We investigated the mechanism of turnover of an actin microfilament system in fibroblastic cells on an electron microscopic level. A new derivative of actin was prepared by labeling muscle actin with biotin. Cultured fibroblastic cells were microinjected with biotinylated actin, and incorporated biotin-actin molecules were detected by immunoelectron microscopy using an anti-biotin antibody and a colloidal gold-labeled secondary antibody. We also analyzed the localization of injected biotin-actin molecules on a molecular level by freeze-drying techniques. Incorporation of biotin-actin was rapid in motile peripheral regions, such as lamellipodia and microspikes. At approximately 1 min after injection, biotin-actin molecules were mainly incorporated into the distal part of actin bundles in the microspikes. Heavily labeled actin filaments were also observed at the distal fringe of the densely packed actin networks in the lamellipodium. By 5 min after injection, most actin polymers in microspikes and lamellipodia were labeled uniformly. These findings suggest that actin subunits are added preferentially at the membrane-associated ends of preexisting actin filaments. At earlier times after injection, we often observed that the labeled segments were continuous with unlabeled segments, suggesting the incorporation of new subunits at the ends of preexisting filaments. Actin incorporation into stress fibers was a slower process. At 2-3 min after injection, microfilaments at the surface of stress fibers incorporated biotin-actin, but filaments in the core region of stress fibers did not. At 5-10 min after injection, increasing density of labeling along stress fibers toward their distal ends was observed. Stress fiber termini are generally associated with focal contacts. There was no rapid nucleation of actin filaments off the membrane of focal contacts and the pattern of actin incorporation at focal contacts was essentially identical to that into distal parts of stress fibers. By 60 min after injection, stress fibers were labeled uniformly. We also analyzed the actin incorporation into polygonal nets of actin bundles. Circular dense foci, where actin bundles radiate, were stable structures, and actin filaments around the foci incorporated biotin-actin the slowest among the actin-containing structures within the injected cells. These results indicate that the rate and pattern of actin subunit incorporation differ in different regions of the cytoplasm and suggest the possible role of rapid actin polymerization at the leading margin on the protrusive movement of fibroblastic cells.