Control of Actin Dynamics in Cell Motility ROLE OF ADF/COFILIN*

A large number of cellular processes, including cytokinesis, en- docytosis, chemotaxis, or neurite outgrowth, is mediated by polymerization of actin filaments. In response to extracellular stimuli, motile protrusions of the plasma membrane, in the form of lamellipodia or filopodia, are driven by the continuous initiation, polar- ized growth, and turnover of actin filaments at the leading edge of the cell.Recent progress has been made in identifying the key players responsible for the spatio-temporal control of actin-based motility and in understanding the molecular mechanism supporting their function. The linkage of the actin cytoskeleton to the signaling pathway is generally mediated by the interaction of a small G-protein (Rac, Cdc42) in its active, GTP-bound form, with a multi- partner “connector” at the plasma membrane. The activation of the connector allows it to recruit the Arp2/3 complex, which initiates branched barbed end growth of actin filaments (see the first minireview in this series (60)) forming the dense reticulated actin network seen in high resolution electron microscopy images of lamellipodia (1–3). Pathogens like Listeria or Shigella or the vaccinia virus, which propel themselves through the cytoplasm by polymerizing actin at their surface (4) , harness the cytoskeletal machinery downstream of the signaling pathway in a constitutive fashion and provide a biochemical approach of the mechanism of actin-based motility. Once actin polymerization is initiated, continuous filament growth causes membrane protrusion (or bacterium propulsion) at a rate of 1–25 m m/min. Growing filaments remain stationary with respect to the substratum (5), demonstrating that actin polymerization

centration for barbed end assembly during movement, so as to sustain rapid steady filament growth. Recent works have shown that actin-binding proteins of the actin-depolymerizing factor (ADF) 1 /cofilin family are the regulatory factors that elicit the rapid turnover-barbed end growth of actin filaments driving the forward movement of the leading edge.
In Vivo Properties of ADF/Cofilin ADF/cofilins have been recognized early as a family of essential, conserved, widespread, small (15)(16)(17)(18) actin-binding proteins (see Ref. 9 for review) playing an important role in cytokinesis (10), endocytosis, and in the development of all embryonic tissues (11), as well as in pathological situations such as ischemia, oxidative or osmotic stresses, and tissue regeneration (12). Genetic studies point out the importance of ADF/cofilin as regulators of actin dynamics in movement and morphogenesis. In budding yeast, temperature-sensitive mutants of cofilin that fail to rapidly depolymerize actin show defects in endocytosis (13). Overexpression of cofilin in Dictyostelium discoideum induces the formation of membrane ruffles and increases the motility of the amoeba (14). Mutations in Drosophila twinstar gene that lower the expression level of ADF lead to defects in cytokinesis and aster migration and to the assembly of misshaped actin structures at the site of formation of the contractile ring (15). Missense mutations in Caenorhabditis elegans unc-60B gene encoding for an isoform of ADF lead to improper actin assembly in myofibrils (16). In plants, ADF controls the pathways of actin-based cytoplasmic streaming (17) and pollen tube growth (18). ADF/cofilins are expressed at high levels in embryonic heart and skeletal muscle, whereas lower amounts of the protein are present in adult tissue (11,19).
Remarkably, the stimulus-responsive function of ADF/cofilin (except for yeast cofilin) is regulated by phosphorylation of a single conserved serine identified in the N-terminal region (20,21). The activity of ADF is induced by dephosphorylation, which occurs rapidly in response to various stimuli known to promote the reorganization of the actin cytoskeleton, such as growth factors (nerve growth factor, insulin), chemotactic peptides, or agents increasing the levels of [Ca 2ϩ ] i and cAMP (9) (for review of works before 1996 see Refs. 22 and 23). Although ADF/cofilin appears diffusely distributed in the cytoplasm of quiescent cells, the activated (dephosphorylated) protein translocates to regions of the cells where actin filaments are highly dynamic, like the leading edge of ruffled membranes, the cleavage furrow of dividing cells, or the neuronal growth cone (11,24,25). Dephosphorylation correlates with increased motility and extension of cellular processes. The phosphorylated and dephosphorylated forms of ADF always coexist in a cell, but the ratio between the two forms varies. Moreover, phosphate turnover is recorded on ADF, indicating that the phosphorylation level is continuously tuned at a rate which is also dependent on signaling (22). The regulation of the phosphorylation level is not fully understood yet and must be complex, because phosphatase inhibitors like okadaic acid or calyculin A lead to dephosphorylation (activation) of ADF. The phosphatase responsible for activation of ADF is unknown. On the other hand, LIM-kinase 1 (whose deletion is linked to cognitive defects associated with Williams syndrome) has been identified as the kinase that phosphorylates and inactivates ADF in vivo (see Refs. 26 -28 for review). LIM-kinase 1 is a downstream effector of the small G-protein Rac, which controls lamellipodium formation. Not only is LIM-kinase 1 activated by Rac, but expression of dominant-negative mutated LIM-kinase 1 blocks Rac-induced formation of lamellipodia. These results are surprising, at face value, because the formation of lamellipodium induced by Rac implies that ADF be activated, i.e. dephosphorylated. The real message from these experiments may be that rapid phosphorylation-dephosphorylation cycles of ADF/ cofilin (hence increased phosphatase activity as well as LIM-kinase activity) are associated with Rac stimulation. In plants, a Ca 2ϩ -dependent kinase phosphorylates ADF on serine 6 (29).

Structure of ADF/Cofilin
Sequence data base searches have allowed identification of a large family of actin-binding proteins containing an ADF homology module (30). The ADF/cofilins, the twinfilins, and the drebrin/ Abp1p proteins represent three phylogenetically distinct subclasses of this family, all three coexisting before the divergence of yeast and animals. Twinfilins contain two repeats of the ADF homology domain, whereas drebrin has an additional C-terminal SH3 domain. Although twinfilins and drebrin have been identified as actin-binding proteins, their functions differ from the better characterized ADF/cofilin. ADF/cofilins from different organisms present a high degree of sequence homology, especially in the N-terminal and C-terminal regions. Many organisms contain more than one variant (except for yeast and amoeba, which have only one cofilin). Although different variants display discrete quantitative functional differences (see next section), the general mechanism of action of the different ADF/cofilins is conserved. Accordingly, the three-dimensional structures of vertebrate ADF (destrin) (31), yeast cofilin (32), and Acanthamoeba cofilin (actophorin) (33), obtained either by NMR or x-ray crystallography, appear very similar. The ADF fold (Fig. 1A) is similar to the fold of either segment 1 or segment 2 of gelsolin or villin, with five central ␤-sheets flanked by three to four ␣-helices. No sequence homology, however, can be found between ADFs and proteins of the gelsolin family, and their functions and biochemical properties are clearly different. The three-dimensional structural similarity between ADF and gelsolin led Hatanaka et al. (31) to propose that ADF and gelsolin segment 1 might interact with actin in the same manner. This hypothesis was upset by mutagenesis studies, which point to a different interface and define regions of the molecule interacting with both G-and F-actin, whereas others interact with F-actin only (34). Further insight into the ADF/actin interface is brought by the electron micrograph observation of ADF-decorated filaments. Image reconstruction (35) shows that ADF interacts with two actin subunits along the long pitch helix, bridging subdomain 1 of one of the actins to subdomain 2 of the second subunit (Fig. 1B). The binding sites of ADF and of gelsolin segment 2 on F-actin show some, but not complete, overlap. The most striking feature of ADF binding to F-actin is the associated massive structural change of the filament, which is visualized by an increase in twist of 5°per subunit (for human ADF) with no change in the axial rise (2.7 nm) or with the radial position of the actin subdomains. As a result, the long pitch helices crossover every 27 nm on average instead of every 36 nm for standard F-actin filaments. The change in twist varies somewhat with the ADF species and may be in relation with differences in thermodynamic stability of the filament decorated by different ADFs. The ADF/ cofilin is unique, among all other F-actin-binding proteins, in inducing this change (see Ref. 36 for review), which indicates that despite the overlap in the footprints of gelsolin segment 1, segment 2, and ADF on F-actin, the interface of ADF with F-actin must be different from that of gelsolin. The observed competition between ADF and gelsolin segment 2 (37) or myosin subfragment 1 (38) for binding to F-actin then should be understood in terms of different structures of the filament linked to the binding of different proteins to actin, rather than in terms of a classical competition for the same site. This view accounts for the different functions of all these proteins. Similarly, ADF does not bind to phalloidin-decorated filaments or to F-ADP-P i -actin filaments (39 -41), which are structurally and mechanically different from F-ADP-actin filaments. Accordingly, ADF accelerates P i release on F-actin (42). Hence ADF reveals the structural variability of the actin filament (43) and may use it to specify functionally different filaments.
In the absence of a high resolution structure of the ADF-G-actin complex, molecular dynamics have been used to propose a structural model, showing some similarity to the actin-gelsolin segment 1 complex (44). How the structure of this complex can be compatible with the reconstruction of ADF-decorated filaments is unclear. Future progress in orienting the cofilin molecule along the filament using appropriately located gold labels and taking into account the mutagenesis studies should help in deriving an atomic model of ADF-F-actin.

ADF/Cofilin Increases the Turnover of Actin Filaments, Which Powers Actin-based Motility
The in vitro biochemical properties of ADF and its effects on actin assembly-disassembly have shed light on its function in actinbased motility and morphogenesis. Early works had shown that ADF caused depolymerization of filaments and had a weak severing activity (see Refs. 9 and 45 for review), both effects being more pronounced at high pH. Further studies showed that ADF elicits the partial depolymerization of F-actin, that is promotes the establishment of a new steady state of assembly, in which a higher concentration of monomeric actin is maintained (45). The underlying mechanism is as follows. First, under physiological ionic conditions, all ADF/cofilins recognize the ADP-bound form of both Gand F-actin with a high specificity, 2 i.e. with a 100-fold higher 2 Although no direct evidence exists for different structures of actin depending on bound nucleotide, it is remarkable that a number of actin-binding proteins display selective binding to either ATP-or ADP-actin. Thymosin ␤4 and profilin interact with ATP-G-actin with 50-and 20-fold higher affinity, respectively, than with ADP-actin, whereas gelsolin binds ADP-actin preferentially.
FIG. 2. Enhancement of actin turnover by ADF. This scheme summarizes important properties of ADF involved in its function. 1) ADF binds cooperatively to F-actin, which results in the coexistence of two populations of filaments: bare filaments and ADF-decorated filaments, which have different dynamic properties. 2) ADF-decorated filaments depolymerize 30-fold more rapidly from their pointed ends than bare filaments, leading to steadystate accumulation of ADF-ADP-G-actin, ATP-G-actin, and a higher rate of barbed end assembly. Minireview: Control of Actin Dynamics in Cell Motility 33828 affinity than ATP-or ADP-P i -bound actin (40,42,45,46). Second, the ability of ADF to interact with both G-and F-ADP-actin, with a slight preference for G-actin, allows it to participate in the assembly of ATP-actin, making use of the associated hydrolysis of ATP. ADF acts at two important kinetic steps of the ATPase cycle in actin assembly, shown in Fig. 2. 1) The rate of dissociation of ADF-F-actin from the pointed ends is 30-fold higher than the rate of dissociation of F-actin (45,47); 2) the dissociation of ADP from the depolymerized ADF-G-actin is 10 -20-fold slower than the dissociation of ADP from G-actin (45,46,48). The combination of these two properties affects actin dynamics at steady state as follows. Because depolymerization from the pointed ends is the rate-limiting step in the treadmilling cycle, the addition of ADF to pure F-actin greatly accelerates filament turnover, up to values comparable with those observed in vivo in lamellipodia. The following changes in the concentrations of monomeric actin species are associated with the faster turnover. The rapid disassembly flux of ADF-F-actin into ADF-ADP-G-actin complex readily leads to a proportional increase in the production of ADP-G-actin because ADF is in rapid equilibrium (k ϩ ϭ 250 M Ϫ1 ⅐s Ϫ1 ; k Ϫ ϭ 20 s Ϫ1 at 4°C) with ADP-G-actin (49) . Because the value of C C B is slightly lower than 0.1 M, barbed end growth is very slow in the absence of ADF and 30-fold faster in the presence of ADF. Hence ADF appears responsible for the fast rate of individual barbed end assembly, which supports actin-based motility. Indirectly, by increasing the concentration of ATP-G-actin, ADF contributes to enhance nucleation of filaments by the Arp2/3 complex. This effect is greatest in the presence of capping proteins. ADF then increases the concentration of ATP-G-actin up to 1 M, and rapid filament turnover is measured in solutions of F-actin containing Arp2/3 and capping proteins (51), conditions that are found in motile cellular extensions (52). The role of ADF in motility of Listeria in acellular extracts has also been observed (see Refs. 45, 53, and 54 for review). Genetic studies in yeast (13) confirm that the enhancement of filament turnover is the physiological function of ADF in vivo. The recent successful reconstitution of actin-based motility of Listeria and Shigella from pure proteins comprising actin, Arp2/3 complex, ADF, and capping protein as essential components (61) is supportive of the model of biased treadmilling that was put forward for actin-based motility (55).
The partial depolymerization of actin induced by ADF is very different from a sequestering effect. A sequestering protein binds G-actin specifically, depolymerizes actin in a fashion linearly dependent on its concentration, and does not affect the turnover of filaments. Depolymerizing actin, like a sequestering factor would do, would fail to account for the stimulating effect of ADF in motility. In summary, the partial depolymerization of F-actin is the manifest consequence of the biological function of ADF, which is to enhance actin dynamics, not to depolymerize actin. The name "actin-depolymerizing factor," derived from early biochemical studies, misleadingly describes the actual function of ADF.
Because of the slow nucleotide dissociation from ADF-actin complex, the major monomeric actin species when F-actin is assembled at steady state in the presence of ADF is not ATP-G-actin but ADF-ADP-G-actin. The pool of ADF-ADP-G-actin represents 1-10 M actin, depending on the ADF species and on the pH. It is easy to appreciate, from the scheme presented in Fig. 2, that the size of this pool can be modulated by changing the rates at which ADF-ADP-G-actin is produced (dissociation from the pointed end) and at which it is consumed (via either ADP dissociation from the complex or dissociation of ADF from ADP-actin). The differences in the steady-state concentrations of ADF-ADP-G-actin observed between different ADFs or between wild-type and mutated ADFs (16,45,47,56,57) are most likely because of differences in the rate constants for those reactions, and the net physiologically relevant effect is a regulation of filament turnover. The effect of pH is particularly interesting. ADF from diverse sources causes a more extensive depolymerization of actin at steady state at pH 8.0 than at pH 7.0 (9,45,49,58). This effect is more or less extensive from one ADF species to the other. The rate of nucleotide exchange on G-actin is known to increase by 1 order of magnitude upon increasing pH. Therefore an increase in pH causes a faster recycling of ADP-Gactin into polymerizable ATP-G-actin. The fact that, despite this faster recycling, a higher amount of depolymerized actin is maintained at high pH in the presence of ADF indicates that the rate of depolymerization of ADF-F-actin from the pointed ends increases upon increasing pH. Consistently, the enhancement of filament turnover by human ADF is 3-fold greater at pH 8.0 than at pH 7.0. 3 The pH dependence of ADF function may be physiologically relevant, in particular in plants, where steep pH gradients are observed at sites of cell growth (e.g. pollen tube) where active actin dynamics are thought to take place.
Interestingly, profilin, another G-actin-binding protein known to play a positive role in actin-based motility, acts in synergy with ADF to further enhance filament turnover. This effect, which was theoretically anticipated (46,55), was experimentally demonstrated (50). The functional properties of profilin complement those of ADF in the treadmilling cycle. By accelerating nucleotide exchange on G-actin, profilin recycles ADP-G-actin into profilin-ATP-G-actin complex, which actively participates in barbed end assembly. Profilin thus enhances the processivity of treadmilling, lowers the pool of ADF-ADP-G-actin, and accelerates actin-based motility in the motility medium fully reconstituted from pure proteins (61).
Analysis of the kinetics of binding of ADF to actin has helped to understand the molecular mechanism of enhancement of filament turnover. As mentioned above, ADF interaction with G-actin is a simple, rapid, reversible bimolecular reaction. Binding to F-actin is more complex. The time courses show a high degree of kinetic cooperativity with a lag followed by an acceleration (42,49), indicating that ADF nucleates a local structural change of the filament which propagates along the polymer, consistent with the change in twist. In addition, ADF dissociates slowly from F-actin. This binding behavior has important implications in the function of ADF. At substoichiometric ratios of ADF to F-actin, ADF does not statistically partially saturate all filaments but fully saturates a small number of filaments, thus generating two populations of energetically different filaments. The ADF-decorated filaments rapidly lose subunits from their pointed ends, whereas the bare filaments actively incorporate actin subunits at their barbed ends. In other words, the treadmilling process takes place not only from one type of end to the other but from one filament type to the other type (51). The result is a fiber-by-fiber renewal of the whole population of filaments by substoichiometric amounts of ADF. The efficiency of turnover is expected to be optimum when the two pools of bare and ADF-decorated filaments are equal, which is in fact observed (45). Interestingly, the dynamic behavior of F-actin in the presence of ADF then becomes very similar to the dynamic instability of microtubules, which also results, albeit through a different molecular mechanism, in a fiber-by-fiber renewal and organization of the meshwork. We suggest that this mechanism of action of ADF may have implications in the control of the morphogenetic organization of actin filaments in neurite extension, axon guidance, muscle fiber assembly in myoblasts, and other developmental processes in which the spatial reorganization of actin filaments is involved. The bell-shaped curve of the ADF concentration dependence of filament turnover observed in vitro indicates that the level of active ADF has to be finely tuned in vivo for maximum efficiency. Consistently, in C. elegans, ADF mutations leading to increased activity of ADF result in defects that are similar to those induced by a lower activity (16). The level of active ADF is controlled by phosphorylation in a simple fashion. The affinity of ADF for G-and F-actin is decreased 20-fold by the serine to aspartate mutation, which mimics phosphorylation (49). Hence phosphorylation of ADF is equivalent to a decrease in the amount of endogenous active protein, without any change in activity per se.

Lowered Thermodynamic Stability of ADF-bound Filaments: Relation with Structure and Length Distribution
The binding of ADF to ADP-G-actin and ADP-F-actin implies that ADF-ADP-actin polymerizes reversibly. Detailed balance implies that the critical concentration for polymerization of ADF-ADP-actin is n-fold higher than the critical concentration for assembly of ADP-actin because ADF binds with a n-fold higher affinity to ADP-G-actin than to ADP-F-actin. As n varies from one ADF species to the other, the stability of the ADF-decorated filaments varies accordingly. For instance, the critical concentration for assembly of ADP-actin is increased 2.5-fold by Arabidopsis thaliana ADF1 and 6-fold by vertebrate ADF (49). The change in structure of the filament associated with ADF binding is expected to reflect the change in thermodynamic stability. The lower thermodynamic stability of ADF-decorated actin filaments, for which structural evidence has recently been provided (62), is also expected to correlate with a change in length distribution because the average length is determined by the thermodynamic properties of assembly. In vitro, the distribution in length is controlled by either one of the two pathways, depolymerization-nucleation-elongation on the one hand and fragmentation-reannealing on the other hand, which are kinetically different but thermodynamically equivalent. The establishment of the steady state length distribution is very slow for pure actin but much faster in the presence of ADF because of the enhanced actin dynamics. The destabilization of filaments by ADF (increase in critical concentration) has to be accompanied by a decrease in average length. ADF had in fact been considered early as a weak severing factor (see Ref. 9 for review), and this view is often offered as the easiest interpretation of kinetic data (42,58). More recent detailed studies of the decrease in average length induced by ADF show that under optimum conditions the change in length is modest and cannot in itself account for the large increase in turnover ADF (42,50,51). At the physiological ADF:actin ratios of 1:10, the measured change in length is very small. The "severing" activity of ADF was also postulated to account for its effect on motility. However, because ADF does not cap one of the ends of the filaments like gelsolin does, a simple severing activity would generate as many polymerizing barbed ends as depolymerizing pointed ends at steady state. The new barbed ends would grow while the new pointed ends would depolymerize, the net rate of barbed end growth per filament being unchanged. A severing activity thus cannot be effective to enhance actin-based motility, which requires a change in the intrinsic kinetic parameters. Hence severing cannot account for the function of ADF. A recent localization study of Arp2/3 and ADF in motile cells (3) shows that ADF is at the rear of lamellipodia but excluded from the narrow zone adjacent to the leading edge where filaments are nucleated, which rules out the possibility that these new ends are created by a severing action of ADF. In conclusion, the severing effect is not a physiological function of ADF but a consequence of its effect on actin dynamics.

Conclusion and Perspectives
Biochemical studies have helped to understand the physiological role of ADF/cofilin in motility and morphogenesis. Many issues remain open for future investigations. Elucidating the regulatory pathways involved in phosphorylation/dephosphorylation of ADF in relation to signaling clearly is a major challenge. Whether the localization of dephosphorylated ADF in motile regions of the cell is linked to local pH changes or to interaction with other factors like the recently described Aip1 protein, which seems to potentiate ADF binding (59), is still unknown. Combined structural and mutagenesis studies should define the interfaces of G-and F-actin with ADF, help understand the structural basis for the functional variability of different ADFs, and provide insight into the different structural states of the actin filament.