Infusions of donor leukocytes have been given to allogeneic bone marrow recipients after transplant to treat leukemia relapse. Treatment with these delayed infusions of donor cells has been called delayed or donor leukocyte infusion (DLI). While graft-vs.-host disease (GVHD) has typically been less severe than expected after DLI, it still remains a significant risk factor. Recently, we used a full major histocompatibility complex (MHC)-mismatched model (C57BL/6 into AKR) to determine how increased immunogenetic disparity affects GVH and graft-vs.-leukemia (GVL) reactions after DLI. In contrast to an MHC-matched model (B10.BR into AKR), GVHD was still observed when MHC-mismatched donor T cells were infused 3 weeks posttransplant. Limiting dilution analysis was used to determine the frequency of alloreactive cytotoxic T lymphocytes (CTL) and interleukin (IL)-2-secreting T helper cells in the spleens of MHC-mismatched recipients 7 days after DLI treatment. GVHD correlated with elevated frequencies of alloreactive T-helper cells. One strategy for reducing the severity of GVHD after DLI is the selective administration of CD4 or CD8 T-subsets. Delayed infusion of purified T-subsets 3 weeks posttransplant resulted in significantly less GVHD than infusion of a mixture of the two subsets. No GVH-associated mortality was observed after DLI with purified donor CD4+ T cells. In GVL studies, MHC-mismatched CD8+ T cells were the most potent antitumor effectors against an acute T cell leukemia. The GVL effect of MHC-mismatched T-subsets was compared with that of MHC-matched subsets. When naive MHC-matched cells were given as DLI, depletion of either T-subset eliminated the GVL effect. CD8+ T cells from MHC-matched donors primed against host alloantigens, however, mediated a CD4 (T-helper)-independent GVL reaction. Together, these results suggest that administration of T-subsets can significantly reduce GVHD after DLI without loss of the beneficial GVL effect.