Hepatocellular carcinoma (HCC) accounts for 90% of primary liver cancers and is the fifth most common malignancy worldwide. HCC typically develops in the cirrhotic liver. Our preliminary results indicated that agrin, a heparan sulfate proteoglycan (HSPG) detected by us for the first time in the liver, accumulates in the basement membranes (BMs) of the cirrhotic liver and HCC. This novel finding prompted us to investigate the role of agrin in the pathogenesis and differential diagnosis of HCC. First, the previously unspecified monoclonal antibody anti-HSPG clone 7E12 was verified as anti-agrin, using mass spectrometry. Our subsequent experiments were carried out on specimens from 131 patients with chronic liver disease and 18 individuals with healthy liver, from 4 rats subjected to cirrhosis/HCC induction and 1 untreated control rat, as well as from cultured cells. In both human and rats, significantly increased expression of agrin in cirrhosis and HCC was demonstrated by immunohistochemistry (IHC), Western blot, and quantitative RT-PCR. By double immunofluorescent studies, agrin was localized to the muscular layer of blood vessel walls, the BM of bile ducts and ductular reaction, the microvessel walls of HCC, and occasionally the BM of hepatocellular tumor cells. Colocalization, gene expression, and mRNA in situ hybridization experiments suggested that the sources of agrin include vascular smooth muscle cells, epithelial cells of bile ducts and ductules, activated mesenchymal cells in the stroma of hepatocellular tumors, and occasionally tumor hepatocytes. Agrin in the BMs of bile ducts and blood vessels is thought to play an important role in the survival of bile duct epithelium and vascular endothelium, respectively. Thus, agrin may contribute to the formation of ductular reaction and HCC neovessels. As opposed to HCC neovessels that were consistently found agrin-positive, normal and cirrhotic sinusoids were always devoid of agrin, raising the possibility that agrin IHC might be useful in the differential diagnosis of benign versus malignant hepatocellular lesions. Agrin IHC was performed on 68 benign lesions (8 large regenerative nodules, 23 low-grade and 7 high-grade dysplastic nodules, 30 liver adenomas) and 29 malignant lesions (8 small HCC, 21 HCC), and was evaluated semi-quantitatively. Based on the results of IHC for agrin as well as CD34, a decision algorithm was devised that differentiated benign and malignant parenchymal lesions with a sensitivity of 93.1% and a specificity of 92.6%. Hence, we propose that agrin IHC might help distinguish between malignant hepatocellular lesions and their benign mimickers.