Influence of gene expression on survival of clear cell renal cell carcinoma

Abstract

Approximately 10%-20% of patients with clinically localized clear cell renal cell carcinoma (ccRCC) at time of surgery will subsequently experience metastatic progression. Although considerable progression was seen in the systemic treatment of metastatic ccRCC in last 20 years, once ccRCC spreads beyond the confines of the kidney, 5-year survival is less than 10%. Therefore, significant clinical advances are urgently needed to improve overall survival and patient care to manage the growing number of patients with localized ccRCC. We comprehensively evaluated expression of 388 candidate genes related with survival of ccRCC by using TCGA RNAseq (n = 515), Total Cancer Care (TCC) expression array data (n = 298), and a well characterized Moffitt RCC cohort (n = 248). We initially evaluated all 388 genes for association with overall survival using TCGA and TCC data. Eighty-one genes were selected for further analysis and tested on Moffitt RCC cohort using NanoString expression analysis. Expression of nine genes (AURKA, AURKB, BIRC5, CCNE1, MK167, MMP9, PLOD2, SAA1, and TOP2A) was validated as being associated with poor survival. Survival prognostic models showed that expression of the nine genes and clinical factors predicted the survival in ccRCC patients with AUC value: 0.776, 0.821 and 0.873 for TCGA, TCC and Moffitt data set, respectively. Some of these genes have not been previously implicated in ccRCC survival and thus potentially offer insight into novel therapeutic targets. Future studies are warranted to validate these identified genes, determine their biological mechanisms and evaluate their therapeutic potential in preclinical studies.

Keywords: biomarkers; clear cell renal cell carcinoma; gene expression; survival.

Figure 1

Outline of overall study design. Data from 515 and 298 patients, respectively, were obtained from TCGA and TCC. A COX regression analysis identified genes with expression levels associated with overall survival. Expression levels of 81 genes were further evaluated using NanoString in 248 cases

Figure 2

Boxplots of nine genes. Nine genes (AURKA, AURKB, BIRC5, CCNE1, MK167, MMP9, PLOD2, SAA1, and TOP2A) were confirmed based on expression level between short‐ and long‐term survivor cases in validation set. All the nine genes were overexpressed in short‐term survivors (aggressive) compared to long‐term survivors (indolent). ****p < 0.0001

Figure 3

Overexpression of these nine genes was associated with poor overall survival in the TCGA and TCC datasets with hazard ratios (HRs) ranging from 1.49 to 2.99 in discovery set

Figure 4

Analysis of survival prognostic risk models: three models (nine genes, stage/grade, and combined) for TCGA, TCC, and Moffitt data. 4A. multivariate Cox regression models showed time‐dependent AUC of 0.731, 0.737, and 0.776 in TCGA. 4B. AUC of 0.783, 0.716, and 0.821 in TCC. 4C. logistic regression model for Moffitt data showed AUC of 0.852, 0.702, and 0.873

Figure 5

Methylation driven expression of SAA1. 5A. A box‐plot for each of the eight CpG‐probes for SAA1 comparing tumor vs normal samples. 5B. A negative correlation between the methylation and the expression level. 5C. A high degree of correlation between methylation and the expression level. 5D. Hypomethylation of these CpG sites leads to an increased expression

Figure 6

Methylation driven expression of PLOD2. 6A. A box‐plot for each of the 21 CpG‐probes for PLOD2 comparing tumor vs normal samples. 6B. A negative correlation between the methylation and the expression level. 6C. A high degree of correlation between methylation and the expression level. 6D. Hypomethylation of these CpG sites leads to an increased expression