This protocol presents a new therapeutic approach to the treatment of patients with otherwise incurable malignant metastatic melanoma. Its objective is to define the safety of escalating doses of an anti-cancer treatment involving intratumoral injections of cells that produce recombinant retroviruses. The experimental treatment is based on the introduction into tumoral cells of a suicide gene coding for the herpes simple virus type 1 thymidine kinase (HSV1-TK). Cells that express HSV1-TK become sensitive to ganciclovir (GCV). GCV has no toxicity for normal cells, but kills cells expressing the HSV1-TK enzyme. Such toxicity is restricted to cells undergoing division. Introduction of the gene into tumoral cells is obtained through the intratumoral injection of murine fibroblasts modified by genetic engineering (M11 cells). These cells continuously produce recombinant defective retroviruses containing the HSV1-TK gene. Retroviruses can integrate their genes only when the cells they infect are undergoing division. Thus, after intratumoral injection of M11 cells, the tumoral cells, but not the quiescent cells of the healthy tissue surrounding them, express the HSV1-TK gene and can be destroyed by GCV. In addition, tumoral cells that do not express the gene, but which are located in the immediate vicinity of the transduced cells, are also destroyed through a "bystander effect," also restricted to cells undergoing division. It is therefore not necessary for all the tumoral cells to express HSV1-TK for all of them to be destroyed. Finally, preliminary data suggest that this localized tumoricidal activity may trigger a more general antineoplastic action, by facilitating a specific antitumoral immune response. The efficacy of the above therapeutic approach has been evidenced with animals in the treatment of brain tumors, of colic adenocarcinoma hepatic metastases and of malignant melanoma. A therapeutic trial on recurrent brain tumors or metastases has begun in the USA, using a similar approach. We propose a phase I-II clinical study of the treatment of metastatic malignant melanoma. The patients enrolled in the study must present a metastatic malignant melanoma that is no longer treatable by conventional therapy (life expectancy of patients < 12 months). Progressively increased doses of M11 cells (1 x 10(8), 2 x 10(8), 3 x 10(8) cells/cm3 of tumor) will be injected transcutaneously in the cutaneous, sub-cutaneous or ganglionary tumoral nodules. For a given dosage, four patients receiving the treatment will be studied. Four additional patients will be enrolled at the higher tolerated dosage. We will study the safety and the tumoricidal effect of the direct intratumoral injection of M11 cells followed by treatment with GCV at a constant, intravenous dosage of 10 mg/kg/d x 14 days.