Aberrant Epigenetic Patterning Defines an Aggressive Molecular Subtype of Adrenocortical Carcinoma and Exposes a Tissue-Specific Therapeutic Vulnerability

Abstract

The adrenal glands are endocrine organs that produce steroid hormones and catecholamines critical for life. Adrenocortical carcinoma (ACC) is a rare cancer of these glands. Up to 75% of patients with ACC develop metastases, for which therapies are limited and ineffective. Standard of care for metastatic disease comprises administration of the adrenolytic agent mitotane +/- cytotoxic chemotherapy, sometimes paired with palliative local therapies, but <10% of patients survive beyond five years. These statistics highlight an urgent need for novel medical therapies for ACC, contingent on a deeper understanding of targetable molecular circuits driving this disease.

Advances in high-throughput profiling of genetic, epigenetic, and transcriptional programs have revolutionized our understanding of molecular predictors of disease states. Through such studies, we identified that tumors of patients with uniformly rapidly recurrent, routinely fatal ACC (comprising ~40% of all ACC) are characterized by an epigenetic signature of DNA hypermethylation directed to CpG islands, “CIMP-high.” We show these genomic regions are protected from methylation in physiological tissues including the adrenal gland, suggesting cancer-specific mechanisms drive aberrant epigenetic patterning. The focus of this dissertation is to characterize CIMP-high ACC as a molecular subtype from a pan-genomic perspective, to develop a strategy enabling prospective biomarker-based identification of CIMP-high ACC, and to investigate the biological consequences of CpG island hypermethylation.

Here, we re-analyze publicly available datasets including The Cancer Genome Atlas study on ACC (ACC-TCGA) to demonstrate CIMP-high ACC is a homogeneous molecular subtype characterized by hyperactivation of three programs: steroidogenic differentiation, coordinated by master adrenal transcription factor SF1; stemness, through somatic alterations driving constitutive activation of Wnt/β-catenin; and proliferation, through somatic alterations in cell cycle regulators. To enable prospective identification of CIMP-high ACC, we leverage ACC-TCGA to identify a locus that is hypermethylated and silenced exclusively in CIMP-high ACC (G0S2) and develop and validate an overnight biomarker assay in an independent cohort of >100 adrenocortical tumors.

In most human cancers, rapid proliferation and CpG island hypermethylation is associated with increased stemness at the expense of differentiation; the convergence of these programs in CIMP-high ACC is paradoxical. We identify that DNA hypermethylation is directed to embryonic targets of a complex known to suppress differentiation, the Polycomb repressive complex 2 (PRC2). PRC2 represses gene expression through EZH2-mediated histone H3 lysine 27 trimethylation (H3K27me3), and DNA methylation at these sites may hamper PRC2 activity. We show CIMP-high ACC exhibit high expression of EZH2/H3K27me3 despite DNA methylation and use high-throughput approaches to demonstrate: the most widely used in vitro model of ACC is CIMP-high, EZH2 assembles in a DNA methylation-sensitive PRC2 complex, EZH2 is globally excluded from hypermethylated regions, and EZH2 catalytic activity does not coordinate DNA methylation. We then show EZH2 catalytic activity is required for sustained proliferation in vitro and identify two novel complexes coordinating transcriptional programming in ACC, SF1/β-catenin and EZH2/β-catenin, that are conserved in mouse models of adrenal carcinogenesis. Finally, we show SF1/β-catenin globally coordinates a cancer-specific and physiological steroidogenic differentiation program that is stabilized in CIMP-high ACC and erased by EZH2 inhibition.

Taken together, our studies illustrate how CpG island hypermethylation defines an ACC molecular class and exposes a tissue-specific therapeutic vulnerability centered on the pharmacologically targetable enzyme EZH2. Ultimately, we hope this work enables prospective molecular subtyping of ACC and illuminates novel strategies for tissue-specific disruption of the aberrant epigenetic wiring supporting this devastating disease.