Allogeneic NK cells may play a therapeutic role in treating patients with AML. We have previously shown that high dose cyclophosphamide (120 mg/kg × 1 day) and fludarabine (125 mg/m2 × 5 days) can clear lymphoid space and induce a surge of endogenous IL-15 to expand haploidentical NK cells obtained from CD3-depleted lymphapheresis products from adult donors. In this initial study, 5 of 19 patients achieved remissions and in vivo NK cell expansion. Limitations of this therapy includeinability of NK cells to expand in most patients,development of PTLD (in one patient) andinadequate disease control.We hypothesized that contaminating T cells could compete for NK cell expansion, that B-cells may contribute to PTLD, and that a 2-step NK cell purification method using CD3 depletion followed by CD56 selection (CliniMacs) may overcome these problems. We tested this in 9 patients with advanced AML. The purified NK cells, activated with 1000 U/ml IL-2 (16–20 hours), were infused 48 hours after the last fludarabine dose. Patients then received subcutaneous IL-2 (10 MU) every other day × 6 doses to expand NK cells in vivo. None of the 9 pts treated on this protocol achieved remission or exhibited evidence of in vivo expansion. Several studies were designed to investigate this unexpected result. First, we found that the more extensive processing resulted in approximately 1/3 the NK cell recovery compared to CD3 depletion alone (38±% viable NK cells vs. 91±2% respectively). In addition, we questioned whether the contaminating B cells and monocytes that were removed in the 2-step depletion strategy had served a critical role in NK cell activation or expansion. Cytotoxicity assays performed against K562 targets showed that the killing was about 3-fold higher with the purified (CD3-CD56+) product compared the CD3-depleted product alone (P=0.001 at E:T of 6.6:1). Proliferation, measured by a 6-day thymidine assay, was higher in proportion to the higher NK cell content. The only difference between the two NK products was their expansion after 14 days of culture, where the CD3-depleted product, with contaminating B-cells and monocytes, gave rise to greater NK cell expansion (14 ±3-fold) compared to the 2-step purified product (4.5±0.9, n=6, P=0.005). If this finding holds true in vivo, the co-infusion of accessory cells may be required for NK cell expansion. We next developed in vitro assays using very low concentrations (0.5 ng/ml) of IL-2 and IL-15 to understand their role in expansion. IL-2 or IL-15 alone induced low proliferation and the combination was synergistic. Lastly, UCB, a rich source of NK cell precursors, was compared to adult NK cells. In a short term proliferation assay, CD56+ NK cells stimulated with IL-2 + IL-15 expanded better from adult donors (61274±12999, n=6) than from UCB (20827± 6959, n=5, P=0.026) but there was no difference after 14 days in expansion culture suggesting that the only difference is in kinetics. However, UCB depleted of T-cells (enriching for NK cell precursors) exhibited higher fold expansion over 14 days under different culture conditions conducive to NK cell progenitors. In conclusion, NK cell expansion in vitro depends on cell source, IL-2 and IL-15 (increased in vivo after lymphoid depleting chemotherapy) as well as accessory cells. The role of these factors to enhance in vivo expansion is under clinical investigation to further exploit the NK cell alloreactivity against AML targets.