1 N-hydroxylation is thought to be an essential step in the haemotoxicity of dapsone (DDS). To investigate both metabolism-dependent and cell-selective drug toxicity in vitro we have developed a three-compartment system in which an hepatic drug metabolizing system is contained within a central compartment separated by semipermeable membranes from compartments containing mononuclear leucocytes (MNL) and red blood cells (RBC). 2 Metabolism of dapsone (100 jAM) by rat liver microsomes resulted in toxicity to RBC cells (47.3 ± 2.1% methaemoglobin), but there was no significant toxicity toward MNL (3.7 ± 1.3% cell death) compared with control values (1.6 ± 0.9%). However, when RBC were replaced with buffer in the third compartment there was significantly greater (P < 0.001) white cell toxicity (17.6 ± 0.6% cell death), demonstrating the protection of MNL by RBC. Metabolism of dapsone by human liver microsomes again resulted in RBC toxicity (12.5 ± 3.3% methaemoglobin) but no significant MNL toxicity (2.9 ± 0.8% cell death). Replacement of RBC resulted in a significant (P < 0.001) increase in MNL toxicity (6.5 ± 0.7% cell death). Addition of synthetic dapsone hydroxylamine (30 pLM) in the absence of a metabolizing system and with no RBC in the third compartment resulted in significant (P < 0.001) toxicity toward MNL (43.36 ± 5.82% cell death) compared with control (1.8 ± 1.1%). The presence of RBC in the third compartment resulted in a significant (P < 0.001) decrease in MNL toxicity (17.6 ± 2.2% cell death), with 40.1 ± 3.7% methaemoglobin in the RBC. 3 Like dapsone, procainamide and hydralazine both undergo bioactivation at a nitrogen centre, but are not toxic to red cells in vivo. Both drugs were bioactivated by rat liver microsomes to species which were toxic toward MNL in the three compartment model (13.2 ± 1.3 and 13.3 ± 2.0% respectively), but non-toxic toward RBC. Replacement of red cells with buffer had no significant effect on the toxicity of either compound towards MNL. No bioactivation could be detected in the presence of human liver microsomes. 4 Primaquine, which is associated with a high incidence of red cell toxicity in vivo, was apparently non-toxic to either red or white cells in the three compartment system. However, in incubations without membranes, in which the target cells were not separated from the drug activating system, primaquine was metabolized to a species toxic toward both MNL and RBC by both rat (13.7 ± 1.4% cell death, 3.2 ± 0.3% methaemoglobin) and human liver microsomes (7.6 ± 1.3% cell death, 3.2 ± 0.2% methaemoglobin). This is consistent with the hypothesis that highly reactive and short-lived metabolites are responsible for toxicity associated with primaquine.Keywords dapsone metabolism toxicity