Association of Villin with Phosphatidylinositol 4,5-Bisphosphate Regulates the Actin Cytoskeleton*

Villin, an epithelial cell actin-binding protein, severs actin in vitro and in vivo. Previous studies report that phosphatidylinositol 4,5-bisphosphate (PIP2) regulates actin severing by villin, presumably by interaction with villin. However, direct association of villin with PIP2 has never been characterized. In this report, we presented mutational analysis to identify the PIP2-binding sites in villin. Villin (human) binds PIP2 with a Kd of 39.5 μm, a stoichiometry of 3.3, and a Hill coefficient of 1. We generated deletion mutants of villin lacking putative PIP2-binding sites and examined the impact of these mutations on PIP2 binding and actin dynamics. Our analysis revealed the presence of three PIP2-binding sites, two in the amino-terminal core and one in the carboxyl-terminal headpiece of human villin. Synthetic peptides analogous with these sites confirmed the binding domains. Circular dichroism and quenching of intrinsic tryptophan fluorescence revealed a significant conformational change in these peptides ensuing in their association with PIP2. By using site-directed mutagenesis (arginine 138 to alanine), we demonstrated the presence of an identical F-actin and PIP2-binding site in the capping and severing domain of villin. In contrast, the mutants lysine 822 and 824 to alanine demonstrated the presence of an overlapping F-actin and PIP2-binding site in the actin cross-linking domain of villin. Consistent with this observation, association of villin with PIP2 inhibited the actin capping and severing functions of villin and enhanced the actin bundling function of villin. Our studies revealed that structural changes induced by association with PIP2 could regulate the actin-modifying functions of villin. This study provided biochemical proof of the functional significance of villin association with PIP2 and identified the molecular mechanisms involved in the regulation of actin dynamics by villin and PIP2.

The association of acidic phospholipids with cytoskeletal proteins has been shown to have several different biological consequences as follows: the formation of linkages between the plasma membrane and the actin cytoskeleton (1); regulation of actin dynamics in the cell (2); targeting cytoskeletal proteins to specific cellular domains (3); and/or allowing for scaffolding of cytoskeletal and signaling molecules at the effector site (4,5). PIP 2 1 modulates many actin regulatory proteins in vitro and includes the following: actin severing and/or capping proteins (6 -8); monomer-binding proteins (9 -11); the actin-nucleating protein, N-WASP/Arp2/3 (12); and actin cross-linking proteins (13,14). It has been hypothesized that PIP 2 induces actin assembly by dissociating capping proteins from filament ends, releasing actin monomers from actin-sequestering proteins, inhibiting actin severing, stimulating actin nucleation, and increasing or decreasing actin cross-linking (1,13,(15)(16)(17). In general, PIP 2 predisposes the actin cytoskeleton to a "polymerization mode" and assists its remodeling as well as its interaction with the plasma membrane. In addition, indirect regulation of cortical actin by PIP 2 has also been suggested (18).
Villin is an actin-binding protein that is expressed in differentiated epithelial tissues where it is associated with the brush border (19 -21). Severin, fragmin, and the vertebrate proteins villin, gelsolin, adseverin, and scinderin belong to the group of actin-severing proteins that are composed of highly conserved domains. Like these proteins villin caps, nucleates, or severs actin filaments (22,23). All these proteins contain segments, which display internal homology with each other; in villin this constitutes the "core" (24,25). In addition, villin contains a small (8.5 kDa) carboxyl-terminal domain called the headpiece (Fig. 1). The villin core retains the Ca 2ϩ -dependent actin severing, capping, and nucleating functions of villin, whereas the headpiece endows villin with the ability to form microfilament bundles (26) (Fig. 1). The villin core also retains several of the ligand-binding sites of villin including tyrosine phosphorylation sites (27), a putative PIP 2 -binding site (28), Ca 2ϩ -binding sites (29), as well as the site of F-actin binding prior to severing (30) (Fig. 1). The headpiece has been shown to contain an F-actin-binding site involved in actin cross-linking (25) (Fig. 1). Ca 2ϩ (31), tyrosine phosphorylation (32), and PIP 2 (30) all regulate the actin modifying activities of villin.
Villin inhibits the catalytic activity of phospholipase C-␥ 1 (PLC-␥ 1 ) by sequestrating its substrate PIP 2 (33). This points to a role for villin in modifying lipid-signaling events. In addition, the actin severing activity of villin is inhibited by PIP 2 in vitro (28). This implies that villin may mediate dynamic changes in the actin cytoskeleton in response to changes in lipid-signaling events. The correlation of these two events suggests that interaction of villin with PIP 2 can regulate epithelial cell structure and function in response to various stimuli. We have shown previously (33) that villin interacts directly with PIP 2 in vitro. The aim of the present study was to investigate the villin-PIP 2 interaction further to include identification of the PIP 2 -binding sites in villin and to investigate the relationship of villin regulation of actin dynamics by phosphoinositides. By using deletion and site-directed mutagenesis, we identify the amino acids that are involved in binding of villin to PIP 2 ( Fig. 1). Our studies reveal the presence of three PIP 2 binding domains consisting of clusters of basic amino acid residues ( Fig.  1). Exploiting the domain structure of villin, we studied the in vitro function of PIP 2 on the actin-modifying functions of villin. The data in this report indicate that two of the lipid binding domains in villin overlap with F-actin-binding sites in villin. Furthermore, our analysis reveals that PIP 2 inhibits actin severing and capping and promotes actin bundling by villin, suggesting that the lipid-binding sites of villin are functional. This is the first detailed study identifying the PIP 2 -binding sites in villin and mapping these PIP 2 -binding sites with the functional characteristics of villin.

Materials
Escherichia coli-BL21 competent cells and QuikChange site-directed mutagenesis kit were from Stratagene; the prokaryotic expression vector pGEX2T was from Amersham Biosciences; glutathione-Sepharose 4B Fastflow was from Amersham Biosciences; GelCode Blue was from Pierce; monoclonal antibodies to villin were from Transduction Laboratories; polyclonal antibodies to villin were commercially generated (BIOSOURCE); monoclonal antibodies to phosphatidylinositol 4,5-bisphosphate were purchased from Assay Design Inc.; phosphatidylcholine (PC) and PIP 2 were purchased from Avanti Lipids; phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol 4-monophosphate (PIP), and phosphatidylinositol (PI) were purchased from Sigma. 96-Well plates (Polysorb) were from Nalge Nunc. The muscle actin polymerization kit and the actin-binding kit were purchased from Cytoskeleton (Denver, CO); all other chemicals were from Sigma or Invitrogen.

Methods
Preparation of Unilamellar Lipid Vesicles-Unilamellar liposomes were prepared from PC and PIP 2 as described before (34,35). Lipids were mixed to give solutions composed of 20% PIP 2 and 80% PC (w/w). Higher concentrations of PIP 2 (40% PIP 2 and 60% PC) were used in the cation-induced aggregation of lipid vesicles. The resuspended lipids were vortexed, sonicated on ice, frozen and thawed five times, and then passed through an Avestin (Ottawa, Canada) LiposoFast extruder 21 times (100-nm pore size). To measure specificity of the phospholipid binding to villin, pure vesicles (100%) containing PC, PE, PS, PIP, PI, or PIP 2 were used.

Protein Conformation Studies
Conformational Studies by CD-CD measurements were made using PIP 2 micelles to minimize light scattering (36). CD spectra were collected from 191 to 260 nm with data collection for 5 s averaging in 1-nm increments with three repeats at 25°C. An AVIV model 62S spectropolarimeter equipped with electronic temperature control and a 1-cm path length quartz cuvette was used for CD studies. Mean molar ellipticity per residue was calculated by ()(10))/(c)(n)(l)); units are degrees cm 2 / decimol. is ellipticity in millidegrees; c is concentration in mol/liter; n is number of amino acid residues per protein; and l is path length in cm. The K2d program was used for estimation of secondary structure from CD data. K2d provides estimated structure and the error estimate (kalel.ugr.es/k2d.htm) (37). A time course of absorbance at 400 nm was done with appropriate controls (38) to demonstrate that villin protein or villin peptides do not aggregate under these conditions. Urea Denaturation Assay-To determine the effects of specific mutations on the overall stability of the villin molecules, fluorescencemonitored urea denaturation was performed on each recombinant protein as described earlier (27).

PIP 2 Binding Assays
Cosedimentation of Wild Type and Deletion Mutants of Villin with Lipid Vesicles-The lipid binding assay was used to determine the association of PIP 2 with villin as described before (33). Briefly, fusion proteins (wild type or deletion mutants, 2 g) were incubated with PIP 2 containing unilamellar vesicles (20% PIP 2 , 80% PC; final concentration 1.7 mg/ml), and after incubation at room temperature for 10 min and on ice for 5 min, the tubes were centrifuged at 100,000 ϫ g for 30 min at 4°C. The supernatant (S), the pellet (P), and the total (T) protein were separated by SDS-PAGE. The gels were stained with GelCode Blue and the amount of protein in T, S, and P was determined by densitometric analysis.
Interaction of Villin Proteins and PIP 2 by Solid Phase Binding Assay-Binding of wild type or mutant villin proteins to PIP 2 containing unilamellar vesicles (20% PIP 2 and 80% PC) was tested as described by Fukami et al. (13). Briefly, 96-well multiplates were coated with villin proteins (5 ng/l, total volume 50 l/well) overnight, blocked with 2% bovine serum albumin (BSA), and incubated with increasing amounts of PIP 2 micelles (0 -2.5 g/well). Bound PIP 2 was detected using a PIP 2 -specific antibody (1:400) and horseradish peroxidase-conjugated secondary anti-mouse IgG (1:3000). A colorimetric reaction was developed using 2,2Ј-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (0.4 mg/ml in 100 mM citrate buffer, pH 5.0) as substrate and absorption (A) was measured at 405 nm in a plate reader, SpectraMax-190 (Molecular Devices). Alternatively, interaction of villin with PIP 2 was determined using what is essentially a sandwich enzyme-linked immunosorbent assay. Briefly, 1:400 dilution of PIP 2 monoclonal antibody was coated on 96-well plates overnight at 4°C. The plates were blocked with 0.5% gelatin, incubated with PIP 2 (0 -2.5 g/well), followed by excess of wild type or mutant villin proteins. The bound villin was detected using rabbit anti-villin polyclonal antibody (1:500), and bound conjugates were detected as described above. To determine binding of villin to different phospholipids, solid phase lipid binding assay was performed using polystyrene flat bottomed 96-well plates (Polysorb) and 100% PS, PE, PC, PI, PIP, or PIP 2 (between 0.5 and 3.0 g/well).
Aggregation of PIP 2 /PC Mixed Vesicles Induced by Divalent Cations-The binding of villin proteins to PIP 2 and other phospholipids was studied using an assay described by Xian and Janmey (36), based on the observation that divalent cations induce aggregation of PIP 2 /PC mixed unilamellar vesicles that is reduced by PIP 2 -binding proteins (40). Aggregation of PIP 2 /PC mixed vesicles (40% PIP 2 and 60% PC) was induced by the addition of 0.6 mM Ca 2ϩ and 6 mM Mg 2ϩ (final concentrations) to a 30 M PIP 2 /PC mixed vesicle sample in a buffer containing 150 mM KCl and 10 mM Tris-HCl, pH 7.4. The samples were analyzed in a Beckman-Coulter DU 640 spectrophotometer, and absorption at 400 nm was recorded every 15 s for 30 min.
Quenching of Intrinsic Tryptophan Fluorescence-The binding of villin proteins with PIP 2 was also examined by recording the quenching of the intrinsic tryptophan fluorescence of villin by PIP 2 as described before (41). Briefly, PIP 2 -containing unilamellar vesicles (final concentrations between 0 and 80 M) were added at 0.3-l increments, and the fluorescence spectra were recorded 5 min after each addition. The excitation wavelength was 290 nm. Data were analyzed as described by Ward (42). The apparent dissociation constant, K d , was calculated by using Equation 1, where ⌬F is the fluorescence quenching at a given PIP 2 concentration; ⌬F max is the total fluorescence quenching of the villin protein saturated with PIP 2 ; and [PIP 2 ] T is the total accessible concentration of PIP 2 . Total accessible concentration of PIP 2 indicates the amount of PIP 2 in the bilayer available for binding to the villin proteins, represents half of the total lipid, and is an approximation of the amount of lipid in the outer exposed leaflet of the bilayer of the unilamellar vesicle (43,44). ⌬F max is estimated by curve fitting of the binding data using Hyperbol.fit program in Microcal Origin. Alternatively, the intrinsic association constant, K a , as well as the stoichiometry of binding, p, were derived using the graphic method of Stinson and Holbrook (84), using Equation 2, where is the fractional binding (⌬F/⌬F max ); p is the stoichiometry of binding; [PIP 2 ] T is the total concentration of PIP 2 ; and [villin] T is the total protein concentration. When 1/(1Ϫ) is plotted against [PIP 2 ] T /, a straight line with a slope of K a and an intercept of [villin] T / is obtained. The stoichiometry of interaction, p, can be calculated by dividing the intercept with the protein concentration as described before (41).

Actin Binding Assay
Cosedimentation of Wild Type and Villin Mutants with F-actin-F-actin (3.0 M) was incubated with varying concentrations of wild type (VIL/WT) or villin mutants lacking putative PIP 2 -binding sites (0 -5 M) in the presence of 2 mM EGTA for 10 min at room temperature and centrifuged for 15 min at 200,000 ϫ g in an ultracentrifuge (Sorval Discovery M120). The supernatant was precipitated with 2 volumes of acetone. The pellet and the acetone-precipitated proteins were analyzed by SDS-PAGE, and gels were stained with GelCode Blue.
Measurement of Actin Polymerization-Depolymerization by Wild Type and Deletion Mutants of Villin-The kinetics of actin polymerization were determined using a muscle actin polymerization kit according to the instructions of the manufacturer as described previously (27). The ability of villin to nucleate actin assembly or to sever actin filaments was determined by its effect on the rate and extent of increase or decrease, respectively, of fluorescence of pyrene-labeled actin. Fluorescence measurements were performed at 25°C using the FluoroMax 3 spectrofluorometer. The excitation wavelength was set at 365 nm, and the emission wavelength was set at 407 nm. Actin nucleation and actin-severing measurements were made in the absence or presence of 40 M PIP 2 as described before (29).
Measurement of Actin Uncapping by Wild Type and Deletion Mutants of Villin-The actin capping activity of wild type and mutant villin proteins was measured essentially as described by Northrop et al. (31) using pyrene-labeled actin as described by Schafer et al. (17). F-actinvillin (wild type or mutant) seeds were prepared by polymerizing unlabeled G-actin (in buffer containing 5 mM Tris-HCl, 0.2 mM ATP, and 2.5 M CaCl 2 , pH 7.5) by adding 150 mM KCl and 1 mM MgCl 2 . 290 nM villin-actin seed were used as nuclei for polymerization with pyrenelabeled G-actin (1.4 M) in a reaction volume of 200 l. The increase in fluorescence was measured over time as described above, in the absence or presence of PIP 2 (100 M). The concentration of calcium (2.5 M) used in the assays has been shown to be saturating for capping but not severing of actin filaments by villin (31).
Electron Microscopic Analysis of Bundling Activity of Villin-The bundling activity of villin was analyzed essentially as described before (32). The divalent cation-free actin was polymerized together with villin constructs in polymerization buffer, pH 7.0, lacking MgCl 2 . F-actin (3 M) was incubated overnight at 4°C with VIL/WT (0.6 M) or VIL/WT and PIP 2 (40 M) in the presence of EGTA (2 mM). Samples were applied to carbon-coated Formvar grids. The grids were stained with 2% aqueous uranyl acetate after a brief fixation in 2% glutaraldehyde and examined at 60 kV in a JEOL 20000 EX-II electron microscope.

Characterization of the Association of Villin with
Phosphoinositides-Association of villin with the lipid bilayer was studied using recombinant villin, which we have shown previously (33) to behave like the native villin protein. Direct interaction between lipid vesicles containing pure phospholipids and fulllength recombinant villin protein (VIL/WT) was examined by solid phase binding assay using different concentrations of PC, PS, PE, or PIP 2 and related phosphoinositides PIP and PI. As shown in Fig. 2A, villin does not associate with unilamellar vesicles containing pure PC, PE, or PS. In contrast, there is significant association of villin with vesicles containing the phosphoinositides, PI, PIP, or PIP 2 , with the following binding profile: PIP 2 Ͼ PIP Ͼ PI. Similar phosphoinositide-specific binding is also known for other actin-binding proteins of the villin family (4,45). Because villin binds to PIP 2 with the highest affinity, all experiments in this report were done examining the effect of the association of villin with PIP 2 .
The interaction of villin with PIP 2 was confirmed by assaying the divalent cation-induced aggregation of mixed phospholipid vesicles containing PIP 2 and PC in the presence or absence of villin (36). Interaction of villin with PIP 2 was examined by preincubating these mixed lipid vesicles with different concentrations of villin protein (0 -10 nM). Aggregation was triggered by addition of divalent cations to this mixture. As shown in Fig.  2B, divalent cations cause aggregation of phosphoinositide vesicles (control), and the addition of villin decreases this aggregation of PIP 2 -containing lipid vesicles in a dose-dependent manner. This is similar to the observation made with a related protein, gelsolin (36). Association of villin with PIP 2 was also observed by determining the cosedimentation of recombinant villin protein with these large PIP 2 -containing vesicles in the absence of divalent cations. By using this approach, we found that greater than 90% (n ϭ 10, p Ͻ 0.01) of the villin protein cosediments with liposomes containing PIP 2 (calculated as percent of total protein; data not shown). These studies demonstrate that villin binds PIP 2 directly (33) either in the form of pure micelles or when packed into mixed lipid bilayer vesicles (containing PC and PIP 2 ). Villin does not associate with mixed lipid vesicles if the molar fraction of PIP 2 in the vesicles is less than 5% and the association of villin with PIP 2 decreases if the molar fraction of PIP 2 in these vesicles is greater than 20%. 2 So the critical concentration of villin association lies between 5 and 20% PIP 2 concentration, similar to that seen with other actin-and PIP 2 -binding proteins (46).
For kinetic analysis the binding of villin to PIP 2 was monitored by recording the quenching of the intrinsic tryptophan fluorescence of villin by PIP 2 . Changes in tryptophan fluorescence upon binding of PIP 2 have been used to characterize binding to PIP 2 of both actin-binding proteins such as profilin (47) and CapG (41) as well as other PIP 2 -binding proteins such as phospholipase C␦ (48) and dynamin pleckstrin homology domain (49). Fig. 3A shows the tryptophan fluorescence emission spectra of VIL/WT. Villin exhibits emission maximum of 337 nm, and PIP 2 induces a dose-dependent and saturable quenching of the intrinsic tryptophan fluorescence of villin without shifting its emission maximal (Fig. 3A). Micelles alone and without villin show no significant emission (data not shown). Quenching of intrinsic tryptophan fluorescence was plotted versus PIP 2 concentration, and the data were analyzed 2 N. Kumar, P. Zhao, and S. Khurana, unpublished observations.

FIG. 2. Association of villin with phosphoinositides.
A, villin binds phosphoinositides and not all phospholipids. Direct interaction of villin with phospholipids was determined using solid phase binding assay. 96-Well multiplates were coated with increasing amounts of phospholipids (100%) including PC, PE, PS, PI, PIP, and PIP 2 (0 -3.0 g/well) overnight, followed by incubation with excess of full-length villin protein. Bound villin was detected using a monoclonal villin antibody and horseradish peroxidase-conjugated secondary anti-mouse IgG. Data represent mean of three independent experiments, each sample run in triplicate. B, villin inhibits aggregation of PIP 2 containing mixed vesicles in a dose-dependent manner. Aggregation of 30 M PIP 2 /PC mixed vesicles (40% PIP 2 and 60% PC) was induced by 0.6 mM Ca 2ϩ and 6 mM Mg 2ϩ in the presence of varying concentrations of VIL/WT (0 -10 nM). Absorbance was measured at 400 nm. Control data were recorded in the absence of villin protein. Data represent mean of three independent experiments, each sample run in triplicate. by using Hyperbol.fit in Microcal Origin using Equation 1. These studies gave a K d value for binding of VIL/WT to PIP 2 micelles as 39.4 M (Fig. 3B). Furthermore, the stoichiometry of PIP 2 binding to villin and its association constant K a were calculated using the equation of Stinson and Holbrook (84) (see Equation 2) as 3.3 and 0.025 M Ϫ1 , respectively (Fig. 3C). This suggests that villin binds three PIP 2 molecules. Whereas it is very likely that villin binds to more than one micelle containing multiple PIP 2 molecules, the calculated stoichiometry allows for comparison with other PIP 2 -binding proteins for which similar measurements of stoichiometry have been made. The Hill coefficient (h) was calculated as 1 suggesting no cooperativity between the three PIP 2 molecules bound to villin (Fig. 3D). Together, these results show the direct association of villin with phosphoinositides and characterize the kinetics of villin-PIP 2 interaction.
Identification of PIP 2 -binding Sites in Villin-To understand the role of PIP 2 in the regulation of the actin modifying activities of villin, we sought to identify the PIP 2 -binding sites in villin. Most actin-binding proteins that associate with PIP 2 contain motifs within their sequence that are recognized by PIP 2 and include a cluster of basic amino acid residues with the consensus motif (K/R)(X) 3-5 (K/R)(X)(K/RK/R) (39). Villin contains three such motifs as follows: (K/R)(X) 5 . Based on these motifs, analysis of the human villin sequence suggested five putative PIP 2 -binding sites (PB1, PB2, PB3, PB4, and PB5). To identify the site(s) of PIP 2 binding in villin, deletion mutants of villin lacking one or more of these sites were constructed as summarized in Fig. 4A. The constructs consisted of internal deletions in human villin cDNA engineered into the prokaryotic expression vector, pGEX-2T. Five deletion mutants (⌬PB1, -PB2, -PB3, -PB4, and -PB5) were generated as Gst fusion proteins in BL21 cells (Fig. 4B). To determine the PIP 2 -binding site(s) in villin, the effect of villin deletion mutants lacking these putative PIP 2 -binding sites on divalent cation-induced aggregation of PIP 2 /PC vesicles was compared with binding of wild type villin to the mixed lipid vesicles. Full-length villin (VIL/WT) decreased the cation-induced aggregation of the lipid vesicles (Fig. 4C). Deletion of putative lipid-binding sites of PB1 resulted in a small decrease (less than 5%, n ϭ 6, p Ͻ 0.01) in the association of villin with PIP 2 . Deletion of PB2 and PB5 re- sulted in the most significant decrease (75 and 50%, respectively) in the ability of the villin protein to associate with PIP 2 (Fig. 4C). Previous studies have suggested the presence of a PIP 2 -binding site(s) in villin that would be consistent with PB1 and/or PB2. However, no PIP 2 -binding site has been described in the COOH-terminal headpiece domain of villin. Our studies demonstrate for the first time the presence of a phospholipidbinding site in the actin cross-linking headpiece domain of villin, namely PB5. Deletion of putative PIP 2 -binding site PB3 and PB4 did not alter the binding of villin protein to PIP 2 , and these mutants behaved like VIL/WT (data not shown).
To confirm that PB1, PB2, and PB5 comprise the PIP 2binding sites in villin, we also examined the direct interaction between wild type and mutant villin proteins and PIP 2 by solid phase lipid binding assay. As shown in Fig. 4D, VIL/WT associates with PIP 2 in a concentration-dependent manner. Deletion of PB1 made a small change in the ability of villin to bind PIP 2 (9.1% decrease, n ϭ 8, p Ͻ 0.01), whereas mutation of either PB2 or PB5 results in a significant decrease in the binding of PIP 2 to villin proteins (41 and 50%, respectively, n ϭ 8; p Ͻ 0.05). Deletion of the two major PIP 2 binding domains in villin, namely PB2 and PB5 together, eliminated the binding of the villin with PIP 2 (n ϭ 8, p Ͻ 0.05). Similarly, deletion of all three PIP 2 -binding sites together, namely PB1, PB2, and PB5, abolished the binding of PIP 2 to villin (n ϭ 8; p Ͻ 0.01). These data confirm that PB1, PB2, and PB5 are the only PIP 2 binding domains in villin. Tryptophan fluorescence emission spectra of villin mutant proteins ⌬PB1, -PB2, and -PB5 were also recorded. A plot of wild type (VIL/WT) and mutant villin protein fluorescence quenching versus PIP 2 concentration showed that compared with VIL/WT saturation was reached at a lower PIP 2 concentration by ⌬PB2 and ⌬PB5 (45 and 60 M respectively, Fig. 5A). In contrast, binding of ⌬PB1 to PIP 2 was saturable at a concentration that was comparable with VIL/WT (between 70 and 80 M). Binding studies based on quenching of villin intrinsic tryptophan fluorescence by PIP 2 suggest a stoichiometry of 1.77, 1.88, and 1.78 for ⌬PB1, -PB2, and -PB5, respectively (Fig. 5, B-D). Taken together, these data show that PB1 (a.a. 112-119), PB2 (a.a. 138 -146), and PB5 (a.a. 816 -824) compose the three PIP 2 -binding sites in human villin. Whereas PB3 and PB4 contain regions with clusters of basic amino acids, these domains were not identified as PIP 2 -binding sites in villin.
PB1, PB2, and PB5 are the PIP 2 -binding Sites in Villin-Because previous characterization of PIP 2 -binding sites in villin depended on the use of synthetic peptides (equivalent to either PB1 and PB2, or PB2 alone) (30), we used two peptides encompassing the two amino-terminal lipid-binding sites in villin, Pep 1 (PB1) and Pep 2 (PB2), respectively. In addition we used a third peptide encompassing the carboxyl-terminal PIP 2 binding domain, PB5 (Pep 5). Equimolar amounts of VIL/WT and peptides were used in these studies (10 nM). Association of VIL/WT or peptides was determined by measuring the decrease in aggregation of lipid vesicles by divalent cations. All three peptides bind PIP 2 and decrease the aggregation of lipid vesicles (Fig. 6). Pep 1 encompassing PB1 showed the least binding, followed by Pep 2 (PB2) and Pep 5 (PB5). The addition of all three peptides together inhibits aggregation of lipid vesicles that is similar to full-length villin. Comparable binding of chicken villin peptides Pep 3 and Pep 4 was noted (data not shown). These data are the first identifying the PIP 2 -binding sites in human villin. Conformational Changes in PIP 2 Binding Domains of Villin-Several recent studies (36,50,51) suggest that binding of PIP 2 to actin-modifying proteins may induce structural changes that may regulate their actin-modifying functions. To determine whether PIP 2 induces similar structural changes in villin, we used both full-length villin as well as villin peptides encompassing the PIP 2 binding domains. Changes in the secondary structure of VIL/WT as well as Pep 1, Pep 2, and Pep 5 were measured using circular dichroism spectroscopy in the far-ultraviolet region, respectively. The CD measurements were made with samples containing 1.5 M protein/peptide in the absence or presence of PIP 2 (100 M). The CD spectra were fitted to the K2d program to quantify the structural changes in the villin protein/peptides. The circular dichroism spectrum of human villin shows relatively high percentage of ␤-sheet (ϳ48%) than ␣-helix (ϳ11%) consistent with earlier reports (52). PIP 2 induced a very small change in the secondary structure of full-length villin ( Fig. 7 and Table I). Whereas large changes in the structure of full-length villin protein were not obvious, significant local structural changes induced by PIP 2 were observed with the peptides. PIP 2 induced a significant change in the secondary structure of the villin peptide, Pep 1, by increasing the relative ␣-helix content of the peptide from 8 to 23% and decreasing the ␤-sheet content from 42 to 31% (Fig.  7B and Table I). There was no significant change in the secondary structure of Pep 2 as a result of PIP 2 binding (Fig. 7C and Table I). Pep 5 corresponding to PB5 in villin contains both ␣-helix and ␤-sheet structure. Binding with PIP 2 results in significant transition in the secondary structure of Pep 5 to a primarily ␤-sheet structure (Fig. 7D and Table I). These data show that whereas PIP 2 does not induce a global change in the secondary structure of the villin proteins, it may induce significant localized changes in the secondary structure of the protein which may regulate both the ligand binding properties (F-actin, PLC-␥ 1 ) as well as the actin-modifying functions of villin.
Quenching of intrinsic tryptophan fluorescence was also used to determine conformational changes in Pep 5 (since it contains a single tryptophan residue) following association with PIP 2 . As shown in Fig. 8A, Pep 5 has an emission maximum at 360 nm. PIP 2 induced a dose-dependent and saturable decrease in fluorescence, without shifting the emission maximum of the peptide. The emission maximum of the peptide is higher than that of full-length villin protein, and saturation is reached at a lower PIP 2 concentration with the peptide compared with VIL/WT (Fig. 3A). Pep 1 (Fig. 8B) and Pep 2 (that do not contain a tryptophan residue) were used as negative controls.
Overlapping and Identical Actin and PIP 2 -binding Sites in Villin-Two of the PIP 2 -binding sites identified in this study correspond to actin-binding sites in villin described previously (32,35,36). PB2 has been described as the site of F-actin binding to villin prior to severing, and PB5 has been identified as the F-actin-binding site involved in the bundling function of villin. The third lipid-binding site, PB1 (a.a. 112-119), lies in

TABLE I Calculated secondary structure of wild type and mutant villin proteins and villin peptides
Calculated secondary structure of full-length (VIL/WT) and mutant villin proteins (⌬PB1, -PB2, and -PB5) was calculated by circular dichroism as described under "Experimental Procedures." Changes in secondary structure of full-length villin protein and villin peptides (Pep 1, Pep 2, and Pep 5) induced by association with PIP 2 (100 M) were also calculated by circular dichroism. Maximum error is the sum of the errors in the prediction of the ␣-, ␤-, and random percentage values divided by 3.  (54) using mutational analysis have revealed that two residues within this motif, namely Lys-822 and Lys-824, are essential for F-actin binding to the villin headpiece as well as determine the actin bundling activity of villin. Similarly Arg-138, Lys-145, and Arg-146 in the villin core have been described as the F-actin-binding residues regulating actin severing by villin (52). Notably, mutation of Arg-138 to alanine was shown to result in significant loss of the actin binding ability of villin core as well as its actin severing activity, suggesting that this site is identical for F-actin side binding as well as severing (52). Consistent with this, deletion of either PB2 or PB5 decreases the binding of villin to F-actin (Fig. 9A). Fig. 9B demonstrates the appropriate controls for this assay, namely cosedimentation of an actin-binding protein ␣-actinin with F-actin, whereas bovine serum albumin (BSA), a protein that does not associate with F-actin, remains in the supernatant under the conditions used in the assay. In this figure, P refers to the pellet and includes proteins that cosediment with F-actin, and S refers to the supernatant fraction and includes proteins that do not associate or pellet with F-actin.
To determine whether the PIP 2 -and actin-binding sites are overlapping or identical, we examined the effect of mutating specific actin-binding residues in villin on PIP 2 binding. We tested wild type villin and its mutants, containing the single site mutations of Lys or Arg to Ala in the F-actin binding domains described above, in a PIP 2 binding assay. As expected, there is significant decrease in binding to PIP 2 by villin deletion mutants ⌬PB2 as well as ⌬PB5 (Fig. 9C). We found no noteworthy changes in the PIP 2 binding activity of the point mutants, K145A, R146A, K822A, or K824A compared with VIL/WT. The only exception to this observation is the point mutant R138A. Mutation of this arginine residue to alanine is known to result in a significant decrease in the binding of villin to F-actin (52); likewise we noted a significant (83.21 Ϯ 2.8, n ϭ 8, p Ͻ 0.01) decrease in the binding of villin to PIP 2 (comparable with ⌬PB2, Fig. 9C) with the villin mutant R138A. These data suggest that Arg-138 represents not only the F-actin sidebinding (52) and actin-severing (52) site but also the PIP 2binding site in the amino-terminal of villin (Fig. 9C). Similar data were obtained in assays measuring cation-induced aggregation of lipid vesicles as well as cosedimentation with lipid vesicles in the presence of wild type and mutant villin proteins (data not shown). Mutually inclusive PIP 2 and actin binding has also been described for other actin-binding proteins, e.g. profilin (9) and gelsolin (55). These observations suggest that the subtle balance between PIP 2 and actin binding may be important in the reorganization of the actin network by villin. Consequently, understanding the possible role of PIP 2 in specifically localizing villin will be of major interest. To answer this question, we determined the actin severing, capping, nucleating, and crosslinking activities of villin in the absence or presence of PIP 2 .
PB1 has been described as the site that stabilizes the binding of F-actin to PB2, whereas PB2 has been described as the site of F-actin binding to villin prior to severing, thus providing a mechanism for the PIP 2 -induced inhibition of the actin severing activity of villin (52). Wild type villin severs actin (Fig.  10A), consistent with previous observations (52). The addition of PIP 2 inhibits the actin severing activity of villin, as reported earlier and as shown in Fig. 10A (30). In addition deletion of either PB1 or PB2 inhibits the actin severing activity of villin. In contrast deletion of PB5 has no effect on the actin severing function of villin. This suggests that deletion of PB1 and PB2, the actin severing domains of villin, or binding of PIP 2 to PB1 and/or PB2 regulates the actin severing function of villin. The identical F-actin side-binding site and the PIP 2 -binding sites in villin explain both the inhibition of F-actin binding and severing by villin in the presence of PIP 2 . Furthermore, the structural changes observed in Pep 1 following its association with PIP 2 (Fig. 7) suggest that PIP 2 -induced structural changes in PB1 could also influence the actin severing activity of villin.
These data provide a molecular basis for the PIP 2 -induced inhibition of the actin severing function of villin.
The ability of villin and its deletion mutants to bind to the plus or barbed end of actin filaments and its effect on the capping activity by PIP 2 micelles was tested by following the polymerization kinetics of pyrene-labeled G-actin from barbed ends under polymerization conditions where there is little or no growth from the pointed end. Proteins that bind to the barbed end of the filament can retard the rate of polymerization nearly 10-fold (56). The concentration of calcium (2.5 M) used in the assays has been shown to be saturating for capping but not severing of actin filaments by villin (56). As shown in Fig. 10B, addition of pyrene-labeled G-actin in the presence of F-actin seeds results in significant and rapid actin polymerization over time. VIL/WT shows decreased rates of polymerization in the presence of 2.5 M Ca 2ϩ (Table II), consistent with capped barbed ends. PIP 2 at 100 M was added to the polymerization reaction containing actin filaments that were completely capped by villin. PIP 2 reversed the actin capping activity of villin and increased the rate of actin polymerization (Table II), suggesting that PIP 2 is uncapping the barbed ends of the actin filaments. Cytochalasin D at 20 nM inhibited actin polymerization, indicating that free barbed ends had been capped in this assay. To determine actin polymerization at the barbed ends, the rate constants were calculated in the linear range of the assay (Table II). The rate constants demonstrate the uncapping of actin filaments in the presence of PIP 2 . In contrast, PIP 2 has no effect on the actin-nucleating functions of villin (Fig. 10C), even at higher concentrations of PIP 2 (data not shown). To- gether, these data suggest that PIP 2 enhances actin polymerization by villin by inhibiting its severing and capping functions; however, PIP 2 does not directly regulate the ability of villin to nucleate actin filaments.
The importance of the COOH-terminal headpiece domain of villin for both the binding to PIP 2 and the bundling of actin filaments prompted us to test the effect of complexes of villin and PIP 2 on actin bundling. PIP 2 forms 6-nm diameter micelles in neutral, low ionic strength aqueous solutions as used in our assays (40). Divalent cations bind PIP 2 through its negatively charged head-groups and at millimolar concentrations induce the formation of large, multilamellar PIP 2 aggregates (57). As seen in Fig. 11B, the actin filaments are covered with PIP 2divalent cation aggregates that appear as clusters of striated filaments; the spacing between striations in the filaments is around 6 nm, comparable with that reported with Mg 2ϩ ions (40). Because divalent cations cause PIP 2 micelles to aggregate, actin was depleted of divalent cations for the actin cross-linking assays. To examine the effect of PIP 2 on the actin crosslinking activity of villin, G-actin was polymerized with buffer containing 50 mM KCl (and no Mg 2ϩ ) to form F-actin filaments (Fig. 12A). F-actin filaments were incubated with wild type villin (Fig. 12, B and BЈ) to study the actin cross-linking function of villin. Wild type villin bundles F-actin as reported earlier (37). In the presence of PIP 2 and villin there is no noticeable change in the morphology of the F-actin bundles (compare insets of Fig. 12BЈ with Fig. 12CЈ); nevertheless, PIP 2 increases the number of F-actin bundles formed in the presence of villin (compare Fig 12, B and C). In the presence of PIP 2 there are fewer free F-actin filaments, and the F-actin bundles appear more branched or cross-linked (Fig. 12, C and CЈ). This finding establishes that the actin bundling and PIP 2 binding activities overlap at the COOH-terminal end of villin and that PIP 2 augments the actin bundling function of villin. Deletion of PB5 results in complete loss of the actin bundling effect of villin (Fig. 12, D and DЈ). PIP 2 also enhances the viscosity of actin filaments in the presence of villin, suggesting increased crosslinking of actin filaments in the presence of PIP 2 (data not shown). Together these data suggest that PIP 2 inhibits the actin capping and severing activities of villin through PB1 and from the slope of the pyrene fluorescence versus time, during the early linear phase after addition of phospholipid, and is shown in Table II   PB2 and enhances the actin cross-linking function of villin through PB5. Such regulation of actin kinetics in vivo by villin-PIP 2 interaction could lead to increased actin polymerization and cross-linking in the vicinity of the plasma membrane.

Wild Type and Mutant Villin Proteins Have a Stable
Conformation-The CD spectra of VIL/WT shows evidence of ␣-helix and ␤-sheet structures ( Fig. 7 and Table I), similar to that reported earlier (52). Comparative study of the secondary structure prediction as calculated by the K2d program shows that the structural integrity was maintained in all mutant proteins of villin (Table I). This suggests that no major conformational changes are induced by mutations in the villin protein, and both the urea denaturation and the CD spectra demonstrate that the overall conformation of villin is preserved after introduction of these mutations. In addition, previous studies with villin 44T and mutants R138A, K145A, and K146A have demonstrated that mutation of these specific residues does not change the conformation of the villin proteins as measured by CD (52). Similarly, mutations in PB5, namely K822D and K824D, have been shown to preserve the conformation of the villin protein as measured by CD (58). The unfolding profiles of the wild type and mutant villin as a function of urea concentration were also recorded by measuring the intrinsic fluorescence emission spectrum. Deletion mutants as well as point mutants of villin express similar unfolding transition as wild type villin (data not shown). Therefore previously published reports (52) as well as our data verify that the villin mutants used in this study maintain the integrity of the wild type protein. DISCUSSION Actin-binding proteins were among the first proteins shown to be regulated by direct interaction with PIP 2 (59). Several recent studies (60,61) have demonstrated that changes in phosphoinositide levels, specifically PIP 2 , and concomitant changes in the actin dynamics are essential to some cell surface events including motility. In fact, peptides encompassing the PIP 2 binding domain of an actin-binding protein of the villin superfamily (gelsolin) have been shown to increase cell motility in fibroblasts (62). These studies then point to the significance of cytoskeletal protein and PIP 2 interactions to cell function. Previous studies have suggested the presence of a PIP 2 binding domain in the NH 2 -terminal domain of villin. These studies were focused on the amino-terminal half of villin because PIP 2 was shown to inhibit actin severing by the NH 2 -terminal half of both villin and a related protein gelsolin (28,30). Up to this point, the characterization of villin-PIP 2 interactions has relied primarily on changes in gelsolin function or association of gelsolin with PIP 2 in the presence of villin peptides that encompass PIP 2 -binding consensus sequences in villin homologous to gelsolin-PIP 2 -binding residues (30). The goal of this study was to better describe the structural basis of the interaction of villin with PIP 2 . Details of the molecular determinants involved in villin-PIP 2 interactions are a prerequisite for studying the physiological role of villin-PIP 2 association in intact cells. Several different phosphoinositide binding domains have been identified in recent years and include the Fab1p, YOTB, Vps27p, EEA1 (FYVE), Phox homology (PX), FERM, and the epsin amino-terminal homology domains. There are distinct structural features of each of these domains (for review see Ref. 63). For instance, PH and epsin amino-terminal homology domains function beneath the plasma membrane, whereas FYVE and PX domains function at vesicular membranes. FYVE and epsin amino-terminal homology domains also prefer single phosphoinositides, whereas PH and PX domains do not display specificity. However, they all share several common features including a positively charged binding core and in some cases a hydrophobic membrane interaction site. Most of the actin regulatory proteins do not employ these binding motifs. Janmey et al. (46) have described a consensus sequence for phosphoinositide binding in cytoskeletal proteins that consists of basic amino acids, (X) 4 (R/K)(X)(R/KR/K). Most of the identified PIP 2 -binding cytoskeletal proteins use basic/aromatic amino acid motifs to bind PIP 2 , which bind PIP 2 with comparable affinity as the more structured motifs. The PIP 2 -binding sites in the actinbinding proteins are composed of a short (Ͻ20 a.a.) peptide stretch displaying a net positive charge and in some cases clusters of basic residues (46). The interaction of actin-binding proteins to PIP 2 has been assigned to the binding of the negatively charged head-group of the phosphoinositide to the basic amino acid residues in the protein of interest. An analysis of the human villin sequence revealed five potential PIP 2 -binding sites with PIP 2 -binding motifs identified in gelsolin, ezrin, profilin, and gCap39 (7). The existence of a common motif among a large variety of PIP 2 -binding proteins is consistent with high affinity and specific binding to PIP 2 . Five such domains were identified in villin and binding of each domain to PIP 2 was examined using mutagenesis. Villin binds specifically to phosphoinositides (PIP 2 , PIP, and PI) and not all phospholipids. These studies suggest that electrostatic interactions determine the association of villin with PIP 2 because villin binds PIP 2 with higher affinity than PI and PIP. The interaction between villin and PIP 2 is not completely charge-based, because villin does not associate with vesicles containing pure PS. Instead, this preferential binding to PIP 2 can be hypothesized to require a special geometry. Furthermore, synthetic peptides corresponding to a PIP 2 -binding site in gelsolin for instance have been shown to undergo a transition from a random coil to an amphipathic ␣-helical structure in the presence of PIP 2 or SDS, suggesting that the gelsolin peptides form electrostatic as well as hydrophobic interactions with acidic lipids (36,50). Previous studies have suggested that villin is less sensitive to PIP 2 than its related protein gelsolin (28). Contrary to this, our kinetic analysis reveals that the dissociation constant for villin binding to PIP 2 (39.4 M) is comparable with that calculated for gelsolin (40.2 M) and CapG (31.9 M) (41), suggesting that villin and structurally and functionally related proteins have comparable affinity for binding to PIP 2 . Interestingly, the K d value for PLC-␥ 1 is much higher than that of villin (1 mM versus 39.4 M) (64) suggesting that villin has a much higher affinity for PIP 2 than PLC-␥ 1 and can therefore sequester the substrate for this lipase. This finding is supported by our previous in vitro and in vivo studies (33,65). These data provide a molecular basis for the role of villin in phosphoinositide-mediated signal transduction in epithelial cells.
Our mutational analysis reveals the presence of three PIP 2 binding domains, which is in contrast with previous predictions for villin, which suggested a single PIP 2 -binding site in villin (25,39). Of the three identified PIP 2 -binding sites, two sites homologous to PB1 and PB2 have been identified in the related protein gelsolin (36,39). Interestingly, all three sites are con-served among this family of actin-binding proteins (Fig. 13). As expected PB1 and PB2 are conserved among the actin-capping and -severing proteins, whereas PB5 is conserved among the actin cross-linking proteins. Thus there is a structure-function relationship between the PIP 2 -binding sites as well as the actin-modifying functions of these proteins. PB2 and PB5 overlap with previously described actin-binding sites in villin (54,66). Similar overlapping PIP 2 and actin-binding sites have been described for other actin-binding proteins including actophorin (67), profilin (9), and gelsolin (55). This suggests that in addition to analogous F-actin binding, these and other related actin-binding proteins might also display similar binding and/or regulation by PIP 2 .
Previous studies (52) have demonstrated PIP 2 -mediated inhibition of actin severing by villin. By using mutational analysis, we demonstrate that the F-actin side binding, actin severing, as well as the PIP 2 -binding sites in villin are identical, namely Arg-138 (52). There are two models that could explain the regulation of the actin-modifying functions of villin by PIP 2 . The first assumes that PIP 2 binds villin as a competitive ligand and could displace F-actin binding. The second assumes that PIP 2 changes the villin structure so that the villin-actin affinity is weakened. Identical PIP 2 and F-actin-binding sites suggests a molecular determinant, which allows villin to associate with either F-actin or PIP 2 , as well as suggests that competitive inhibition is a mechanism that could explain inhibition of actin severing by villin in the presence of PIP 2 . The only problem with this model is that the binding affinity of villin for F-actin is higher (K d ϭ 4.4 M) (32) than for PIP 2 (K d ϭ 39.5 M). The significance of the identical PIP 2 and F-actin-binding residue in PB2 (Arg-138) is therefore unclear at this point. This paradox supports the second model, namely that binding of PIP 2 could induce a structural change in the villin protein that could decrease the affinity of villin for actin. Such a conformational change has been reported in villin in the presence of Ca 2ϩ (68). A change in conformation (from random coil to ␣-helical conformation) is known for the PIP 2 binding domains in gelsolin (36,50). Our studies using circular dichroism revealed that PIP 2 induces localized structural changes in PB1 but not PB2. A lack of structural changes in PB2 does not suggest that conformational change may not be the mechanism for PIP 2induced inhibition of actin severing by villin. PB1 does not contain the major actin-severing residue in villin but is part of the actin severing domain in villin. Binding of F-actin to PB2 is stabilized by PB1; further deletion of this domain disrupts actin severing by villin (Fig. 10A). Conformational changes in PB1 following its association with PIP 2 , as noted by us, may still disrupt the association of villin with F-actin thus inhibiting actin severing in the presence of PIP 2 . Taken together, these data suggest that actin severing by villin in the presence of PIP 2 may be regulated by PIP 2 -induced conformational changes in the villin protein resulting in a decrease in the binding affinity of villin for actin.
Regulation of the actin cross-linking functions of actin-binding proteins by PIP 2 is less well understood. For instance PIP 2 has been shown to enhance the actin cross-linking function of ␣-actinin (69) as well as inhibit the bundling activity of ␣-actinin (16). In this regard, studies done with the actin depolymerizing protein cofilin and its association with PIP 2 are relevant (70). Cofilin is an actin-depolymerizing protein, but in the presence of PIP 2 cofilin forms actin bundles. Recent studies (71) have shown that self-association of cofilin is regulated by PIP 2 , and it is suggested that cofilin oligomers inhibit the actin severing and promote the actin bundling activity of this protein. In fact bundling activity is not restricted to covalently linked oligomers but is displayed by spontaneously aggregated cofilin molecules as well (72). These findings suggest that simply increasing the concentration of cofilin may form actincofilin bundles. Such a scenario can be speculated for the increase in the actin bundling function of villin in the presence of PIP 2 . We hypothesize that PIP 2 interactions with PB5 could result in dimerization or oligomerization of villin proteins, thus enhancing actin cross-linking by villin in the presence of PIP 2 . Furthermore, circular dichroism studies reveal that Pep 5, which encompasses the actin cross-linking and F-actin-binding site in villin, has an ␣-helical structure that changes to a ␤-sheet structure following its association with PIP 2 . The significance of this observation is not clear at this point, except to suggest that there is a conformational change in this domain and that this alternatively folded state of villin enhances the actin cross-linking function of villin. We speculate that structural changes induced in this domain of villin may facilitate additional actin binding or oligomers formed by PIP 2 may facilitate additional actin binding at neighboring subunits.
Association of villin with PIP 2 can have several different physiological consequences, and we hypothesize any of the following scenarios. (i) Villin is translocated to the plasma membrane upon cell stimulation (65) and is concomitantly activated by tyrosine phosphorylation (33). PIP 2 could function in recruiting villin to the cell periphery, presenting it to the tyrosine kinase. Consequently, understanding the possible role of PIP 2 in specifically localizing villin will be of major interest. (ii) Association of villin with PIP 2 regulates the PLC-␥ 1 -signaling events (33). Because villin has a higher affinity for PIP 2 than PLC-␥ 1 , villin could sequester PIP 2 and release PIP 2 following tyrosine phosphorylation (in response to receptor activation) and thus play a role as a signal-generating actin-regulatory protein (33). (iii) Villin is localized to the leading edge of intestinal epithelial cells (73,74). Association of villin with PIP 2 could regulate actin assembly in the vicinity of the plasma membrane by inhibiting severing and capping, thus accelerating actin assembly, and by increasing actin bundling it can stabilize the cell surface protrusion, thus regulating cell shape change as well as cell motility.
Interaction of villin with PIP 2 may lead to changes in cell structure and function in a spatially defined manner by providing a local environment that recruits and activates the actin-modulatory functions of villin. Recent evidence suggests that PIP 2 promotes membrane-cortical cytoskeleton interactions (18) and that plasmalemmal actin polymerization promotes cell spreading (75). In this context, interaction between villin and PIP 2 could promote growth and stabilization of peripheral actin structures. First, inhibition of villin by PIP 2 should reduce actin severing; second, it should promote local growth by uncapping filament plus ends; third, bundling and cross-linking of actin bundles can stabilize the peripheral actin structures and cytoskeleton stability. Villin-PIP 2 interaction could result in increased actin polymerization in the vicinity of the plasma membrane, and actin bundles stabilized by villin-PIP 2 interactions could support and stabilize protrusion and invagination or domains of the plasma membrane thus having a profound effect on cell shape and function. Villin-PIP 2 interaction might be a mechanism for the membrane-associated actin polymerization at the leading edge or may be associated with membrane ruffles. This idea is consistent with many reports linking PIP 2 to the reorganization of the actin cytoskeleton through the regulation of actin-binding proteins. Experimental manipulation of cellular phosphoinositide levels regulates the formation of cell surface structures as well as the motile phenotype of the cell (76 -79). Most noteworthy, gelsolin and PIP 2 localize in lamellipodia during epidermal growth factor-induced motility (80), and overexpression of PI 4-phosphate 5-kinase induces the formation of vesicle-associated actin structures similar to the Listeria-induced actin comet tails (81), suggesting PIP 2 plays a key role in regulating cell motility. Interestingly, there are several studies that allude to a role for villin in cell migration (73,74,82,83).
The fact that there is one identical and two overlapping sites for PIP 2 and actin binding in villin suggests that PIP 2 could selectively displace actin from the actin-severing site, thus inhibiting the severing function of villin while binding to the other two sites could result in the formation of a villin-PIP 2 complex that could bind and stabilize the actin filaments. The existence of multiple PIP 2 -binding sites in villin in addition to its other ligand binding functions (F-actin, G-actin, Ca 2ϩ , and tyrosine phosphorylation) adds another level of complexity as well as another limit to the regulation of actin-modifying functions of villin. Future experiments designed to determine the spatial relationship between villin and PIP 2 would help understand the regulation of villin function by PIP 2 and the concomitant changes in the actin dynamics in epithelial cells.