Dynamic turnover of the actin network drives cell motility and muscle contraction. Alpha-actinin, one of several actin-binding proteins essential for cytoskeletal function, is a ubiquitous protein that cross-links actin filaments in muscle and non-muscle cells 2324252627. The protein is found at cell adhesion sites, focal contacts, and along actin stress-fibers in migrating cells. Alpha-actinin can localize to the plasma membrane, where it cross-links the cortical actin, aids in membrane displacement, and links transmembrane receptors with the cytoskeleton. Alpha-actinin is the major thin filament cross-linking protein in the muscle Z-discs. Mutations to the Drosophila melanogaster alpha-actinin gene disrupt the Z-discs and are generally lethal 26. Translocation of alpha-actinin from the cytosol to the plasma membrane may occur indirectly by interactions with the cytoplasmic tails of transmembrane receptors. Alpha-actinin associates with several plasma membrane associated proteins including ICAM-1, ICAM-2, beta1-integrin, beta2-integrin, L-selectin, vinculin, and zyxin. The peptides that interact with alpha-actinin tend to be basic, alpha-helical, and appear to interact with the conserved acidic surface of the alpha-actinin rod 28.

Alpha-actinin may interact with phospholipid membranes directly 29. Static light scattering experiments, employing monolayers and bilayers of varied charge composition, demonstrate that alpha-actinin reconstitutes into the hydrophobic core of lipid bilayers containing negatively charged phospholipids 30. Phosphoinositides, such as phosphatidylinositol 3,4,5-trisphosphate (PIP₃) and phosphatidylinositol 4,5-bisphosphate (PIP₂), differentially regulate alpha-actinin flexibility and function 2731323334. Binding of phosphoinositides to alpha-actinin occurs through the calponin homology domain and has been localized to amino acids 168–184 of striated muscle species 32. Phosphatidylinositol 3-kinase may directly bind to alpha-actinin through its p85 subunit 35. In the presence of diacylglycerol and palmitic acid, alpha-actinin can form microfilament-like complexes with actin 36.

Alpha-actinin is an anti-parallel homodimeric rod with extensive homology to spectrin and dystrophin 2837. The 30–40 nm long dimer consists of two identical polypeptide chains, divided into three functional domains: an actin-binding region at the amino-terminus, a central alpha-actinin segment (rod), and a carboxyl-terminus containing two EF hands (generally a 12 residue loop flanked on both sides by a 12 residue alpha helix) (Figure 1). The actin-binding region contains the amino terminal calponin-homology (CH) domain and the carboxyl-terminal calmodulin-homology (CaM) domain. The relatively rigid central rod domain (242 × 31–49 Å), derived from four spectrin repeats, defines the distance between cross-linked actin filaments and mediates interactions with receptors and signaling proteins.

Electron and cryo-electron microscopy have provided low-resolution (15 Å) images of the intact alpha-actinin molecule 3839. Unfortunately, only the rod domain (residues 274–746, Table 1) has been successfully crystallized for high-resolution structural studies 28. The segments implicated in lipid-binding by our algorithm, amino acid residues 281–300 (1st spectrin repeat) and residues 720–739 (4th spectrin repeat), lie at the head/tail junctions of opposite ends of the isolated monomer in the crystallized rod domain (Figure 1; Table 2) 14. The site experimentally implicated in phosphatidylinositide binding, amino acids 168–184, is absent from the crystallized construct 2831. This segment was not identified as a lipid-binding candidate by our computer algorithm, presumably because the amino acid sequence (TAPYRNVNIQNFHLSWK) forms an extended loop or coil 40.

In the dimeric rod, the predicted lipid-binding regions from constituent monomers lie close, but not confluent, to one another. The left-handed ninety-degree trans-rod twist places the dimer's two amino-terminal lipid-binding segments, residues 281–300, on a common face while separating the carboxyl-terminal segments. Amino acid residues 281–300 and 720–739 are largely alpha-helical and solvent exposed. Whether this accessibility is maintained in the intact alpha-actinin molecule is not clear from the low-resolution structural studies since the region of the protein that joins the 47 kDa head to the rod domain appears to be quite flexible 38.

Alpha-actinin is an acidic protein with a pI of 6.0. Membrane binding is not calcium-dependent but the protein may undergo conformational changes in response to salts, cations, and lipids 3041. The native alpha-actinin rod is globally electrostatically negative; however, the ends containing the predicted lipid-binding sites are less acidic than the middle core (Figure 1,

panel c

Arp2 (actin-related-protein), in a complex with six other proteins including Arp3, promotes branched growth of actin filaments. Immunoelectron microscopy localizes the Arp2/3 complex to the Y-branch, the point where a daughter actin filament branches off at a seventy-degree angle from the parent filament 424344. The Arp2/3 complex attaches to the side of the parent actin filament through the interactions between three of its five ancillary proteins (p16, p34 and p40) and actin subunits. Activation of the Arp2/3 complex requires the presence of nucleation-promoting factors and a pre-existing filament 4546. Nucleation factors such as WASP/Scar (Wiskott-Aldrich Syndrome), in turn, require activation through chemotactic signaling pathways that guide cellular movement. WASP promotes the binding of the Arp2/3 complex to the side of a pre-existing filament and may transfer the first actin subunit to the nascent filament's rapidly growing barbed end. Vinculin may also bind to the Arp2/3 complex, in a phosphatidylinositol-dependent manner, during membrane protrusion 47.

The Arp2/3 complex is a 220 kDa stable assembly of two actin-related proteins and five novel protein subunits 4849. Arp protein sequences are homologous to actin, and subunit p40 (gene name ARPC1) resembles a beta-propeller protein. The other 4 subunits of the complex (gene names ARPC2 through ARPC5) share little sequence homology to known proteins. The maximum dimensions of the complex are 150 × 140 × 100 Å (Figure 2) 49.

The low-resolution 'kidney bean' structure revealed for the Arp2/3 complex by electron microscopy is in general agreement with the inactive crystallographic complex 4849. It is thought that ATP binding induces a modest rigid body rotational conformational change, together with a more dramatic translation, that activates the Arp2/3 complex (Figure 2,

panel d

Our algorithm predicts that amino acid residues 185–202 of Arp2 are involved in mediating lipid interactions. The isolated segment partially inserts into lipid aggregates with an apparent K_d of 1.1 μM 16. In the crystal structure, this segment is primarily alpha-helical (72 %) and lies near the center of the Arp2/3 complex (Figure 2,

panel d

Both p21 and p40 have substantial areas of positive surface charge. These regions are relatively remote from the Arp2's predicted lipid interface in the inactive complex. The computationally predicted lipid interaction site is itself electrostatically neutral but surrounded by strong negative potentials in the assembled complex (Figure 2,

panel e

Capping protein is crucial for actin filament assembly. Activated Cap binds to the barbed end of actin with high affinity (K_d = 1nM) and at a 1:1 stoichiometry forming a mechanical 'cap' that prevents the addition or loss of actin monomers 5051. The sarcomeric isoform of capping protein, which is composed of two polypeptide chains (CapZ α1-β1), localizes to the Z-line of muscle through an interaction with alpha-actinin 52. The non-sarcomeric isoforms are localized at the sites of membrane-actin contact 53545556. Capping protein 'caps' the Arp1 mini-filament in the dynactin complex, directly interacts with twinfillin, and indirectly affects the Arp2/3 complex via the CARMIL protein 57585960. Residues at the carboxyl-termini of each CapZ chain (α 259–286 and β 266–277) are essential for actin binding.

CapZ is an elongated, tightly assembled, heterodimeric alpha/beta protein with overall dimensions of 90 × 50 × 55 Å 61. The two subunits, which may have arisen from gene duplication, are structurally homologous creating a pseudo two-fold symmetry perpendicular to the long axis of the molecule (Figure 3). Each subunit contains three domains and an additional carboxyl-terminal extension. Three anti-parallel helices in an up-down-up arrangement (helices 1–3) form the amino-terminal domain. The middle domain is composed of four beta strands (strands 1–4 for the alpha subunit; three beta strands 1–3 for the beta subunit), containing two reverse turns. The carboxyl-terminal domain comprises an anti-parallel beta sheet formed by five consecutive beta strands (strands 5–9), flanked on one side by a short amino-terminal helix (helix 4) and a long carboxyl-terminal helix (helix 5). The beta strands of each subunit form a single 10-stranded anti-parallel beta-sheet in the center of the molecule. A 'jellyfish' model has been proposed for Cap function in which the carboxyl-terminal helical regions of the protein are mobile and extend outward to engage the barbed end of actin 61.

Phosphatidylinositol 4,5-bisphosphate (PIP₂) regulates CapZ function by dissociating the protein from the barbed ends of actin filaments 5962. This effect appears to be due to the direct binding of dispersed PIP₂ to CapZ. High concentrations of other anionic phospholipids also inhibit the ability of CapZ to effect actin polymerization 63. In some phosphatase and kinase structures, nitrate ions have been found near the phosphate binding sites mimicking the transition state 6465666768. Sulfate ion also may serve as a marker for phospholipid binding sites. The crystal structure of CapZ beta-1 contains four nitrate ions 61. Only two nitrate ions appear to bind to the protein with high specificity; one nitrate is associated with Lys95 while the other interacts with the dipole of helix 5 (Figure 3,

panel a

The sequences predicted to mediate lipid binding by our algorithm, amino acid residues 134–151 and 215–232 of the CapZ-β1 subunit, lie adjacent to one another in the crystal structure 61. Residues 134–151 primarily form beta-sheet whereas residues 215–232 are part of Helix 5. Both segments are solvent-accessible despite contributing residues to the strong dimer interface (e.g., Lys136, Glu221 and Asn222). Although CapZ is predominantly electrostatically negative, the proposed lipid-binding interface varies from neutral to positive (Figure 3,

panel c

Talin is an abundant cytoskeletal protein that binds to the cytoplasmic tails of integrin beta subunits, to actin filaments, to other actin-binding proteins, and to phospholipids 126970717273747576. In fibroblasts, the binding of talin to membranes may induce the formation of focal adhesions or trigger actin assembly by activating integrins or layilin, respectively. In platelets, activated talin translocates from the cytoplasm to the membrane where it co-localizes with the GPIIb/IIIa complex 76.

Talin is a member of the 4.1 superfamily of FERM proteins, a group of membrane-associated proteins that includes the erythrocyte membrane protein 4.1, the ezrin, radixin, moesin, and merlin proteins, and some tyrosine phosphatases 77. A common feature of FERM domain proteins is extensive intramolecular head-tail interactions that mask binding sites on the head 7879. Association of extracellular matrix ligands with integrins triggers the binding of the second messenger phosphatidylinositol 4,5-bisphosphate (PIP₂) to the head domain, altering its conformation to allow talin to bind to the cytoplasmic tails of integrin receptors 78. Binding occurs through a largely hydrophobic area centered on the b5 strand and also involves residues of the b6 strand, the carboxyl-terminal half of helix H5 and the b4-b5 loop. During outside-in integrin signaling, talin binds to other partners on the cytoplasmic face of adhesion complexes, and in particular vinculin, which then binds directly to actin and induces actin bundling 8081. The incorporation of talin into zwitterionic phospholipid bilayers is low but improved in the presence of negatively charged phospholipids (K = 2.9 × 10⁶ M^-1) 13. Talin is able to bind in vitro to phosphatidylinositol, phosphatidylinositol 4-monophosphate, and PIP₂. However, within a phospholipid bilayer, binding is restricted to PIP₂.

Talin is a flexible 235 kDa 51 nm dumbbell-shaped homodimer (Figure 4) 8283. Calpain cleavage before amino acid residue 434 yields 2 major domains, an N-terminal 47 kDa FERM head and a carboxyl 190-kDa rod domain. The rod domain, which is responsible for actin interaction and nucleation, contains low-affinity integrin binding sites as well as actin and vinculin binding sites 8485. The isolated 47 kDa FERM-containing domain retains the lipid-binding capacity of intact talin and includes a primary integrin-binding site 71. Talin binds to phospholipids using both hydrophobic and electrostatic forces with a strong preference for negatively charged aggregates 86.

FERM domains are cysteine-rich modules that bind phosphoinositides via amino acid sequences with a high percentage of basic and polar amino acids. FERM domains contain three modules arranged in a clover shape: F1, F2 and F3 87. The F3 module of talin, which structurally resembles a phosphotyrosine-binding domain, is formed by a single carboxyl-terminal helix that partly encloses one edge of an internally hydrophobic beta sandwich 88. A consensus sequence for PIP₂ binding has been described (K/R)XXXKX(K/R)(K/R) but exceptions are frequent 89.

The computationally predicted lipid-binding site, amino acids 385–406, has a calculated hydrophobicity of 0.029, high amphipathicity, and a hydrophobic moment of 0.3 131590. At pH 7.4 the total free energy of binding (ΔG₀) is approximately -9.4 kcal/mol, a value that compares favorably with that determined for myristylolated membrane-anchoring peptides. Residues 385–406 lie within helix 5 and thus contribute substantially to the binding site for the integrin beta3 tail. This proximity suggests a mechanism for the PIP2 induced conformational change that permits tail binding 78.

Vinculin is a conserved regulator of cell-cell adhesion (cadherin-mediated) and cell-matrix focal adhesions (integrin/talin-mediated). In its resting state, vinculin is held in a closed conformation through interactions between its head (Vh) and tail (Vt) domains. Vinculin activation, associated with junctional signaling, generates an open conformation that binds in vitro to talin, alpha-actinin, paxillin, actin, the Arp2/3 complex, and to itself 479192939495.

Talin and phospholipids activate vinculin. Talin binds to Vh through high-affinity vinculin-binding sites present in its central rod domain. Talin binding stimulates conformational changes in the amino-terminal helical bundle of Vh, displacing the tightly bound Vt 95. Talin also increases the activity of phosphatidylinositol phosphate kinase-1 γ, generating PIP₂ 96979899. The binding of phosphatidylinositol 4,5-bisphosphate to Vt, in turn, disrupts the Vh-Vt interaction freeing vinculin to bind talin, actin, VASP or the Arp2/3 complex 100. Vinculin can readily insert into the hydrophobic core of mono/bilayers containing acidic (phosphatidic acid, phosphatidylinositol and phosphati-dylglycerol), but not neutral (phosphatidylcholine and phosphatidylethanolamine), lipids 101102. Vinculin can also undergo covalent modification by lipids in vivo or bind acidic phospholipids through its carboxyl-terminal domain (amino acids 916–970) 103104105106. The latter process may inhibit the intramolecular association between the amino and carboxyl terminal regions of vinculin and/or expose a binding site for protein kinase C 107108.

Vinculin is a large (1,066 amino acid), structurally dynamic protein with overall dimensions of 100 × 100 × 50 Å in its autoinhibited conformation (Figure 5) 92939495. The protein is composed of eight four-helix bundles that divide the protein into five distinct domains; an 850 amino acid head (Vh), a 200 amino acid tail (Vt) and 3 intervening linkers (Vh2, Vh3, Vt2). The sequences implicated in lipid binding by our algorithm, amino acid residues 935–978 and 1020–1040, contribute to helices 2 through 5 of Vt. Segment 935–978 includes residues involved in Vt-Vh interactions (Arg 945, Arg 978) as well as those mediating phosphatidylinositol binding. Phosphatidylinositol 4,5-bisphosphate appears to bind to a basic "collar" surrounding the carboxyl-terminal arm (residues 910, 911, 1039, 1049, 1060 and 1061), and a basic 'ladder' along the edge of helix 3 (residues 944, 945, 952, 956, 963, 966, 970, 978, 1008 and 1049) (Figure 5,

panel a