Identification of tumor tissue origin by a microRNA-based molecular assay

Abstract

11036 Background: Hundreds of thousands of patients are diagnosed each year with metastatic cancer. For ∼10% of these, the tumor primary site is never identified, and they are defined as Cancer of Unknown Primary (CUP). Identification of tumor tissue-of- origin has significant therapeutic implications and presents a major diagnostic challenge. In previous work we showed that by combining expression profiles of tissue-specific microRNAs with a biologically-motivated classification scheme, tumor tissue-of-origin can be identified with high accuracy. Here we describe the development of this approach into a practical diagnostic assay. Methods: We developed protocols for extraction of high-quality RNA that retain the microRNA fraction from FFPE archival tissue samples. Proprietary, highly specific qRT-PCR was used to profile microRNA expression levels in hundreds of samples. Results: A training set of nearly 400 primary and metastatic tumors samples with known primary sites, representing 25 different tumor types, was used to define a standardized diagnostics assay (miRview mets). The assay uses a qRT-PCR protocol to measure a panel of 48 microRNA biomarkers. The assay was validated on a test set of nearly 200 primary and metastatic tumors whose primary sites were blinded. The classification protocol identifies either a single, high-confidence origin or two possible low-confidence predictions. Overall, correct primary site was identified for 83% of the tumors. For 70% of the cases a single high-confidence prediction was made; these cases had a higher accuracy: 90% of the primary sites predicted with high confidence were accurately identified. Conclusions: Previous studies highlighted the tissue-specificity of microRNA expression. We developed this potential into a diagnostic assay that identifies tumor origins with high accuracy. This assay provides an important new tool for diagnosing tumor tissue origin. [Table: see text]