The cloning of the PIG-A gene has facilitated the unraveling of the complex pathophysiology of paroxysmal nocturnal hemoglobinuria (PNH). Of current major concern is the mechanism by which a PNH clone expands. Many reports have suggested that an immune mechanism operates to cause bone marrow failure in some patients with PNH, aplastic anemia, and myelodysplastic syndromes. Because blood cells of PNH phenotype are often found in patients with these marrow diseases, one hypothesis is that the PNH clone escapes immune attack, producing a survival advantage by immunoselection. To test this hypothesis, we examined the sensitivity of blood cells, with or without PIG-A mutations, to killing by natural killer (NK) cells, using 51Cr-release assay in vitro. To both peripheral blood and cultured NK cells, PIG-A mutant cells prepared from myeloid and lymphoid leukemic cell lines were less susceptible than their control counterparts (reverted from the mutant cells by transfection with a PIG-A cDNA). NK activity was completely abolished with concanamycin A and by calcium chelation, indicating that killing was perforin-dependent. There were no differences in major histocompatibility (MHC) class I expression or sensitivity to either purified perforin or to interleukin-2-activated NK cells between PIG-A mutant and control cells. From these results, we infer that PIG-A mutant cells lack molecules needed for NK activation or to trigger perforin-mediated killing. Our experiments suggest that PIG-A mutations confer a relative survival advantage to a PNH clone, contributing to selective expansion of these cells in the setting of marrow injury by cytotoxic lymphocytes.