Investigating the ID3/SLC22A4 as immune-related signatures in ischemic stroke

Abstract

Background: Ischemic stroke (IS) is a fearful disease that can cause a variety of immune events. Nevertheless, precise immune-related mechanisms have yet to be systematically elucidated. This study aimed to identify immune-related signatures using machine learning and to validate them with animal experiments and single cell analysis.

Methods: In this study, we screened 24 differentially expressed genes (DEGs) while identifying immune-related signatures that may play a key role in IS development through a comprehensive strategy between least absolute shrinkage and selection operation (LASSO) regression, support vector machine (SVM) and immune-related genes. In addition, we explored immune infiltration using the CIBERSORT algorithm. Finally, we performed validation in mouse brain tissue and single cell analysis.

Results: We identified 24 DEGs for follow-up analysis. ID3 and SLC22A4 were finally identified as the better immune-related signatures through a comprehensive strategy among DEGs, LASSO, SVM and immune-related genes. RT-qPCR, western blot, and immunofluorescence revealed a significant decrease in ID3 and a significant increase in SLC22A4 in the middle cerebral artery occlusion group. Single cell analysis revealed that ID3 was mainly concentrated in endothelial_2 cells and SLC22A4 in astrocytes in the MCAO group. A CIBERSORT finds significantly altered levels of immune infiltration in IS patients.

Conclusions: This study focused on immune-related signatures after stroke and ID3 and SLC22A4 may be new therapeutic targets to promote functional recovery after stroke. Furthermore, the association of ID3 and SLC22A4 with immune cells may be a new direction for post-stroke immunotherapy.

Keywords: ID3; SLC22A4; bioinformation; immune; ischemic stroke.

Figure 1

Flowchart about the entire working processes of this study. IS: ischemic stroke; Con: controls; DEGs: differently expressed genes.

Figure 2

Difference analysis of identification criteria: P < 0.05 and |log₂FC| >0.585. (A) Volcano plots for DEmRNAs in discovery set (71 upregulated and 60 downregulated). (B) Volcano plots for DEmRNAs in validation set (166 upregulated and 135 downregulated). (C) Venn diagram showing the common 24 DEmRNAs between discovery set and validation set. (D) Heatmap plot showing the common 24 DEmRNAs in discovery set. (E) Heatmap plot showing the common 24 DEmRNAs in validation set.

Figure 3

Triple ceRNA network construction. (A) Venn diagram showing the common miRNAs between validation set GSE110993 and predicted miRNAs. (B) Venn diagram showing the DEGs between DEmRNAs and mRNAs. (C) CeRNA network in IS, the octagon represents lncRNA, the diamond represents miRNAs, and the ellipse represents mRNAs. (red represents upregulated, and green represents downregulated).

Figure 4

Enrichment plots from GSEA. (A) The DEGs positively correlated with the NF-κB signaling pathway. (B) The DEGs positively correlated with the IL-17 signaling pathway. (C) The DEGs positively correlated with the TGF-β signaling pathway. (D) The DEGs positively correlated with the TNF signaling pathway. (E) The DEGs negatively correlated with the primary immunodeficiency signaling pathway. (F) The DEGs negatively correlated with the B cell receptor signaling pathway.

Figure 5

Comprehensive strategy to select the better DEIRG in IS. (A) 12 differentially expressed genes are represented by LASSO coefficient profiles. (B) Twelve differentially expressed genes were examined for binomial deviance profiles. (C) ROC curve for analysing LASSO regression model accuracy in the discovery set. (D) ROC curve for analysing LASSO regression model accuracy in the validation set. (E) ROC curve for ID3 in the discovery set. (F) ROC curve for SLC22A4 in the discovery set. (G) ROC curve for ID3 in the validation set. (H) ROC curve for SLC22A4 in the validation set. (I) ROC curve for SVM model accuracy in the discovery set. (J) Venn diagram for showing a comprehensive strategy among DEGs (pink circle), LASSO regression (light green circle), SVM models (purple circle), immune-related genes (light blue circle). DEIRS: differentially expressed immune-related signatures.

Figure 6

Expression validation in vivo models. (A) ID3 expression patterns in IS patients and controls in discovery set. (B) SLC22A4 expression patterns in IS patients and controls in discovery set. (C) The mRNA levels of ID3 in mouse brain tissues. (D) The mRNA levels of SLC22A4 in mouse brain tissues. (E, F) The protein levels of ID3 in mouse brain tissues. (G, H) The protein levels of SLC22A4 in mouse brain tissues. (I) The immunofluorescence levels of ID3 and SLC22A4 in mouse brain tissues. MCAO group and sham group, number of mice per group n=6.

Figure 7

Immune infiltration characteristics. (A) A bar plot shows the relative percentage of 22 immune cell subsets. (B) Comparison of immune cells infiltrating IS patients and controls. (C, D) A Spearman correlation of immune cell subsets and ID3; SLC22A4. The color and size of the dots indicate the strength of the correlation.

Figure 8

Mouse brains’ scRNA-seq demonstrates transcriptome atlas. (A) UMAP plot for visualizing clustering profiles between MCAO and sham groups. (B) The UMAP plots display clustering of single cells by types. (C) The proportion of cells in each sample for each cluster is shown in a bar plot. (D) Visualizing the number of cells in each sample for each cluster. (E, F) Violin plots showing ID3 and SLC22A4 expression levels in the above 13 clusters.

Figure 9

Correlation analysis based on the single cell level. (A–C) Correlation analysis of ID3 with Claudin5, Occludin and ZO1 in vascular endothelial cells. (D–F) Correlation analysis of SLC22A4 with GFAP, S100β and EAAT1 in astrocytes.