Natural killer (NK) cells belong to the lymphocyte lineage; however a myeloid origin has been debated in the past based on nascent experimental evidence. We studied the in vitro development of human NK cells from UCB-derived CD34+ cells following culture with cytokines (IL15, IL7, SCF, FLT3L, IL3) on a murine fetal stromal cell line EL08.1D2 (Blood , 2006; 108: 3824–3833). We investigated the differential requirement of CD34+ subsets for stromal cell support. Limiting dilution experiments showed that CD34+ cells negative for phenotypic markers of NK commitment (CD7, CD161, integrin B7, CD122, CD45RA) absolutely require stromal cells and/or addition of hydrocortisone (HC) to differentiate into functional NK cells. Without stromal cells or HC those progenitors give rise to myeloid lineage cells, but not NK cells. Thus, we hypothesized that stromal cells could instruct myeloid precursors to convert to the NK lineage. Indeed, CD56+ cells generated in stroma supported cultures frequently co-express CD33 and CD13. To determine whether myeloid cells developing from CD34+ cells after 2–3 wk cultures could give rise to NK cells, we FACS sorted the CD56−CD33+CD13high and CD56−CD14+ populations. Such CD33+CD13high and CD14+ cells express macrosialin (CD68) and acquire lyzozyme (by FACS), confirming their myeloid characteristics. Sorted cells cultured further in cytokines alone (IL15, IL7, SCF, FLT3L) did not give rise to NK cells. However, in the presence of cytokines, stromal cells and HC, NK cells were generated. To exclude the possibility of NK cell contamination, CD33+CD13high and CD14+ cells were isolated from cultures of CD34+ cells in conditions not supportive of NK cell development (GM-CSF, IL3, FLT3L, SCF, without stroma, IL15 or IL7). Such cells gave the same results as above (i.e., NK cells developed only with stroma and HC). In additional studies, a fraction (∼16%) of CFU-GM colonies isolated from methylocellulose cultures could generate NK cells only in the presence of stromal cells, HC and cytokines, but not cytokines alone. As more of a definitive marker of the monocytic lineage, we used the surface expression of M-CSF receptor (CD115) on hematopoietic precursors. CD56−CD117+CD115+ and CD56−CD117+CD115− fractions were FACS sorted from 2–3 wk cultures of CD34+ cells. While both populations could differentiate into NK cells, only the CD115+ monocytic precursors required stromal cells. Quantitatively the CD117+CD115− cells were the main source of NK cells in this culture system. Notably the NK cells derived from CD115+ precursors were remarkably different, showing significantly higher expression of Killer Immunoglobulin-like Receptors (KIR: CD158a, CD158b and CD158e) than their CD115− derived counterparts (52% vs 15% KIR+, n=3, p=0.002). With respect to the repertoire of HLA-specific inhibitory receptors, NK cells derived from monocytic precursors resemble the dominant fraction of peripheral blood NK cells, including potentially alloreactive NK cells (KIR+CD94/NKG2A−). Collectively we present evidence that NK cells can be derived from developmental intermediates of the monocytic lineage and this differentiation pathway is dependent upon interaction with stroma. Our data indicate that the developmental trajectory shapes the pattern of inhibitory receptor expression on mature NK cells. Such findings have bearing on our understanding of NK cell biology, post transplant NK cell reconstitution and could explain the paucity of recognized immature NK cell leukemias coinciding with the occurrence of AML variants with NK specific antigen expression.