Sera from 230 hepatocellular carcinoma patients were tested for antinuclear antibodies by anticomplement immunofluorescence in 16 types of transformed, diploid or primary cells of human, monkey, chimpanzee or rat origin. As controls, we tested 85 sera from patients with chronic liver diseases, 48 sera from patients with nonhepatic cancers and 164 sera of normal controls. Exactly 11.2% of all cancer patients but only 3.6% of noncancer patients had complement-fixing antinuclear antibody that reacted with all substrates. Only sera from hepatocellular carcinoma reacted with subsets of the tumor cell substrates. These sera reacted with hepatocellular carcinoma cells and nonhepatic cancer cells (antitumor) or only with one or more of the human hepatocellular carcinoma cell lines, PLC/PRF/5, Hep3B and Mahlavu, that were derived from HBsAg-positive patients (antihepatocellular carcinoma). Three of these reacted only with hepatitis B virus DNA-positive cells (PLC/PRF/5 and Hep3B) that contained "hepatitis B-associated nuclear antigen," 1 reacted only with hepatitis B virus DNA-negative Mahlavu cells, 1 reacted with PLC/PRF/5 and Mahlavu and 3 reacted with all 3 cells. The nuclear antigen in Mahlavu was expressed as a homogeneous fluorescence that spared the nucleoli, was present in a lower percentage of cells than hepatitis B-associated nuclear antigen and was more thermostable than hepatitis B-associated nuclear antigen. However, it resembled hepatitis B-associated nuclear antigen in kinetics of expression and susceptibility to digestion with DNase, RNase and proteinase K. The nature of the nuclear antigens in the hepatocellular carcinoma cells is poorly understood but one possibility is that they may represent the expression of viral or tumor-related genes.(ABSTRACT TRUNCATED AT 250 WORDS)