It has been suggested that the development of sickle RBC membrane defects might be related to abnormal amounts of membrane-associated heme (a term we use in its generic sense to include hemoglobins, hemichromes, and free heme). Techniques previously used to measure membrane heme, however, would not distinguish between what is truly membrane-associated and what is merely trapped in RBC ghost preparations. Consequently, we have examined extensively washed inside- out membranes (IOM) prepared from normal and sickle RBC. Approximately 25% of the sickle ghost heme is lost upon conversion to IOM, but sickle IOM still have a significant excess (1.6 +/- 0.3 nmol heme/mg membrane protein compared with 0.7 +/- 0.2 nmol/mg for normal IOM, P less than .001). Amounts of ghost heme are only poorly predictive of amounts of IOM heme (r = .664). Preparation of IOM by using isotonic lysis with saponin yields virtually identical amounts of IOM heme. Small amounts of heme (less than 15%) can be displaced from IOM by using manipulations that elute spectrin, displace electrostatically bound proteins, or cleave the cytoplasmic portion of band 3. Treatment of IOM with dithiothreitol (DTT), however, displaces the most heme (35%), and this is almost reproduced (25% displacement) by the treatment of intact RBC with DTT before IOM preparation. Sequential treatment with all manipulations still leaves about 40% of the heme in sickle IOM, which indicates a compartment more intimately associated with the membrane. At least part of this is free heme without globin, as evidenced by abnormal binding of radiochloroquine to sickle IOM. Conversely, some IOM-associated globin is globin without heme because the measurement of globin per se markedly overpredicts amount of IOM heme. There is a strong correlation between RBC density and amounts of either ghost or IOM heme. Finally, the amount of membrane thiol oxidation (as measured by thiol-disulfide-exchange chromatography) does not correlate at all with ghost heme (r = .105), but it correlates well with IOM heme (r = .877, P less than .001). These data demonstrate that there are abnormal amounts of heme truly associated with sickle RBC membranes, and they are consistent with the hypothesis that this membrane-associated heme participates in the pathobiology of the sickle RBC membrane, particularly those aspects perhaps related to thiol oxidation.