGTGGAC; reverse primer: TTTGAGGGTGCAGCGAACTT. The expression level of the target gene was analyzed by the 2−ΔΔCT method. Amplified PCR products were quantified and normalized to GAPDH as a control.
Western blotting
The total protein from the tissues or cells described above was extracted and quantitatively analyzed. Protein homogenates were run on SDS‒PAGE gels and transferred to PVDF membranes. After the membrane was blocked with 5% milk, it was incubated overnight in the primary antibody (Additional file
3: Table S3). β-actin was conducted as an internal reference. The secondary antibody was incubated for 90 min at 37 °C. Membranes were subjected to enhanced chemiluminescence for signal intensity quantification. A chemiluminescence imaging system (ChemiScope 6100, China) was adopted for analysis.
Functional enrichment analysis of ANXA3
The GSVA package was used for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses. Correlations between ANXA3 expression and pathways were analyzed. Important pathways among them were screened. Correlations between ANXA3 expression and all genes were analyzed. A gene set of 22 immune cells was utilized for immune cell analysis.
Statistical analysis
Quantitative analysis was performed using Prism 8.0. A Pearson's coefficient correlation test was used to measure correlations. The level of significance was set at *p < 0.05. The error bar represents mean ± standard deviation.
Discussion
Targeting pyroptosis to modulate IS damage is a potential therapeutic strategy [
19]. Based on random forest analysis, GSVA, and Pearson correlation analysis, this study identified three DEGs that might be associated with pyroptosis in IS: ADM, ANKRD22, and ANXA3. ADM and ANXA3 were significantly overexpressed in the MCAO/R model, and the fold difference in ANXA3 expression was greater. Pyroptosis-related proteins were highly expressed in the MCAO/R model. Silencing ANXA3 could reverse the expression of NLRC4 and AIM2 pyroptotic inflammasomes, suggesting that inhibition of ANXA3 might inhibit pyroptosis through the NLRC4/AIM2 axis in PC12 cells.
Bioinformatics is an efficient biomarker screening method [
20]. Li et al. screened immune-related genes in IS based on bioinformatics methods [
21]. Machine learning techniques are increasingly used in the medical field due to their high accuracy [
22]. Random forest is a type of machine learning and is a common way to screen important marker genes [
23]. Chen et al. identified DEGs associated with ferroptosis in IS through bioinformatics analysis of the GSE16561 and GSE22255 datasets [
24]. Martha et al. used a random forest algorithm to predict the prognostic biomarkers of stroke in patients with IS [
25]. Through weighted gene coexpression network analysis and machine learning, the ADM, ANXA3, SLC22A4, and VIM genes were predicted to be serum markers associated with immune cell infiltration for the diagnosis of IS [
26]. In this study, based on random forest, GSVA, and Pearson correlation analyses, three genes possibly related to IS pyroptosis were screened: ANXA3, ANKRD22, and ADM.
ANKRD22, a nuclear-encoded mitochondrial protein, can be involved in gastric mucosal injury and various cancer processes, such as colorectal cancer and prostate cancer [
27‐
29]. In a study of Parkinson's disease, ANKRD22 obtained through bioinformatics screening was shown to regulate neuronal development by regulating cell viability and IL-6 expression [
30]. ADM, which acts as a vascular inhibitor, is upregulated in major depressive disorder, suggesting that ADM may be associated with neuroinflammation-related diseases [
31]. ANXA3 has been studied in various neuroinflammatory diseases, such as schizophrenia, traumatic brain injury, and cerebral small vessel disease [
32‐
34]. Current studies suggest that ANKRD22, ADM, and ANXA3 may serve as markers associated with pyroptosis in IS. ANXA3 is an upregulated protein in the brains of IS patients and rats with MCAO/R [
35,
36]. ADM and ANXA3 may be potential immune-related markers for the diagnosis of IS [
26]. Similar to the above studies, our study found that ADM and ANXA3 were upregulated in the brain tissue and peripheral blood of the rats in the MCAO/R group.
Studies on the regulation of pyroptosis by the NLRP3 inflammasome in IS are more frequently reported than those on the NLRC4/AIM2 inflammasome. For example, TMEM59, a type I transmembrane protein, is protective against IS by inhibiting pyroptosis and inflammatory responses via NLRP3/ASC/cleaved caspase-1 [
37]. Hypoxia-inducible factor-1α (HIF-1α) may modulate the inflammatory response through the NLRP3 inflammasome complex, thereby affecting pyroptosis after stroke [
38]. Medioresinol, a PGC-1α activator, reduces cerebral infarct volume and blood‒brain barrier permeability through the PPARα/GOT1 axis, inhibits endothelial cell pyroptosis, and promotes long-term neurobehavioral recovery [
39]. Recent studies have shown that the NLRC4 inflammasome complex mediates the inflammatory response and pyroptosis of microglia in vitro and in vivo under ischemic conditions [
8]. Long noncoding RNA MEG3 promotes cerebral ischemia‒reperfusion injury by targeting the miR-485/AIM2 axis to increase pyroptosis [
40]. Similarly, in this study, we found that inflammasome (NLRP3, NLRC4, and AIM2 proteins) and pyroptosis-related downstream proteins (GSDMD-N, caspase-8, pro-caspase-1, cleaved caspase-1, IL-1β, and IL-18) were upregulated in the MCAO/R model. ANXA3 pathway inhibition can protect against brain MCAO/R injury [
41]. More importantly, we found that inhibition of ANXA3 might inhibit pyroptosis by inhibiting the NLRC4/AIM2 pathway but not the NLRP3 pathway.
Toll-like receptor binding, inflammasome complex, MYD88-dependent Toll-like receptor signaling pathway, regulation of interleukin 1-mediated signaling pathway, and TGF beta signaling pathway are all pathways closely related to neuroinflammatory and pyroptotic responses. For example, the Toll-like receptor binding pathway can be involved in neutrophil infiltration and polarization in IS and is also associated with cytokine secretion and phagocytic activity of microglia in the central nervous system [
42,
43]. Blockade of the TLR2/TLR4-MyD88 pathway inhibits neuroinflammation in mice with MCAO/R [
44]. Through a cascade of signaling pathways in TLRs, NLRP3 is activated to induce the production of IL-1β and IL-18, leading to pyroptosis [
45]. Activation of the TGF-β/NLRP3/caspase-1 signaling pathway stimulates pyroptosis, which ultimately exacerbates sepsis-induced acute kidney injury [
46]. Therefore, ANXA3 is enriched in the pathways mentioned earlier, suggesting that ANXA3 may be related to inflammation and immune regulation pathways in IS. Further correlation analysis of immune cell infiltration showed that the low ANXA3 group was enriched in CD8 T cells, M2 macrophages, and naive CD4 T cells. The above results suggest that ANXA3 may be a key gene related to inflammation, pyroptosis, and immunity.
However, we did not further explore the detailed mechanism by which inhibition of ANXA3 suppresses MCAO/R-induced pyroptosis, which is a limitation of our study. We plan to do further mechanistic exploration using gene chips and other in vivo and in vitro experiments in future studies.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.