The plasma membrane and the underlying cortical actin cytoskeleton undergo continuous dynamic interplay that is responsible for many essential aspects of cell physiology. Polymerization of actin filaments against cellular membranes provides the force for a number of cellular processes such as migration, morphogenesis, and endocytosis. Plasma membrane phosphoinositides (especially phosphatidylinositol bis- and trisphosphates) play a central role in regulating the organization and dynamics of the actin cytoskeleton by acting as platforms for protein recruitment, by triggering signaling cascades, and by directly regulating the activities of actin-binding proteins. Furthermore, a number of actin-associated proteins, such as BAR domain proteins, are capable of directly deforming phosphoinositide-rich membranes to induce plasma membrane protrusions or invaginations. Recent studies have also provided evidence that the actin cytoskeleton-plasma membrane interactions are misregulated in a number of pathological conditions such as cancer and during pathogen invasion. Here, we summarize the wealth of knowledge on how the cortical actin cytoskeleton is regulated by phosphoinositides during various cell biological processes. We also discuss the mechanisms by which interplay between actin dynamics and certain membrane deforming proteins regulate the morphology of the plasma membrane.
A. The Actin Cytoskeleton
The actin cytoskeleton is a dynamic structure that plays a fundamental role in diverse processes in all eukaryotic cells. Actin-dependent cellular processes are typically associated with membrane dynamics, and the coordinated polymerization of actin filaments against cellular membranes provides the force for these processes. Probably the most thoroughly characterized of such processes are cell migration and endocytosis. During cell migration, precisely coordinated polymerization of actin filaments against the plasma membrane induces the formation of plasma membrane protrusions and consequent advancement of the leading edge of the cell (see Fig. 1). Similar membrane protrusions, in the absence of retraction of the trailing edge of the cell, drive cell elongation during morphogenetic processes such as neuronal outgrowth (54, 153, 295). During endocytosis, actin polymerization provides the force for generating membrane invaginations and for the scission of the endocytic vesicles from the plasma membrane (168).
In addition to regulating membrane dynamics, actin filaments together with myosin filaments form contractile structures in cells. These include myofibrils of muscle cells, stress fibers found in many nonmuscle cell types, and the contractile ring that contributes to the separation of the two daughter cells during cytokinesis (18, 264). The force in the contractile structures is generated through ATP-dependent sliding of myosin motor protein along the actin filament scaffold, which is in contrast to membrane dynamics where the force is produced by actin polymerization.
Actin is a globular protein that contains a nucleotide (either ATP or ADP) associated with the cleft located between the two lobes of the protein. Under physiological conditions, actin monomers (G-actin) assemble into polar, helical filaments (F-actin). Because actin filaments are polar structures, they have two structurally and biochemically distinct ends that are called “barbed end” and “pointed end.” Under steady-state conditions, the net addition of subunits to actin filaments occurs at the barbed end. Irreversible hydrolysis of the bound ATP subsequently destabilizes the filament and results in net dissociation of actin subunits at the pointed end. This ATP-dependent polymerization of actin filaments at their barbed ends and depolymerization at the pointed ends results in “filament treadmilling,” which plays a central role in above-mentioned actin-based motile processes in cells (284, 294) (see also Fig. 1).
The structure and dynamics of the actin cytoskeleton in cells are regulated by a vast number of actin-binding proteins. These include proteins that promote nucleation of new actin filaments (Arp2/3 complex, formins, Cobl, Spire, leiomodin) and proteins that enhance actin filament severing and depolymerization (e.g., ADF/cofilins, gelsolin) (53, 274). Also, proteins that regulate actin filament polymerization either by interacting with actin monomers (e.g., profilin, twinfilin, β-thymosins) or with filament barbed ends (e.g., capping protein, Eps8, Ena/VASP) play crucial roles in actin dynamics in cells. Finally, actin-dependent cellular processes typically rely on the organization of actin filaments to desired three-dimensional structures through a large number of actin filament bundling and cross-linking proteins (e.g., α-actinin, fascin) (57, 76, 79, 204, 277). Fundamental studies by the Carlier laboratory revealed that the minimal set of actin-binding proteins required for actin-based motility in vitro consists of ADF/cofilin, profilin, activated Arp2/3 complex, and capping protein (223). However, most cellular processes driven by actin filament treadmilling depend on a significantly larger number of actin-associated proteins and their regulators.
The activities of actin-binding proteins are controlled through various signaling pathways to ensure proper spatial and temporal regulation of actin dynamics in cells. The most thoroughly characterized regulators of actin-binding proteins are the Rho-family small GTPases. In mammals, these include, e.g., RhoA that induces the formation of contractile stress fibers, Rac1 that drives the formation of lamellipodial actin filament network at the leading edge of motile cells, as well as Cdc42 and Rif that induce the formation of thin actin-rich filopodial protrusions at cell periphery (137, 289) (see sect. iii).
In addition to small GTPases, membrane phosphoinositides play a key role in regulating the dynamics of the actin cytoskeleton. Phosphoinositides regulate the activities of Rho GTPases, but they also directly associate with actin-binding proteins to regulate their interactions with actin.
B. General Features of Phosphoinositides and Their Metabolism
Phosphatidylinositol and its phosphorylated derivatives, phosphoinositides, are among the most versatile regulatory molecules in eukaryotic cells. They have strikingly diverse roles in the modulation of many cellular events, such as membrane trafficking, intracellular signaling, cytoskeleton organization, and apoptosis (23, 74, 266). Phosphoinositides contribute also to the pathogenesis of human diseases by functioning as signaling lipids during inflammation, cancer, and metabolic diseases (393).
1. Different phosphoinositide species
Phosphatidylinositol is composed of a d-myo-inositol-1-phosphate linked via its phosphate group to diacylglycerol. Phosphatidylinositol can be reversibly phosphorylated at the D-3, D-4, or D-5 positions of the inositol ring, generating seven distinct phosphoinositide species [PI(3)P, PI(4)P, PI(5)P, PI(3,4)P2, PI(3,5)P2, PI(4,5)P2, and PI(3,4,5)P3] that provide a membrane-binding platform for a variety of proteins (Fig. 2). Phosphoinositides and the enzymes that synthesize them are present in various subcellular compartments including the plasma membrane, endosomes, endoplasmic reticulum, Golgi, nuclear matrix, heterochromatin, and the sites of active RNA splicing (64, 133, 370).
Phosphatidylinositol (PI) and its phosphoinositide derivatives constitute ∼10% of the total cellular lipids in most cells. Among them, PI is the most abundant in eukaryotic cells. The most abundant phosphorylated derivative of phosphatidylinositol is PI(4,5)P2, which comprises 0.3–1.5% of the phospholipids at the plasma membrane of mammalian erythrocytes (97, 131), lymphocytes (254), and hepatocytes (365). The total concentration of PI(4,5)P2 in human myeloid cells is ∼30 μM (32). Furthermore, studies using green fluorescent protein (GFP)-PH domain constructs suggest that the effective concentration of PI(4,5)P2 in various cells is 2–30 μM (239). Although the total amount of PI(4,5)P2 in cells is relatively constant, local dynamic changes in PI(4,5)P2 concentration have been observed at the sites of phagocytosis and in actin-rich protrusions such as membrane ruffles (26, 58, 219). Also PI(3)P appears to be a relatively abundant phosphoinositide in cells, as it has been reported to comprise 2.5 and 0.2% of total phosphoinositides in yeast and mammalian cells, respectively. In contrast, other phosphoinositides are present only at low concentrations in cells. PI(3,5)P2 comprises <0.1 and <0.05% of total phosphoinositides in yeast and mammalian cells, respectively (25, 81). PI(3,4,5)P3 is practically undetectable in resting cells, and even after receptor stimulation the cellular concentration of PI(3,4,5)P3 is at least 25 times lower than that of PI(4,5)P2 (211). Despite their low abundances, each of these phosphoinositide derivatives has a distinct set of biological activities that are typically mediated by proteins that recognize and selectively bind to their phosphorylated head groups (64, 210). Several different phosphoinositide-binding domains have been characterized. They bind to specific phosphoinositide isomers and, in turn, can induce changes in the subcellular localization, posttranslational modification, or activity of proteins containing such a domain (see sect. iiB).
2. Subcellular localizations of phosphoinositides and regulation of their metabolism
Phosphorylations of the D-3, D-4, and D-5 positions of the inositol ring are carried out by distinct classes of phosphatidylinositol kinases that are differentially localized within cells. Therefore, each of the seven phosphoinositides displays its own distinctive subcellular localization, and they not only have an important function in determining specificities of membrane interaction, but also in defining and maintaining organelle identity. For example, PI(4,5)P2 and PI(3,4,5)P3 are primarily enriched at the plasma membrane, where they control many important reactions including generation of intracellular second messengers, exocytosis, endocytosis, and the reorganization of the actin cytoskeleton (71, 74, 269). In contrast, PI(3)P is highly enriched on early endosomes and multivesicular bodies (118, 119, 173). PI(4)P is abundant at the Golgi apparatus, and PI(3,5)P2 locates at late endosomes and lysosomes (67, 248). In addition, the presence of different phosphoinositide species and phosphoinositide-binding actin-binding proteins in the nucleus suggests that phosphoinositides also regulate the still poorly understood dynamics of nuclear actin (155, 374).
Interactions of PI(4,5)P2 with various proteins at the membrane can also cause sequestering/clustering of PI(4,5)P2, resulting in microdomains in which PI(4,5)P2 and specific proteins are concentrated. These microdomains may be necessary for the assembly of distinct multimolecular complexes that control protein activity, specify organelle identity, or regulate membrane trafficking and receptor signaling. For example, actin-binding proteins MARCKS, ezrin, dynamin, MIM, and IRSp53 can sequester PI(4,5)P2 on the membrane (40, 49, 114, 315).
Phosphoinositide kinases display a high degree of specificity in their substrate preference as well as in the phosphatidylinositol phosphorylation position. These enzymes are ubiquitously expressed and are most abundant in cellular membranes including the Golgi, lysosomes, endoplasmic reticulum, plasma membrane, and a variety of vesicles including Glut4-containing vesicles, secretory vesicles, and coated pits (78, 328). Phosphoinositide kinases can be categorized into three general families: phosphoinositide 3-kinases (PI3Ks), phosphoinositide 4-kinases (PI4Ks), and phosphoinositide 5-kinases (PI5Ks) (321).
PI(4,5)P2 is generated by type I (PIPK-Is) or type II (PIPK-II) kinase isoforms (α, β, γ) which utilize PI(4)P and PI(5)P as substrates, respectively. PIPK-Is localize on the plasma membrane and are thought to account for the majority of PI(4,5)P2 synthesis, whereas PIPK-IIs are predominantly localized at intracellular sites (78). PI3Ks are a family of related intracellular signal transducer enzymes capable of phosphorylating the 3 position hydroxyl group of the inositol ring of phosphatidylinositol (142). PI3Ks control essential cellular functions such as cytoskeletal dynamics, signal transduction, and membrane trafficking (142, 218). Upon activation, PI3Ks phosphorylate PI(4,5)P2 to produce PI(3,4,5)P3, a second messenger that binds a subset of PH, FYVE, PX, C1, C2, and other lipid-binding domains in downstream targets to recruit them to the activation center at the membrane. In contrast, PI4Ks convert phosphatidylinositol to PI(4)P by phosphorylating the inositol ring at the D-4 position. The generated PI(4)P can be further phosphorylated by both PI3Ks and PIPKIs to yield PI(3,4)P2 or PI(4,5)P2, respectively. In this way, the PI4Ks play a central role in signaling by feeding into multiple pathways (Fig. 2). Phosphoinositides can also be dephosphorylated by a similar multitude of inositol lipid phosphatases that have a wide range of substrate and positional specificities as well as distinct cellular localizations (321). Some but not all phosphoinositides are subject to hydrolysis by phospholipase C (306) that releases water-soluble inositol phosphates and leaves diacylglycerol (DAG) in the membranes. Ins(1,4,5)P3 and DAG serve as important signaling molecules, the former by releasing Ca2+ from internal Ca2+ stores and the latter by activating protein kinase C (306).
II. Regulation of Actin Dynamics by Phosphoinositides
The organization and dynamics of the actin cytoskeleton are regulated by membrane phosphoinositides at several levels. First, many actin-binding proteins directly interact with phosphoinositides, which regulate the activity and/or subcellular localization of these proteins. Second, phosphoinositides control the subcellular localization of larger scaffolding proteins that are involved in the interplay between the actin cytoskeleton and plasma membrane or intracellular membrane organelles. Finally, proteins controlling the activity of Rho family small GTPases are in many cases regulated by plasma membrane phosphoinositides.
Among different PIs, PI(4,5)P2 is the best-characterized regulator of the actin cytoskeleton. PI(4,5)P2 interacts directly with several actin-binding proteins and regulates their activities (23, 141, 330, 332, 397). Typically, PI(4,5)P2 inhibits those actin-binding proteins that promote actin filament disassembly and activates proteins that induce actin filament assembly. As a consequence, PI(4,5)P2 is considered to promote the formation of actin filament structures beneath the plasma membrane and other phosphoinositide-rich membrane organelles (Fig. 3) (141). This is also supported by a number of studies demonstrating that an increase in the plasma membrane PI(4,5)P2 levels induces actin filament assembly in mammalian cells (313, 398) while sequestration of PI(4,5)P2 leads to a defect in the cortical actin cytoskeleton (304). In addition to regulating actin-binding proteins, PI(4,5)P2 can also bind to ion channels and change their conformation, thereby regulating channel activities (349). Here, we discuss the biological roles and actin interactions of certain extensively characterized actin-regulating proteins.
A. Actin-Binding Proteins That Contain Noncanonical Phosphoinositide-Binding Domains
1. Proteins involved in actin filament nucleation
Actin filament growth begins with the formation of actin nucleus. In cells, this process is greatly accelerated by actin filament nucleating proteins (301). Among them, actin-related protein complex 2/3 (Arp2/3) nucleates new actin filaments while remaining anchored to the sides of preexisting filaments, thus creating branched actin networks (124). However, some controversy remains concerning the stability of Arp2/3-induced filament branches in cells and the existence of a branched actin filament network at protrusive structures in cells (180, 197, 231). Formins, in contrast to Arp2/3, nucleate unbranched filaments, creating actin cables, contractile rings, and stress fibers (125). Other recently identified filament nucleation-promoting factors are Spire, leiomodin, cordon bleu, and JMY. They all contain actin-binding WH2 (WASP homology 2) domains, which are important for their actin filament nucleation activity. Similarly to formins, these proteins promote the formation of unbranched filaments (3, 52, 302, 408). The activities of actin-nucleating proteins are regulated by small GTPases, Wiskott-Aldrich syndrome protein (WASP) family proteins, phosphorylation, and phosphoinositides.
The main regulators of the Arp2/3 are the WASP family proteins (WASP, N-WASP, WAVE, and WASH). WASP and N-WASP are relatively large (molecular mass ∼55 kDa) proteins that contain an NH2-terminal basic (B) motif, a G protein binding domain (GBD), and a VCA domain that binds to and activates Arp2/3 (140, 192, 298, 303, 312). In isolation, these proteins exist in an autoinhibited, inactive state. Inactivation involves interactions between GBD and the VCA domain. The constitutively active VCA domain is released when GTP-Cdc42 binds to GBD and when PI(4,5)P2 interacts with the B motif (310, 359). The B-region of N-WASP binds simultaneously several PI(4,5)P2 molecules. Interestingly, the activation threshold of N-WASP corresponds to a small change in PI(4,5)P2 surface density, suggesting that subtle changes in PI(4,5)P2 concentration at the plasma membrane may induce a “switch” for N-WASP-mediated Arp2/3 actin polymerization (285). Surprisingly, PI(4,5)P2 appears to display differential effects on WASP and N-WASP activity. PI(4,5)P2 vesicles alone stimulate actin polymerization by N-WASP but not by WASP. Moreover, in the presence of Cdc42, PI(4,5)P2 inhibited the effect of WASP in actin polymerization (363). In contrast to N-WASP and WASP, WAVE is constitutively active in vitro. WAVE-2 was reported to bind PI(3,4,5)P3, which is also necessary for the WAVE-2-induced formation of lamellipodia (271).
Actin-depolymerizing factor (ADF)/cofilins are small (molecular mass 15–20 kDa) and abundant proteins found in all eukaryotes. ADF/cofilins interact with both monomeric and filamentous actin, preferring ADP-actin (16). ADF/cofilins promote rapid actin dynamics by depolymerizing (38) and severing actin filaments (8, 43, 287). In addition to promoting actin filament disassembly in cells (147, 178, 200), ADF/cofilins may also contribute to stimulus-responsive actin filament assembly by creating new polymerization-competent filament barbed ends through their severing activity (117). ADF/cofilin-induced actin dynamics are also linked to many diseases. For example, in mammary tumors, the activity status of cofilin is directly related to the invasion, intravasation, and metastasis (381).
The activities of ADF/cofilins are regulated by phosphorylation, pH, and interactions with other proteins (84, 373). Furthermore, the actin-binding and depolymerization activities of ADF/cofilins can be efficiently inhibited by PI(4,5)P2. A recent study demonstrated that the activity of cofilin is spatially and temporally regulated by PI(4,5)P2 in cells (371). Acute reduction of PI(4,5)P2 from the plasma membrane resulted in a release of the membrane-bound pool of cofilin to sever/depolymerize F-actin.
ADF/cofilins bind PI(4,5)P2, PI(3,4)P2, and PI(3,4,5)P3 with relatively high affinity, but they do not interact with IP3, which is the inositol headgroup of PI(4,5)P2 (126, 272, 401). Despite extensive studies, the PI(4,5)P2 binding site of ADF/cofilins and the exact mechanism of the interaction have remained controversial (126, 193, 272, 372, 401). NH2-terminal peptides corresponding to residues D9-T25 and P26-V36 of chick cofilin were shown to inhibit the ability of PI(4,5)P2 to abrogate the cofilin-actin interaction, suggesting that both peptides bind PI(4,5)P2 (193). Combined mutagenesis and native gel electrophoresis studies suggested that yeast cofilin interacts with PI(4,5)P2 through a relatively large positively charged surface, which also overlaps with the actin binding site of the protein (272). In marked contrast to these studies, a recent NMR work suggested that the PI(4,5)P2 binding site of cofilin is located at a specific area at the COOH-terminal region of cofilin and that cofilin may interact both with the inositol headgroup and with the acyl-chain of the lipid (126). Thus further studies are necessary to reveal the exact mechanism by which ADF/cofilins interact with PI(4,5)P2, and to elucidate why cofilin does not bind IP3.
Profilin is a ubiquitous protein that plays a central role in actin dynamics in all eukaryotes. This small (∼15 kDa) globular protein binds at least three types of ligands: monomeric actin, phosphoinositides, and proteins that contain poly-l-proline stretches, including members of the Ena/VASP, formin, cyclase-associated protein, and WASP/WAVE families (340, 347, 384, 390). Profilin functions in both Arp2/3 and formin-dependent actin nucleation pathways and is required for proper cell migration and division (165, 199, 293). In the absence of free filament ends, profilin acts as an ATP-actin monomer sequestering protein. However, in the presence of free barbed ends, it promotes actin polymerization. Thus, depending on the cellular context, profilin may either inhibit or promote actin filament assembly (283, 390). Furthermore, at least some profilin isoforms also promote the nucleotide exchange on actin monomers and thus further induce actin filament treadmilling in cells (391).
Profilin interacts with phosphoinositides and appears to have slightly higher affinity for PI(3,4,5)P3 than for PI(4,5)P2 (201, 256). However, both PIs are capable of disrupting the actin-profilin complex and inhibit profilin's actin monomer sequestering activity (202, 227). Multiple regions of profilin have been implicated in PI(4,5)P2 binding, and the interaction appears to be largely electrostatic (307). In human profilin I, two distinct regions are vital for interactions with PI(4,5)P2. One of these overlaps with the actin-binding surface, whereas the other overlaps with the poly-l-proline binding site (198). Similarly to ADF/cofilin, fluctuations in the concentration of PI(4,5)P2 at the membrane may cause profilin to shuttle between a membrane-bound and actin-bound form, and thus it is a potent mediator of external signals to microfilaments (229, 371). The in vivo role of profilin's PI(4,5)P2 binding has remained largely elusive. Expression of a profilin mutant that is defective in interactions with PI(4,5)P2 and proline-rich ligands enhanced neurite outgrowth (199), suggesting that PI(4,5)P2 may regulate neuronal differentiation through interaction with profilin.
Twinfilin is an evolutionarily conserved regulator of actin dynamics composed of two ADF-H domains. All twinfilins characterized to date interact with monomeric actin and prevent the assembly of the bound monomers to filament ends (277). Like ADF/cofilins, twinfilins bind ADP-actin with higher affinity and ATP-actin (273). In addition to sequestering actin monomers, mammalian twinfilins also cap ADP-actin filament barbed ends with high affinity, and yeast twinfilin severs actin filaments at low pH (138, 258, 278). Furthermore, all twinfilins bind heterodimeric capping proteins, and this interaction is essential for twinfilin's correct subcellular localization and activity in vivo (93, 281). However, despite the wealth of knowledge concerning the biochemical activities of twinfilin, the cellular role of twinfilin is still poorly understood.
Twinfilin binds phosphoinositides in vitro with the highest affinity for PI(4,5)P2 and PI(3,4,5)P2. Interaction with PI(4,5)P2 downregulates the actin monomer sequestering activity of yeast and mouse twinfilins in vitro (281, 375). In addition, PI(4,5)P2 inhibits the filament severing activity of yeast twinfilin (258).
5. Heterodimeric capping protein
Capping protein, a heterodimeric protein composed of α-and β-subunits, is a key cellular component regulating actin filament assembly and organization. It binds to the barbed ends of the filaments and works as a “cap” by preventing the addition and loss of actin monomers at the end. Capping protein is expressed in various eukaryotic cells, where it regulates cell motility, morphogenesis, and endocytosis. In vertebrate striated muscle, a sarcomere-specific isoform of capping protein at the Z-disk caps the barbed ends of the thin filaments, leading to the name CapZ (57). Capping protein is active in the absence of calcium and under a wide range of buffer conditions. However, a number of proteins interact with capping protein and regulate its activity. These include, e.g., V-1/myotrophin, CKIP-1, CD2AP, and CARMIL (22, 31, 36).
Interestingly, the activity of capping protein is inhibited by polyphosphoinositides (186). However, the exact mechanism by which PI(4,5)P2 regulates capping protein is still controversial. While some studies suggest that PI(4,5)P2 rapidly and efficiently removes capping protein from the filament barbed ends in a concentration-dependent manner (176, 325), another study provided evidence that although PI(4,5)P2 prevents capping protein binding to filament barbed ends, it does not efficiently dissociate capping protein form filament ends (186). Despite the controversy concerning the uncapping versus anticapping activity of PI(4,5)P2, a conserved set of basic residues on the surface of capping protein was identified to be important for its interaction with PI(4,5)P2 (176). The PI(4,5)P2 binding site of capping protein overlaps with the actin-binding sites, providing an explanation for how PI(4,5)P2 inhibits the filament barbed end binding of capping protein. Interestingly, elevation of PI(4,5)P2 concentration at the plasma membrane by PIPK-I-overexpression was shown to result in inhibition of actin binding by capping protein, suggesting that regulation of capping protein by PI(4,5)P2 is also important in vivo (398).
Gelsolin superfamily consists of seven different proteins: gelsolin, adseverin, villin, capG, advillin, supervillin, and flightless I, which contain three or six homologous copies of gelsolin-like domain. The founding member of the family, gelsolin, is a relatively large (molecular mass 82–84 kDa) protein that exists as a cytoplasmic as well as a circulating plasma isoform. Gelsolin displays strong actin filament severing and barbed end capping activities, and it contributes to multiple cellular functions including motility and morphogenesis (195, 336). Gelsolin consists of six structurally similar domains, G1 through G6, that fold into a compact conformation that does not interact with actin (33, 34, 351, 367). Upon activation, the F-actin binding duo (G2-G3) and the two G-actin sequestering/F-actin capping domains (G1 and G4) are revealed allowing actin filament severing and/or capping to proceed (33, 34, 351). The activity of gelsolin is regulated by Ca2+, pH, and phosphoinositides (238).
Phosphoinositides, in particular PI(4,5)P2, inhibit interactions of gelsolin with F-actin and induce dissociation of gelsolin from actin filaments (164, 394). It has been shown that in N-cadherin adhesions of rat fibroblasts, local PI(4,5)P2 concentration regulates actin assembly by uncapping gelsolin from the filament barbed ends (85). At least three regions of gelsolin are found to cooperatively associate with PI(4,5)P2 (94). Two of these sites locate near the polypeptide hinge between G1 and G2 and the third PI(4,5)P2-binding site is present in G5-G6 (94, 216). Calcium increases the affinity of full-length gelsolin for PI(4,5)P2. Studies with the NH2- and COOH-terminal halves of gelsolin showed that PI(4,5)P2 binding occurred primarily at the NH2-terminal half and that Ca2+ exposed its PI(4,5)P2 binding sites through a conformational change in the COOH-terminal half. Mild acidification promotes PI(4,5)P2 binding by directly affecting the NH2-terminal sites (216). It is important to note that calcium activation of the actin-modifying properties of gelsolin is sensitive to ATP and that interactions of ATP and PI(4,5)P2 with gelsolin share several similar characteristics (367). Both molecules bind more strongly to G4-G6 in the absence of calcium than in its presence, and PI(4,5)P2 has been reported to have an overlapping binding site with ATP (367).
Although originally viewed in the context of regulation of gelsolin function, the phosphoinositide-gelsolin interaction is also suggested to be involved in lipid signaling events. Gelsolin has been isolated in complexes with an increasing array of lipases and kinases, including phospholipases C-γ1, C-δ, and D, PI3K, and c-src (14, 17, 49, 337, 343). These results suggest that gelsolin may alter lipid signaling pathways, either through direct binding to these proteins or through joint binding to clustered phosphoinositides. Moreover, gelsolin and capG overexpression in NIH3T3 cells has significant effects on signaling through cytokine-stimulated pathways activating either phospholipase C-β or phospholipase C-γ (350). PI(4,5)P2 also enhances phosphorylation of gelsolin by the Src kinase in vitro (70). In addition to binding to PI(4,5)P2, gelsolin and other family members bind to lysophosphatidic acid with high affinity (243) that blocks gelsolin's severing activity and removes gelsolin from filament ends. Importantly, lysophosphatidic acid acts cooperatively with phosphoinositides in inhibiting gelsolin's interaction with actin, and this provides another means for the regulation of gelsolin activity in cells.
Villin, found in microvilli of absorptive epithelium, is a member of the gelsolin family. In addition to standard gelsolin-type activities, villin has a number of other biochemical activities (134, 175). It was reported to nucleate actin, cross-link and sever actin filaments, and cap filament barbed ends. These activities of villin are regulated by Ca2+, tyrosine phosphorylation, and phosphoinositides (175, 187, 379, 405).
Villin interacts with phosphoinositides with the following binding affinity: PI(4,5)P2 > PIP > PI (175). The actin capping and severing activities of villin are inhibited by PI(4,5)P2, whereas the actin filament bundling of villin is enhanced by PI(4,5)P2 (188). In addition, villin sequesters PI(4,5)P2 to inhibit PLC-γ1 activity (282). The PI(4,5)P2 binding sites in villin consist of a cluster of basic residues with the motif x4R/KxR/KR/K (where x is any residue), and these sites are conserved among related actin-binding proteins such as gelsolin, scinderin, adseverin, CapG, supervillin, advillin, dematin, and pervin. Three PI(4,5)P2 binding sites that have been identified in human villin locate between residues 112–119, 138–146, and 816–824 (188). The PI(4,5)P2 binding sites of villin overlap with its actin-binding sites. Consequently, binding of PI(4,5)P2 to the two NH2-terminal sites inhibits the actin filament severing activity of villin. Furthermore, it was proposed that the most COOH-terminal PI(4,5)P2 binding site could regulate dimerization or oligomerization of villin, thus enhancing actin cross-linking by villin in the presence of PI(4,5)P2 (188).
α-Actinin is a ubiquitous actin-binding protein with multiple roles in different cell types. It forms an antiparallel rod-shaped dimer with one actin-binding domain at each end of the rod and is thus capable of bundling/cross-linking actin filaments in multiple nonmuscle cell types and their cytoskeletal frameworks. In contrast, skeletal, cardiac, and smooth muscle isoforms of α-actinin localize to the Z-disk and analogous dense bodies, where they form a latticelike structure that stabilizes the contractile apparatus. Besides binding to actin filaments, α-actinin associates with a number of cytoskeletal and signaling molecules as well as cytoplasmic domains of transmembrane receptors and ion channels, rendering it important in structural and regulatory roles in cytoskeletal organization and muscle contraction (338).
The activity of α-actinin can be regulated by phosphoinositides, although the exact mechanism of this regulation is still somewhat unclear. Mutations decreasing the phosphoinositide affinity of α-actinin lead to a slower dissociation rate of actin/α-actinin complex, resulting in excessive bundling of actin filaments and lethargic actin cytoskeleton in cells (105). PI(4,5)P2 and PI(3,4,5)P3 bind to the calponin homology 2 domain of α-actinin, regulating its interactions with actin filaments and integrin receptors (105, 106). PI(3,4,5)P3 inhibits and disrupts α-actinin's actin bundling activity, whereas PI(4,5)P2 can only inhibit this activity. As PI(4,5)P2 appears to increase the stability of the α-actinin while PI(3,4,5)P3 decreases its stability, it was proposed that PI(4,5)P2 and PI(3,4,5)P3 differentially regulate α-actinin function by modulating the structure and flexibility of the protein (111). These data are supported by recent protease-sensitivity experiments, which demonstrated that PI(4,5)P2 binding decreases the proteolysis of α-actinin, suggesting a role in stabilizing the structure of the protein. In contrast, PI(3,4,5)P3 binding enhanced α-actinin proteolysis, indicating an increase in the flexibility of the protein (341). The in vivo importance of the interaction between α-actinin and PI(4,5)P2 was demonstrated by studies on Xenopus oocytes and hippocampal neurons, where decrease in the PI(4,5)P2 concentration or inactivation of the PI(4,5)P2-binding site of α-actinin lead to the detachement of α-actinin from the plasma membrane and subsequent inhibition of the NMDA receptor activity (247).
Vinculin is a conserved and abundant cytoskeletal protein involved in linking the actin cytoskeleton to the membrane at the sites of adhesion (207, 249). Therefore, vinculin plays an important role in processes such as cell migration, development, and wound healing. Loss of vinculin function has been associated with cancer, cardiovascular diseases, and lethal defects in embryogenesis. The molecular mass of vinculin is ∼115 kDa, encompassing a globular head domain, a flexible neck region, and a tail domain (407). The recruitment and assembly of vinculin into focal adhesions is downregulated by strong intermolecular interactions between its head and tail domains (249, 280). Vinculin binds a number of cytoskeletal and adhesion proteins, including actin, talin, α-actinin, α-catenin, β-catenin, vinexin, ponsin, Arp 2/3 complex, vasodilator-stimulated phosphoprotein (VASP), and paxillin (150, 404, 407).
Vinculin also binds acidic phospholipids, including phosphatidylserine (PS), PI, and PI(4,5)P2 (91, 267, 279). The binding of PI(4,5)P2 to vinculin tail domain disrupts interactions between vinculin head and tail domains, thus freeing vinculin to bind talin, actin, VASP, or the Arp2/3 complex (120, 151, 265, 407). Similarly to PI(4,5)P2 regulation of α-actinin, PI(4,5)P2 binding of vinculin also appears to affect the dynamics of actin binding. In the absence of PI(4,5)P2 binding, vinculin cannot be released from actin, and therefore, its dissociation from focal adhesions is slower, leading to slow turnover rate of focal adhesions (44, 344). The COOH-terminal tail domain of vinculin consists of five amphipathic helices (12, 13). These helices are connected by short loops and adopt an antiparallel topology with their hydrophobic residues buried inside the bundle. The last helix is followed by a COOH-terminal arm that terminates into a five-residue hydrophobic hairpin. Extensive deletions within the COOH-terminal region of vinculin introduced structural perturbations and reduced PI(4,5)P2 binding. A significant decrease in PI(4,5)P2 binding was also observed for vinculin variants that perturb interactions between the NH2-terminal strap and helix bundle, suggesting that a rearrangement of this NH2-terminal strap may be required for PI(4,5)P2 binding (279). Thus PI(4,5)P2 binding of vinculin's COOH-terminal region may require a conformational change in the NH2-terminal strap to promote a higher affinity PI(4,5)P2 association. In addition to actin binding, interactions with PI(4,5)P2 were suggested to promote adhesion-site targeting of vinculin (120). However, recent studies provided evidence that mutations in the lipid-binding sites of the vinculin tail region do not significantly affect vinculin's subcellular localization (44, 324).
Talin is a ubiquitous cytosolic protein that is especially abundant in focal adhesions. It is capable of linking integrins to the actin cytoskeleton either directly or indirectly by interacting with vinculin and α-actinin. Talin is a large protein containing an NH2-terminal head (∼50 kDa) and a COOH-terminal rod domain (∼220 kDa). The NH2-terminal head of talin is globular, containing a FERM (four-point one, ezrin, radixin, moesin) domain composed of three lobes, F1, F2, and F3 or phosphotyrosine-binding (PTB) domain (63, 115). The COOH-terminal rod domain of talin is highly elongated, containing a series of helical bundles separated by linkers (122).
PI(4,5)P2 is a known talin activator, and it promotes talin binding to β-integrins (234), resulting in the formation of a ternary PI(4,5)P2/talin/integrin complex to promote integrin activation and clustering (56). PI(4,5)P2 binds to the COOH-terminal region of the FERM domain of talin and inhibits its association with the talin rod, thereby exposing the integrin-binding site of the molecule (123). Talin also interacts with the PI(4,5)P2-producing enzyme PIPK-Iγ, which regulates PI(4,5)P2 synthesis at synapses and at focal adhesions. PIPK-Iγ was shown to be recruited to focal adhesions by interactions with talin (75).
11. ERM family proteins
ERM proteins (e.g., Band 4.1, ezrin, and radixin) can directly link the actin cytoskeleton to the plasma membrane (28). Therefore, these proteins play a fundamental role in many cellular processes such as morphogenesis, cytokinesis, phagocytosis, and stabilization of intercellular junctions (149). A characteristic feature of ERM proteins is that they contain a FERM domain, coiled-coil region, and a COOH-terminal A/FBD (ERM actin and FERM binding) domain. In the inactive state, the NH2-terminal FERM domain interacts with the COOH-terminal A/FBD domain. Separation of this autoinhibition interaction is mediated by threonine phosphorylation and PI(4,5)P2 binding, leading to the exposure of the COOH-terminal actin-binding site (28).
The FERM domain is composed of three subdomains (A, B, and C), and one of them (subdomain C) contains a PTB/PH domain fold which, surprisingly, is not responsible for phosphoinositide binding of the FERM domain. The cocrystal structure of the FERM domain with the PI(4,5)P2 headgroup analog IP3 revealed that the binding site is composed of positively charged cleft between the subdomains A and C (132). However, mutagenesis studies suggested that also other regions of the FERM domain may contribute to phosphoinositide binding (19). Interestingly, a recent study revealed that the interaction of ERM proteins with PI(4,5)P2-containing membranes can be accurately regulated, because phospholipase C-mediated hydrolysis of PI(4,5)P2 in lymphocytes resulted in rapid release of ERM proteins from the plasma membrane (135).
Spectrin is a component of submembrane compartments, where it can cross-link transmembrane proteins, signaling proteins, membrane lipids, and the actin cytoskeleton to form complexes that enhance mechanical stability of the membranes (11). The α- and β-spectrins form antiparallel heterodimers that assemble into tetramers, which can be linked to protein scaffolds composed of F-actin and protein 4.1 at their distal ends. The ternary complex composed of spectrin, F-actin, and protein 4.1 forms the skeletal network of erythrocytes governing the stability and elasticity of the membranes.
Spectrin also interacts with phosphoinositides and appears to harbor two separate phosphoinositide binding sites. The pleckstrin homology (PH) domain of spectrin interacts with PI(4,5)P2 and PI(3,4,5)P3 with a moderate affinity (152, 174). Additional PI(4,5)P2 binding site is located in the calponin homology (CH) domain2 of βI- and βII-spectrin (6). Interestingly, recent studies suggested that the terniary structure of spectrin undergoes considerable changes upon interactions with phospholipids that might have implications for the function of the protein (65). The PH domain of spectrin was shown to target nonerythroid spectrin to PI(4,5)P2-rich domains in Golgi membranes (72, 225). Furthermore, inhibition of PI(4,5)P2 synthesis resulted in phosphorylation and redistribution of spectrin to the cytoplasm and in fragmentation of the Golgi apparatus (335). Spectrin also binds to phosphatidylethanolamine (PE) and PS (7, 65, 128, 305), and the binding regions for these lipids are located near the ankyrin binding site (7, 148).
B. Canonical Phosphoinositide-Binding Domains Found in Actin-Binding/Modulating Proteins
A number of protein domains evolved to interact with certain phosphoinositide species with a high specificity. These domains are found in a variety of signaling proteins as well as in proteins involved in the regulation of the cytoskeleton. The specific lipid binding allows these conserved domains to regulate the spatiotemporal targeting of regulatory/scaffolding proteins to the correct sites of action where they can recruit/activate other proteins to carry out their specific functions (Fig. 4) (210).
1. PH domain
The PH domain was first identified from pleckstrin, which is the major PKC substrate in platelets. Subsequently, this domain was shown to interact with phosphoinositides (136). PH domain is an ∼100-residue globular domain that is present in >200 copies in the human proteome (136, 210). It is found in numerous proteins including protein kinases, GTPase-activating proteins, guanosine nucleotide exchange factors, and lipid transport proteins (61). Different PH domains display specificity for either PI(4,5)P2, PI(3,4)P2, PI(3)P, or PI(3,4,5)P3, allowing the targeting of different PH domain-containing proteins to distinct cellular compartments. PH domains usually bind phosphoinositides through a “canonical” binding site, which is composed of the β1-β2 and β3-β4 strands, and the variable loops connecting them (95, 136, 317) (Fig. 4). However, in some PH domains, such as in β-spectrin, the phosphoinositide binding site is formed by two loops: β1-β2 and β5-β6 (15, 152). PH domains typically interact only with the phosphoinositol headgroup, and the binding does not involve significant membrane insertion of the domain (95), with the exception of PH-PLCδ1 (102).
2. PTB domains
The PTB domains display a high degree of structural similarity to the PH domains, and they are found in many cytoskeletal adaptor proteins including talin. Despite the structural similarity to PH domains, only PTB domains of tensin, Dab1, and ARH have been reported to bind phosphoinositides (212).
3. PX domains
The PX domain was originally found from phagocyte NADPH oxidase (phox) complex and named the phox or PX domain (296). PX domains are small, ∼130 amino acid modules that specifically bind to 3-phosphorylated phosphoinositides [PI(3)P and PI(3,4)P2]. These lipids are enriched at the endosomal membranes, and thus PX domains target proteins to the endosomal compartments. PX domains are present in, for example, sorting nexins (SNX), which are relatively large proteins involved in protein sorting, vesicular trafficking, and phospholipid metabolism (323, 331). Some PX domains, such as the ones from p40phox and SNX3, bind PI(3)P with high affinity, whereas most PX domains display only relatively low affinity for PI(3)P and thus other membrane binding domains are required to target these PX domain proteins to the correct membrane compartment (170, 299, 395). Structural studies on several PX domains revealed a common phosphoinositide-binding site where the tips of the three β-sheets and α-helices of the domain form a pocket where the PI(3)P headgroup can bind (27, 226) (Fig. 4). In addition to the phosphoinositide headgroup binding by electrostatic interaction, membrane insertion contributes to the phosphoinositide binding in PX domains (27).
4. FYVE domains
Fab1, YOTB, Vac1, and EEA1 (FYVE) domains are ∼70-residue protein motifs that interact with a relatively low affinity with PI(3)P and participate in vacuolar sorting and endocytosis (345). Almost 30 FYVE domain proteins are present in humans and among them are also proteins known to modulate the actin cytoskeleton. These include, for example, frabins, which are Cdc42 GEFs (262). The three-dimensional structure of FYVE domain consists of a zinc finger composed of a small α-helix and two β-hairpins. FYVE domains also contain a hydrophopic motif that inserts into the membrane bilayer (253) (Fig. 4). Accordingly, FYVE domains bind more strongly to PI(3)P embedded in membrane than to free PI(3)P (194). Because FYVE domains bind PI(3)P with low affinity, isolated domains are unable to localize to PI(3)P-rich membranes. Thus additional sequences that facilitate the dimerization of the domain or its interactions with other proteins such as small GTPase Rab5 are required for efficient localization of FYVE domain proteins to PI(3)P-rich membranes in vivo (82, 203).
III. Regulation of Small GTPases by Phosphoinositides
The small GTPases (also known as GTP-binding proteins or G proteins) form a heterogeneous group of proteins that regulate a wide variety of cellular processes by acting as molecular switches that activate and inactivate numerous downstream signaling pathways. The activities of small GTPases are regulated by guanosine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs). These proteins control the cycling of GTPases between the inactive GDP-bound and active GTP-bound forms. The activities of GAPs and GEFs are subsequently regulated by upstream signals, e.g., the growth factor receptors (161). Interestingly, many GEFs and GAPs contain phosphoinositide binding modules, such as PH, FYVE, or BAR domains, thus facilitating the activation/inactivation of the small GTPase signaling cascades at spatially restricted membrane environments. For example, Sos1 and Vav, which are GEFs for Rac1 GTPase, are activated through PI(3,4,5)P3 binding to their PH domain (Fig. 5) (68). Moreover, some GTPase interacting proteins contain catalytic domains capable of metabolizing phosphoinositides. This provides feedback loops to GTPase-phosphoinositide signaling and can modulate the subcellular localizations of proteins that interact with both GTPases and phosphoinositides (74). Finally, a recent study revealed that Rho family GTPases bind PI(4,5)P2 also directly through their polybasic region, demonstrating that membrane phosphoinositides regulate small GTPases at several different levels (402).
Based on structural features, the small GTPases can be divided into five groups: Rab, Ran, Ras, Rho, and Sar1/Arf families (356). From these, the members of the Rho and Arf families are most directly involved in regulation of the dynamics and organization of the actin cytoskeleton.
A. Rho Family GTPases
The RhoA GTPase has a pronounced role in the formation and regulation of focal adhesion complexes and contractile actomyosin bundles such as stress fibers (288, 308). RhoA induces actin polymerization at focal adhesions by activating the Dia1 formin and inhibits actin filament disassembly by initiating a signaling cascade that leads to phosphorylation and subsequent inactivation of the ADF/cofilin family of actin filament severing/depolymerizing proteins through the action of LIM kinases (146, 230, 383). Furthermore, RhoA promotes contractility by activating the myosin light-chain kinase through ROCK kinase (364). As described above, many Rho GEFs and GAPs contain PH domains, and thus their activities and/or subcellular localizations are likely to be regulated by phosphoinositides. In addition, active GTP-bound RhoA can stimulate PIPK-I activity through ROCK, thereby regulating controlled synthesis of PI(4,5)P2 (5, 55, 276).
Small GTPase Rac1 plays an important role in many developmental processes and cancer cell metastasis by promoting cell locomotion through the formation of lamellipodia and membrane ruffles (137). Rac1 promotes the polymerization of branched actin filament networks by activating the Arp2/3 complex through interactions with a multisubunit WAVE protein complex. WAVE family members are large actin-binding proteins that regulate the activity of the Arp2/3 complex (358). Similarly to RhoA, Rac1 also inhibits actin filament disassembly by inactivating ADF/cofilins through PAK and LIM kinases (83, 156, 228). Rac1 activity is regulated by phosphoinositides. Furthermore, Rac1 interacts with PIPK-I independently of its guanine-nucleotide bound stage (362) and thus promotes the synthesis of PI(4,5)P2 at the plasma membrane (47).
Cdc42 has a critical role in promoting cell polarity in various organisms (137). In mammalian cells, activated Cdc42 signaling induces the formation of filopodia by promoting actin polymerization through N-WASP and/or Ena/VASP proteins, and membrane deformation through IRSp53 (185, 214, 290, 311). Like Rac1, Cdc42 also inhibits actin filament disassembly by activating PAK kinases (83). Regulators of Cdc42, such as the frabin family GEFs, contain PH and FYVE domains, thus linking Cdc42 activity/localization to phosphoinositide signaling (262).
B. Arf Family GTPases
In mammals, the Arf (ADP ribosylation factor) family of GTPases is composed of at least 22 members that include 6 Arf proteins, 2 Sar proteins, and 14 Arl proteins (166, 320, 356). The members of this GTPase family play important roles in many routes of membrane trafficking, and their activities are often linked to the actin cytoskeleton.
Most Arfs localize to the Golgi apparatus and certain endosomal compartments, but Arf6 localizes primarily to the plasma membrane and endosomes. Arfs are involved in the rearrangements of the actin cytoskeleton by inducing PI(4,5)P2 production both at the plasma membrane and Golgi (261). Furthermore, Arf6 was shown to be involved in the recruitment of Rac1 to the membrane ruffles and modulation of its activity (80, 304). Arf6 can alter plasma membrane phospholipid composition by activation of PIPK-I through a direct interaction (4, 145, 183). Arf also binds to phosopholipase D (PLD), which cleaves phosphatidylcholine (PC) to generate phosphatidic acid (PA) (30). PA synergizes with Arfs to enhance the activity of PIPK-I kinases and thus increase in the production of PI(4,5)P2.
IV. Membrane Deformation by Actin Modulating Proteins
The number of proteins that have been shown to directly influence membrane curvature has expanded in recent years, and it has become clear that many essential cellular functions rely on the formation of membrane domains characterized by a specific degree of curvature. Both the actin cytoskeleton and membrane-associated proteins alone can induce membrane curvature (240), but recent findings provided evidence that the two systems are often tightly interconnected to drive local membrane deformation in cells. In this section we discuss the functions of BAR and ENTH/ANTH domains as well as dynamin, because these membrane-deforming modules have been linked to actin dynamics. In addition, tandem C2 domains as well as components of the ESCORT-III complex were recently shown to induce membrane deformation, but they are not discussed here because these proteins have not been directly linked to actin dynamics (160).
A. BAR Domain Protein Superfamily
The BAR (Bin-Amphiphysin-Rvs) domain protein superfamily is composed of a vast number of multidomain proteins that share a structurally homologous coiled-coil membrane binding/deforming module, the BAR domain. These proteins are typically intimately linked to actin dynamics, because in addition to the membrane deforming BAR domain they typically contain actin-binding WH2 (WASP homology 2) domains, Rho GAP/GEF domains, PDZ and SH3 domains that link them to actin-regulatory networks. In cells, these proteins localize to regions characterized by dynamic interplay between membranes and cytoskeletal components.
The BAR domains are homodimeric protein modules that typically interact with cellular membranes through electrostatic interactions between the positively charged regions located close to the poles of the dimeric domains and the negatively charged phospholipid headgroups. BAR domains bind membranes cooperatively to make rigid scaffolds, which force membranes to bend according to their intrinsic curvature of the domain scaffold, thus inducing the formation of membrane tubules (110, 113). Interestingly, the intrinsic curvature varies between different types of BAR domains (Fig. 6). Variability in the curvature of membrane tubules induced by different BAR domains also arises from additional motifs that insert into the bilayer as well as from tilting of the BAR domain array relative to the axis of the deformed membrane tubule (109, 112, 315). Some BAR domain-containing proteins also harbor other lipid-binding domains that specify their functions. These include PX and PH domains that are thought to enhance the membrane binding of BAR domain proteins and provide specificity towards certain phosphoinositides (210). The specific degree of curvature and mechanistic variability within the BAR domain family may serve to distinguish their functions in cells possibly by localizing these adaptor molecules to specific sites of action. Interestingly, it was recently shown that the combination of PX and BAR domains in sorting nexin 9 (SNX9) displays variability in the degree of intrinsic curvature (380), possibly allowing it to bind vesicles of different sizes.
1. BAR/N-BAR domains
The founding member of this expanding family of membrane-binding modules is the canonical BAR domain. The discovery of the BAR domain structure opened new avenues in understanding how rigid proteins can force membranes to adopt appropriate curvature through strong electrostatic interactions and scaffolding properties of the domain (291). The BAR domain consists of two kinked α-helical monomers that dimerize in an antiparallel fashion to form a “banana”-shaped molecule (291, 360). BAR domains contain a group of positively charged residues in their concave face that interacts with the negatively charged headgroups of phospholipids. Mutations in these positively charged residues result in lack of membrane binding/deforming activity, demonstrating that electrostatic interactions are crucial for the membrane curvature-generating activity of the BAR domains (291). Structural and biochemical studies revealed that some BAR domains (e.g., endophilin and amphiphysin) additionally contain disordered NH2-terminal regions that form amphipathic α-helices upon membrane binding and can serve as anchors that penetrate into the membrane bilayer (112). The insertion of these ∼25-residues-long amphiphatic α-helices increases the affinity of so-called N-BAR domains to membranes and induces local bending by perturbations in the lipid monolayer, thus enhancing the membrane deformation activity of these domains (10, 24, 96, 112).
BAR domain and N-BAR domain proteins are typically involved in endocytosis. For example, SNX9, endophilin, and Tuba interact with N-WASP through their SH3 domains and are thus capable of linking BAR domain-induced membrane tubulation to N-WASP-Arp2/3 complex induced actin filament assembly (182, 275, 334). Whether the physiological functions of BAR and N-BAR domain proteins are to sense membrane curvature and subsequently recruit the actin polymerization machinery to the proper sites in cells during endocytosis, or if these proteins also actively generate membrane curvature during endocytic internalization remains to be determined.
2. F-BAR domains
Database searches for BAR domain homologs revealed the existence of a protein module known as FER-CIP4 homology (FCH) domain (9), which is present in many proteins involved in cytoskeletal dynamics. These domains turned out to display strong membrane deforming activity both in vitro and in vivo, and hence, they were named F-BAR domains (FCH-BAR domains) (158). It was also demonstrated that the minimal region required for membrane deformation is longer than the FCH region, and therefore, the domain is also known as EFC (extended FC) domain (366). The three-dimensional structures of these domains revealed a banana-shaped dimeric bundle composed of six α-helices. However, F-BAR domains are more elongated and display a more shallow degree of intrinsic curvature compared with the BAR/N-BAR domains (139, 333). This is in good agreement with observations that F-BAR domains typically induce wider membrane tubules compared with the ones induced by BAR/N-BAR domains. Cryoelectron micrographic studies of membranes decorated with F-BAR domains revealed that they drive membrane tubulation by forming a scaffold around the membrane, by using both lateral and distal interdomain contacts. Furthermore, these studies proposed that F-BAR domains can bind flat membranes using residues that are dispensable for membrane deformation (109).
Although most F-BAR domains appear to interact with membranes through their concave surfaces and induce positive membrane curvature (i.e., bind to the outer leaflet of the membrane tubules), a recent study revealed that the F-BAR domain protein srGAP2 induces the formation of membrane protrusions in cells and promotes negative membrane curvature in vitro (129). Although the exact mechanism by which srGAP2 F-BAR domain is capable of inducing opposite membrane curvature compared with other so far characterized F-BAR domains is currently unclear, this study shows that the F-BAR domain is a more versatile regulator of membrane dynamics than previously considered (129).
In full-length proteins, the F-BAR domain is typically associated with SH3 or RhoGAP domains to link F-BAR domain induced membrane deformation to cytoskeletal dynamics. For example, the F-BAR domain protein Toca1 binds and activates N-WASP through its SH3 domain, whereas WRP binds another Arp2/3 activator, WAVE1 (143, 339). Furthermore, srGAP1 controls neuronal migration by functioning as a GAP for Cdc42 in the Slit-Robo pathway (392). Interestingly, a recent study revealed that F-BAR domain proteins Toca1 and FBP17 promote membrane curvature-dependent actin polymerization by recruiting N-WASP and WIP (WASP interacting protein) in an appropriate orientation to membranes of correct curvature (357). Thus these F-BAR domain proteins appear to coordinate actin filament assembly to membrane curvature sensing/generation during cellular processes involving the formation of plasma membrane invaginations.
3. I-BAR domains
Due to lack of detectable sequence similarity to other BAR domains, the I-BAR domain was first discovered as a conserved domain that was found at the NH2-terminal regions of five mammalian proteins including IRSp53 and missing-in-metastasis (MIM). Therefore, this domain was initially named the IM domain (IRSp53/MIM homology domain) (396). Earlier studies showed that expression of the IM/I-BAR domain induced the formation of F-actin-rich plasma membrane protrusions in cells and proposed that this domain would cross-link actin filaments (250, 396). Interestingly, the crystal structures of IRSp53 and MIM IM/I-BAR domains revealed a gently curved α-helical antiparallel dimer, which displayed significant structural similarity to canonical BAR domains. However, instead of being banana-shaped like BAR and F-BAR domains, the IM/I-BAR domains are “cigar”-shaped (208, 250) (Figs. 4 and 6). Subsequent studies revealed that IM/I-BAR domains bind phosphoinositides with high affinity and are capable of deforming phosphoinositide-rich membranes into tubules. The binding of IM/I-BAR domains to phosphoinositides involves electrostatic interactions between the positively charged poles located at the distal ends of the I-BAR dimer and the negatively charged headgroups of phosphoinositides (237, 314, 315, 348). Importantly, the IM/I-BAR domain utilizes a convex face to bind membranes and thus it drives membrane deformation in the opposite direction compared with BAR/N-BAR and F-BAR domains (237, 315, 348). Thus this domain was renamed inverse BAR (I-BAR) domain.
Similarly to other BAR domains, the lipid binding of I-BAR domains does not seem to be specific towards certain phospholipid, although the I-BAR domain of MIM favored binding to PI(4,5)P2. Interestingly, similarly to N-BAR domains, the binding of a subset of I-BAR domains (at least the ones of mammalian MIM and ABBA) involves membrane insertion of an NH2-terminal flexible amphipathic region composed of ∼30 residues of the I-BAR domain. Penetration of this region into the membrane bilayer enhances the membrane binding/deforming activity of these domains (315).
I-BAR domain proteins are tightly linked to cytoskeletal dynamics. IRSp53 interacts with many actin-binding proteins such as Eps8, WAVE, and N-WASP, whereas MIM interacts with cortactin, which is a regulator of Arp2/3-dependent actin filament assembly in cells (215, 327). Furthermore, most I-BAR domain proteins contain a COOH-terminal actin monomer binding WH2 domain that at least in MIM appears to be central for the cellular role of the protein (J. Saarikangas and P. Lappalainen, unpublished data). It is also worth noting that although the isolated I-BAR domains are capable in inducing membrane tubulation in vitro in the absence of actin filaments, in cells the efficient formation of filopodia-like membrane protrusions by I-BAR domains depends on the presence of actin filaments (237, 314, 399).
B. ENTH/ANTH Domains
ENTH/ANTH (epsin/AP180 NH2-terminal homology) domains are typically found in proteins involved in clathrin-mediated endocytosis, although they are also implicated in vesicular trafficking from the trans-Golgi network. The ENTH/ANTH domains are often located at the NH2-terminal regions of the proteins, whereas the COOH-terminal regions of the ENTH/ANTH domain proteins harbor clathrin-binding or clathrin adaptor-protein binding motifs (157).
In mammals, the ENTH domain is present in four proteins: epsins 1–3 and epsinR (157, 209). In budding yeast, there are five ENTH domain proteins, which have been linked to the regulation of the actin cytoskeleton in addition to vesicular trafficking (387). Most ENTH domains interact with cellular membranes by binding PI(4,5)P2, but mammalian epsinR binds PI(4)P and budding yeast Ent3p, as well as Ent5p, bind almost exclusively to PI(3,5)P2 (90, 103, 108, 159, 251, 342).
In a purified in vitro system, the epsin ENTH domain is capable of deforming membranes into narrow tubules, and when overexpressed in cells, it induces membrane invaginations (103). The three-dimensional structure of the ENTH domain is composed of an α-helical fold that contains a distorted NH2-terminal extension (Fig. 4). Interestingly, when ENTH was cocrystallized with I(1,4,5)P3, which mimics the PI(4,5)P2 headgroup, the “distorted” NH2-terminal extension folded into an α-helix (helix-0) (103). Mutation of critical residues or deletion of the helix-0 resulted in lack of PI(4,5)P2 binding, demonstrating the critical role of this ENTH domain region in lipid binding (103). Furthermore, mutational analyses provided evidence that the presence of an intact ENTH domain with PI(4,5)P2 binding activity is critical for receptor-mediated endocytosis (103, 159, 263). The helix-0 of ENTH domains displays amphipathic properties and is considered to be the driving force behind ENTH-mediated membrane deformation (103). The insertion of helix-0 into the membrane monolayer could induce local bending, which would result in the formation of a membrane bud that could subsequently be stabilized by clathrin.
ANTH domains are structural homologs to the ENTH domain, but they bind PI(4,5)P2 with a different mechanism. The ANTH domain is present in many proteins that link actin dynamics to endocytosis. These include, for example, Sla2/HIPr1, which interacts with plasma membrane via its ANTH domain, and with actin filaments via its talin homology domain. In addition, this protein also binds clathrin (87–89, 205). The differences between the PI(4,5)P2 binding mechanisms of ANTH and ENTH domains were discovered from the crystal structure of the ANTH domain of AP180. The structure revealed a similar fold compared with ENTH domains, but the ANTH domain contained three additional α-helices (Fig. 4). Importantly, the PI(4,5)P2 binding site was different compared with one of ENTH domains (104, 232). ANTH domains bind PI(4,5)P2 via a small positively charged patch of amino acids without membrane penetration (103, 104). The functionality of this structural difference is also demonstrated by the lack of membrane deformation activity in ANTH domains (103, 342).
Dynamins are large GTPases that are linked to membrane scission events. They are found, for example, in caveolae, clathrin-coated vesicles, and phagosomes, but also in membrane ruffles and lamellipodia (184, 297). Many dynamin-like molecules are found in the human proteome, but only the best characterized classical dynamins 1, 2, and 3 are discussed here. In addition to their large GTPase domain, dynamins also contain many targeting domains that are believed to recruit the protein to the desired site of action. These include, for example, phosphoinositide-binding PH domains as well as proline-rich (PRD) and SH3 domains.
Dynamin is distinguished from other GTPases by its tendency to oligomerize at low salt and capability to bind and tubulate negatively charged membranes (354). Although the PH domain of dynamin has relatively low affinity to PI(4,5)P2, the oligomerization of dynamin into a helical coat provides high avidity to the membranes. After recruitment to the site of action, dynamin polymerizes into a coat surrounding the membrane invagination and a subsequent GTP hydrolysis triggers a conformational change that lengthens the dynamin coat. This is believed to be responsible for the scission of the membrane tubule into vesicles, although some controversy concerning the actual mechanism of membrane scission by dynamin exists (297). The PH domain is essential for the function of dynamin in endocytosis as a single point mutation in this domain causes a dominant negative effect (1, 206, 368). Interestingly, a recent study suggested that the role of PH domain in dynamin is not only to localize the protein to correct sites, but also to cluster PI(4,5)P2 to promote membrane scission (21).
Interactions with other proteins that are responsible for the elongation of membrane invaginations in endocytosis, mainly BAR domain proteins, are believed to be important for the correct function of dynamin in cells (69). Dynamin also interacts with certain actin binding proteins such as cortactin (241) and syndapin (300). Furthermore, dynamin-2 was reported to regulate actin filament organization and stability in concert with cortactin both in vitro and in cultured osteosarcoma cells (257, 326).
V. Phosphoinositide Signaling in Actin-Dependent Cellular Processes
As described in the previous sections, the biochemical activities and subcellular localizations of many actin-binding and signaling proteins are regulated by phosphoinositides. Furthermore, many membrane deforming proteins, such as the ones harboring a BAR domain, function in cells together with the actin cytoskeleton. In this section, we discuss the cellular processes that rely on the tightly controlled interplay between phosphoinositide-rich membranes and the actin cytoskeleton.
Endocytosis is a fundamental cellular event that is responsible for nutrient uptake of cells, regulation of signaling pathways, and morphological processes. During endocytosis, external cargos are internalized via complex molecular machinery. This machinery is composed of core molecules regulating the sequential endocytic events from cargo recognition to the scission of the internalized cargo-containing vesicle. The endocytotic machinery is tightly regulated by a set of adaptor proteins, and it utilizes forces generated by membrane deforming proteins, GTP hydrolysis, and actin polymerization. Physiological importance of this well-oiled machinery is demonstrated by the fact that mutations in many genes encoding endocytotic proteins result in severe diseases (252), and by the links between defects in endocytotic processes and many neurological malfunctions (41, 42).
Currently, at least 11 different endocytotic routes have been established, including the clathrin-mediated, caveolin-mediated, and Arf6-dependent endocytotic routes as well as macropinocytosis and phagocytosis (77). Due to space limitations, we will only highlight the clathrin-mediated endocytosis and phagocytosis as two different examples of internalizing extracellular particles. Both processes are driven by a tight and controlled interplay between the plasma membrane phosphoinositides and the actin cytoskeleton.
1. Clathrin-mediated endocytosis
Elegant live-cell imaging studies on budding yeast (167) and mammalian cells (244) elucidated the temporal requirements of different protein machineries during endocytosis. On the basis of these and other studies, the clathrin-mediated endocytosis (CME) can roughly be divided into four phases, each characterized by the appearance of a specific set of proteins (169). However, the wide range of molecules that bear similar functions during CME indicates that this process might apply distinct sets of tools to suite different purposes, e.g., according to cargo requirements. It is also important to note that actin is absolutely required for the early stages of endocytosis in budding yeast to provide the force for the plasma membrane against turgor pressure, whereas in mammalian cells actin appears to be required only for the vesicle scission during later steps of endocytosis (2).
CME begins with the recognition of the cargo followed by the recruitment of clathrin. This is mediated by so-called adaptor proteins such as AP180 that binds to PI(4,5)P2 through its ANTH domain. Together with AP2, it induces the recruitment and assembly of a curved clathrin lattice to the plasma membrane (104, 352). The formation of membrane invagination is subsequently aided by membrane deforming proteins such as epsins that, upon binding to PI(4,5)P2, insert an amphiphatic helix into the membrane (103). Also, actin is present already during the first steps of the internalization of the clathrin-coated pits, although it does not appear to be necessary for the formation of the clathrin-coated buds (400). Actin polymerization is regulated at the CME by cortactin, N-WASP, and Arp2/3 (20, 37, 246). Actin filaments are linked to the clathrin pits by adaptor proteins such as HIP1R (87, 88, 169). The vesicle neck constriction is facilitated by BAR and possibly F-BAR domain proteins, such as amphiphysin, endophilin, SNX9, and FBP17, that have the capacity to both drive and stabilize existing membrane curvature at the vesicle neck as well as to recruit other important molecules such as dynamin and the actin polymerization machinery to the site of vesicle scission (309, 333, 389). Subsequently, dynamin molecules oligomerize into a helical lattice around the vesicle neck and, upon GTP hydrolysis-mediated conformational change, provide force for vesicle scission (233, 297). In addition to dynamin, forces produced by actin polymerization also appear to be essential for the vesicle scission. It is thus most likely a cooperative event, since dynamin also binds to cortactin (241) and activates Arp2/3-mediated actin polymerization, which reaches its peak at the time of the scission (167, 245). A recent theoretical study based on the known endocytotic events in yeast proposed a mechanochemical model for the progression of endocytosis where membrane curvature is coupled to biochemical reactions through positive-feedback loops to facilitate the progression of cargo internalization (221). Following internalization, the endocytotic factors rapidly dissociate from the vesicle membrane. This is believed to be facilitated by dephosphorylation of PI(4,5)P2 by inositol 5-phosphatases such as synaptojanin, which interacts with dynamin during the last stages of endocytosis (62, 242). In summary, CME represents a complex interplay between regulation of the plasma membrane PI(4,5)P2 and the actin cytoskeleton.
As a part of the innate and adaptive immunity, phagocytosis is a crucial defense mechanism against invading microorganisms. Macrophages, neutrophils, and dentritic cells are the major phagocytotic cells in mammals. They carry out engulfment of large extracellular particles, mainly microbes, which are subsequently trapped in intracellular vacuoles and processed for antigen presentation (101). The formation of a phagosome is triggered upon direct or indirect recognition of the microorganism. This is followed by the clustering of the recognition receptors that initiates a signaling cascade activating the small GTPases Rac1 and Cdc42 (39). This results in rearrangement of the actin cytoskeleton that drives the formation of the phagosome.
PI(4,5)P2 accumulates to early phagosomes via local activation of PIPK-I and plays a crucial role in actin remodeling through recruitment/activation of actin-binding proteins at phagosome membrane (26, 58). However, shortly after the initiation of the phagosome formation, PI(4,5)P2 is hydrolyzed by PLC, resulting in the disassembly of actin from the base of the phagosomes, a process which is essential to complete the internalization. Thus PI(4,5)P2 hydrolysis and actin disassembly show similar spatial and temporal kinetics during phagocytosis and, importantly, actin detachment from the forming phagosomes can be inhibited by the impairment of PI(4,5)P2 hydrolysis in macrophages (48, 328). Phagocytotic receptors also activate PI3K, which phosphorylates PI(4,5)P2 to PI(3,4,5)P3 (377). Accumulation of PI(3,4,5)P3 is important for the sealing of the phagocytotic cup through the action of an unconventional myosin, MyoX (60). This myosin motor protein localizes to the phagocytic site through its PI(3,4,5)P3 specific PH domain and is required for membrane extension that seals up the phagocytic cup (60). Finally, after sealing, the PI(3,4,5)P3 is depleted through the inositol 5-phosphatase completing the internalization of the particle.
B. Cell Motility
Cell migration is critical for wound healing, immune responses, and a number of developmental processes. It also plays an important role during many pathological conditions, such as cancer cell invasion. Cells can actively move by using at least two fundamentally distinct mechanisms: 1) lamellipodia-dependent cell migration, which is typical for cell movement on two-dimensional environments, and 2) membrane blebbing-dependent migration that occurs when cells invade into the three-dimensional matrix.
1. Lamellipodia-dependent cell migration
Lamellipodia-dependent cell migration is driven by extension of actin-rich protrusions, filopodia and lamellipodia, at the leading edge of motile cells (204). Filopodia are thin (diameter of 100–300 nm) plasma membrane protrusions that are packed with a tight bundle of unbranched actin filaments. Filopodia have been described as “antennae” that cells use to probe their environment, thus serving as pioneers in ameboid-type cell migration (130, 236). In contrast to filopodia, the lamellipodia consist of an actin filament network generated by the Arp2/3 complex and its activators. Coordinated polymerization of the actin filament network against the plasma membrane induces the formation of lamellipodial protrusions that drive the advancement of the leading edge of the cells. This is followed by formation of adhesions and retraction of the cell's tail (204).
Phosphoinositides play a key role in lamellipodia-dependent cell migration by inducing actin filament assembly at the plasma membrane and by regulating the direction of cell movement during chemotaxis. Studies using neutrophils and the amoeba Dictyostelium discoideum demonstrated that PI(3,4,5)P3 accumulates at the front of chemotactic cells, where it induces a rearrangement of the actin cytoskeleton (Fig. 7). Rapid production of PI(3,4,5)P3 in response to activated chemoattractant receptors is promoted by PI3K (369). Furthermore, the regulated hydrolysis of PI(3,4,5)P3 by phosphatase and tensin homolog (PTEN) at the retracting tail called the uropod is believed to suppress lateral pseudopod formation, thus keeping cells on track (388). However, although numerous studies have indicated that the polarized PI(3,4,5)P3 gradient is important for directional cell migration, it does not appear to be essential for chemotaxis in all conditions (144, 181).
Importantly, recent studies identified PIPK-Iβ and -γ to be concentrated at the uropod of neutrofils during chemotaxis. These kinases contribute to rear retraction, cell polarization, and directed movement by producing PI(4,5)P2 and by acting as scaffolds for uropod proteins such as ERM proteins (196, 224).
PI(3,4,5)P3 gradients promote chemotaxis mainly by inducing actin filament assembly through activation of the Rho family GTPases. For example, in neutrophils, elevated PI(3,4,5)P3 levels polarize the cell towards the chemoattractant by recruiting the atypical Rac1 GEF, DOCK2, which binds PI(3,4,5)P3 via its DHR-1 domain and phosphatidic acid via a cluster of polybasic amino acid residues (191, 268). DOCK2 activates small GTPase Rac1, which in turn activates Arp2/3-mediated actin filament polymerization to drive cell migration towards the chemoattractant (268).
Also PI(4,5)P2 plays an important role in directional cell migration. As described in previous sections and displayed in Figure 3, PI(4,5)P2 typically inactivates proteins that inhibit actin filament polymerization or promote actin filament disassembly, whereas, e.g., WASP family proteins that induce actin filament assembly through the Arp2/3 complex are activated by PI(4,5)P2. Thus actin filament assembly is enhanced close to the membranes containing high levels of PI(4,5)P2. However, recent studies by the Condeelis laboratory (260, 371) provided evidence that also PI(4,5)P2 hydrolysis plays an important role in directional migration of carcinoma cells. These studies proposed that regulated hydrolysis of PI(4,5)P2 by PLC-γ leads to induction of actin filament assembly through activation of ADF/cofilins and consequent production of new actin filament barbed ends through actin filament severing (260, 371).
2. Membrane blebbing
Membrane blebbing occurs in many cellular events such as apoptosis and cytokinesis and has recently been also recognized as a form of cell locomotion (45, 92). The role of membrane blebbing in cell migration has raised substantial interest since many cancer cells, which utilize protease-independent amoeboid-like movement to invade in three-dimensional extracellular matrices, display bleblike structures (316). Blebs are described by a local disruption between the cortical cytoskeleton and the plasma membrane. The bleb expands ∼1–2 μm from the plane of the plasma membrane and has a characteristic life cycle of ∼1 min. Local disruption of the membrane cytoskeleton is launched by specific extracellular cues. The force for the expansion of the plasma membrane is created by the internal hydrostatic pressure of the cell and does not involve forces produced by actin polymerization. ERM family of membrane-actin cytoskeleton linker proteins such as ezrin are subsequently recruited to the bleb, and the expansion is terminated by the local polymerization of actin filaments. Finally, the bleb retracts through ROCK/RhoA-induced contractile actomyosin forces (45, 46, 92). Formation of membrane blebs is also linked to phosphoinositide signaling, because membrane blebbing can be induced by depleting/inactivation of proteins that act as linkers between phosphoinositide membrane and the actin cytoskeleton (e.g., the ERM family proteins) or by hydrolyzing/depleting PI(4,5)P2. Therefore, in the light of present evidence, it seems likely that elevated PI(4,5)P2 concentration at the plasma membrane maintains strong actin-membrane interactions through the ERM proteins and thus suppresses the formation of membrane blebs, whereas loss of these components [cortical F-actin, ERM proteins, PI(4,5)P2] predisposes cells to blebbing (46, 304, 332, 382).
3. Podosomes and invadopodia
Podosomes and invadopodia are adhesive actin-rich membrane protrusions facing the cell substratum. Both structures are involved in the degradation of extracellular matrix and display similar molecular architectures composed of Rho family GTPases, dynamin, actin nucleating proteins (Arp2/3, formins), actin nucleation-promoting factors (cortactin, N-WASP, WIP), and actin cross-linking and adhesion proteins (α-actinin, vinculin, zyxin). Podosomes and invadopodia display also similarities to focal adhesions, and their formation is often detected in the vicinity of focal adhesions (50, 270). However, there are also important differences between podosomes and invadopodia that justify the different nomenclature. These structures are typically present in different type of cells; podosomes are found in osteoclasts, monocytes, endothelial cells, macrophages, and dentritic cells, whereas invadopodia are mainly present in transformed cancer cells. There are also differences in the number and size of these structures, as podosomes tend to be smaller yet more frequent and dynamic compared with invadopodia (121, 217).
Phosphoinositides have been implicated to regulate these structures, although further studies are required to elucidate the exact mechanisms of phosphoinositide regulation in podosomes and invadopodia. In osteoclast podosomes, both PI(4,5)P2 and PI(3,4,5)P3 appear to regulate the actin cytoskeleton through Rho GTPases, gelsolin, and WASP (50). Interestingly, a recent study demonstrated that PI(3,4)P2 is concentrated to podosomes/invadopodia of c-Src transformed NIH3T3 cells through local PI3K activity. Importantly, sequestration of PI(3,4)P2 with a specific PH domain inhibited the formation of podosomes/invadopodia (270). In this regard, Tsk5/FISH is an interesting candidate for the initiation of podosomes/invadopodia formation. This adaptor protein localizes to podosomes/invadopodia via its PI(3,4)P2-specific PX domain and is required for their formation (222, 329). Tsk5/FISH can recruit N-WASP to specific membrane domains and consequently initiate Arp2/3-mediated actin assembly, providing an important new link between PI(3,4)P2 signaling and actin filament assembly (270).
During cytokinesis, cells undergo a series of actin-driven morphological changes, which are regulated by similar signaling pathways to the ones involved in cell migration. However, whereas migrating cells are unipolar, dividing cells are bipolar and constricted from the middle by a beltlike structure known as the cleavage furrow (18, 189). Both Arp2/3 complex- and formin-mediated actin polymerization are required for cytokinesis (288). The polymerization of actin filaments at each end of the dividing cell maintains tension while the contractile actomyosin structure at the cleavage furrow drives the separation of the cell into two daughter cells.
The phosphoinositides PI(4,5)P2 and PI(3,4,5)P3 play fundamental roles during cytokinesis. Their generation and hydrolysis are both spatially and temporally under tight regulation by PI kinases and phosphatases during this process (163). When Dictyostelium discoideum rounds up for cytokinesis, PI(3,4,5)P3 is not present at detectable levels. At this stage, PI3K is inhibited and high levels of PI(3,4,5)P3 phosphatase (PTEN) are present. During subsequent cell elongation, PI3K localizes to the poles of the dividing cell to promote a local increase of PI(3,4,5)P3 (Fig. 7). The increased PI(3,4,5)P3 levels at the cell poles induce actin polymerization through activation of Rho family GTPases to result in cell elongation by pushing the two poles apart (162, 163).
Also PI(4,5)P2 plays an important role in the correct progression of cytokinesis. PI(4,5)P2 localizes specifically to the cleavage furrow, and inhibition of PI(4,5)P2 production results in defects in cell division (Fig. 7) (29, 86, 100, 406). Accordingly, PTEN, a phosphatase that catalyzes dephosphorylation of PI(3,4,5)P3 to generate PI(4,5)P2, specifically localizes to the cleavage furrow (255). The exact role of PI(4,5)P2 at the cleavage furrow is not known, but it may regulate the activities of several actin binding proteins such as profilin and ADF/cofilins and anillin. Anillin is an actin, myosin II, and septin binding protein that is recruited to the cleavage furrow by PI(4,5)P2 through its PH domain (98, 99, 177, 346). Finally, a good candidate for a protein regulating the actin cytoskeleton during cell division in response to PI(4,5)P2 is moesin, an ERM domain-containing protein that is activated upon binding to PI(4,5)P2 (28). Recent studies revealed that moesin regulates cortical rigidity and cell rounding during cytokinesis in cultured Drosophila cells (190).
D. Cell Polarity
Polarized cells have asymmetric distribution of cellular organelles and structures, and this is important for the proper differentiation, proliferation, and morphogenetic processes of many cell types. Consequently, loss of cell polarity can lead to developmental disorders and cancer. Recent studies revealed that spatial localization of phosphoinositides PI(4,5)P2 and PI(3,4,5)P3 to distinct compartments in epithelial cells is crucial for the development and maintenance of epithelial polarity (Fig. 7) (116, 220, 235). By using PH domains that specifically recognize different phosphoinositides, PI(4,5)P2 was shown to predominantly localize to the apical domain, whereas PI(3,4,5)P3 localized to the basolateral domain. PI(4,5)P2 recruits annexin-2 to the apical domain, which in turn localizes Cdc42 to this site to induce the formation of apical actin network. Cdc42 is also responsible for the localization of the Par6/aPKC complex to the apical domain that is required for polarity establishment in three-dimensional cell culture model (235). Interestingly, a recent study demonstrated that aPKC-λ is also essential for basolateral restriction of PI(3,4,5)P3 in polarized epithelial cells (355). However, more studies are required to reveal the exact role of PI(3,4,5)P3 at the basolateral regions of epithelial cells.
VI. Interplay Between the Actin Cytoskeleton and Phosphoinositides in Pathogenesis
Abnormally regulated phosphoinositide signaling is a hallmark of many cancers (393). PI(3,4,5)P3 phosphatase PTEN is a tumor suppressor that is commonly mutated in cancers (319). Mice harboring only one allele of PTEN are predisposed to cancer (73, 353). Furthermore, mutations and gene amplifications of PI3Ks have been identified in many cancer patients, further demonstrating the central role of PI(3,4,5)P3 in cancer progression (171, 172, 403).
Given the pronounced role of PI(3,4,5)P3 and PI(4,5)P2 in maintaining epithelial polarity, these lipids may be involved in epithelial to mesenchymal transition (EMT) of cancer cells (179). During EMT, epithelial cells lose polarity and cell-cell adhesions, become motile, and invade into surrounding tissues. Interestingly, overexpression of PTEN in chick embryo mesodermal cells inhibits EMT (213). The exact mechanisms by which the elevated PI(3,4,5)P3 levels can induce EMT are still largely unclear. One possibility is that an increase in PI(3,4,5)P3 can attribute to the increased plasma membrane localization of molecules that promote cell motility. These include, for example, Vav and Tiam which bind PI(3,4,5)P3 through their PH domains and activate Rac via their GEF domain (see Fig. 4), thus reorganizing the actin cytoskeleton to promote cell invasion and metastasis.
B. Bacterial Pathogens
Pathogenic bacteria are masters in developing strategies that allow them to harness the cellular machineries to promote their replication and spreading. Consequently, regulation of the host cell phosphoinositides is a key mechanism facilitating the infection process of many bacteria including Legionella pneumophila, Salmonella enterica, Shigella flexneri, Listeria monocytogenes, and Yersinia ssp. Many bacteria use their own PI-metabolizing enzymes or alternatively can recruit and activate host cell enzymes (385). Here, we highlight the entry/adhesion mechanisms of the pathogenic bacteria Listeria, Salmonella, and Escherichia coli EPEC/EHEC, which engage remodeling of the host cell actin cytoskeleton through phosphoinositide signaling.
1. Listeria monocytogenes
Listeria is a food-borne pathogen that can travel through the intestinal epithelium to reach the blood circulation and finally the liver, which is the main site of infection. Upon infection, Listeria can cause gastroenteritis, meningitis, and abortions (59). The entry of Listeria into host cells occurs through receptor-mediated “zippering” phagocytosis by binding of bacterial surface proteins InlA or InlB to their host cell receptors E-cadherin or Met, respectively (376). Both InlA- and InlB-mediated entry mechanisms result in rearrangements of the actin cytoskeleton, yet the role of phosphoinositides in regulation of the actin cytoskeleton during InlB entry has been more extensively characterized. Binding of InlB to Met receptor in hepatocytes induces phosphorylation of the cytoplasmic tail of Met, recruitment of downstream targets (Gap1, Cbl, Shc), and activation of the PI3K. The local production of PI(3,4,5)P3 by this kinase is thought to recruit WAVE/N-WASP proteins to the entry site possibly through activation of Rho family GTPases. This activates the Arp/2/3 complex-mediated actin polymerization to drive the internalization of the bacterium (154, 259). Interestingly, also many components of the clathrin-mediated endocytosis machinery have been found at the sites of Listeria internalization (376). This might be mediated by recruitment of the clathrin adaptor protein A1 through the local production of PI(4)P by PI4Ks, which are essential for Listeria entry into host cells (292).
2. Salmonella enterica
Salmonella is a food-borne pathogen that causes diarrhea. It induces massive membrane ruffling by injecting its effectors SopB, SopE, and SopE2 into host cells using a type III secretion system. SopE and SopE2 activate Rac and Cdc42 directly, whereas SopB is a phosphoinositide phosphatase that hydrolyzes PI(4,5)P2 at the bacterial entry site (361). SopB can also activate RhoG through SGEF (286). Together with other Rho family GTPases activated by SopE and SopE2, RhoG is responsible for much of the actin cytoskeleton rearrangements that facilitate bacterial entry (286). The local depletion of PI(4,5)P2 by SopB may also destabilize the membrane cytoskeleton linkage, and thus promote sealing of the phagosomes and mediate bacterial uptake.
3. Enteropathogenic and enterohemmorhagic E. coli
Two closely related bacteria, enteropathogenic (EPEC) and enterohemmorhagic (EHEC) E. coli, are food-borne pathogens that adhere to the intestinal epithelia promoting the formation of actin-rich protrusions called pedestals. As a consequence, infection causes a loss of microvilli from intestinal epithelial cells and diarrhea (35, 107). Both EPEC and EHEC strains use the type III secretion system to translocate intimin receptor Tir into the host cell, where it inserts to the plasma membrane. Interactions of the bacterial membrane protein Intimin and the extracellular segments of Tir induce clustering of Tir that initiates a cytosolic signaling cascade leading to the formation of the pedestals (35). In addition to Tir, EHEC also needs another secreted factor EspFU for the pedestal formation. In the host cell, Tir interacts with phosphoinositide-binding I-BAR domain proteins IRSp53 and/or IRTKS (378, 386), which in turn recruit EspFU to the bacterial interaction site through their SH3 domains. One EspFU molecule is capable of binding simultaneously five N-WASP molecules and is thus a powerful activator of Arp2/3 complex, which induces actin polymerization to promote pedestal formation (51, 318).
In contrast to EHEC, EPEC Tir protein directly recruits N-WASP activator Nck to induce Arp2/3 complex-mediated actin polymerization at the site of the bacterial attachment (127). A recent study demonstrated that PI(4,5)P2 and PI(3,4,5)P3 accumulate to the site of EPEC adhesion through the activation and recruitment of PI4PK and PI3K. Acute depletion of PI(4,5)P2 also compromised the pedestal integrity. Interestingly, the accumulation of phosphoinositides occurred in parallel with actin polymerization (322). This suggests that changes in the local phosphoinositide composition might play an important, although currently unknown, role in regulating the host cell actin cytoskeleton during EPEC and possibly also during EHEC attachment. These phosphoinositides may, for example, regulate the activation of N-WASP and/or other factors such as the I-BAR domain proteins that regulate the actin polymerization machinery at the plasma membrane in response to phosphoinositides.
VII. Conclusions and Future Perspectives
Recent studies have revealed that a number of cellular processes, such as endocytosis, phagocytosis, cell migration, and morphogenesis, as well as cytokinesis rely on tight interplay between membrane phosphoinositides and the underlying actin cytoskeleton. Phosphoinositides, especially PI(4,5,)P2, regulate the activities of a large number of actin-binding proteins to enhance actin polymerization and to inhibit actin filament disassembly at cellular membranes. In addition, phosphoinositides, especially PI(3,4,5)P3, interact with the regulators of Rho family GTPases to induce the formation of specialized actin filament arrays in cells. Many actin-binding proteins also either contain membrane-deforming domains or associate with proteins that contain such domains. These domains can either generate positive (BAR domain, most F-BAR domains, dynamin, ENTH domains) or negative (I-BAR domains, SrGAP2 F-BAR domain) membrane curvature, and thus contribute to the formation of plasma membrane invaginations or protrusions in cells. However, whether the physiological role of these domains in the context of full-length proteins is to generate membrane curvature, or to sense membrane curvature and consequently recruit the actin polymerization machineries at the curved membranes, remains to be elucidated in the case of each protein family. Furthermore, the exact mechanisms by which various actin-binding proteins interact with phosphoinositides are largely unknown. Thus future research is needed to reveal these interaction mechanisms as well as to determine the biological roles of various actin-binding protein-PI(4,5)P2 interactions. In this regard, new imaging methods and the development of inducible phosphoinositide depletion systems are bound to expand our knowledge of these systems in the near future.
J. Saarikangas was supported by a fellowship from Helsinki Graduate School in Biotechnology and Molecular Biology. H. Zhao and P. Lappalainen were supported by grants from the Academy of Finland and Finnish Cancer Research Organization.
We thank A. Pykälainen for critical reading of the manuscript.
Address for reprint requests and other correspondence: P. Lappalainen, Institute of Biotechnology, PO Box 56 (Viikinkaari 9), 00014 University of Helsinki, Finland (e-mail:).
- Copyright © 2010 the American Physiological Society