We have studied the efflux of dapsone hydroxylamine from normal and diabetic erythrocytes by the use of a two-compartment (1 and 2) in vitro dialysis system, in order to model the in vivo blood supply to the bone marrow. When both types of erythrocytes were dialysed against mononuclear leucocytes, the hydroxylamine crossed the membrane and caused significantly greater white cell death compared with dialysis of leucocytes against untreated erythrocytes. However, in the case of both normal and diabetic cells, the presence of the glutathione depletor diethyl maleate (DEM) caused a marked reduction in movement of hydroxylamine from compartment 1 to 2. Diethyl dithiocarbamate (DDC), a methaemoglobin accelerant, caused a marked reduction in movement of hydroxylamine from erythrocytes (diabetic and normal) in compartment 1 to 2 which led to a significant reduction in white cell death compared with the absence of DDC (18.3 +/- 5.5 vs 34.8 +/- 8.1%, P < 0.05). Dapsone recovery from compartment 1 rose significantly in the presence of DDC compared with control in both erythrocyte types. In contrast, recovery of dapsone from normal erythrocytes incubated in compartment 1 was significantly reduced by the presence of DEM compared with control, although there was no difference between control and DEM-treated diabetic cells. Dapsone analysis in compartment 2 revealed a significant increase in dapsone recovery in both diabetic (11.3 +/- 1.1%) and normal (11.9 +/- 1.1%) erythrocytes in the presence of DDC compared with diabetic (3.3 +/- 0.4%) and normal control (4.8 +/- 2.0%, P < 0.001). The presence of DEM in compartment 1 caused a significant fall in dapsone recovery in compartment 2 (3.7 +/- 0.26) compared with control(4.7 +/- 0.36%,P < 0.05). Hence, dapsone hydroxylamine is capable of leeching out of normal and diabetic erythrocytes, traversing a semipermeable membrane and causing toxicity to human mononucleocyte cells in vitro. This process may be one of the first stages in immune-mediated agranulocytosis.