Leukemic blast progenitors in acute myeloblastic leukemia (AML) undergo terminal divisions and/or self-renewal, which can be studied by the methylcellulose culture method and suspension culture, respectively. Using these methods, we have shown that busulfan (BU) was more effective against the self-renewal of blast progenitors than against the terminal divisions. As reported previously, cytosine arabinoside (Ara-C) is more effective against self-renewal and adriamycin (ADR) is more effective against terminal divisions. To determine the reason why antileukemic drugs show different sensitivity against self-renewal and terminal divisions, we studied the effects of these drugs on normal human bone marrow cells. We have shown that BU and Ara-C suppressed the colony formation induced by interleukin-3 (IL-3) more effectively than that by granulocyte colony-stimulating factor (G-CSF). On the other hand, ADR suppressed the colonies induced by G-CSF more effectively. In normal hemopoiesis, IL-3 has been shown to stimulate the growth of more primitive hemopoietic stem cells, while G-CSF has been documented to stimulate the growth of granulopoietic committed precursors at a more differentiated stage. Therefore, BU and Ara-C are considered to be more effective against the immature cells in normal bone marrow and ADR is effective against the more differentiated cells. We suppose that the different differentiation stage of target cells may explain the difference of the effects on self-renewal and terminal divisions between BU, Ara-C and ADR.