Genomic and transcriptomic analyses of thyroid cancers identify DICER1 somatic mutations in adult follicular-patterned RAS-like tumors.

Abstract

Papillary thyroid carcinoma (PTC) is the most common type of thyroid cancer (TC). Several genomic and transcriptomic studies explored the molecular landscape of follicular cell-derived TCs, and BRAFV600E, RAS mutations, and gene fusions are well-established drivers. DICER1 mutations were described in specific sets of TC patients but represent a rare event in adult TC patients. Here, we report the molecular characterization of 30 retrospective follicular cell-derived thyroid tumors, comprising PTCs (90%) and poorly differentiated TCs (10%), collected at our Institute. We performed DNA whole-exome sequencing using patient-matched control for somatic mutation calling, and targeted RNA-seq for gene fusion detection. Transcriptional profiles established in the same cohort by microarray were investigated using three signaling-related gene signatures derived from The Cancer Genome Atlas (TCGA). The occurrence of BRAFV600E (44%), RAS mutations (13%), and gene fusions (13%) was confirmed in our cohort. In addition, in two patients lacking known drivers, mutations of the DICER1 gene (p.D1709N and p.D1810V) were identified. DICER1 mutations occur in two adult patients with follicular-pattern lesions, and in one of them a second concurrent DICER1 mutation (p.R459*) is also observed. Additional putative drivers include ROS1 gene (p.P2130A mutation), identified in a patient with a rare solid-trabecular subtype of PTC. Transcriptomics indicates that DICER1 tumors are RAS-like, whereas the ROS1-mutated tumor displays a borderline RAS-/BRAF-like subtype. We also provide an overview of DICER1 and ROS1 mutations in thyroid lesions by investigating the COSMIC database. Even though small, our series recapitulates the genetic background of PTC. Furthermore, we identified DICER1 mutations, one of which is previously unreported in thyroid lesions. For these less common alterations and for patients with unknown drivers, we provide signaling information applying TCGA-derived classification.