In a serum-free medium we have established two new human breast carcinoma cell lines from a single primary tumor. Cultures were maintained on chemically defined medium CDM3 or on minor modifications of this medium, Dulbecco's modified Eagle medium-Ham's F12 supplemented with epidermal growth factor, insulin, transferrin, estradiol, hydrocortisone, triiodothyronine, cyclic AMP, phosphoethanolamine, ethanolamine, fibronectin, fetuin, ascorbic acid, bovine serum albumin, and trace element salts including selenite (Petersen and van Deurs, Cancer Res., 47: 856-866, 1987). Primary cultures comprised both NADPH-neotetrazolium reductase-positive carcinoma cells and NADPH-neotetrazolium reductase-negative cells of stromal appearance, as well as normal epithelial cells (Petersen and van Deurs, Cancer Res., 46: 2013-2020, 1986). In subsequent passages the cells were monitored exclusively using the tumorigenicity assay on nude mice. Two cell lines, one nontumorigenic, HMT-3909S1, and one tumorigenic, HMT-3909S8, were selected from the primary cultures. Selection of S8 through subline S4 required transient supplementation of CDM3 with fetal calf serum. Permanent lines S1 and S8 were maintained on serum-free medium. Further characterization of the two cell lines in terms of normal breast gland differentiation (Petersen and van Deurs, Differentiation, 39: 197-215, 1988) was carried out using immunocytochemistry, immunochemistry, electron microscopy, and cytogenetics. S8 appeared to be identical with the NADPH-neotetrazolium reductase-positive carcinoma cells of the primary cultures, with a particular subpopulation of carcinoma cells in the tumor of origin, and with the tumorigenic cells of the nude mice. This subline was aneuploid, typically epithelial in morphology, and expressed keratins K8 and K18 and the glycoprotein MAM-6, typical of luminal epithelial cells in the normal breast gland. Subline S1 appeared more like the elongated cells in the primary cultures and like a second subpopulation of cells in the carcinoma of origin. However, S1 cells were in fact epithelial, since they expressed keratins. Also, S1 cells seemed to be a triploidation of a cell with close resemblance to S4, while only few cytogenetic differences were found between S4 and S8, suggesting an origin of S1 and S8 via S4 from a single hypothetical stem cell.