Comprehensive integration of single-cell RNA and transcriptome RNA sequencing to establish a pyroptosis-related signature for improving prognostic prediction of gastric cancer

Abstract

Cell pyroptosis, a Gasdermin-dependent programmed cell death characterized by inflammasome, plays a complex and dynamic role in Gastric cancer (GC), a serious threat to human health. Therefore, the value of pyroptosis-related genes (PRGs) as prognostic biomarkers and therapeutic indicators for patients needs to be exploited in GC. This study integrates single-cell RNA sequencing (scRNA-seq) dataset GSE183904 with GC transcriptome data from the TCGA database, focusing on the expression and distribution of PRGs in GC at the single-cell level. The prognostic signature of PRGs was established by using Cox and LASSO analyses. The differences in long-term prognosis, immune infiltration, mutation profile, CD274 and response to chemotherapeutic drugs between the two groups were analyzed and evaluated. A tissue array was used to verify the expression of six PRGs, CD274, CD163 and FoxP3. C12orf75, VCAN, RGS2, MKNK2, SOCS3 and TNFAIP2 were successfully screened out to establish a signature to potently predict the survival time of GC patients. A webserver (https://pumc.shinyapps.io/GastricCancer/) for prognostic prediction in GC patients was developed based on this signature. High-risk score patients typically had worse prognoses, resistance to classical chemotherapy, and a more immunosuppressive tumor microenvironment. VCAN, TNFAIP2 and SOCS3 were greatly elevated in the GC while RGS2 and MKNK2 were decreased in the tumor samples. Further, VCAN was positively related to the infiltrations of Tregs and M2 TAMs in GC TME and the CD274 in tumor cells. In summary, a potent pyroptosis-related signature was established to accurately forecast the survival time and treatment responsiveness of GC patients.

Keywords: Cell pyroptosis; Gastric cancer; Immune infiltration; Prognostic signature; Single-cell RNA sequencing.

Graphical abstract

Fig. 1

The analysis flow of this study.

Fig. 2

Identification of cell type by scRNA-seq analysis. A T-SNE distribution of 32 independent cell clusters. B T-SNE distribution of cell types identified by marker genes. C The total number of each cell type identified by marker genes. D The distribution of cell types in different specimens. E Violin plots represent the expression levels of the marker genes for the eleven cell types. F Bubble diagram shows the top 10 marker genes in each annotated cell types.

Fig. 3

Differentially expressed pyroptosis-related genes (PRGs) in gastric cancer at single cell level. A Venn diagram of differentially expressed PRGs and marker genes. B Heatmap showed the expression of 9 PRGs in each annotated cell types. C The GO analysis of the differentially expressed PRGs. D The KEGG analysis of the differentially expressed PRGs.

Fig. 4

Scoring the cell type based on the expression of PRGs. A T-SNE distribution of pyroptosis score of cells by using AUCell function. B Violin plots of pyroptosis score of cell type. C T-SNE distribution of high-pyroptosis and low-pyroptosis group. D Volcanic plot of differentially expressed genes (DEGs) between GC and adjacent tissues. E Venn diagram of DEGs between GC and adjacent tissues and specifically expressed genes in high-pyroptosis group. F The results of GO and KEGG analysis of DEGs in high-pyroptosis group.

Fig. 5

Construction and validation of prognostic models based on PRGs. A LASSO Cox regression analysis of the association between deviance and log(λ). B LASSO Cox regression analysis of the association between coefficients of genes and log(λ). C Heatmap showed the differences of 6 PRGs between high risk and low risk patients. D The survival status of GC patients ranked by risk score. E The survival time of GC patients ranked by risk score. F Kaplan-Meier analysis between high-risk groups and low-risk groups. G Time-dependent ROC curve of risk score predicting the 1-, 3-, and 5-year overall survival (OS). H Details of the nomogram. I The calibration curve for predicting 1-, 3-, and 5-year OS. J The quick response code of online dynamic nomogram K The OS discrepancy between the high-risk and low-risk groups and the ROC curve of risk score predicting the 1-, 3-, and 5-year OS using the GSE62254 dataset. L The DFS discrepancy between the high-risk and low-risk groups and the ROC curve of risk score predicting the 1-, 3-, and 5-year DFS using the GSE62254 dataset.

Fig. 6

GSEA analysis and enrichment analyses of differentially expressed genes between high-risk groups and low-risk groups. A-B The GSEA analysis of differentially expressed genes between high-risk groups and low-risk groups. C The results of GO analysis. D The results of KEGG analysis. E The results of GO analysis. F The results of KEGG analysis.

Fig. 7

The correlation between the risk signature and tumor immune microenvironment. A The correlation between the risk score and infiltrated immune cells. B The correlation between the 6 PRGs and infiltrated immune cells.

Fig. 8

Drug sensitivity and immune infiltration level analysis based on the risk model. A The results of drug sensitivity analysis between high-risk and low-risk groups. B The correlation between the PRGs and CD274. C The expression of immune checkpoints between high-risk and low-risk groups.

Fig. 9

The expression of MKNK2, RGS2, TNFAIP2, VCAN, SOCS3, C12orf75, CD274, CD163 and FoxP3 in GC. A The expression of MKNK2, RGS2, TNFAIP2, VCAN, SOCS3 and C12orf75 in 54 paired GC and adjacent normal tissues. B The representative IHC images of MKNK2, RGS2, TNFAIP2, VCAN and SOCS3 in GC and adjacent normal tissues (scale bar: 50 µm and 20 µm). C The representative IHC images of VCAN, CD274, CD163 and FoxP3 in the same GC tissues (scale bar: 20 µm). D The correlation between VCAN and CD274, CD163 and FoxP3 in the GC tissues. Student’s t-test was used to determine statistical significance: * ** p < 0.001.