Identification of a pyroptosis-related prognostic signature in breast cancer

Abstract

Background: The relationship between pyroptosis and cancer is complex. It is controversial that whether pyroptosis represses or promotes tumor development. This study aimed to explore prognostic molecular characteristics to predict the prognosis of breast cancer (BRCA) based on a comprehensive analysis of pyroptosis-related gene expression data.

Methods: RNA-sequcing data of BRCA were collected from The Cancer Genome Atlas (TCGA) and Gene Expression Ominibus (GEO) datasets. First, pyroptosis-related differentially expressed genes (DEGs) between normal and tumor tissues were identified from the TCGA database. Based on the DEGs, 1053 BRCA patients were divided into two clusters. Second, DEGs between the two clusters were used to construct a signature by a least absolute shrinkage and selection operator (LASSO) Cox regression model, and the GEO cohort was used to validate the signature. Various statistical methods were applied to assess this gene signature. Finally, Single-sample gene set enrichment analysis (ssGSEA) was employed to compare the enrichment scores of 16 types of immune cells and 13 immune-related pathways between the low- and high-risk groups. We calculated the tumor mutational burden (TMB) of TCGA cohort and evaluated the correlations between the TMB and riskscores of the TCGA cohort. We also compared the TMB between the low- and high-risk groups.

Results: A total of 39 pyroptosis-related DEGs were identified from the TCGA-breast cancer dataset. A prognostic signature comprising 16 genes in the two clusters of DEGs was developed to divide patients into high-risk and low-risk groups, and its prognostic performance was excellent in two independent patient cohorts. The high-risk group generally had lower levels of immune cell infiltration and lower activity of immune pathway activity than did the low-risk group, and different risk groups revealed different proportions of immune subtypes. The TMB is higher in high-risk group compared with low-risk group. OS of low-TMB group is better than that of high-TMB group.

Conclusion: A 16-gene signature comprising pyroptosis-related genes was constructed to assess the prognosis of breast cancer patients and its prognostic performance was excellent in two independent patient cohorts. The signature was found closely associated with the tumor immune microenvironment and the potential correlation could provide some clues for further studies. The signature was also correlated with TMB and the mechanisms are still warranted.

Keywords: Breast Cancer; Prognosis; Pyroptosis; Tumor immune microenvironment; Tumor mutational burden.

Fig. 1

The flowchart of data collection and analyses

Fig. 2

Expressions of the 39 pyroptosis-related DEGs and the interactions among them. A The heatmap of the pyroptosis-related differential expressed genes (DEGs) between the normal and the tumor tissues. P values were showed as: *p < 0.05, **p < 0.01, ***p < 0.001. B The mutational status of every DEG and the TMB of samples. Different colors indicate different mutation types

Fig. 3

Patients stratified into two clusters based on the DEGs. A 1053 BRCA patients were divided into two clusters according to the consensus clustering matrix (k = 2). B Kaplan–Meier overall survival (OS) curves of the two clusters. C Gene Set Variation Analysis (GSVA) analysis indicated that immune related passways enriched more in C2 than in C1

Fig. 4

Identification of risk signature in the TCGA cohort. A Least absolute shrinkage and selection operator (LASSO) regression of the OS-related genes. B Cross-validation for tuning the parameter selection in the LASSO regression

Fig. 5

Validation of risk signature in the TCGA and GEO cohorts separately. A Kaplan–Meier curves for the OS of patients in the high- and low-risk groups in the TCGA cohort. B ROC curves showed the predictive efficiency of the risk scores in the TCGA cohort. C Kaplan–Meier curves for the OS of patients in the high- and low-risk groups in the GEO cohort. D ROC curves showed the predictive efficiency of the risk scores in the GEO cohort. Kaplan–Meier curves for the OS of low-risk group is better than that of high-risk group in stage I-II (E, p < 0.001) and stage III-IV (F, p < 0.001) patients in the TCGA cohort

Fig. 6

Univariate and multivariate Cox regression analyses and nomogram construction. A Univariate analysis for the TCGA cohort. B Multivariate analysis for the TCGA cohort. C After the multivariate analysis, age, stage, riskscore were significant for the prognosis and selected to construct a nomogram to facilitate the prognosis prediction. The corresponding score of each factor (age, stage and riskscore) was calculated in the nomogram and the total score could be used to predict the OS of BRCA patients. D A calibration curve was plotted to indicate the consistency between the actual observed prognosis value and the value predicted by the nomogram. E ROC curves showed the predictive efficiency of the nomogram in the TCGA cohort. F Compared with age and stage in the prognosis prediction, the signature gave an advantage in C-index

Fig. 7

Differences of clinical characteristics and functional analysis. A Some clinical characteristics such as age, T are statistically different between high- and low-risk groups. (*p < 0.5, ***p < 0.001). B Bubble graph for GO enrichment (q-value: the adjusted p-value). C Bubble graph for KEGG pathways enrichment analysis

Fig. 8

Differences of immune cells, pathways and subtypes between different risk groups. A, B Comparison of the enrichment scores of 16 types of immune cells and 13 immune-related pathways between low- and high-risk groups in the TCGA cohort p values were showed as: *p < 0.05, **p < 0.01, ***p < 0.001. C BRCA patients in the TCGA cohort were accordingly divided into 5 different immune subtypes. The classifications of immune subtypes were statistically different between high- and low-risk groups

Fig. 9

Correlations with checkpoints. The signature was negatively correlated with most checkpoints such as TIGIT and LAG3. Several genes in the signature such as CXCL1 and IL27RA were positively correlated with nearly all checkpoints

Fig. 10

The correlation between the signature and TMB. A The correlation between riskscore of our signature and TMB (R = 0.21, p = 1.2e− 10). B The TMB is higher in high-riskscore group compared with low-riskscore group (p = 2.8e− 06). C OS of low-TMB group is better than that of high-TMB group (p < 0.001). D OS of low-TMB+ low-risk group is better than that of high-TMB+ high-risk group (p < 0.001)