Screening and validation of atherosclerosis PAN-apoptotic immune-related genes based on single-cell sequencing

Abstract

Background: Carotid atherosclerosis (CAS) is a complication of atherosclerosis (AS). PAN-optosome is an inflammatory programmed cell death pathway event regulated by the PAN-optosome complex. CAS's PAN-optosome-related genes (PORGs) have yet to be studied. Hence, screening the PAN-optosome-related diagnostic genes for treating CAS was vital.

Methods: We introduced transcriptome data to screen out differentially expressed genes (DEGs) in CAS. Subsequently, WGCNA analysis was utilized to mine module genes about PANoptosis score. We performed differential expression analysis (CAS samples vs. standard samples) to obtain CAS-related differentially expressed genes at the single-cell level. Venn diagram was executed to identify PAN-optosome-related differential genes (POR-DEGs) associated with CAS. Further, LASSO regression and RF algorithm were implemented to were executed to build a diagnostic model. We additionally performed immune infiltration and gene set enrichment analysis (GSEA) based on diagnostic genes. We verified the accuracy of the model genes by single-cell nuclear sequencing and RT-qPCR validation of clinical samples, as well as in vitro cellular experiments.

Results: We identified 785 DEGs associated with CAS. Then, 4296 module genes about PANoptosis score were obtained. We obtained the 7365 and 1631 CAS-related DEGs at the single-cell level, respectively. 67 POR-DEGs were retained Venn diagram. Subsequently, 4 PAN-optosome-related diagnostic genes (CNTN4, FILIP1, PHGDH, and TFPI2) were identified via machine learning. Cellular function tests on four genes showed that these genes have essential roles in maintaining arterial cell viability and resisting cellular senescence.

Conclusion: We obtained four PANoptosis-related diagnostic genes (CNTN4, FILIP1, PHGDH, and TFPI2) associated with CAS, laying a theoretical foundation for treating CAS.

Keywords: PAN-apoptosome related genes; bioinformatics; carotid atherosclerosis; diagnostic genes; single-cell sequencing.

Figure 1

Identification of DEGs and WGCNA analysis. (A) Volcano map of DEGs. (B) Heatmap of DEGs. (C) Differences in PORGs scores in the CAS sample and control group. (D) The scale-free distribution was best captured by gene correlations when power = 7, according to soft threshold analysis. (E) The cluster dendrogram of co-expression in CAS. (F) Heatmap of correlations between different gene modules and traits, with p.adj values in parentheses and correlation coefficients outside parentheses for different modules.

Figure 2

Quality control of single-cell data. (A) Number of genes, number of cells, and percentage of mitochondria sequenced for dataset GSE159677. (B) Number of genes, number of cells, and percentage of mitochondria sequenced for single-cell self-test data.

Figure 3

Clustering and annotation of self-test data and GSE159677 dataset. (A) 2000 highly variable genes in self-assay data. (B) 2000 highly variable genes in the GSE159677 dataset. (C, D) PCA plots of single-cell data. (E, F) Dataset GSE159677 is divided into 16 clusters and 9 cell types. (G, H) Self-test data is divided into 7 clusters and 4 cell types.

Figure 4

DEGs in CAS. (A) Annotation of CAS and normal cells in the GSE159677 dataset. (B) Heatmap of pan-apoptotic gene high expression group vs. low expression group. (C) Single-cell annotation of pan-apoptotic gene high-expression and low-expression groups. (D) Intersecting genes of module genes, ScDEGs1, ScDEGs2, and training set differential genes. (E, F) Enrichment analysis of GO (E) and KEGG (F). DEGs, differentially expressed genes; CAS, carotid atherosclerosis; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Figure 5

PANoptosis-related diagnostic model for CAS. (A) Gene coefficient maps for 67 genes. (B) LASSO regression analysis on 67 genes. (C) Number ofoptimal classification trees. (D) Importance ranking of characterization genes. (E) Intersecting genes screened by Lasso regression analysis and RF algorithm. (F) PANoptosis-related diagnostic model. (G) SVM confusion matrix. (H) PCA analysis of CAS and normal samples. (I–K) Diagnostic models in the validation set (I), SVM confusion matrix (J), and PCA analysis of CAS and normal samples (K).

Figure 6

Construction and verification of Nomogram diagnostic model. (A) Nomogram of 4 diagnostic genes in GSE100927. (B) Nomogram of 4 diagnostic genes in GSE43292. (C) Calibration curve of GSE100927. (D) ROC curve of GSE100927. (E) DCA curve of GSE100927. (F) Calibration curve of GSE43292. (G) ROC curve of GSE43292. (H) DCA curve of GSE43292.

Figure 7

Immune microenvironment of CAS. (A) Comparison of CAS and normal samples using a heat map of the 22 immune cell subpopulations. (B) Expression of different immune cells in CAS and normal. (C) Heatmap of the correlation of CNTN4, FILIP1, PHGDH, and TFPI2 with different immune cells. (D–G) Single-gene GSEA of CNTN4 (D), FILIP1 (E), PHGDH (F) and TFPI2 (G). ns p > 0.05, ** represents p less than 0.01, *** represents p less than 0.001, **** represents p less than 0.0001.

Figure 8

Expression validation of the diagnostic genes. (A) Differences in the expression of four genes in the training and validation sets of CAS versus the normal group. (B) Differential expression of four genes in the single-cell dataset GSE159677 and the self-test dataset. ** represents p value less than 0.01, *** represents p less than 0.001, **** represents p less than 0.0001.

Figure 9

Validation of hub genes by qRT-PCR. The relative expression levels of CNTN4 (A), FILIPLI (B), FHGDH (C), and TFPI2 (D) in control and carotid atherosclerosis samples identified by RT-PCR, with GAPDH as a reference, ***P < 0.001. CAS, carotid atherosclerosis.

Figure 10

Cell senescence and cell viability analysis of HASMCs. (A) qPT-PCR of over-expression of CNTN4, FILIP1, PHGDH, and TFPI2; (B, C) SA-β-Gal staining showed the over-expression of CNTN4, FILIP1, PHGDH and TFPI2 reduced senescence of HASMC treated with H₂O₂; (D, E) Live/Dead staining showed that CNTN4, FILIP1, PHGDH, and TFPI2 increased cell viability of HASMC in a high lipid damage environment compared to the blank control groups. *** represents p less than 0.001.