CpG Island Methylator Phenotype in Adrenocortical Carcinoma: Fact or Fiction?

Abstract

Adrenocortical tumors are common and most are asymptomatic, benign adrenocortical adenomas (ACAs), but a significant number of patients undergo adrenalectomy to exclude an adrenocortical carcinoma (ACC) diagnosis (1). Patients with ACC have a poor but variable prognosis (2). The Weiss criterion has been used for localized adrenocortical tumors to histologically distinguish between ACA and ACC. A Weiss score of 0–2 is considered benign, whereas a Weiss score of 3–9 is considered malignant. Molecular or genomic prognostic markers that can clearly distinguish between low-risk and high-risk ACC tumors for recurrence have important clinical ramifications, especially when considering adjuvant therapy with mitotane in patients with stage I and II ACC (3). Over the last decade, with the application of advanced high throughput genomic technology, our understanding of adrenocortical tumorigenesis has improved. For example, through genome-wide gene expression profiling, it was found that overexpression of the IGF 2 (IGF2), underexpression of the nonprotein coding RNA H19, activation of Wnt- β-catenin signaling, and inactivation of the tumor suppressor p53 are associated with ACC. Through genome-wide microRNA profiling, it has been observed that overexpressed microRNA, such as miR-483-5p, and underexpressed microRNA, such as miR-195 and miR-335, can distinguish ACC from ACA and serve as diagnostic and prognostic markers (4). In 2012, several studies on the epigenetic changes associated with adrenocortical tumors have been reported. Epigenetic changes refer to heritable modifications to the genome that change gene expression but do not alter the DNA sequence (5). Rechache et al. (6) reported the first genome-wide DNA methylation profiling in normal adrenal cortex, and benign and malignant adrenocortical tumors. Methylation profiling of normal, ACA, primary ACC, and metastatic ACC showed unique methylation patterns, and showed that primary and metastatic ACC samples were globally hypomethylated compared with normal and benign samples. Through integrated analysis of genome-wide gene expression and methylation profiling, 52 hypermethylated and down-regulated genes in ACC were identified, which suggests that gene methylation status may be an important regulator of gene expression in ACC. Fonseca et al. (7) also analyzed the genome-wide methylation profiles of normal, benign, and malignant adrenocortical samples. They identified 212 CpG islands in the promoter regions that were significantly hypermethylated in ACC and further validated six of the top 50 genes that might be important in the pathogenesis of ACC. In this issue of the JCEM, Barreau et al. (8) report on their analysis of DNA methylation in adrenocortical tumors from a distinct point of view. Their study focused on the methylation profiles of the CpG islands in the promoter regions. CpG islands are genomic regions of larger than 500 bp, with more than 55% of the nucleotides composed of CpG dinucleotides (9). Unlike the two previous studies, they examined only tumor tissues, which included 84 ACA and 51 ACC samples. Their analysis also focused on the molecular classification of ACC based on patient outcome. The investigators report two important results. First, ACCs were more globally hypermethylated than ACAs at the CpG islands in the promoter regions, which was consistent with findings in previous studies (6, 7). Second, ACC samples could be categorized into two different groups based on their methylation profiles. One of them exhibited only slightly higher methylation than ACAs, which they named the “non-CIMP (CpG island methylator phenotype)” group; the other one exhibited higher methylation than ACAs and the “non-CIMP” group, which they termed the “CIMP” group. The “CIMP” group was further divided into “CIMP-low” and “CIMP-high” subgroups based on their differential hypermethylation profile. CIMP was first described in colorectal cancer to depict a tumor-specific DNA hypermethylation phenotype that can distinguish different tumor groups (10). There have been several reports on CIMP in gastric, ovarian, liver, and lung cancers and leukemia, suggesting that CIMP might be a universal phenotype in cancer (11). This is the first paper that describes the presence of CIMP in ACC. Consistent with their groupings, CIMP and non-CIMP patients showed significantly different prognosis, with longer survival in non-CIMP patients. Previously, the same group reported, based on gene expression analysis, that ACAs could be distinguished from ACCs, and that ACCs could be divided into two groups—one with a good prognosis and the other with a poor prognosis (12). The poor prognosis group was further divided into three subgroups, which were associated with inactivated p53, activated β-catenin, or an “unknown” molecular characteristics group. It is interesting to see that for each group categorized by gene expression, the methylation-based groupings were also different. These data suggest that differential methylation in the CpG promoter regions of genes might be functionally important and classification based on methylation may be a marker for prognosis in patients with ACC. There are some limitations to the study. The authors used the Infinium HumanMethylation27 Beadchip to study the methylation patterns of CpG islands in the proximal promoter regions of 14,475 genes (8). However, the coverage of this platform is not completely genome-wide. There are 20,000 to 25,000 protein-coding genes in the human genome (13). Only 70% of the gene promoters are associated with a CpG island, and only 6.8% of CpG dinucleotides are in CpG islands (14, 15). For example, in colon cancer, it has been shown that neither the promoter nor the CpG islands, but sequences up to 2 kb distant from the CpG islands (CpG island shores) have the most methylation changes (16). Thus, by looking only at the CpG islands at the promoter region of a subset of genes, a lot of information could be missed, and thus the conclusion of the presence of CIMP in ACC is not conclusive. Although CIMP has been described in various cancers, there is also not yet a consensus definition for it. Some investigators have suggested several criteria to define CIMP, which include a quantitative analysis, a group of genes that are hypermethylated in high frequency, and that the methylation should be tumor-specific (17). Barreau et al. (8) in their study used quantitative methylation profiling, described a differential methylation phenotype in ACC samples, and divided them into two groups (non-CIMP and CIMP), and then two more subgroups (CIMP-low and CIMP-high). However, rather than featuring a group of marker genes that can define a CIMP phenotype, the methylation patterns in these groups were more universal. Because only adrenocortical tumor samples were analyzed in their study, it would also be important to make sure that the methylation differences observed were specific to ACC by including normal adrenal cortex samples in the analysis. Finally, through integrated analysis of methylation profiling and gene-expression analysis, a list was compiled of the top 25 genes, which showed inverse correlation between their methylation and expression. Among other genes with tumor suppressor function, H19 was identified, which has consistently been observed to be underexpressed and hypermethylated in ACCs. This served as a validation for their analysis. It would be informative if the fold changes in methylation and gene expression were provided for each top gene identified. Rechache et al. (6) performed a similar comparison in ACC and ACA samples and identified 52 genes that were hypermethylated and down-regulated, but there appears to be no overlap between the top genes identified in these two studies. The discrepancy may be due to the different platforms used for the methylation profiling because the HumanMethylation450 BeadChip used in the study by Rechache et al. (6) provides a more comprehensive coverage of the genome with 17 times more CpG sites than the soon-to-be discontinued HumanMethylation27 BeadChip used in the study by Barreau et al. (8). Overall, the study performed by Barreau et al. (8) advances our knowledge on the role of epigenetic changes in ACC and suggests for the first time the presence of CIMP in ACC. They show differential methylation phenotype in ACC samples, which was associated with prognosis. Future studies are needed to confirm this discovery and to understand the role of CIMP in ACC.