Low expression of Fas by different tumors including osteosarcoma, correlates with poor prognosis. We found that osteosarcoma lung metastases from patients expressed negligible amounts of Fas, but primary tumors often expressed high Fas levels. The reason for this discrepancy is unknown. We hypothesized that because FasL is constitutively expressed in the lungs, Fas-positive (Fas(+)) tumor cells entering the lungs would bind with FasL and die from Fas-induced apoptosis, resulting in the "selection" of Fas-negative (Fas(-)) cells, which would eventually form metastases. To test this hypothesis, we injected K7 osteosarcoma cells, which express functional Fas in vitro, into mice and confirmed that its bone tumors were Fas(+), but lung metastases were Fas(-). Next, to inhibit Fas signaling without affecting Fas expression, we transfected these cells with a FADD-dominant negative (FDN) plasmid and developed K7/FDN cells. Metastases formed by K7/FDN cells contained Fas(+) tumor cells. Moreover, K7/FDN cells were retained in the lungs longer and formed more lung metastases than K7 cells. In addition, the incidence of lung metastases in FasL-deficient mice injected with K7 cells was higher than that in wild-type mice. Metastases from FasL-deficient mice but not from wild-type mice contained Fas(+) tumor cells. Based on that, we conclude that Fas(-) osteosarcoma cells are selected during lung metastases formation and that inhibition of Fas signaling in tumors or lack of FasL in the host environment allows the proliferation of Fas(+) osteosarcoma cells in the lungs and promotes metastases growth. Therefore, Fas may be considered as a new therapeutic target for osteosarcoma treatment.