I. INTRODUCTION

Top Previous Next References

Actin microfilaments provide the "rails" along which myosin "motors" perform work in a variety of cellular functions. A major review of myosin motors (263) has been published, and they are excluded from this review. Microfilaments cooperate with microtubules via microtubule-associated proteins (MAPs) during the transport of vesicles and organelles, and this interesting aspect of microfilament function was recently reviewed (34). Actin filaments also interact with intermediate filaments, a function that may play an important role in enabling extracellular stimuli to be transmitted to key targets like ribosomes and chromosomes deep within the cell. This field is an emerging one. It will be covered by Quinlan et al. (235) and is beyond the scope of this review.

	II. STRUCTURE AND DYNAMICS OF MONOMERIC ACTIN

Top Previous Next References

Actin is only found in eukaryotes. It comprises a highly conserved family of proteins that fall into three broad classes: -, -, and -isoforms. It is mainly located in the cytoplasm, but it is also present in the nucleus where it may or may not have motor-associated functions. The highest concentrations (~20% of the total protein) of actin are in striated muscles; however, significant quantities of actin are present in nonmuscle cells where it plays a variety of roles including myosin-independent changes of cells shape, motor-based organelle transport, regulation of ion transport, and receptor-mediated responses of the cell to external signals.

Few people in the field of actin remember that Christine Oriol-Audit was the first to report the crystallization of actin in the absence of an actin-binding protein (212).¹ These crystals were achieved in the presence of polyethylene glycol but were too small to be useful for X-ray diffraction at that time. If today's high-powered X-ray beam facilities were available then, they may well have yielded the atomic structure of actin in the absence of an ABP. Perhaps it was a cruel twist of fate that she sadly died early in 2001 without knowing that the first structure of uncomplexed actin was soon to be published (216). The principal obstacle to crystallizing actin was its propensity to spontaneously self-assemble under solvent conditions conventionally used to grow protein crystals.

In their pioneering book on the biophysics of protein polymerization, Fumio Oosawa and Sho Asakura (210) predicted that actin, like tubulin, would form helically tubular crystalline assemblies. We subsequently showed that these structures could be formed in the presence of the trivalent lanthanide ions, particularly Gd³⁺ (3, 87). These lanthanide-induced actin tubes (composed of a single layer of monomers) and microcrystals (bilayered sheets) provided the first views of the shape of the actin monomer (3, 73). The computed average image of the monomer was "pear" shaped, having a "large" and a "small" domain. Electron diffraction studies revealed structure out to at least 14.5-Å resolution, but this was simply not good enough to provide reliable molecular details (74). In solution, actin filaments can form supramolecular assemblies called paracrystals (99) as well as forming liquid crystalline arrays (67).

Although the cocrystallization of actin with bovine pancreatic DNase I was first reported in the late 1970s (176), 13 years passed before the first atomic resolution (2.4 Å) structure of actin was reported (137). The actin monomers used in these crystals were lightly digested with trypsin to remove the COOH-terminal three residues before crystallization. Because the crystal structure of DNase I had already been determined at 2.5-Å resolution (272), it was a relatively simple task to subtract its structure from the actin-DNase I complex to obtain a clear view of the actin monomer. The resulting structure (illustrated in ribbon form in Fig. 1A) produced a number of surprises. For example, the large and small domains described at low resolution contained nearly equal numbers of residues and were nearly equal in size. Both the NH₂ and COOH termini were located in the same subdomain. DNase I makes crystal contacts across the "top" of the nucleotide cleft, thus explaining how it inhibits nucleotide exchange. However, since these were cocrystals, did this structure represent a monomer conformation when it was not complexed to DNase I?

View larger version (42K): [in this window] [in a new window]   Fig. 1. A: actin monomer ribbon structure of G-actin from DNase I cocrystals at 2.8-Å resolution (137). The numbered amino acid residues were chosen because of their relevance to this review. B: actin monomer ribbon structure of uncomplexed rabbit skeletal muscle actin at 1.54-Å resolution (217).

Monomeric (G-)actin has dimensions of ~67 × 40 × 37 Å and a molecular mass of nearly 43,000 Da. The overall features of the structure include four quasi-subdomains (here they are numbered 1-4 but an alternative nomenclature of IA, IB, IIA, and IIB has an historical basis, Ref. 266), each having a repeating motif comprising a multi-stranded -sheet, a -meander, and a right-handed -unit. About 40% of the structure is -helical. In Figure 1A, the monomer is oriented so that the top has about the same orientation it would have in the filament with its "pointed" end directed at the top of the page and the "barbed" end toward the bottom.

A tightly bound nucleotide lies in a deep cleft in the center of G-actin. This site is usually occupied by ATP or ADP-P_i rather than ADP, which binds with ~10-fold lower affinity to Mg-G-actin and ~200-fold weaker binding with Ca-G-actin (147). ATP binds as a complex with either Mg²⁺ [dissociation constant (K_d) 1.2 nM] or Ca²⁺ (K_d 0.12 nM) (79, 270). Mg²⁺, the dominant cation in vivo, determines how tightly or weakly the nucleotide binds. Six or seven relatively low-affinity divalent cation-binding sites have been identified that appear to be important for actin paracrystal (94, 271), and actin microcrystal formation (73), but their physiological importance is not well understood. A notable feature of the structure is a two-stranded "hinge" at residues 140 and 338 joining the large and small domains. Tirion et al. (283) describe a "propeller" motion between the two domains that produces an opening and closing of the nucleotide cleft. This putative slewing motion is permitted because of this hinge and is viewed as a rigid body movement between the subdomains.

There are several other atomic structures of actin. In 1993, McLaughlin's group at the Medical Research Council (MRC) in Cambridge (186) reported the structure of actin complexed with gelsolin segment 1. This segment binds to the bottom of the monomer, making contacts with subdomains 1 and 3 (see discussion below). Unlike the DNase I-actin structure, the COOH-terminal three residues were not cleaved before crystallization. However, the DNase I binding loop was not well resolved, presumably because it was not sufficiently stabilized in the absence of DNase I. In the same year Schutt et al. (259) reported the structure of spleen actin (a -isoform) complexed to profilin at a resolution of 2.55 ÅA. This group had been refining their structure for many years and demonstrated that, although there was broad similarity to the structure of McLaughlin et al. (186), there were significant differences that they attributed to the sequence differences between the actin isoforms.

The Schutt/Lindberg consortium that captured actin in a different conformation subsequently reported a fourth actin-profilin structure. They again used bovine -actin cocrystallized with bovine pancreatic profilin (60). They named the two structures the open and closed states and used them to argue that the transition between the two states represented a structural change that could convert their "ribbon" polymer into conventional F-actin.

In 2001, the inevitable happened. Dominguez et al. at the Boston Biomedical Research Institute succeeded in producing small, high-quality crystals of actin that were not complexed to ABPs (Fig. 1B). This finding was reported at the Boston meeting of the Biophysical Society colleagues (216) and was subsequently published in Science (217). They achieved the result that had eluded so many others by modifying the COOH-terminal cysteinyl (Cys-374) with a rhodamine label that suppressed the tendency of actin to polymerize.

Dominguez and colleagues (217) were also the first to crystallize actin with ADP present in the nucleotide-binding cleft. This was important because we know that the critical concentration of ADP-G-actin is about an order of magnitude higher than for ATP or ADP-P_i-G-actin (241). Kabsch et al. (137) had also produced crystals of ADP-G-actin and showed that the ADP structure was not remarkably different from the ATP form. However, the Kabsch crystals were formed in the presence of ATP, and the nucleotide was cleaved to ADP subsequent to crystallization.

The structure of Dominguez et al. significantly differs from the structure of Kabsch et al., particularly in subdomain 2 where DNase I binds. However, it could be argued that this Boston/Heidelberg difference is because ADP was present before the crystals were formed, i.e., before the crystal contact points were established.

A significant structural difference actin crystallized alone and complexed with an ABP protein is relatively large coil-to--helix conversion and a 10° rotation at the top of subdomain 2. Although it is not clear if these differences represent native cellular G-actin, Dominguez et al. are currently completing the analysis of an uncomplexed ATP-G-actin structure that should satisfy their critics.

The substantial conformational change reported by Otterbein et al. (217) is well supported by other laboratories including our own. Dedova et al. (81) recently demonstrated, using fluorescence spectroscopy, that significant (1-4 Å) changes occur in subdomains 1 and 2 when cofilin and DNase I bind either separately or as a ternary complex. Several other authors have suggested that actin can undergo allosteric conformational changes, and this is discussed in more detail in section IV. The major challenge for structural biologists now lies in relating the structural changes to functional changes.

Actin is an essential and ubiquitous cytoskeletal component of all eukaryotic cells and is absent from prokaryotic cells. However, although it has long been suspected that the evolutionary origins of actin lie in their prokaryotic ancestors, until 2001 no one had reported a credible candidate. The first suggestion that MreB may be an ancestral actin gene was published 10 years ago (30), and recently (136), it was localized in Bacillus to form distinct filaments located close to the cell surface. The atomic structure of these filaments was more recently reported by Fusinita van den Ent and colleagues (289) at the MRC, Cambridge, who provided convincing evidence that MreB is indeed an ancestral actin gene. This actin look-alike is present in all nonspherical bacteria.

Amino acid sequence homology to actin is limited to 15% (289), and although its overall size and shape strongly resemble actin, there are differences. The sequence corresponding to the actin DNase I-binding loop in subdomain 2 is larger, and the COOH terminus, located in subdomain 1, is 20 residues longer. A 2.1-Å resolution crystal structure reveals that, like actin, MreB has two major domains separated by a nucleotide-binding cleft. Also, like actin, each domain has two subdomains with essentially the same topology. The structure of MreB is shown in Figure 2.

View larger version (66K): [in this window] [in a new window]   Fig. 2. Crystal structure of MreB showing the four-domain motif seen in the actin monomer structure (136). The NH2 and COOH termini are located in subdomain 1A (equivalent to subdomain 1 of actin) (F. van den Ent, personal communication).

MreB can self-assemble into 51-Å-wide filaments that are about half the diameter of actin microfilaments, but unlike F-actin, they are linear (nonhelical) polymers resembling the linear polymers seen in actin crystalline sheets first reported by ourselves (87) and later by others (3). The axial repeat for MreB microfilaments is 51.1 Å, somewhat less than the axial repeat of F-actin (55 Å) (Fig. 3). These polymers form spirals beneath the bacterial cell wall, and although their precise function remains elusive, it is likely that MreB filaments behave like microfilament analogs and control the shape of bacteria. However, we know nothing of how assembly of this structure is regulated, or whether it binds to motor proteins. Figure 3 illustrates the striking similarities between the crystal structure of protofilaments of actin (right) and MreB (left). Nearly identical molecular orientations and contacts are seen between monomers for the two proteins. Conversion of a pair of actin protofilaments into F-actin is achieved by simply twisting a pair of actin protofilaments to conform to the actin filament symmetry (see below).

View larger version (42K): [in this window] [in a new window]   Fig. 3. A comparison of crystal structure of MreB (left) with a model of actin protofilaments (right) derived from linear polymers along a single strand of F-actin. Five monomers are illustrated in both protofilaments. Although the axial repeats between monomers differ, the molecular contacts between monomers in protofilaments of MreB are essentially the same as the contact between actin monomers along the two strands of F-actin (van den Ent, personal communication).

The field has made substantial progress in understanding the structure of F-actin. Nearly 40 years ago Jean Hanson and Jack Lowy were the first to describe the helical nature of actin filaments (116). In this model, F-actin can be viewed either as a single-start, left-handed tightly wound helix of monomers (the so-called genetic helix) or as a two-start, right-handed long-pitch helix. The conventional view of F-actin is the two-start helix because it is probable that the monomer-monomer affinity is stronger along the two long-pitch strands than between those strands (128).

No atomic structure has yet been determined for F-actin, although highly plausible models have been proposed based on the model of Holmes et al. (128). Figure 4 illustrates five monomers in a filament. Viewed along the genetic helix, monomers rotate by 166° and have an axial translation of 27.5 Å. There are 13 monomers in 6 turns with a pitch of 59 Å yielding a filament diameter of ~90-100 Å (128). Contacts between the monomers along this helix occur across the diameter of the filament. The two-start long-pitch helix is right-handed with a somewhat variable half pitch of 360-390 Å comprising 12-14 monomers per half turn. In this model each monomer is still surrounded by four others. The center-to-center distance between monomers along this helix is ~55 Å. These filaments have a distinct structural polarity that was first noticed when the filaments were "decorated" with myosin fragments, but it can also be seen in good images of F-actin.

View larger version (49K): [in this window] [in a new window]   Fig. 4. The Holmes model of F-actin showing six monomers, three in each of the two-start long-pitch helix originally described by Holmes et al. (128).

Thus, despite the passage of nearly 40 years and the best efforts of laboratories in Germany, the United Kingdom, the United States, Japan, and elsewhere, the atomic structure of F-actin remains elusive. A major part of the problem is the inherent disorder in linear aggregations of filaments. Even if they could be aligned with the same polarity, it is difficult to pack F-actin complexed to phalloidin (a 7-residue peptide that dramatically stabilizes actin filaments) so there are precise contacts between filaments. Devices like the innovative trumpet-shaped quartz tubes used by David Popp et al. (233) achieved greatly improved filament alignment that yielded the first low-angle X-ray diffraction data. More recently, Yuichiro Maeda (who also worked in Heidelberg and has since returned to the RIKEN SPring8 Institute synchrotron facility in Japan) and colleagues achieved even better alignment of F-actin without the help of phalloidin using very high (13.5 Tesla) magnetic fields. They already have structural data out to 5-Å resolution (207), and we can expect them to report further progress in the near future.

In the meantime, we must content ourselves with the available models of F-actin based on data from low-resolution electron microscope images. The original model (128) was created by fitting the atomic structure of the actin monomer (137) to models based on the limited-resolution X-ray diffraction data from aligned filaments. This model is illustrated in Figure 4. Subsequently, Michael Lorenz et al. (167) refined the Holmes (128) model using data from actin mutations. In the same year Tirion et al. (283) described the propeller motion of the two major domains about the hinge and then refined the F-actin model to take into account domain motion. Thus it emerged that the actin monomer was not the rigid, static (double strand of pearls) structure so commonly depicted. Monomers in this model are positioned slightly tangential to the filament axis, with subdomain 1 (the major myosin S-1 binding site) being located at the highest radius.

F-actin displays an arrowhead-like appearance when decorated with myosin subfragment 1 (S-1) (131, 188). For this reason the opposite ends of the filament have been named the "barbed" and "pointed" ends. These ends correspond to exposed subdomains 1 and 3 and subdomains 2 and 4, respectively.

Quite a different model was proposed by Schutt, Lindberg, and colleagues (259) based on crystallographic principles and known examples from other structures in biology. In profilin-actin crystals, a "ribbon" structure can be seen in which monomer subdomains 1 and 2 lie close to the ribbon axis. Alternating monomers displayed front and back views. They proposed that monomers were reoriented to produce F-actin. In essence, this group turned the model of Holmes et al. (128) inside out by rotating the actin monomer about 180° in the plane of the filament axis. The effect of this was to locate subdomains 1 and 2 closer to the filament axis, leaving subdomains 3 and 4 at higher radii. These authors further suggested that the subdomains were able to undergo significant movements relative to each other and that this could feasibly be built into a mechanism of contraction (258).

Low-resolution (25-30 Å) electron microscope studies have revealed that the actin filament can exist in multiple conformations depending on the type of bound cation and nucleotide, the isoform of actin (213, 214), and the presence of other proteins bound to actin (182, 218). Thus it is now more common to view F-actin as a dynamic, responsive structure than a passive structural element.

The most authoritative dissertation on the biophysics of F-actin is Oosawa and Asakura's slim volume, Thermodynamics of the Polymerization of Protein (210). Despite its age, this book continues to provide a comprehensive theoretical background for understanding the assembly of actin into filaments and higher order assemblies. The conditions under which actin monomers self-assemble in vitro are well documented (211), although it is fair to say they are less well understood for in vivo assembly.

Polymerization is essentially a condensation reaction. The main features of this process are 1) a slow initial association to a dimer that is more likely to rapidly dissociate to monomers than to assemble; 2) the formation of a stable trimer that represents the nucleus of polymerization, a state where actin assembly is more likely than is disassembly; and 3) the elongation phase during which actin monomers are rapidly assembled. Solvent conditions that promote polymerization include high ionic strength (KCl concentrations >50 mM), neutral or slightly acidic pH, high Mg²⁺ (rather than high Ca²⁺), and elevated temperature (9, 111, 294), in other words, the conditions found in cells.

In addition to these three processes, actin filaments are in a continuous state of assembly/disassembly. As a consequence, in the steady state, a small but finite concentration of actin monomers will be present in any filament population (140, 211). This concentration of free monomers in equilibrium with a population of actin filaments is referred to as the critical concentration and is dependent on solvent conditions, particularly on the presence of certain ABPs and/or actin ligands such as phalloidin (80) or latrunculin A (193).

Elongation involves association and dissociation of monomers from the filament. These processes can occur at either end of the filament, but association predominantly occurs at the barbed end and dissociation at the pointed end. The nucleotide binding site of monomeric actin is almost exclusively associated with ATP in vivo. Conversely, the majority of F-actin subunits contain bound ADP. While the G-actin bound ATP is readily exchangeable with solvent nucleotides, the ADP of F-actin is essentially nonexchangeable (121). Hydrolysis of bound ATP was originally thought to be tightly coupled to the polymerization process (210, 300); however, subsequent investigations (44, 46, 77, 221) revealed that a time lag exists between the incorporation of ATP-G-actin onto the filament end and hydrolysis of the bound nucleotide. The kinetics of ATP hydrolysis and release of its product, P_i (43), suggests that the bound ATP of newly added actin subunits is hydrolyzed through two distinct sequential steps as illustrated in the following reaction scheme

The rate of ATP hydrolysis was calculated to be 10 times higher than the rate of P_i release, suggesting that ADP-P_i-F-actin is the major intermediate in the nucleotide hydrolysis process. Conversion of ATP to ADP-P_i and then to ADP + P_i is not random but primarily occurs at the growing (barbed) ends of filaments (47).

Depolymerization of actin is not simply the opposite of polymerization, principally because actin cannot regenerate ATP from ADP and P_i (45). Instead, dissociated ADP-actin subunits rapidly exchange their bound ADP for ATP in solution (200), a process that is accelerated by profilin (see below). Polymerization of actin is not dependent on nucleotide hydrolysis. It is possible under special circumstances to form F-actin from G-actin containing either no bound nucleotide (141, 227) or a nonhydrolyzable analog of ATP (63). However, nucleotide hydrolysis is required for the normal function of F-actin.

The structural polarity of the actin filament and the irreversible nature of ATP hydrolysis during actin assembly have implications for the rate and direction of filament growth at opposite ends of F-actin. The critical concentration for the pointed end is 12- to 15-fold higher than for the barbed end under physiological conditions (301). This difference may result in the unidirectional growth of the actin filament due to a continual flux of actin subunits from the pointed to the barbed end of the filament. This reaction is called "treadmilling" (300).

Under in vivo conditions one might expect that all or most of the actin exists in an assembled (filamentous) form. Slow addition at the barbed end and even slower dissociation at the pointed end of the filaments produces a rate of treadmilling (300) of monomers (~2 µm/h) that is ~200-fold slower than observed in vivo. Monomers move progressively along the filament from the barbed end (the end associated with Z-disks of the sarcomere, or the cell membrane where ruffling occurs in motile cells) toward the free pointed end of the filament (located in the middle of the sarcomere or oriented away from the cell membrane). ABPs present in vivo regulate different aspects of the assembly/disassembly process. These include filament stabilizers (e.g., tropomyosin), capping proteins (e.g., CapZ, tropomodulin), ABPs that promote branching (e.g., Arp2/3), and ABPs that sequester G-actin and thus maintain a pool of monomers in solution (e.g., thymosin 4, profilin).

Finally, filament lengths are affected by fragmentation and annealing. Fragmentation can arise from thermal motion in vitro, but in vivo much of this will be constrained by the other contents of the cytoplasm. Annealing occurs when an existing filament binds to the appropriate end of a second filament. This has been observed directly using fluorescent-phalloidin-labeled single filaments (Ishiwata, personal communication) and has been quantified (2.2 µM¹·s¹) under similar conditions (145). Arp2/3 (discussed below) is believed to create branch points in actin microfilaments by "capturing" existing filaments.

	III. ACTIN BINDING PROTEINS

Top Previous Next References

Actin binds a substantial number of proteins collectively called ABPs. Some years ago, Pollard and Cooper (231) identified a large number of ABPs, and recently we counted 162 distinct and separate proteins without including their many synonyms or isoforms. No doubt more will be identified. Many of the known ABPs bind to the same loci on the surface of actin and therefore can be expected to compete. A few bind with positive cooperativity and tend to form ternary complexes (see below) but rather more bind with negative cooperativity. In myofibrils, at least eight sarcomeric proteins bind to the thin filaments. At least 12 ABPs are membrane-associated proteins, and another nine are membrane receptors or ion transporters. Thirteen ABPs cross-link actin filaments, whereas others enable filaments to interact with other elements of the cytoskeleton. Microfilaments probably do not interact directly with microtubules and/or intermediate filaments but do so via linker proteins.

Actin also binds ~30 other ligands including drugs and toxins. Thus the sheer number of ligands that have a significant affinity for actin strongly suggests there is probably a large number of binding sites that cover much of the exposed surface of the molecule. A comprehensive list is available from Actin (266), the Guidebook to the Cytoskeletal and Motor Proteins (151), and most recently from the two-volume Molecular Interactions of Actin (88). Attempts to classify these ABPs leave many "orphans" that do not fit into families, so any attempt to group them is bound to be somewhat arbitrary. Classifications according to the function of ABPs or their consensus sequences can also be problematic.

Several types of ABPs facilitate disassembly and assembly. Below is a review of the most common ABPs. Classification of these ABPs can be reduced to seven groups. 1) Monomer-binding proteins sequester G-actin and prevent its polymerization (e.g., thymosin 4, DNase I). (Note: C. E. Schutt and Uno Lindberg are completing an extensive review of Actin Monomer Binding Proteins to be published in the Protein Profile series of Oxford University Press.) 2) Filament-depolymerizing proteins induce the conversion of F- to G-actin (e.g., CapZ and cofilin). 3) Filament end-binding proteins cap the ends of the actin filament preventing the exchange of monomers at the pointed end (e.g., tropomodulin) and at the barbed end (e.g., CapZ). 4) Filament severing proteins shorten the average length of filaments by binding to the side of F-actin and cutting it into two pieces (e.g., gelsolin). (Note: H. Hinssen is completing an extensive review on the Gelsolin Family to be published in the Protein Profile series of Oxford University Press.) 5) Cross-linking proteins contain at least two binding sites for F-actin, thus facilitating the formation of filament bundles, branching filaments, and three-dimensional networks (e.g., Arp2/3). 6) Stabilizing proteins bind to the sides of actin filaments and prevent depolymerization (e.g., tropomyosin). 7) Motor proteins that use F-actin as a track upon which to move (e.g., the myosin family of motors). Here we consider only groups 1-4.

ABPs are not limited to one class, for example, gelsolin is capable of severing and capping the barbed end of actin filaments, and the Arp2/3 complex can nucleate filament formation, elongate filaments, and establish branch points in actin networks (69). In this review we have selected eight ABPs that are biologically relevant and have atomic structures or good working models.

The rapid assembly of actin filaments is the principal driving force behind many forms of cell locomotion. Cells can migrate at rates up to ~0.5 µm/s (264). This means that filaments must have a net rate of elongation that is slightly less than 200 monomers/s. Principally this occurs at their barbed ends where MgATP-actin assembles about five times faster than MgADP-actin (149). The exchange of ATP for ADP in filaments is so slow (of the order of days, Ref. 201) it can be ignored. At steady state, filament length is constant, but there is a slow turnover (~2 µm/h) (264) of monomers due to a treadmilling process that involves the hydrolysis of ATP to ADP-P_i-actin and then a slower dissociation of P_i leaving most of the monomers in the filament with bound MgADP. Thus, if pure actin filaments treadmill very slowly and if actin assembly is the driver of cell motility, then the intrinsic rate of treadmilling must increase in vivo. This is the task of the ABPs.

ADF/cofilin was first discovered and purified in 1980 from embryonic chick brain extracts by Bamburg et al. (20). Since then, the family has grown to include a number of related proteins, including invertebrate depactin (named because it depolymerizes actin) (173); porcine ADF or destrin (destroys F-actin) (191); cofilin (cosediments with filamentous actin) (1), Acanthamoeba actophorin (236); Dictyostelium coactosin (82), Drosophila twinstar or D-61 (91); unc-60A and unc-60B from Caenorhabditis (185); Xenopus Acs (or XAC1 and XAC2) (2); and finally Toxoplasma ADF (4). All of these share considerable (30-40%) amino acid sequence identity. Furthermore, two other major protein families are related to ADF through the presence of an ADF homology domain. One protein has a duplication of this domain and is consequently called twinfilin; the other contains a single ADF homology domain linked to another motif and encodes the drebin family of proteins (16).

Despite this somewhat confusing array of homologs, vertebrates have genes for only two forms, ADF and cofilin. The names alone suggest that one depolymerizes F-actin (ADF) while cofilin cosediments with F-actin, but actually both can elevate the levels of monomeric actin and both can bind to F-actin. Only one isoform of ADF is known in mammals (and birds), whereas two are known for cofilin. ADF and cofilin are clearly different but related proteins.

Cofilin is diffusely distributed in the cytoplasm of quiescent cells. However, in active cells, it translocates to cortical regions where the actin cytoskeleton is highly dynamic and drives the ruffling of membranes of motile cells (18, 309), the cleavage furrow of dividing cells (197), the advancing of neuronal growth cones (18, 198), and myofibrillogenesis (206). ADF and cofilin are not necessarily expressed at the same level in all tissues. In adults, ADF is relatively highly expressed in nerves, intestine, kidney, and testes (18), whereas cofilin levels are higher in hematopoietic tissues, bone osteoclasts, and fibroblasts (309). In baby hamster kidney cells, ADF constitutes 0.4% of the soluble protein, while cofilin accounts for 1.3% (148). The total ADF and cofilin amounts to 20 µM in cells where the total actin concentration is more than three times higher (67 µm) (148) (see Table 1).

Recently, Vartiainen et al. (291) clarified the ADF/cofilin nomenclatures for vertebrates. Using mice, they proposed that cofilin 1 is expressed in most tissues of embryonic and adult cells, cofilin 2 is expressed in muscle cells, and ADF is limited to epithelial and endothelial cells. More importantly, they defined the functional similarities and differences between them.

Porcine destrin was the first structure in this family of proteins to be solved by NMR spectroscopy (118). Subsequently, atomic structures of three more members of the ADF/cofilin family have been determined. A ribbon representation of yeast cofilin is shown in Figure 5A solved at 1.8 Å (95). Crystal structures have also been determined for both Acanthamoeba actophorin (161) and a plant ADF from Arabidopsis thaliana (32). No atomic structure has been published for any vertebrate cofilin or for any muscle ADF/cofilin, but the atomic coordinates for a high-resolution NMR structure for chick cofilin from embryonic skeletal muscle were recently deposited in the BioMagRes Bank (http://bmrb.wisc.edu) under BMRB accession number 5177 (11).

View larger version (36K): [in this window] [in a new window]   Fig. 5. A: the atomic structure of yeast cofilin solved at 1.8 Å showing the NH2 and COOH termini. Putative sites for binding F-actin and G-actin (purple helix) are located in the COOH-terminal region of the molecule. The solid spheres identify residues that are essential for actin binding. [Modified from Fedorov et al. (95).] B: the crystal structure of yeast cofilin (red) was fitted using molecular dynamics to the actin monomer structure (grey) using energy minimization. This approximate orientation is preserved in views of subsequent actin-ABP complexes. (Model kindly provided by Willy Wriggers.)

Indirect methods have been used to examine the sites on cofilin that bind to actin. Despite the best efforts of several groups, these two proteins have not yet been cocrystallized, and consequently, unlike DNase I, gelsolin segment 1, and profilin, the precise contact sites between cofilin and actin are not known. Structural homologies between cofilin, gelsolin, and profilin strongly suggest they bind to the same region of G-actin (i.e., to subdomains 1 and 3). However, the functional products of systematic mutagenesis of yeast cofilin indicate that the cofilin-actin interaction is distinctly different.

Competitive binding of a synthetic dodecapeptide (residues 104-115 of vertebrate cofilin) has implicated this region in the interaction of cofilin with monomeric but not with filamentous actin (308). This peptide is largely conserved in other members of the ADF/cofilin family including destrin and depactin, suggesting that it may be a consensus sequence essential for actin binding and depolymerizing activities.

Mutagenesis of surface residues for both yeast (158) and chick (154) cofilin have implicated the NH₂-terminal region in G-actin binding. Zero-length cross-linking of G-actin to vertebrate cofilin by 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (308) revealed that either Lys-112 and/or Lys-114 of cofilin is in direct contact with actin. Mutations of these cross-linked residues (192), or the corresponding Arg-96 and Lys-98 of yeast cofilin, totally abolished actin binding. Thus a significant fraction of the surface residues has been implicated in the binding of cofilin to G- and F-actin (Fig. 5A, circles) and within these areas, eight or nine residues (solid spheres) are considered essential. Despite the absence of cocrystals, the cofilin structure has been "docked" onto the Kabsch et al. (137) actin structure (Fig. 5B) using energy minimization techniques (304). This figure provides a clear view of the structural relationship between these two proteins and is presented so it can be compared with the binding sites for profilin and DNase I (see below).

Bamburg (16) recently stressed the importance of switching from an ADF/cofilin system to a tropomyosin-based regulation of F-actin function. Tropomyosin binds along the grooves of the long pitch helix of F-actin. In myofibrils, where tropomyosin is constitutively expressed, the binding of tropomyosin and cofilin to actin filaments is mutually exclusive and competitive (311). Thus the depolymerizing activity of cofilin is inhibited by the presence of muscle and nonmuscle tropomyosins (17). Tropomyosin is located deeper into the cortex from the site of active actin assembly and its presence probably inhibits cofilin binding (28). Cofilin and phalloidin also compete for binding to F-actin (120), even though their binding sites are separated by 15-20 Å. McGough et al. (182) proposed that an allosteric change in the twist of F-actin induced by cofilin may inhibit phalloidin from binding between monomers close to the filament axis.

The blue residues in Figure 6 represent the probable minimal molecular contacts for ADF/cofilin on actin. The principal motif shown in Figure 5, A and B (a 4-stranded -sheet surrounded by 4 -helices), is also found in an unrelated sequence of gelsolin segment 1. It is therefore not surprising that cofilin competes with gelsolin segment 1 and profilin (85) for binding to G-actin. Both profilin and gelsolin segment 1 bind to the barbed end of the actin monomer (subdomains 1 and 3), even though only Glu-167 and Tyr-169 (see Table 1) are common binding sites. Because cofilin competes with both proteins, it is possible that it too interacts with residues in the 162-176 loop of actin subdomain 3 (see Fig. 1A).

View larger version (65K): [in this window] [in a new window]   Fig. 6. The molecular contact sites between actin (shown here as Van der Waals representations) and ABP sites. "Front" (A) and "back" (B) views of the monomer are shown where the binding sites are colored: row 1: yellow, actin monomer-monomer interaction sites; light green, DNase I; and blue, ADF/cofilin; row 2: dark green, thymosin 4; and red, profilin; row 3: pink, gelsolin segment 1; and black, vitamin D-binding protein; row 4: red, CapZ sites. There is significant overlap between the actin binding sites for ADF/cofilin, profilin, gelsolin segment 1, and CapZ and are therefore shown on separate figures. These are the minimal contacts and are based on chemical cross-linking and mutagenesis data.

The structural sites to which cofilin binds along the outside of the filament are not known in atomic detail, so the blue residues in Figure 6 are based on chemical cross-linking and other data. At low resolution (25 Å), ADF/cofilin bridges subdomains 1 and 3 of an upper actin to subdomains 1 and 2 of a lower actin in a filament. The major axis of cofilin makes a 30° angle with the plane normal to the helical axis of the filament. It is centered (axially) at about the position of subdomain 2 of the lower actin subunit and radially at the cleft between subdomains 1 and 3 of the upper actin subunit. Electron cryomicroscopy and helical reconstructions of F-actin decorated with cofilin revealed its ability to reduce the angle of rotation between subunits along the short-pitched left-handed genetic helix by 3-5° (182). There is no effect on the axial rise per actin subunit, and consequently, the long-pitch helical repeat of F-actin is decreased from ~360 to 270 Å, and the diameter increases slightly. The change of helical pitch can even be seen in negatively stained filaments in the absence of cofilin (Fig. 7a) and in the presence of stoichiometric cofilin (Fig. 7b). We are grateful to Roger Craig for providing these micrographs.

View larger version (93K): [in this window] [in a new window]   Fig. 7. Electron micrographs of negatively stained (uranyl acetate) rabbit skeletal muscle actin. a: F-actin only. b: F-actin + chick cofilin (molar ratio 1:1). c: F-actin + human gelsolin (1:0.2). (Micrographs provided by Roger Craig.)

A) TREADMILLING. The ADF/cofilin family is almost single-handedly credited for the high rate of treadmilling of monomers in actin filaments in vivo. This is achieved in part by increasing (by 30-fold) the off rate (9 s¹, Ref. 240) at the pointed ends of filaments without changing the off rate at the barbed ends (42). However, in the presence of profilin (see sect. IIIB), the rate of pointed-end disassembly is even faster (125 s¹, Ref. 85). The resulting elevated concentration of cofilin-G-actin in the cytoplasm can be rapidly recycled at the barbed end of filaments, provided that ATP can replace the actin-bound nucleotide. This concept is illustrated in Figure 8. In vivo, the concentration of ADF/cofilin is low relative to other ABPs such as profilin and thymosin 4 (see discussion below). ADF/cofilin "decoration" of actin filaments is probably restricted to their "aged" ends where ATP has been converted to ADP, i.e., some distance from the membrane surface where the filaments are actively growing (see below).

View larger version (37K): [in this window] [in a new window]   Fig. 8. A cartoon representation of F-actin in the absence (left) and presence (right) of cofilin. Cofilins (red crescents) preferentially bind to ADP-containing monomers in the filament and induce a small change in the angular rotation of monomers in the filament with a consequent slight increase in filament diameter and decrease in filament length. Cofilin accelerates the off rate at the pointed ends of the filaments. Actin monomers are represented as having a small and a large domain and are color coded according to their bound nucleotide (brown for ATP, orange for ADP-Pi, and yellow for ADP).

B) DEPOLYMERIZATION. At steady state, increasing the rate of dissociation at the pointed end would not depolymerize actin filaments if the released monomers were able to reassemble at the barbed end. However, if a barbed-end capping protein (e.g., CapZ) blocks reassembly, then cofilin will depolymerize actin filaments.

C) NUCLEATION OF POLYMERIZATION. ADF/cofilin can nucleate the assembly of actin, and this function is likely to be especially important in the presence of barbed-end blockers like CapZ. The ability to nucleate assembly is probably pH dependent and may vary between different isoforms of ADF/cofilin.

D) ADP-ACTIN. Binding of ADF/cofilin to actin is regulated, at least in part, by the state of the nucleotide bound to actin. It binds to ADP-actin with about two orders of magnitude greater affinity than to ATP-actin (Table 1) or ADP-P_i-actin, and this is true for both the G- and F-actin (42) at pH 7.8. Thus cofilin not only promotes the disassembly of ADP-actin monomers from filaments, but it also binds to released ADP-actin monomers and inhibits the exchange of their bound nucleotide (202). The off rate at the pointed ends of filaments is probably the rate-limiting step in the recycling of disassociated actin subunits (281). The extent of depolymerization by ADF/cofilin depends on several factors; for example, different isoforms may have differing pH sensitivities (240, 183), and other ABPs can influence the binding of cofilin to actin because they share binding sites.

E) SEVERING. Marie-France Carlier et al. (42) argue that the kinetic effects of ADF/cofilin are specific for the ends of filaments and therefore it cannot sever. However, the rate of filament fragmentation would not have to be very great to increase the disassembly rate from ~125 s¹ in vitro to the rate required to match the speed of motility (~200 s¹). Evidence based on light microscopy (119) suggests that filaments are rapidly depolymerized by ADF at slightly alkaline pH. Also, there is a difference in severing activities of recombinant and native cofilin where the native form is more active than the recombinant cofilin. Recombinant cofilins can vary in their severing activity and can give the impression that they have only weak severing activity (132). Thus it remains unclear whether ADF/cofilin can sever actin filaments like true severing proteins or whether the highly cooperative binding of cofilin at substoichiometric ratios makes the filaments "brittle" at the points where the decorated and undecorated regions meet (19). Recently, Ed Egelman and colleagues (104) reported filament severing in the presence of two cofilins (yeast) per actin subunit. These were seen by electron microscopy as two distinct sites in image reconstructions. This raises the question of whether one or two cofilins bind to actin monomers, a feature which may depend on the isoform of cofilin. Thus the question is whether cofilin accelerates treadmilling by severing or depolymerizing F-actin. Current opinion seems to favor the severing ability of cofilin.

The cellular concentrations of ADF/cofilin are substantially lower than the concentration of actin, so it is unlikely there is sufficient ADF/cofilin present in cells (20 µM, Ref. 148) to bind to all polymerized and unpolymerized actins (65 µM; see Table 1), even taking into account its low affinity for ATP- or ADP-P_i-actins. Furthermore, the binding of ADF/cofilin to filaments is cooperative (182). If there is not enough ADF/cofilin to act as a monomer-sequestering protein, this role probably belongs to profilin and thymosin 4 (see below).

Relatively few ABPs bind to the pointed compared with the barbed end of a filament. Cofilin can compete with 1) spectrin that stabilizes short-actin oligomers (see below) (254); 2) tropomodulin that caps the pointed end of tropomyosin-coated actin filaments in muscle and nonmuscle cells (97); 3) DNase I, which binds very tightly to the pointed end of actin monomers but only weakly to actin filaments (see sect. IIIE); and 4) new pointed-end ABPs that are likely to be identified from proteins that currently are either orphans or have not yet been identified in gene arrays and two-dimensional gel electrophoresis.

A number of isoforms of cofilin have been described, and it is possible that not all members of the family perform the same functions under the same conditions (e.g., divalent cations and pH) (128). Some cofilins bind tightly to actin monomers while others do not (Nosworthy, unpublished observations). Some cosediment with F-actin while others have only a low affinity (120). Some promote actin assembly while others rapidly depolymerize it (128). Human cofilin seems to be a better nucleator of assembly whereas ADF is a better polymerizing agent, and this is reflected in their respective K_d values (42, 240) (Table 1).

If different isoforms of cofilin have different effects on microfilament assembly, are they differentially expressed within a cell? Mouse myocyte cells lines express significant amounts of ADF. Elevated expression of actin causes a downregulation of ADF and decreases spreading of these cells without changing the levels of cofilin (256). Regulating ADF/cofilin expression is likely to be slow (on a time scale of tens of minutes/hours) and, apart from being energetically wasteful, it is unlikely to be an effective way of enabling microfilaments to respond quickly to environmental stimuli. However, because most ABPs bind with moderate to low affinity (in the micromolar range), elevated levels of ADF/cofilin could, over time, subtly shift the balance of ABPs and, by mass action, alter the state of assembly of cytoskeletal microfilaments.

The ability of cofilin to bind G-actin is also regulated by phosphorylation of Ser-3, which is conserved in most members of the ADF/cofilin family. When ADF/cofilin is phosphorylated, there is a sharp fall in its affinity for actin (190). ADP exchange in G-actin is strongly inhibited by cofilin, but once cofilin is phosphorylated, the released ADP-actin monomer can exchange with cytoplasmic ATP, and it is now ready for reincorporation at the barbed end of a growing filament. Typically, this occurs at the interface of the microfilaments and the membrane at the leading edge of the moving cell. Thus cellular microfilament turnover can potentially be regulated by cycles of phosphorylation and dephosphorylation. As we will see below, profilin plays an important role here.

We know that in yeast, phosphorylation is not a regulatory factor whereas it is in most vertebrates. An important question is, How much of total cellular ADF/cofilin is phosphorylated? For most cells, we simply do not have this information, but in amoeba ~30% of actophorin is phosphorylated (29). On balance, it seems likely that, since not all of ADF/cofilin is constitutively active, there may be mechanisms other than phosphorylation that control actin assembly-disassembly in vertebrates.

Phosphorylation of vertebrate cofilin is achieved by LIM-kinase proteins 1 and 2 (LIMK-1 and LIMK-2). LIMK-1 is predominantly neuronal, whereas LIMK-2 is more widespread (26, 187). Although the signaling pathway for activation of the LIM-kinases is not yet fully understood, it is known that LIMK-1 is under the control of the small GTPase Rac (307), whereas LIMK-2 is regulated by the GTPase cdc42 and rho (21) (see discussion below). LIMKs are themselves regulated by phosphorylation of a Thr residue (90). In Acanthamoeba, activation of LIM-kinase by Pak1 (91) results in phosphorylation of Ser-1 (the Acanthamoeba version of Ser-3) of actophorin (29). When expressed in cultured cells, LIMK-1 induces actin reorganization and reverses cofilin-induced depolymerization (307). Expression of inactive LIMK-1 results in the accumulation of F-actin filaments (7). Phosphorylation of cofilin appears to increase with pH-induced activation of cofilin, consistent with a compensating homeostatic mechanism (27).

Although NH₂-terminal Ser residues (Ser-1, -3, -4, and/or -6 in different species) are recognized as the site for phosphorylation, it is not clear whether the resulting inhibition of actin binding involves a steric or conformational change. Recently, Blanchoin et al. (29) examined this question by crystallizing Acanthamoeba actophorin (ADF/cofilin) in the phosphorylated state. Their structure was essentially identical to unphosphorylated ADF/cofilin, and they concluded that inhibition of activity by LIM-kinase was not due to a conformational change at the ADF/cofilin-actin interface. Their reported steric blocking of actin binding was consistent with molecular dynamics predictions (Fig. 5B) and with site-directed mutant studies. Frustratingly, Ser-1 residue in these crystals could not be resolved.

Until recently, the mechanism by which cofilin is dephosphorylated was not well understood. Slingshot, a protein phosphatase with F-actin binding ability, was shown by Niwa et al. (204) to dephosphorylate cofilin in cultured cells and in cell-free assays. In Drosophila, loss of this enzyme results in a dramatic increase in cellular levels of both F-actin and phosphorylated cofilin.

The activity of ADF/cofilin can be regulated by the membrane lipids phosphatidylinositol 4-phosphate (PIP) and phosphatidylinositol 4,5-bisphosphate (PIP₂) (309). These bind to and inhibit the actin-binding domain of ADF/cofilin at residues 104-115 and the NH₂-terminal region (154) and suggest that transmembrane signaling by PIP₂ can regulate the function of ADF/cofilin. Little is known of the molecular nature of this binding site.

The ability of ADF/cofilin to assemble or disassemble F-actin is pH dependent in vitro (311). Acidic conditions (less than pH 6.8) enhance the ability of ADF/cofilin to stabilize F-actin while at more alkaline pH (>7.3), cofilin can rapidly depolymerize F-actin. Furthermore, the critical concentration of the cofilin-actin complex is known to be pH dependent, being low (~2 µM) at pH 6.5 and significantly higher (7 µM) at pH 8.2 (240). Conversely, it has been suggested that the severing activity of cofilin is pH independent (89). The physiological significance of this pH dependency is that actin filaments may be stabilized in the vicinity of membrane Na⁺-H⁺ antiports (16). A more recent report (27) focused on the role of pH in vivo. Using pH-sensitive fluorescent probes introduced into mouse fibroblasts (Swiss 3T3 cells), they (27) showed that by lowering intracellular pH, ADF colocalized with actin filaments and that when the pH was increased, ADF partitioned more with monomeric actin. In contrast, cofilin is much less responsive to altered pH, and there are no reports of structural changes induced in either ADF or cofilin, despite the fact that its structure has been determined by both NMR spectroscopy (11) and crystallography (158).

All vertebrate ADF/cofilin (destrin) sequences contain a putative nuclear localization sequence (NLS) that enables it to migrate from its normal location in the cytoplasm to the nucleus under conditions of cellular stress (203). Heat shock and other forms of stress probably affect a few residues immediately preceding the NLS, which expose it for subsequent binding to a nuclear transport factor and thus passage through the nuclear pores (32). Transport is an active process, and actin accompanies the cofilin into the nucleus, although the significance of this is not clear.

The profilin family of ABPs is found in eukaryotic cells from Acanthamoeba through to human and in many but not all species there are two or more profilin genes (230). Profilins are small proteins with an approximate molecular mass of 19 kDa (6). They are among the most highly expressed (20-100 µM, Ref. 37) of the cytoplasmic proteins and are distributed throughout the cytoplasm. Like cofilin and DNase I, profilin can be denatured in 8 M urea and then renatured by dialysis (138), a property used during its purification.

X-ray diffraction and NMR atomic resolution structures are available for no less than six isoforms of profilin. All are very similar. Profilin has also been cocrystallized with actin in two different conformations, the so-called "closed" and "open" states (60). It was on the basis of these two structures that Schutt et al. (259) developed their model of F-actin (discussed above). Within the profilins, the NH₂-terminal sequence is conserved whereas the COOH-terminal region of certain isoforms resembles gelsolin (see below). The atomic structure of the profilin--actin complex is illustrated in Figure 9. Profilin binds to subdomains 1 and 3 at loci that substantially overlap the binding sites of gelsolin segment-1 (259). Its main crystal contacts with actin are as follows: subdomain 1, 113, 354, 355, 361, 364, 369, 371, 373, 375; subdomain 3, 166, 167, 169, 171-173, 284, 286-288, 290 (see Fig. 6).

View larger version (62K): [in this window] [in a new window]   Fig. 9. A ribbon diagram of the atomic (2.55-Å resolution) structure of profilin (red) cocrystallized with monomeric actin where the orientation of actin is about the same as seen in Figure 5B (259).

Profilin has several cellular functions. It is essentially a high-affinity (K_a = 10⁷ M¹) monomer-binding protein (225). It is best known for its ability to promote the exchange of nucleotide in actin monomers released from filaments (109) and, because the high (millimolar) concentrations of Mg-ATP in the cytoplasm, it catalyzes the exchange of ADP for ATP. It achieves this by binding to subdomains 1 and 3 near the hinge between the two major domains (see Figs. 6 and 9) and modulates the opening of the nucleotide cleft. Profilin enhances filament turnover in the presence of cofilin (85) because the two proteins act at opposite ends of a filament, with profilin adding ATP-actin to the barbed end and cofilin dissociating ADP-actin from the pointed end (see Fig. 10). Profilin also inhibits the hydrolysis of ATP bound to actin, thus maintaining actin monomers in a state where they retain a high affinity for the growing barbed end of filaments (6).

View larger version (35K): [in this window] [in a new window]   Fig. 10. A cartoon representating the cooperative roles of ADF/cofilin (red crescents) and profilin (blue rhomboids) in regulating the turnover of actin monomers (yellow) in a microfilament containing ATP and ADP. Cofilin accelerates the dissociation of monomers from the pointed ends of filaments. Phosphorylation of cofilin (dark red) dissociates it from ADP-actin and profilin promotes the exchange of ADP for ATP that facilitates the addition of profilin-ATP-actin at the barbed end.

While profilin acts as an effective buffer for high monomer concentrations in cells, it can promote polymerization by transporting monomers to the fast-growing barbed ends of filaments. Profilin binds to a site on actin (see row 2 in Fig. 6) that is inaccessible in the Holmes model of the filament (128) (see Fig. 4). Thus, after delivering a monomer to the growing end of the filament, it must either move to a new site that does not block assembly or completely dissociate. This can occur even when actin is complexed to thymosin (see below) because of the high exchange rates for both thymosin and profilin (219). It is not clear why some cells employ thymosin to buffer actin monomers while others use profilin. Profilin binds only very weakly to the pointed ends of filaments. Dissociation of profilin from actin is stimulated by PIP and PIP₂ (108); thus profilin may play a role in transmitting signals between the cell membrane and the actin cytoskeleton. Profilin binds to Arp2 in the Arp2/3 complex (see sect. IIIG).

These functions are schematically illustrated in Figure 10. Here profilin is shown as blue rhomboids that bind to the barbed end of F-actin. The blue arrows indicate the cycle of profilin-actin interactions.

Gelsolin belongs to a superfamily of ABPs expressed in all eukaryotes. The grouping includes (but is not limited to) gelsolin, villin, adseverin (also known as scinderin), CapG, flil, and severin (aka fragmin). All contain at least one of the 120-amino acid structural repeats and many have three or six gelsolin repeats (157). For example, villin (92.5 kDa) is a six-domain ABP that regulates actin assembly in microvilli. It has a Ca²⁺-dependent NH₂-terminal half (180), whereas CapG (aka gCap39) is a smaller, three-domain gelsolin analog that cannot sever F-actin but responds to Ca²⁺ and PIP₂ under conditions where gelsolin is ineffective (312).

Unlike ADF/coflin and profilin, an isoform of intracellular gelsolin can circulate in plasma where it severs and caps actin filaments released into the circulation (e.g., following cell death). The resulting actin monomers and oligomers are then strongly sequestered by vitamin D-binding protein (DBP) and ultimately removed from the circulation in the liver (124). Plasma gelsolin is slightly larger (83 kDa) than cytosolic gelsolin and is derived by alternative splicing of a single gene resulting in a short peptide appended to its NH₂ terminus (155). Otherwise, the sequences of plasma and intracellular gelsolins are identical.

Gelsolin is an 80-kDa protein consisting of two tandem homologous halves (segments 1-3 and 4-6), each containing threefold repeats (Fig. 11). Its structure in the absence of Ca²⁺ has been determined at 2.5-Å resolution (35). The isolated NH₂-terminal half can bind to (cap) two actin monomers and can sever F-actin without the need for free Ca²⁺. In contrast, the COOH-terminal half binds a single actin and is Ca²⁺ dependent. The molecule is compact and globular having dimensions of ~85 × 36 × 55 Å. Segment 1 cocrystallized with monomeric actin is shown in Figure 12A. It has two bound Ca²⁺ (186). Gelsolin segments 4-6 also bind to monomeric actin, and their relationship to the actin monomer is illustrated in Figure 12B.

View larger version (57K): [in this window] [in a new window]   Fig. 11. A ribbon diagram of the atomic structure of inactive gelsolin showing segment 1 (red) (containing the NH2 terminus), segments 2 and 3 (green, yellow), and segments 4-6 (magenta, green, gold). (Figure kindly provided by Bob Robinson.)

View larger version (39K): [in this window] [in a new window]   Fig. 12. A: crystal ribbon structure of gelsolin segment 1 (red) and actin. B: gelsolin segments 4, 5, and 6 and actin. These figures are shown in the same orientation as Figs. 5B and 9A. Colors are those used in Fig. 11. (Figures kindly provided by Bob Robinson.)

McGough et al. (181) had earlier suggested a mechanism in which segments 2 and 3 bind to the actin filament while segment 1 wedges itself between two adjacent actin monomers along the longitudinal axis. This step requires a rearrangement of the interface between segments 1 and 2 to lengthen the linker between these segments. Segments 4-6 reach across the filament to bind a monomer in the other strand. This step would also require extension of the convoluted linker between segments 3 and 4.

The following year, on the basis of crystallographic data, Robinson et al. (243) concluded that Ca²⁺ binding induces a conformational rearrangement in which segment 6 is flipped over and translated by ~40 Å relative to segments 4 and 5. The structural reorganization tears apart the continuous -sheet core of segments 4 and 6. This exposes the actin-binding site on segment 4, enabling severing and capping of actin filaments to proceed (Fig. 13). Severing of F-actin occurs only when sufficient actin-actin bonds are broken (36). The severing of F-actin by gelsolin is illustrated in Figure 7C where only short segments of actin filaments are observed in the presence of recombinant gelsolin (electron micrograph courtesy of Roger Craig, University of Massachusetts Medical Center, Worcester, MA). Here the molar ratio of gelsolin (segments 1-3) to actin is 0.05:1.

View larger version (53K): [in this window] [in a new window]   Fig. 13. A cartoon representation of three monomers in a filament showing a possible rearrangement of gelsolin segments involved in cleavage of the actin filament.

The crystal structure of the F-actin binding domain of severin, the gelsolin homolog from Dictyostelium discoideum, has recently