Prognostic Classifier Based on Genome-Wide DNA Methylation Profiling in Well-Differentiated Thyroid Tumors

Abstract

Context: Even though the majority of well-differentiated thyroid carcinoma (WDTC) is indolent, a number of cases display an aggressive behavior. Cumulative evidence suggests that the deregulation of DNA methylation has the potential to point out molecular markers associated with worse prognosis. Objective: To identify a prognostic epigenetic signature in thyroid cancer. Design: Genome-wide DNA methylation assays (450k platform, Illumina) were performed in a cohort of 50 nonneoplastic thyroid tissues (NTs), 17 benign thyroid lesions (BTLs), and 74 thyroid carcinomas (60 papillary, 8 follicular, 2 Hürthle cell, 1 poorly differentiated, and 3 anaplastic). A prognostic classifier for WDTC was developed via diagonal linear discriminant analysis. The results were compared with The Cancer Genome Atlas (TCGA) database. Results: A specific epigenetic profile was detected according to each histological subtype. BTLs and follicular carcinomas showed a greater number of methylated CpG in comparison with NTs, whereas hypomethylation was predominant in papillary and undifferentiated carcinomas. A prognostic classifier based on 21 DNA methylation probes was able to predict poor outcome in patients with WDTC (sensitivity 63%, specificity 92% for internal data; sensitivity 64%, specificity 88% for TCGA data). High-risk score based on the classifier was considered an independent factor of poor outcome (Cox regression, P < 0.001). Conclusions: The methylation profile of thyroid lesions exhibited a specific signature according to the histological subtype. A meaningful algorithm composed of 21 probes was capable of predicting the recurrence in WDTC.