Studies were conducted on the hypothesis that melanoma metastasis might be initiated through the generation of hybrids comprised of cells of the primary tumor and tumor-infiltrating leukocytes. Fusion hybrids were generated in vitro between weakly metastatic Cloudman S91 mouse melanoma cells and normal mouse or human macrophages. Hybrids were implanted s.c. in the tail and mice were monitored for metastases. Controls included parental S91 cells, autologous S91 x S91 hybrids, and B16F10 melanoma cells. Of 35 hybrids tested, most were more aggressive than the parental melanoma cells, producing metastases sooner and in more mice. A striking characteristic was heterogeneity amongst hybrids, with some lines producing no metastases and others producing metastases in up to 80% of mice. With few exceptions, hybrids with the highest metastatic potential also had the highest basal melanin content whereas those with the lowest metastatic potential were basally amelanotic, as were the parental melanoma cells. A spontaneous in vivo supermelanotic hybrid between an S91 tumor cell and DBA/2J host cell was one of the most metastatic lines. Hybrids with the highest metastatic potential also exhibited markedly higher chemotaxis to fibroblast-conditioned media. Histologically, the metastatic hybrids demonstrated vascular invasion and spread to distant organs similar to that of metastatic melanomas in mice and humans. Thus previous findings of enhanced metastasis in leukocyte x lymphoma hybrids can now be extended to include leukocyte x melanoma hybrids. Whether such hybridization is a natural cause of metastasis in vivo remains to be determined; however the fusion hybrids with genetically-matched parents described herein so closely resembled naturally-occurring metastatic melanoma cells that they could serve as useful new models for studies of this complex and deadly phenomenon.