We have recently shown that conservation and differentiation of primitive human hematopoietic progenitors in in vitro long-term bone marrow cultures (LTBMC) occurs to a greater extent when hematopoietic cells are grown separated from the stromal layer than when grown in direct contact with the stroma. This finding suggests that hematopoiesis may depend mainly on soluble factors produced by the stroma. To define these soluble factors, we examine here whether a combination of defined early-acting cytokines can replace soluble stroma-derived biologic activities that induce conservation and differentiation of primitive progenitors. Normal human Lineage-/CD34+/HLA-DR- cells (DR-) were cultured either in the absence of a stromal layer ("stroma-free") or in a culture system in which DR- cells were separated from the stromal layer by a microporous membrane ("stroma-noncontact"). Both culture systems were supplemented three times per week with or without cytokines. These studies show that culture of DR- cells for 5 weeks in a "stroma-free" culture supplemented with a combination of four early acting cytokines (Interleukin-3 [IL-3], stem cell factor [SCF], leukemia-inhibitory factor [LIF], and granulocyte colony-stimulating factor [G-CSF]) results in a similar cell expansion as when DR- cells are cultured in "stroma-noncontact" cultures supplemented with the same cytokines. However, generation of committed progenitors and conservation of the more primitive long-term bone marrow culture initiating cells (LTBMC-IC) was far superior in "stroma-noncontact" cultures supplemented with or without IL-3 than in "stroma-free" cultures supplemented with IL-3 alone or a combination of IL-3, LIF, G-CSF, and SCF. These studies indicate that human BM stroma produces soluble factors that can either alone or in synergy with defined cytokines (1) conserve primitive LTBMC-IC, (2) induce early differentiation of a fraction of the primitive progenitors, and (3) prevent their terminal differentiation. We show here that these stroma-derived factors are not likely to be the known early acting cytokines IL-3, SCF, LIF, or G-CSF. Characterization of the stroma-derived factor(s) may have important implications for clinically relevant studies, such as in vitro stem cell expansion in cancer treatment and gene therapy.