A graft-versus-leukemia effect has been well documented to prevent relapse in chronic myelogenous leukemia (CML) after allogeneic marrow transplantation. One type of lymphocytes that may contribute to this effect are natural killer cells (NK), which after activation with interleukin (IL)-2, exhibit a broad range of cytolytic activity against allogeneic and autologous cells. We have previously demonstrated that IL-2-activated NK (ANK) can be generated from blood of patients with CML and are benign in origin. Their proliferation and function, however, diminish with disease progression in CML, suggesting a role in tumor surveillance. We studied the effect of IL-2-activated NK (ANK) on normal and malignant primitive and committed progenitors in a novel long-term bone marrow culture (LTBMC) assay. Because ANK destroy marrow stromal layers, the use of classic stroma-dependent long-term cultures is not possible. Therefore, we used the stroma noncontact LTBMC system developed in our laboratory to analyze the effect of autologous ANK cells on primitive hematopoietic progenitors. Autologous ANK (CD56+/CD3-) were generated from the peripheral blood of 10 patients with chronic phase CML and from six normal individuals by culturing CD5/CD8-depleted mononuclear cells for 14 days in 1,000 U/mL IL-2. At the same time ANK cultures were initiated, sorted normal (CD34+/DR+) marrow populations were plated in Transwell inserts of the stroma noncontact culture. On day 15, hydrocortisone, which rapidly inhibits ANK function, was removed, and autologous ANK were added to the Transwell inserts with fresh LTBMC medium without hydrocortisone but supplemented with 1,000 U/mL IL-2. After 48 hours, the number of colony-forming cells (CFC) was enumerated in methylcellulose culture. To determine the effect of ANK on more primitive long-term culture-initiating cells (LTCIC), the IL-2-supplemented LTBMC medium was replaced with fresh hydrocortisone containing LTBMC medium, and cultures were maintained for an additional 5 weeks. We demonstrate that autologous ANK did not suppress normal CFC or LTCIC. In contrast, ANK from eight patients with CML with potent cytotoxicity against NK-sensitive (K562) NK-resistant (Raji) tumor targets exhibited an ANK dose-dependent suppression of both CFC and LTCIC. Interestingly, ANK from two patients with CML who exhibited diminished cytotoxicity also did not suppress autologous CFC and LTCIC. These studies indicate that ANK with potent major histocompatibility complex unrestricted cytotoxic activity suppress malignant hematopoiesis. This effect was not mediated by soluble factors and was absolutely dependent on direct cell-to-cell contact. We further demonstrate that the beta2 integrin receptor is involved in ANK recognition of CML targets. These observations support the use of autologous ANK therapy to prevent relapse of CML after autologous marrow transplantation or use of ANK to purge CML marrow for autologous transplantation.