The property of monomeric globular actin (G-actin) to polymerize into noncovalent helical filaments (F-actin) is fundamental to its biological activity in all eukaryotic cells (1). In nonmuscle cells actin filaments are dynamic and undergo self-assembly and disassembly to extents and at turnover rates that are finely regulated by associated proteins and that vary from one place to another in the cell according to the different motile functions in which actin is involved. The changes in the

more »

... ge length of filaments will cause changes in the viscoelastic properties of the cytoplasm. It is therefore important to understand the unique structural properties of the actin molecule that are at the origin of its ability to polymerize and to elucidate the mechanism of polymerization itself in order to anticipate possible modes of regulation. The hydrolysis of actin-bound ATP, which is linked to filament assembly, was early shown to complicate the classical thermodynamics of actin polymerization developed by Oosawa (2) and to introduce the possibility of subunit flux in the filament, due to different critical concentrations at the two ends (3). In the past decade, more progress has been accomplished. We now understand better how ATP hydrolysis affects the dynamics of actin filaments; we also know more about the binding of ATP, the structure of the ATP binding site, and the mechanism of ATP hydrolysis. Very recently, the structure of the actin monomer has been solved at atomic resolution which provides new bases for the approach of structure-function relationship in actin and for analyzing its interaction with actin binding proteins. The present paper will focus on these new points and update previous reviews on the same subject (4-7). Other reviews cover the related fields of actin binding proteins (5,8), actomyosin interaction in muscle contraction (9, 10) and in nonmuscle cells (11, 12), actin ADP-ribosylation by Clostridium toxins (13), and the cellular control of cortical actin assembly by external stimuli (14). Actin Filament Structure The actin filament can be described by either a one-start left-handed genetic helix of 5.9 nm pitch or by a two-start right-handed helix of 72 nm pitch (15). Torsional motion of actin subunits in the filament has been reported (16, 17) . Accordingly disorder appears on images of actin filaments in the electron microscope. The variable twist can be accounted for by a model of cumulative angular disorder (18) that appears to be modulated by actin binding proteins (19). Reconstructions of actin filaments from images of negatively stained or frozen hydrated specimens lead to a structure in which the strongest actin-actin contacts are along the small pitch genetic helix (18, 19) . Other works, however, offer the contrasting view that bonds along the long pitch two-strand helix are the strongest, as suggested by the occasional separation of the two strands (20). ATP and Divalent Cation Binding to Actin G-actin binds tightly one ATP and one divalent metal ion that can be Ca2+ or Mg2+ and is thought to be M e in the cell. The affinity of this metal ion was recently shown to be in the IO9 M" range (21-23). The dissociation of tightly bound Ca2+ is slow (24) and rate-limiting in the overall process of nucleotide exchange (25). Actin also binds mono-or divalent cations (Mg2+, Ca", K+ . . .) to a series of lower affinity sites with affinities in the 105-104 M-' range (23, 26). How binding of cations to these different sites affects the conformation of actin (27), the binding of ATP, and the polymerization properties has been a subject of intense debate in recent years. Both kinetic and physical measurements of metal ion and nucleotide interactions with G-actin failed to provide a consensus answer to the issue of the proximity of tightly bound metal ion and ATP on actin (see Refs. 25 and 30 for review). The picture emerging now is that rapid binding of either Mg2+, Ca", or K+ to low affinity sites represents the "monomer activation step," i.e. the preliminary step in actin polymerization (23, 28, 29); on the other hand, slow M e exchange for Ca2+ at the tight binding site is associated with a different conformational change of the protein (23). Recently use of P-y Cr-ATP, an exchange inert analog of Mg-ATP, led to the conclusion that the tightly bound divalent metal ion directly interacts with the Pand y-phosphates of ATP on actin and that metal-ATP is bound in the A configuration (30). These last two results are now confirmed by the recent three-dimensional structure of actin (31). The actin ligand therefore is the metal-ATP complex as described by the following scheme, \ \ 1 A + M + N A -M -N A -N + M where A is G-actin, M the tightly bound metal ion (Ca2+ or Mg2+), and N the nucleotide. The reported 3-5-fold higher affinity of Ca2+ versus M g + for actin means that Ca-ATP binds 3-5-fold more tightly than Mg-ATP. Examination of the rate parameters for the binding of ATP, metal ion, and metal-ATP (Table I) shows that nucleotide exchange follows different routes according to whether Ca-ATP or Mg-ATP is bound. Ca-ATP dissociates at a much slower rate than via the consecutive dissociation of Ca2+ ion and nucleotide, which accounts for the dependence of the rate of ATP dissociation on Ca2+ ion (25); in contrast, the rate constant for dissociation of Mg-ATP is practically as slow as Mg2+ dissociation. Therefore, the fast nucleotide exchange observed on Mg-ATP-actin implies that ATP can dissociate from actin with Mg2+ remaining bound to actin. In conclusion actin can be either a Ca-or a Mg-ATPase, which we will see affects in turn the dynamics of actin filaments. Three-dimensional Structure of Actin Crystals were obtained from the 1:l complexes of actin with profilin (32) and DNase I (31), two relatively small proteins that prevent actin from polymerizing and whose three-dimensional structures are already known. Crystals of profilin-actin (obtained in the absence of divalent metal ion) show a flat zigzagging ribbon structure similar to a flattened filament 1