Gelsolins, prepared from a number of different sources, showed similar severing activity on F-actin in vitro or on stress fibers of detergent-extracted cells but differed in their effects on actin in stress fibers of microinjected cells. When human gelsolin isolated from plasma was injected into cells in a Ca(++)-containing buffer, stress fibers were degraded, the cellular morphology was changed, and numerous actin patches appeared. These effects were particularly striking when the Ca(++)-insensitive N-terminal proteolytic fragment of this gelsolin was injected. By contrast, Ca(++)-sensitive gelsolins isolated from human platelets, pig stomach smooth muscle and pig plasma showed no comparable activity. Furthermore, the Ca(++)-independent N-terminal proteolytic fragments prepared from these gelsolins also had no effect despite their in vitro actin severing activity. Most striking was the finding that human plasma gelsolin expressed in E. coli did not degrade stress fibers, in contrast to the same protein isolated from plasma; nor was there any stress fiber disruption observed with the N-terminal half of human gelsolin expressed in Escherichia coli. The different behavior of these gelsolins in cells cannot be explained by sequence diversity between plasma and cytoplasmic forms, nor by variability in the Ca++ sensitivity of the preparations. It suggests the presence of factors, as yet unidentified, that may regulate gelsolin activity in the cytoplasm of living cells and discriminate between gelsolins of different origin. Such discrimination could be achieved as a result of post-translational modification of the gelsolin; only in this way can differences between apparently identical proteins isolated from human plasma and expressed in E. coli be reconciled.