Ex vivo T cell depletion of donor marrow grafts in humans and mice has virtually eliminated severe graft-versus-host disease (GVHD). However, as a consequence of T cell depletion, sustained donor cell engraftment is likely compromised. Since the majority of T cell depletion techniques also deplete natural killer (NK) cells, we investigated the role of donor NK cells in engraftment and GVHD in a murine model. Using a monoclonal antibody directed against an NK-specific epitope, we have selectively depleted NK cells while preserving donor marrow T cells. In an established model of engraftment, selective NK depletion demonstrated that removal of donor NK cells did not impair the engraftment process under conditions in which donors and recipients are major histocompatibility complex-disparate. In contrast, recipients of anti-Thy 1.2 plus complement (C')-treated marrow grafts had a significantly higher incidence of either partial engraftment or graft rejection as compared with recipients of selective NK-depleted donor grafts or control grafts. In addition, we have observed that NK-specific depletion of donor marrow/splenocyte inocula does not alter the incidence of GVHD. Recipients of NK-depleted donor grafts developed lethal acute GVHD, whereas recipients of anti-Thy 1.2-depleted donor grafts did not (P less than 0.0001). Interestingly, NK 1.1-depleted donor graft recipients had a significantly increased mortality in comparison with control groups receiving C'-treated grafts (P = 0.04) or anti-Thy 1.2 plus C'-treated grafts (P less than 0.05). Thus, NK depletion may reduce immunosurveillance, thereby increasing the risk of posttransplant infection. We conclude from these results that donor NK cells play an insignificant role in engraftment as well as in the afferent phase of GVHD, but may be important in immunosurveillance when murine bone marrow is transplanted across the major histocompatibility barrier.