Abstract

The mechanism by which paroxysmal nocturnal hemoglobinuria (PNH) clones expand to manifest symptoms is still unknown. PNH cells with PIG-A mutations do not synthesize glycosylphosphatidylinositol (GPI), resulting in deficiency of a series of GPI-linked membrane proteins. Blood cells with PNH phenotype (GPI cells) expand in patients with bone marrow (BM) failure syndromes including PNH, aplastic anemia, and myelodysplastic syndromes. The diseases share an immune-mediated BM injury possibly by cytotoxic lymphocytes such as natural killer (NK) cells and cytotoxic T lymphocytes (CTL). It is suggested that the BM injury allows PNH clones survive selectively. Indeed, we have shown that human leukemic cells (K562) decrease their sensitivity to NK cells in vitro when K562 cells acquire PIG-A mutations (Nagakura et al, Blood 2002). In the present study, we show that the decrease in NK sensitivity of PIG-A mutant cells is ascribable to the deficiency of GPI-linked membrane proteins, ULBPs. ULBPs bind cytomegalovirus protein UL-16, consist of ULBP1-3, and emerge in cell membrane when cells are infected or transformed. They trigger off a NK activation signal, which overrides an inhibitory signal from MHC class I (Cosman et al, Immunity 2001). ULBPs also activate CTL besides NK cells. As target cells, a pair of GPI+ control and GPI mutant cell lines was prepared from a GPI K562 cell line bearing a PIG-A mutation by transfection with a PIG-A cDNA and a vector alone, respectively. Flow cytometry detected ULBP1-3 on the surface of GPI+ but none of GPI K562 cells. As effector cells, we used cultured human NK cells (KHYG-1) deficient in Fc γR type III (CD16). GPI cells were more resistant than GPI+ cells to the killing by NK cells in the 51Cr-release assay. GPI+ cells decreased their sensitivity to NK cells to the level of GPI cells in the presence of antibodies to both ULBP1 and ULBP2, while antibodies to ULBP3 exerted no effects. None of the antibodies to ULBPs showed any effects on the killing of GPI cells, while antibodies to NKG2D, which is a NK receptor for both ULBPs and MICA/B, inhibited the killing of both GPI+ and GPI cells. Thus, GPI cells that lack membrane ULBPs show a survival advantage in the setting of immune attack in vitro. Of clinical interests, ULBPs were detected on the cell surface of GPI+ but none of GPI granulocytes of patients with PNH. GPI+ granulocytes of healthy individuals were negative for ULBPs. There appears pressure to induce membrane ULBPs in patients with PNH, leading to BM injury. These findings suggest that the failure of membrane expression of ULBPs permits selective expansion of PNH clones in patients with PNH and other BM failure syndromes.

Author notes

Corresponding author