Résumé : BACKGROUND: Malignant gliomas of the central nervous system remain associated with dismal prognoses because of their diffuse invasion of the brain parenchyma. Very few experimental models that mimic clinical reality are available today to test potentially new therapies. The authors set up experimental in vivo glioma models of anaplastic astrocytomas of human and rat origins and anaplastic oligodendroglioma of human origin. Standard hospital chemotherapies were employed to test the validity of these models. METHODS: Three glioma cells lines obtained from the American Type Culture Collection (i.e., human Hs683 and U373 cells and rat C6 cells) were implanted into nude mouse brains (Hs683 and U373 cells) and rat brains (C6 cells). The astrocytic nature, as opposed to the oligodendrocytic nature, of the Hs683 and U373 models was investigated by using quantitative (computer-assisted microscopy) immunohistochemical characterizations of nestin, vimentin, glutathione-S-transferase alpha (GSTalpha), GSTmu, GSTpi, and p53 expression. Comparative genomic hybridization (CGH) was employed to investigate 1p19q losses. Chronic administrations of carmustine (BCNU), fotemustin, or temozolomide were assayed in the xenografted U373 and Hs683 models. Both BCNU-related chemotherapy and surgery were assayed in the C6 model. RESULTS: The quantitative phenotypic analyses pointed to the oligodendroglial nature of the Hs683 cell line and the astrocytic nature of the U373 cell line. The Hs683 cells exhibited 1p19q losses, whereas the U373 cells did not. BCNU, fotemustin, and temozolomide dramatically increased the time of survival of the Hs683 oligodendroglioma-bearing mice, whereas temozolomide only induced a weak but nevertheless statistically significant increase in the U373 glioma-bearing mice. In the C6 rat glioma model, surgery and BCNU chemotherapy were more efficient than either treatment alone. CONCLUSIONS: The in vivo models of gliomas of the central nervous system developed in the current work best mimicked clinical reality. They can be used either to identify new therapies against human gliomas or to optimize existing therapies.