Integrative Bioinformatics Analysis Demonstrates the Prognostic Value of Chromatin Accessibility Biomarkers in Clear Cell Renal Cell Carcinoma

Abstract

Clear cell renal cell carcinoma (ccRCC) accounts for 75%-85% of renal cell carcinoma (RCC) and has a poor 5-year survival rate. In recent years, medical advancement has promoted the understanding of the histopathological and molecular characterization of ccRCC; however, the carcinogenesis and molecular mechanisms of ccRCC remain unclear. Chromatin accessibility is an essential determinant of cellular phenotype. This study aimed to explore the potential role of chromatin accessibility in the development and progression of ccRCC. By the combination of open-access genome-wide chromatin accessibility profiles and gene expression profiles in ccRCC, we obtained a total of 13,474 crucial peaks, corresponding to 5,120 crucial genes and 9,185 differentially expressed genes. Moreover, two potential function modules (P2 and G4) that contained 129 upregulated genes were identified via the weighted gene co-expression network analysis (WGCNA). Furthermore, we obtained five independent predictors (FSCN1, SLC17A9, ANKRD13B, ADCY2, and MAPT), and a prognostic model was established based on these genes through the least absolute shrinkage and selection operator-proportional hazards model (LASSO-Cox) analysis. This model can stratify the ccRCC samples into a high-risk and a low-risk group, from which the patients have distinct prognosis. Further analysis demonstrated a completely different immune cell infiltration pattern between these two risk groups. This study also suggested that mast cell resting is associated with the prognosis of ccRCC and could be a target of immunotherapy. Overall, this study indicated that chromatin accessibility plays an essential role in ccRCC. The five prognostic chromatin accessibility biomarkers and the prognostic immune cells can provide a new direction for the treatment of ccRCC.

Keywords: carcinogenesis; chromatin; clear cell renal cell carcinoma; computational biology; gene expression profiling; weighted gene co-expression network analysis.

Figure 1

The workflow of this study.

Figure 2

Selection of crucial peaks and genes. (A) The distribution of correlation coefficient between peaks and genes. Neg, correlation coefficient ≤0; Pos1, 0 ≤ correlation coefficient ≤ 0.1; Pos2, 0.1 ≤ correlation coefficient ≤ 0.2; Pos3, 0.2 ≤ correlation coefficient ≤ 0.3; Pos4, 0.3 ≤ correlation coefficient ≤ 0.4; Pos5, 0.4 ≤ correlation coefficient ≤ 0.5; Pos6, 0.5 ≤ correlation coefficient ≤ 0.6; Pos7, 0.6 ≤ correlation coefficient ≤ 1. (B) The location distribution of crucial peaks in the genome. (C) The location distribution of crucial genes in the genome. (D) The volcano plot of the DEGs between 250 tumor tissues and 40 normal tissues. Red, upregulated genes; blue, downregulated genes. (E) The volcano plot of the differentially expressed crucial genes between 250 tumors and 40 normal tissues. Red, upregulated genes; blue, downregulated genes. (F) The distribution of DEGs among ATAC-seq data, RNA-seq data, and transcription data. Group A, ATAC-seq data; group B, RNA-seq data; group C, transcription data.

Figure 3

WGCNA for the crucial peaks. (A) Cluster dendrogram of crucial peaks. (B) Module–module correlation heatmap of 44 modules in crucial peaks. Red, high correlation; blue, low correlation. (C) The distribution of crucial peaks among six groups. (D) The distribution of DEGs and corresponding crucial peaks among six groups.

Figure 4

WGCNA for the crucial genes. (A) Cluster dendrogram of crucial genes. (B) Module–module correlation heatmap of 17 modules in crucial genes. Red, high correlation; blue, low correlation. (C) Relationships between mRNA modules and clinical characteristics. (D) The distribution of crucial genes among four groups. (E) The distribution of DEGs among four groups.

Figure 5

Functional analysis and PPI construction of upregulated genes. (A) The Sankey plot. (B) The Venn diagram. (C) GO analysis. (D) KEGG pathway analysis. (E) PPI network.

Figure 6

Construction of the prognostic gene model. (A) Selection of optimal parameter (lambda) in LASSO analysis. (B) LASSO coefficient profiles of the 11 prognostic genes. (C) Multivariate Cox regression analysis determines the five prognostic genes as independent predictors in ccRCC. *p < 0.05; **p < 0.01, ***p < 0.001.

Figure 7

Risk score analysis. (A–C) Gene model risk score distribution, survival times and status, and heatmap of the expression of five prognostic genes in patients with ccRCC. The samples were classified into the low-risk and the high-risk groups based on the cutoff value of risk scores. (D) Kaplan–Meier curves of OS for patients with ccRCC based on the prognostic model. (E) Time-dependent ROC curves of the prognostic model. (F) PCA plot. (G) t-SNE analysis.

Figure 8

The univariate and multivariate Cox regression analyses. (A) Univariate Cox regression analyses. (B) Multivariate Cox regression analyses. (C) ROC curves of the risk score and clinical characteristics for 1-year survival.

Figure 9

The immune cell infiltration analysis between the high-risk and the low-risk groups. (A) The bar chart of the proportion of immune cells in the high-risk group. (B) The bar chart of the proportion of immune cells in the low-risk group. (C) The correlation analysis between immune cells and risk scores in the high-risk group. (D) The correlation analysis between immune cells and risk scores in the low-risk group. (E) Violin plots showing the different abundance of the immune cells between the high-risk (red) and the low-risk (green) groups.

Figure 10

Comprehensive analysis for the immune cells. (A) Venn diagram. (B–G) Correlation analysis between the six hub immune cells and risk scores. (H) Kaplan–Meier analysis of mast cells resting.