Graft-versus-host disease (GvHD) caused by alloreactive T cells within the graft is a major drawback of allogeneic BMT, but depletion of T cells leads to higher rates of relapse, opportunistic infections and graft failure. Therefore, selective removal of GvHD-inducing alloreactive T cells might be beneficial. We describe here the separation of alloresponsive T cells, based on carboxyfluorescein succimidyl ester labeling, in vitro allostimulation and FACS-sorting. In vivo effects of the separated cell populations were investigated in the context of allogeneic BMT in murine models: in vitro resting T cells were shown to survive in the allogeneic host and retain immunoreactivity against 'third-party' antigens. As demonstrated in two different transplantation models, elimination of proliferating cells significantly reduces GvHD but offers no advantages to using T-cell-depleted bone marrow alone concerning engraftment and tumor control. Transplanting T cells that proliferate in response to tumor antigens in vitro may narrow down the spectrum of antigens recognized by T cells and therefore reduce GvHD while maintaining graft-facilitating function and tumor control. Therefore, selecting tumor-reactive T cells on the basis of their proliferative response in vitro may be beneficial for the recipient, less time consuming than T-cell cloning and still reduce the extent of GvHD.