Differential diagnosis of hepatocellular carcinoma from metastatic tumors in the liver using microRNA expression.

Distinguishing hepatocellular carcinoma from metastatic tumors in the liver is of great practical importance, with significant therapeutic and prognostic implications. This differential diagnosis can be difficult because metastatic cancers in the liver, especially adenocarcinomas, may mimic the morphology and immunoexpression of hepatocellular carcinoma. Biomarkers that are specifically expressed in either hepatocellular carcinoma or metastatic adenocarcinoma can therefore be useful diagnostic tools. To find such biomarkers, we studied microRNA expression in 144 tumor samples using custom microarrays. Hsa-miR-141 and hsa-miR-200c, microRNAs that promote epithelial phenotypes, had significantly higher levels in non-hepatic epithelial tumors. In contrast, endothelial-associated hsa-miR-126 showed higher expression levels in hepatocellular carcinomas. Combinations of these microRNAs accurately identified primary hepatocellular carcinoma from metastatic adenocarcinoma in the liver. These findings were validated using quantitative real-time PCR to measure microRNA expression in additional samples. Thus, the tissue-specific expression patterns of microRNAs make them useful biomarkers for the diagnosis of liver malignancies.

MicroRNAs accurately identify cancer tissue origin.

MicroRNAs (miRNAs) belong to a class of noncoding, regulatory RNAs that is involved in oncogenesis and shows remarkable tissue specificity. Their potential for tumor classification suggests they may be used in identifying the tissue in which cancers of unknown primary origin arose, a major clinical problem. We measured miRNA expression levels in 400 paraffin-embedded and fresh-frozen samples from 22 different tumor tissues and metastases. We used miRNA microarray data of 253 samples to construct a transparent classifier based on 48 miRNAs. Two-thirds of samples were classified with high confidence, with accuracy >90%. In an independent blinded test-set of 83 samples, overall high-confidence accuracy reached 89%. Classification accuracy reached 100% for most tissue classes, including 131 metastatic samples. We further validated the utility of the miRNA biomarkers by quantitative RT-PCR using 65 additional blinded test samples. Our findings demonstrate the effectiveness of miRNAs as biomarkers for tracing the tissue of origin of cancers of unknown primary origin.