Gelsolin forms ternary complexes with two actin monomers in the presence of Ca2+, which nucleate actin polymerization and cap the barbed ends of filaments. It has therefore been assumed that the two actins are oriented in a similar manner to the terminal subunits in the genetic helix of F-actin. We have tested this using chemical cross-linking with N,N'-1,4-phenylenedimaleimide. For all conditions tested, we identified as the only cross-linked dimeric species an actin dimer indistinguishable from the lower actin dimer of 86 kDa. This lower dimer was previously identified in the initial phase of actin polymerization and also when actin paracrystals are chemically cross-linked [Millonig, R., Salvo, H. & Aebi, U. (1988) J. Cell Biol. 106, 785-796]. It probably defines a contact between adjacent monomers oriented in an antiparallel orientation. In contrast, when F-actin is cross-linked by the same reagent, an upper dimer of apparent molecular mass 115 kDa is formed, which corresponds to adjacent monomers in the genetic helix. The formation of this upper dimer was specifically inhibited by addition of gelsolin to F-actin. Evidence is presented for a Cys374-Cys374 cross-link in the lower dimer. Isolated lower dimer binds to gelsolin in a 1:1 stoichiometry, but it inhibits nucleation of polymerization by gelsolin. Other gelsolin constructs that bind two actin subunits (e.g. the N-terminal half of the molecule, which has severing and capping but no nucleating activity) also form only lower dimer when cross-linked with N,N'-1,4-phenylenedimaleimide. Only segment 2-6 (gelsolin fragment devoid of the N-terminal segment 1) induces an upper dimer orientation of the two actins under nucleating conditions. Our evidence suggests that the two actins associated with gelsolin are not fixed in the orientation of adjacent subunits in F-actin; instead they have a flexible orientation with respect to each other, which permits cross-linking into a stable antiparallel form that does not correspond to the presumed nucleating conformation.