The structure of the macrophage actin skeleton

Actin, one of the most abundant and highly conserved proteins in nature, is a major determinant of cell shape and movement. In the cytoplasm, actin moves between pools of monomeric subunits (G actin) and double helical filaments (F actin). Filaments, in turn, can be further organized into a variety of interactive structures, ranging from anisotropic parallel bundles to isotropic orthogonal networks. The formation of these assemblies, and their diversity, which can subserve many different functions, depend on the actions of a large repertoire of cytoplasmic actin-binding proteins (Stossel et al. 1985; Pollard & Cooper, 1986). Since the actin structures are dynamic, they may continuously assemble, disassemble and re-orient within the cell. In this paper, we will briefly review the structure of the macrophage actin cytoskeleton, present evidence that it changes in response to agonist stimulation, and describe how the properties of several actin-modulating proteins provide a model for the dynamic regulation of actin structures in these highly motile cells.

The organization of actin filaments in macrophage cytoplasm has been studied by removal of the plasma membrane either by detergent solubilization or ‘peeling’ of the apical (non-attached cell surface) plasma membrane of cells attached to glass. Fig. 1 shows a branching filament network in the edge of a macrophage, which was allowed to adhere to a coverslip, and was extracted with detergent. Labelling with the SI subfragment of myosin reveals that >95% of these filaments are actin (Fig. 2). Unlike actin filaments polymerized in vitro from purified actin, filaments in the cell cortex are short (about 0·5 μm in length) and frequently intersect along their lengths with striking perpendicularity. The frequency of intersections defines the pore size of the network, a value we have measured, and determined to be about 04 μm. There are no free filament ends within the body of the network. Actin filaments always attach their ‘barbed’ and ‘pointed’ ends (with respect to the binding of myosin S1) on the sides of other actin filaments, forming T- and Y-shaped branches. Actin filaments are concentrated in the periphery of the cell, and the relationship between the three-dimensional actin skeleton and cell plasma membrane, as well as other membrane-bounded compartments within the cell, can be examined by selectively peeling portions of apical plasma membrane from adherent cells (Fig. 3). Careful examination of such replicas in the electron microscope shows that actin filaments interact with the membrane in two ways: through lateral interactions, and through end-on interactions, predominantly with attachment at the ‘fast’ polymerizing (barbed) end of the filament but occasionally (10% of the time) at the ‘slow’ or pointed end of the filament (Fig. 4, Table 1).

A perpendicular network structure of the actin filaments could be the basis for the gel-like consistency of cytoplasm if the filaments were indeed cross-linked at their intersections. In vitro, actin filaments can be induced to form a network closely resembling that observed in the macrophage cortex by addition of the filament crosslinking protein actin-binding protein (ABP) (also called filamin) (Davies et al. 1982). ABP is a homodimer of 270K (K = 10³M_r) subunits, and electron microscopy of the protein shows that each subunit is a flexible 80-nm long strand (Hartwig & Stossel, 1981). The actin-binding site is on one end and the self-association site is on the opposite end of the subunit strand. Three sets of observations indicate that ABP is primarily responsible for the orthogonal structure of actin filaments in the cell. First, ABP accounts for the bulk of actin cross-linking activity in macrophage extracts. Second, as mentioned above, actin assembled in the presence of ABP in vitro forms a rigid gel composed of perpendicular filament branches (Hartwig et al. 1980; Niederman et al. 1983). Third, ultrastructural studies by immunogold labelling with anti-ABP immunoglobulin G (IgG) show that ABP is located at points of filament intersection in the cell cortex (Hartwig & Shevlin, 1986), directly demonstrating that actin filaments in the cortex are cross-linked into a gel. At present, there is no evidence for a direct regulation of ABP-actin interaction by known intracellular messengers. The actin-ABP interaction can, however, be regulated indirectly by proteins that can control actin’s polymerization state (Yin et al. 1980).

Under physiological salt conditions, actin monomers polymerize into filaments. Assembly of actin in vitro occurs in at least three steps. First, monomers must become activated, a process that appears to involve a conformational change forming ‘F monomers’. Two or three actin monomers then slowing aggregate to form stable complexes called nuclei. This is the rate limiting step in the assembly process. The last step, elongation, is more rapid, and involves addition of actin monomers onto the ends of the nuclei. Monomer addition occurs preferentially at the barbed end of filaments decorated with heavy meromyosin arrowheads. The concentrations of free monomers required to maintain the steady state (known as the critical monomer concentration) are, under physiological salt conditions, 0·1 μm at the barbed end and l·5 μm at the pointed end. Therefore, filaments formed in purified actin solutions exist in equilibrium with about 1 μM-monomeric actin reflecting an average contribution of the two ends.

Given that the rate of actin assembly is dependent on the nucleation step and the extent of polymerization from the two ends of the filaments, actin assembly can be regulated by proteins that alter these parameters. There is ample evidence that receptor-ligand interactions, such as binding to the chemotactic peptide, f-Met-Leu-Phe (fMLP), induce a rapid and transient polymerization of cytoplasmic actin in human leukocytes and a reorganization of the cytoskeleton (Howard & Meyer, 1984; Howard & Oresajo, 1985). This net increase in cytoplasmic actin polymerization occurs selectively at the barbed ends of the filaments because it is blocked by the barbed end capping cytochalasins (Hartwig & Stossel, 1979; Maclean-Fletcher & Pollard, 1980). The intracellular signal(s) that initiates and regulates actin polymerization are still unknown. fMLP induces hydrolysis of polyphosphatidylinositides, release of intracellular Ca²⁺, activation of protein kinase C, and metabolism of arachidonic acid. The fMLP-induced rapid actin polymerization is mediated through a guanine-nucleotide-binding protein, because it is inhibited by pertussis toxin (Sha’afi & Molski, 1987). Both protein kinase C activation and the rise in intracellular Ca²⁺ to micromolar concentration occur coincidentally with the initiation of actin polymerization and precede the time of maximal F actin content in neutrophils. However, several recent studies using Quin 2 to buffer Ca²⁺ transients in cells have shown that the increase in intracellular Ca²⁺ is not necessary for actin polymerization, but does have an accelerating effect on the net rate of actin polymerization after fMLP exposure (Sha’afi et al. 1986; Sheterline et al. 1986). Howard & Wang (1987), using human neutrophils, show that A23187, a Ca²⁺ ionophore, causes a significant increase in F actin content but requires free Ca²⁺ concentrations higher than that measured in fMLP-activated cells. Phorbol myristate acetate (PMA), a protein kinase C activator, induces a slow and smaller rise in F actin content. Therefore, it appears that neither a rise in intracellular Ca²⁺ concentration nor activation of protein kinase C alone, or in combination, are sufficient to explain the fMLP-induced changes in cytoskeletal organization.

The mechanism(s) mediating these transient F actin increases is also not known. To increase F actin content, cells must either add monomers onto the ends of existing filaments, or create new filaments, or do both. For addition to occur on the ends of pre-existing filaments, their barbed ends which are normally blocked in unactivated cells must become unblocked. New actin nuclei could also be added to the system, either through the fragmentation of existing filaments or by the activation of protein(s) functioning as actin nuclei. Furthermore, there is also a need to access actin monomers for polymerization.

Gelsolin (Yin, 1988) and profilin (Korn, 1982) are two well-characterized actin-modulating proteins, which are present in large quantities in macrophages. The interaction of each protein with actin is modulated by polyphosphoinositides (Janmey & Stossel, 1987; Janmey et al. 1987; Lassing & Lindberg, 1985), and that of gelsolin is further regulated by Ca²⁺ (Yin & Stossel, 1980). Since Ca²⁺ and polyphosphoinositide levels change transiently following agonist stimulation, these proteins may have a pivotal role in modulating cytoskeletal changes.

Gelsolin is a potent actin-modulating protein first identified in rabbit lung macrophages (Yin & Stossel, 1979) and subsequently found in many mammalian cells. The name derives from its ability to mediate the transition of cytosolic extracts from a gel phase to a sol phase, through a reduction in the actin filament length distribution. Cytoplasmic gelsolin consists of a single 80000 M_r polypeptide (Kwiatkowski et al. 1985). The interactions of gelsolin with actin are multiple and complex (Bryan & Hwo, 1986; Janmey et al. 1985; Chaponnier et al. 1986; Yin et al. 1988). Gelsolin has three main effects on actin. First, it breaks the non-covalent bond between actin-actin monomers within an actin filament, severing the filament. Second, gelsolin binds to a barbed filament end. This has the effect of raising the critical concentration for actin association into filaments, thus causing net actin depolymerization. Third, gelsolin nucleates actin assembly, again promoting formation of short actin filaments. Elongation from gelsolin-actin nuclei is on the slow growing end and therefore not likely to be important in cells.

The immediate effect of actin filament shortening is a decrease in their ability to be cross-linked into a gel network, by dramatically increasing the amount of a crosslinking protein required to join filamentous elements into a network, inhibiting actin gelation and solating pre-formed gels (Yin et al. 1980). In addition, filaments severed are capped by gelsolin. These could serve as a potential source of actin nuclei, provided that gelsolin can be removed from the barbed ends of the filaments. Our in vitro data would suggest that gelsolin is activated by Ca²⁺ to interact with actin and create short actin nuclei, while uncapping is effected not simply by removal of Ca²⁺, but also by polyphosphoinositides. Actin monomers required for explosive growth of filaments will be provided by dissociation of profilin-actin complexes by polyphosphoinositides. Profilin is an actin monomer binding protein, which, on binding actin, impairs its ability to assemble into filaments (Korn, 1982; DiNuble & Southwick, 1985). Profilin can therefore decrease the steady-state F actin concentrations, and may explain why close to 50% of the actin in the cell extract of unactivated cells is not polymerized in spite of conditions favouring polymerization.

Several pieces of in vivo evidence suggest that gelsolin may indeed be involved in regulating actin assembly-disassembly in cells:

(1) Gelsolin is present in cells at high concentrations. There is one molecule of gelsolin per 100 actin monomers in macrophages. It is the major Ca²⁺-dependent actin-severing protein, and also accounts for the bulk of actin nucleating and filament end capping activity in the cell cytoplasm.

(2) In vitro studies have shown that although gelsolin requires micromolar Ca²⁺ to bind actin, forming a 2 actin: 1 gelsolin complex, subsequent chelation of Ca²⁺ with EGTA releases only one actin. The resultant EGTA-resistant 1:1 complexes (GA₁) do not fragment actin filaments but cap their ends even at submicromolar Ca²⁺ concentrations. Therefore, once gelsolin blocks the barbed end of a filament in vitro, it can no longer be dissociated from actin by removal of Ca²⁺.

(3) Dissociation of the GA₁ complex can also be demonstrated in a number of cell types. Freshly isolated macrophages contain very little GA₁. lonomycin increases GA₁ complex formation, while fMLP causes dissociation of GA₁ complexes. Therefore, while gelsolin—actin interaction in cells can be induced by Ca²⁺, other factors, activated by fMLP, can dissociate GA₁ complexes (Chaponnier et al. 1987). Likewise, thrombin, which promotes actin polymerization in platelets, induces GA₁ complex formation and subsequent dissociation (Lind et al. 1987). Furthermore, it has been shown by immunogold labelling of ultrathin sections that gelsolin reversibly translocates to the plasma membrane of platelets with the same time course as that observed for formation and dissolution of GA₁ complexes (K. Chambers & J. H. Hartwig, unpublished results).

Taken together, these results suggest that cells possess a mechanism not directly involving Ca²⁺ for dissociating EGTA-resistant actin-gelsolin complexes subsequent to agonist stimulation. Since polyphosphoinositide 4,5-bisphosphate and monophosphate (PIP₂ and PIP, respectively) can functionally uncap actin filaments and dissociate GA₁ complexes in vitro, they are likely candidates as second regulators of gelsolin function. Besides having an effect on gelsolin, PIP₂ can also promote actin assembly by dissociating actin-profilin complexes.

As a working hypothesis, we propose that explosive polymerization of actin filaments can occur through the effect of PIP₂ on gelsolin and profilin (Fig. 5). In cells at rest, gelsolin is a soluble protein and is not bound to actin. When cells are activated, the concentration of PIP₂ in the plasma membrane falls as it is converted to diacylglycerol and inositol 1,4,5 triphosphate by activated phospholipase C. The latter product, in turn, mobilizes Ca²⁺ from internal stores and begins a ‘gelsolin activation cascade’. Gelsolin-Ca²⁺ binds to the side of filaments, severing the cortical actin filament network, and rapidly produces numerous gelsolin-capped actin oligomers. Because these oligomers are short, they can diffuse freely within the cytoplasm. Oligomers can therefore move towards the plasma membrane and contact PIP and PIP₂, as their concentrations are restored following an initial fall. This leads to dissociation of gelsolin from actin oligomers, exposing nuclei with free barbed ends. Simultaneously, PIP₂ releases actin monomers from profilin—actin complexes (Lassing & Lindberg, 1985; Lind et al. 1987). The increased availability of both polymerization-competent actin monomers and the number of nuclei with free barbed ends result in rapid polymerization. Ca²⁺ potentiates the response by allowing gelsolin to sever actin prior to a rise in PIP₂ concentration, creating a large number of nuclei, and a large pool of actin monomers.

The multifunctionality of gelsolin, and its differential regulation by Ca²⁺ and polyphosphoinositides, suggest that there are distinct functional and regulatory domains in gelsolin. The domain structure of gelsolin has been analysed by limited proteolysis (Bryan & Hwo, 1986; Chaponnier et al. 1986; Kwiatkowski et al. 1985), and the picture that emerges is that gelsolin contains at least three actin binding sites located on the peptides CT14N, CT28N, and CT38C (Fig. 6). CT14N and CT38C, located on opposite halves of gelsolin and each containing an actin binding site, bind actin monomers to form 2 actin: 1 gelsolin complexes. The third site, CT28N, located in the N-terminal half binds to F actin but not actin monomers. Since the N-terminal half of gelsolin (containing CT14N and CT28N) can sever actin filaments as effectively as intact gelsolin, but its subfragments do not, there must be a high degree of interaction between CT14N and CT28N. Recently, we have demonstrated that CT28N binds stoichiometrically to the side of actin filaments. Binding is inhibited by polyphosphoinositides with a dose response similar to that observed for the inhibition of severing, suggesting that CT28N initiates severing by allowing gelsolin to bind to filament sides (Yin et al. 1988). Severing by the N-terminal half is no longer Ca²⁺-regulated, in contrast to that of intact gelsolin. Instead, the C-terminal half of gelsolin contains a Ca²⁺-regulated actin-binding domain. Therefore, the stringent Ca²⁺ requirement for actin severing found in intact gelsolin is not due to a direct effect of Ca²⁺ on the severing domain, but indirectly through an effect on domains in the C-terminal half of the molecule. Ca²⁺ induces a conformational change in gelsolin, causing it to assume a more elliptical shape (Soua et al. 1985) and to expose a protease-sensitive site in the middle of the molecule (Chaponnier et al. 1986; Bryan & Hwo, 1986; Rouayrenc et al. 1986). Therefore, a reasonable model is that the C-terminal half of gelsolin covers up the N-terminal severing sites, and this block is relieved by Ca²⁺ through a documented conformational change in the C-terminal half (Kwiatkowski et al. 1985; Hwo & Bryan, 1986).

Less is known about how polyphosphoinositides regulate gelsolin severing. Because polyphosphoinositides also inhibit severing by the N-terminal half of gelsolin and CT28N binding to F actin, it probably binds to CT28N. Gelsolin binds phenyl-Sepharose in the presence of micromolar Ca²⁺, but is eluted by EGTA (Soua et al. 1985). Therefore, a hydrophobic domain on gelsolin may also be inaccesssible in EGTA. In this way, gelsolin resembles other proteins such as calpactin (Glenney et al. 1987), which requires Ca²⁺ to bind phospholipids. In fact, gelsolin contains putative phospholipid-binding consensus sequences identified for the calpactin-like family of proteins (Geisow & Walker, 1986).

The primary structure of gelsolin, deduced from human gelsolin cDNA clones (Kwiatkowski et al. 1986, 1988) supports the existence of duplicated actin-binding domains in the two halves of the molecule, because there is a corresponding strong tandem repeat in their amino acid sequence. Within each half, there is an additional threefold repeat. These repeated domains may have arisen from a gene duplication event, and diverged subsequently to adopt their respective unique functions. The site on actin to which gelsolin binds has been localized by chemical cross-linking studies to the first 12 amino acids of actin (Doi et al. 1987) and it will be interesting to determine whether the various actin binding domains have identical binding sites on actin.

This work was supported by USPHS Grants GM36507 and HL29113, NSF Grant DCB8517973 and Grants from the Council for Tobacco Research, USA, the Edwin S. Webster Foundation and the Whittaker Health Sciences Fund. H. L. Yin is an Established Investigator of the American Heart Association.