BACKGROUND/OBJECTIVE: Triple-negative breast cancer (TNBC) is a heterogeneous collection of breast tumors with numerous differences including morphological characteristics, genetic makeup, immune-cell infiltration, and response to systemic therapy. DNA methylation profiling is a robust tool to accurately identify disease-specific subtypes. We aimed to generate an epigenetic subclassification of TNBC tumors (epitypes) with utility for clinical decision-making.

METHODS: Genome-wide DNA methylation profiles from TNBC patients generated in the Cancer Genome Atlas project were used to build machine learning-based epigenetic classifiers. Clinical and demographic variables, as well as gene expression and gene mutation data from the same cohort, were integrated to further refine the TNBC epitypes.

RESULTS: This analysis indicated the existence of four TNBC epitypes, named as Epi-CL-A, Epi-CL-B, Epi-CL-C, and Epi-CL-D. Patients with Epi-CL-B tumors showed significantly shorter disease-free survival and overall survival [log rank; P = 0.01; hazard ratio (HR) 3.89, 95% confidence interval (CI) 1.3-11.63 and P = 0.003; HR 5.29, 95% CI 1.55-18.18, respectively]. Significant gene expression and mutation differences among the TNBC epitypes suggested alternative pathway activation that could be used as ancillary therapeutic targets. These epigenetic subtypes showed complementarity with the recently described TNBC transcriptomic subtypes.

CONCLUSIONS: TNBC epigenetic subtypes exhibit significant clinical and molecular differences. The links between genetic make-up, gene expression programs, and epigenetic subtypes open new avenues in the development of laboratory tests to more efficiently stratify TNBC patients, helping optimize tailored treatment approaches.