We studied the correlation between the formation of brain metastasis and the malignant growth potential of seven human melanoma cell lines, isolated from lymph node metastases (A375-SM, TXM-1, DM-4) or from brain metastases (TXM-13, TXM-18, TXM-34, TXM-40), and the potential of three variants of the mouse K-1735 melanoma. Growth rates in different concentrations of fetal bovine serum and colony-forming efficiency in semisolid agarose were measured, and the tumorigenicity and metastatic ability were determined in nude mice (for the human melanoma cell lines) or in C3H/HeN mice (for the K-1735 variants). The ability to form brain metastasis was tested by injection of cells into the carotid artery. A high colony-forming efficiency in agarose, especially at concentrations of agarose greater than 0.6%, corresponded with high tumor take rates, rapid tumor growth rates, and metastatic colonization of the lungs of the recipient mice. For the human melanomas, the lymph node metastasis-derived cells were more tumorigenic and metastatic than the brain metastasis-derived cells. In the K-1735 mouse melanoma, the tumorigenic and metastatic behavior of the cells after i.v. and s.c. injection corresponded with growth in agarose cultures. However, for growth in the brain after intracarotid injection, the different melanoma cell lines showed similar frequencies of tumor take, regardless of tumorigenicity in other sites of the recipient mice, although mice given injections of brain metastasis-derived cells survived longer than mice given injections of lymph node metastasis (human melanoma) or lung metastasis (K-1735 M-2)-derived cell lines. The results from the human and mouse melanoma cell lines show that the brain metastasis-derived cell lines were not more malignant than the lymph node or lung metastasis-derived cells. These data imply that the production of brain metastasis is not always the final stage of a metastatic cascade.