A Novel Nine Apoptosis-Related Genes Signature Predicting Overall Survival for Kidney Renal Clear Cell Carcinoma and its Associations with Immune Infiltration

Abstract

Background: This study was designed to establish a sensitive prognostic model based on apoptosis-related genes to predict overall survival (OS) in patients with clear cell renal cell carcinoma (ccRCC). Methods: Obtaining the expression of apoptosis-related genes and associated clinical parameters from online datasets (The Cancer Genome Atlas, TCGA), their biological function analyses were performed through differently expressed genes. By means of LASSO, unadjusted and adjusted Cox regression analyses, this predictive signature was constructed and validated by internal and external databases (both TCGA and ArrayExpress). Results: A total of nine apoptosis-related genes (SLC27A2, TNFAIP2, IFI44, CSF2, IL4, MDK, DOCK8, WNT5A, APP) were ultimately screened as associated hub genes and utilized to construct a prognosis model. Then our constructed riskScore model significantly passed the validation in both the internal and external datasets of OS (all p < 0.05) and verified their expressions by qRT-PCR. Moreover, we conducted the Receiver Operating Characteristic (ROC), finding the area under the ROC curves (AUCs) were all above 0.70 which indicated that riskScore was a stable independent prognostic factor (p < 0.05). Furthermore, prognostic nomograms were established to figure out the relationship between 1-, 3- and 5-year OS and individual parameters for ccRCC patients. Additionally, survival analyses indicated that our riskScore worked well in predicting OS in subgroups of age, gender, grade, stage, T, M, N0, White (all p < 0.05), except for African, Asian and N1 (p > 0.05). We also explored its association with immune infiltration and applied cMap database to seek out highly correlated small molecule drugs. Conclusion: Our study successfully constructed a prognostic model containing nine hub apoptosis-related genes for ccRCC, helping clinicians predict patients' OS and making the prognostic assessment more standardized. Future prospective studies are required to validate our findings.

Keywords: apoptosis-related genes; clear cell renal cell carcinoma; overall survival; prognosis; signature.

FIGURE 1

The 127 differentially expressed apoptosis-related genes identified in ccRCC; (A) Heatmap of these 127 differently expressed apoptosis-related genes; (B) Volcano plot of these 127 differently expressed apoptosis-related genes including 39 down-regulated and 88 up-regulated genes; Red nodes mean the up-regulated apoptosis-related genes, log2 (FC) > 1 and p < 0.05. Green nodes represent the down-regulated apoptosis-related genes, log2 (FC)< −1 and p < 0.05. (C) GO enrichment of 127 differently expressed apoptosis-related genes; (D) KEGG pathway analysis of 127 differently expressed apoptosis-related genes.

FIGURE 2

Protein–protein interaction network and its top 3 key modules analysis; (A) Protein–protein interaction network of differentially expressed apoptosis-related genes; (B) key module 1 from PPI network; (C) key module 2 from PPI network; (D) key module 3 from PPI network.

FIGURE 3

Construction of a prognostic model (riskScore) by means of LASSO and adjusted Cox regression analysis; (A,B) LASSO coefficients profiles of the prognostic apoptosis-related genes; The partial likelihood deviance plot presented the minimum number corresponds to the covariates. utilized for adjusted Cox regression analysis; (C) The forest plot of differently expressed apoptosis-related genes were identified to be significantly related to OS of ccRCC patients by adjusted Cox regression analysis. Green points represent negative correlations, whereas red points represent positive correlations.

FIGURE 4

Evaluation and external verification of nine apoptosis-related genes established model; (A) Survival curves of OS from high- and low-risk groups of high- and low-risk divided by riskScore in the TCGA dataset; (B–D) 1-, 3- and 5-year ROC in the TCGA dataset; (E)Survival curves of OS from high- and low-risk groups of high- and low-risk divided by riskScore in the external validation dataset (ArrayExpress); (F–H) 1-, 3- and 5-year ROC in the ArrayExpress dataset; (I,J) The distribution of risk scores for samples, the survival status of patients and the expression heatmap of nine differently expressed genes in the TCGA dataset and the ArrayExpress dataset.

FIGURE 5

Internal verification of nine apoptosis-related genes signature; (A) Survival curves of OS from high- and low-risk groups of high- and low-risk divided by riskScore in the internal validation dataset 1 (test 1); (B–D) 1-, 3- and 5-year ROC in the internal validation dataset 1 (test 1); (E) Survival curves of OS from high- and low-risk groups of high- and low-risk divided by riskScore in the internal validation dataset 2 (test 2); (F–H) 1-, 3- and 5-year ROC in the internal validation dataset 2 (test 2); (I,J) The distribution of risk scores for samples, the survival status of patients and the expression heatmap of nine differently expressed genes the internal validation dataset 1 (test 1) and in the internal validation dataset 2 (test 2).

FIGURE 6

The evaluation of independent prognostic factor; (A,B) Univariate and multivariate Cox regression analysis of the whole training dataset (TCGA); (C,D) Univariate and multivariate Cox regression analysis of the internal validation dataset 1 (test 1); (E,F) Univariate and multivariate Cox regression analysis of the internal validation dataset 2 (test 2); (G,H) Univariate and multivariate Cox regression analysis of the external validation dataset (ArrayExpress).

FIGURE 7

Nomogram and calibration plots based on our established signature and clinical characteristics in both TCGA and ArrayExpress databases; (A,B) Nomogram for predicting probabilities of patients with ccRCC with 1-, 3- and 5-year OS in the TCGA and ArrayExpress databases respectively; (C,D) The calibration plot of the nomogram for agreement test between 1-, 3- and 5-years OS prediction and real outcome in the TCGA and ArrayExpress databases separately.

FIGURE 8

Verification of the expression and prognostic significance of nine critical apoptosis-related genes in ccRCC; (A)Survival analysis of APP; (B) Survival analysis of CSF2; (C) Survival analysis of DOCK8; (D) Survival analysis of IFI44; (E) Survival analysis of IL4; (F) Survival analysis of MDK; (G) Survival analysis of SLC27A2; (H) Survival analysis of TNFAIP2; (I) Survival analysis of WNT5A; (J) qRT-PCR verification of these nine critical apoptosis-related genes in ccRCC.

FIGURE 9

Clinical parameters stratified by our established riskScore for OS; (A) Age>65 stratified by riskScore for OS; (B) Age<=65 stratified by riskScore for OS; (C) Female stratified by riskScore for OS; (D) Male stratified by riskScore for OS; (E) Grade1-2 stratified by riskScore for OS; (F) Grade3-4 stratified by riskScore for OS; (G) M0 stratified by riskScore for OS; (H) M1 stratified by riskScore for OS; (I) N0 stratified by riskScore for OS; (J) N1 stratified by riskScore for OS; (K) African stratified by riskScore for OS; (L) Asian stratified by riskScore for OS; (M) White stratified by riskScore for OS; (N) Stage I-II stratified by riskScore for OS; (O) Stage III-IV stratified by riskScore for OS; (P) T1-2 stage stratified by riskScore for OS; (Q) T3-4 stage stratified by riskScore for OS.

FIGURE 10

Associations between apoptosis-related genes signature and tumor-infiltrating immune cells (TIICs) in ccRCC; (A) The expression of TIICs of all ccRCC samples. (B) Correlation matrix of all TIICs proportions; (C) Dendritic cells resting between high- and low-risk patients with ccRCC; (D) Macrophages M0 between high- and low-risk patients with ccRCC; (E) Macrophages M1 between high- and low-risk patients with ccRCC; (F) Mast cells resting between high- and low-risk patients with ccRCC; (G) Monocytes between high- and low-risk patients with ccRCC; (H) Plasma cells between high- and low-risk patients with ccRCC; (I) T cells follicular helper between high- and low-risk patients with ccRCC; (J) T cells regulatory (Tregs) between high- and low-risk patients with ccRCC; (K) The difference of immune cell infiltration abundances in ccRCC subtypes.