Comprehensive genomic analysis of adrenocortical carcinoma reveals genetic profiles associated with patient survival

Abstract

Background: Adrenocortical carcinoma (ACC) is one of the most lethal endocrine malignancies and there is a lack of clinically useful markers for prognosis and patient stratification. Therefore our aim was to identify clinical and genetic markers that predict outcome in patients with ACC.

Methods: Clinical and genetic data from a total of 162 patients with ACC were analyzed by combining an independent cohort consisting of tumors from Yale School of Medicine, Karolinska Institutet, and Düsseldorf University (YKD) with two public databases [The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO)]. We used a novel bioinformatical pipeline combining differential expression and messenger RNA (mRNA)- and DNA-dependent survival. Data included reanalysis of previously conducted whole-exome sequencing (WES) for the YKD cohort, WES and RNA data for the TCGA cohort, and RNA data for the GEO cohort.

Results: We identified 3903 significant differentially expressed genes when comparing ACC and adrenocortical adenoma, and the mRNA expression levels of 461/3903 genes significantly impacted survival. Subsequent analysis revealed 45 of these genes to be mutated in patients with significantly worse survival. The relationship was significant even after adjusting for stage and age. Protein-protein interaction showed previously unexplored interactions among many of the 45 proteins, including the cancer-related proteins DNA polymerase delta 1 (POLD1), aurora kinase A (AURKA), and kinesin family member 23 (KIF23). Furthermore 14 of the proteins had significant interactions with TP53 which is the most frequently mutated gene in the germline of patients with ACC.

Conclusions: Using a multiparameter approach, we identified 45 genes that significantly influenced survival. Notably, many of these genes have protein interactions not previously implicated in ACC. These findings may lay the foundation for improved prognostication and future targeted therapies.

Keywords: adrenocortical carcinoma; mRNA; mutation; protein–protein interaction; survival.

Graphical abstract

Figure 1

Messenger RNA (mRNA) expression in normal adrenal tissue (NAT), adrenocortical adenoma (ACA), and carcinoma. Differentially expressed genes were identified from transcriptomic data for the Gene Expression Omnibus (GEO) cohort of NAT (n = 10), ACA (n = 22), and adrenocortical carcinomas (ACCs; n = 33). (A) ACCs plotted against principal component one (PC1) and two (PC2) based on mRNA expression. ACCs showed a higher intracluster heterogeneity and also clustered further away compared with ACAs and NAT. (B) Venn diagram showing the overlap between the three sets of differentially expressed genes: ACA versus NAT, ACC versus NAT, and ACC versus ACA. (C) Volcano plot showing all differentially expressed genes between ACC and ACA with samples P < 0.05 and fold change >1.5 in green and all other samples in red. (D) The 50 most differentially expressed genes based on P value between ACC and ACA, clustered by expression similarity. FC, fold change.

Figure 2

Association between messenger RNA (mRNA) expression and overall survival for differentially expressed genes. (A) Schematic illustration of identification of 461 genes associated with survival in The Cancer Genome Atlas (TCGA) cohort and differentially expressed between adrenocortical carcinoma (ACC) and adrenocortical adenoma (ACA) in the Gene Expression Omnibus (GEO) cohort. (B–F) Kaplan–Meier survival curves for ACC cases from the TCGA cohort for high and low mRNA expression levels determined as above or below the median. High mRNA expression of (B) PBK (PDZ-binding kinase); (C) CCNB2 (G2/mitotic-specific cyclin-B2); (D) CDK1 (cyclin-dependent kinase 1); (E) APSM (assembly factor for spindle microtubules); and (F) PTTG1 (PTTG1 regulator of sister chromatid separation) was associated with worse survival.

Figure 3

Tumor mutational burden (TMB). (A) Scatterplot of TMB for each adrenocortical carcinoma (ACC) sample from the Yale–Karolinska–Düsseldorf (YKD; teal) and The Cancer Genome Atlas (TCGA; red) cohorts showing similar levels in the two cohorts. (B) Comparison of median TMB across all TCGA cancer sets with integration of the YKD cohort. The YKD and TCGA–ACC cohorts are indicated. The median TMB of the YKD cohort was significantly lower compared with four of the TCGA cohorts: uterine corpus endometrial carcinoma (UCEC, P < 0.0001), melanoma (SKCM, P < 0.0001), colon adenocarcinoma (COAD, P = 0.0094), and stomach adenocarcinoma (STAD, P = 0.031). The same four cohorts had significantly higher median TMB compared with TCGA–ACC (UCEC, P < 0.0001; SKCM, P < 2 0.0001; COAD, P = 0.0010; STAD, P = 0.0077). (C) Boxplot illustrating higher TMB in stage III-IV tumors than in stage I-II tumors in the YKD and TCGA cohorts combined. Sample with the highest TMB removed from analysis (TMB 1841). (D) A high TMB was associated with worse survival even after adjusting for stage. (E) The top 15 genes with most total mutations and corresponding mutation type in the YKD cohort (left) and the TCGA cohort (right), respectively. The bars correspond to the number of mutations detected.

Figure 4

Survival and protein–protein interaction (PPI) of the 45-gene signature. (A) Survival curve showing significantly worse survival for patients carrying a mutation in at least one of the genes in the 45-gene signature for patients of all stages and (B) only for the patients with stage III or stage IV disease. (C) PPI analysis of the 45 genes adding TP53 as an additional node showed 14 significant interactions with TP53.