CapZ ("capping protein") is a heterodimeric actin capping protein that blocks actin filament assembly and disassembly at the fast growing (barbed) filament ends and is proposed to function in regulating actin filament dynamics as well as in stabilizing actin filament lengths in muscle and nonmuscle cells. We show here that erythrocytes contain a nonmuscle isoform of capZ (EcapZ) that is present exclusively in the cytosol and is not associated with the short actin filaments in the erythrocyte membrane skeleton. This is unlike other cell types where capZ is associated with cytoskeletal actin filaments and suggests that cytosolic EcapZ may be inactive, or alternatively, that the barbed ends are capped by adducin, a membrane skeleton protein that was shown recently to cap actin filament barbed ends in vitro [Kuhlman, P. A., Hughes, C. A., Bennett, V., & Fowler, V. M. (1996) J. Biol. Chem. 271, 7986]. To distinguish between these possibilities, we purified EcapZ from erythrocyte cytosol and characterized its biochemical and functional properties. Two-dimensional gel electrophoresis and western blotting reveals the EcapZ subunit composition to be alpha1beta2, as described for capZ from many other nonmuscle cells, with no evidence for posttranslational modifications. Purified EcapZ is fully functional in blocking actin elongation from barbed filament ends (Kcap approximately 1-5 nM) as well as in nucleating actin polymerization. Furthermore, cytosolic EcapZ binds to actin filament barbed ends, indicating that sequestering of EcapZ by a cytosolic inhibitory factor or insufficient amounts of EcapZ in cytosol also cannot account for its absence from the membrane skeleton. To test directly whether the barbed ends of the erythrocyte actin filaments were already capped, we measured binding of purified EcapZ to isolated membranes. Purified EcapZ does not cosediment with membranes prepared by hypotonic lysis in the presence of magnesium, suggesting that the barbed ends of the erythrocyte actin filaments are capped under these conditions but not by EcapZ. In contrast, purified EcapZ stoichiometrically reassociates with all the actin filament barbed ends in membranes prepared by hypotonic lysis in 5 mM sodium phosphate, pH 8.0 (5P8), conditions in which the barbed filament ends were previously reported to be uncapped. Comparison of the amounts of adducin associated with membranes prepared in the presence and absence of magnesium reveals that 60-80% of the adducin dissociates from the membrane during hemolysis and washing in 5P8 buffer, suggesting that the barbed ends become artifactually uncapped due to loss of adducin. The erythrocyte actin filaments may thus represent a specialized class of membrane-associated actin filaments that are capped by adducin instead of capZ.