The 3D Structure of Villin As an Unusual F-Actin Crosslinker

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Structure Article

The 3D Structure of Villin as an Unusual F-Actin Crosslinker

Cheri M. Hampton,1,3 Jun Liu,1,4 Dianne W. Taylor,1 David J. DeRosier,2 and Kenneth A. Taylor1,* 1Institute of Molecular Biophysics, Florida State University, Tallahassee, FL 32306-4380, USA 2W.M. Keck Institute for Cellular Visualization, Rosenstiel Basic Medical Sciences Research Center, Department of Biology, Brandeis University, Waltham, MA 02254, USA 3Present address: Howard Hughes Medical Institute, Columbia University Medical Center, P&S Black Building 2-221, 650 W. 168th Street, New York, NY 10032, USA 4Present address: Department of Pathology and Laboratory Medicine, The University of Texas–Houston Medical School, 6431 Fannin, MSB 2.228, Houston, TX 77030, USA *Correspondence: [email protected] DOI 10.1016/j.str.2008.09.015

SUMMARY Calcium binding to the type 2 (intradomain) site within G6 re- leases this inhibition (Choe et al., 2002). Further calcium effects Villin is an F-actin nucleating, crosslinking, severing, involve domain rearrangements that release ‘‘latch’’ mechanisms and capping protein within the gelsolin superfamily. between G1 and G3 to activate the G1 actin-binding site, and We have used electron tomography of 2D arrays of between G4 and G6 to allow G4 to participate in actin capping villin-crosslinked F-actin to generate 3D images re- (Choe et al., 2002). Further interaction of G1 and G4 with the 2+ vealing villin’s crosslinking structure. In these polar filament via type 1 (interdomain) Ca ion coordination between arrays, neighboring filaments are spaced 125.9 ± actin and gelsolin leads to severing and capping. The severing mechanism requires micromolar [Ca2+](Yin and Stossel, 1979). 7.1 A˚ apart, offset axially by 17 A˚ , with one villin cross- Villin, in contrast, has only three established calcium-binding link per actin crossover. More than 6500 subvolumes sites (Hesterberg and Weber, 1983a) and requires [Ca2+]as containing a single villin crosslink and the neighbor- high as 200 mM to activate severing. Such concentrations are ing actin filaments were aligned and classified to not usually found in live cells, but might occur in apoptotic cells. produce 3D subvolume averages. Placement of a Tyrosine phosphorylation of villin can reduce or even eliminate complete villin homology model into the average den- the calcium-binding requirement for severing, suggesting that sity reveals that full-length villin binds to different phosphorylation may be the primary regulator of villin conforma- sites on F-actin from those used by other actin-bind- tion (Kumar and Khurana, 2004). ing proteins and villin’s close homolog gelsolin. The crystal structure of gelsolin’s domain G1 in complex with G-actin (McLaughlin et al., 1993) revealed G1 bound in the hydrophobic pocket between actin subdomains 1 and 3 where it caps INTRODUCTION the F-actin barbed end. The crystal structure of G1-G3 in combination with calcium and G-actin identified the unique G2-G3 Villin, an 95 kD actin crosslinking protein, is a member of the actin side-binding position (Burtnick et al., 2004) that competes gelsolin superfamily (Bretscher and Weber, 1980). Gelsolin family with a-actinin for F-actin binding (Way et al., 1992). Electron mi- proteins regulate F-actin length by severing filaments and/or cap- croscopy (EM) difference mapping of G2–G6 decorated F-actin ping the barbed ends. Each member has three or six homologous in the presence of calcium localized the G2-binding site on the copies of a 120 amino acid (14 kD) domain, the gelsolin-like do- side of F-actin, but the rest of the molecule was disordered main. Gelsolin and villin each have six such domains, denoted (McGough et al., 1998). The C-terminal half, G4–G6, has similarly here as G1–G6 and V1–V6, respectively. Villin has 45% amino been crystallized with calcium and G-actin and shows that G4 acid sequence identity with gelsolin over the six domains (Finidori binds to G-actin in a similar fashion as G1 (Choe et al., 2002). et al., 1992), suggesting that their tertiary structures are very The crystal structure of full-length gelsolin in the inactive, cal- similar. Gelsolin cannot crosslink F-actin whereas villin can, a cium-free state revealed a compact, autoinhibited form in which property attributed to the presence of a small, 76 amino acid G6 and G2 interact to form the above-mentioned ‘‘tail helix latch’’ ‘‘head piece’’ domain at villin’s C terminus (Glenney et al., 1981). that prevents the actin-binding regions from contacting actin Villin can nucleate actin polymerization, crosslink filaments, (Burtnick et al., 1997). Small-angle X-ray scattering of gelsolin sever F-actin at high [Ca2+] or following tyrosine phosphorylation, revealed global conformational changes with increasing calcium and cap the barbed ends of filaments. Much of what is inferred and highly flexible interdomain linkers (Ashish et al., 2007). about villin function and structure comes from a comparison Structural data for villin are limited. Hydrodynamic and spec- with gelsolin, its closest homolog. Gelsolin’s function is regulated troscopic studies have shown that villin in solution undergoes a by calcium binding at up to eight different sites. At submicromolar large calcium-induced conformational shift to become more [Ca2+], gelsolin folds into a compact autoinhibited state which asymmetric, with an overall length increase from 84 to 123 A˚ places the tail helix of the C-terminal G6 domain close to G2, cre- (Hesterberg and Weber, 1983b). These data led Hesterberg and ating a steric block to G2 binding of F-actin (Burtnick et al., 1997). Weber to suggest a ‘‘hinge mechanism’’ where calcium binding

Figure 1. ElectronMicrographs ofF-Actin:Vil- lin Crosslinks (A) Electron micrograph of a well-ordered segment from tomogram 1 (Table 1). Protein is white. (B) Fourier transform of the raft shown in (A). The meridional and equatorial axes are shown in black dashed lines; layer lines and row lines are in gray. (C) Projection image of the 3D global average of 6500 aligned villin crosslink repeats. (D) Averaged 2D projection image of ice-embedded specimens, for comparison.

Two types of 2D arrays are formed when fimbrin crosslinks F-actin on a lipid monolayer, which suggests polymorphism among fimbrin crosslinks (Volk- mann et al., 2001). In one type, adjacent filaments are in axial register, whereas in the second, adjacent filaments are dis- placed axially by 55 A˚ . This raises the question of whether the addition of villin creates a localized tightening of the villin structure coupled to an crosslinks would favor one form of fimbrin crosslink over the outward shift of another domain (presumably the head piece) into other. solution. Recently, two potential latch residues have been identi- Here we employ electron tomography of F-actin:villin 2D arrays fied based on sequence homology to gelsolin, Asp467 and to learn in 3D how villin interacts with and crosslinks F-actin in low Asp715, that potentially form salt bridges with a cluster of basic [Ca2+]. The 2D arrays provide a format to study the actin-bound residues in villin domain V2 (Kumar and Khurana, 2004). How- structure of proteins that interact weakly with actin and have ever, no clear demonstration of villin in a folded, autoinhibited a tendency to form bundles either in vitro or in vivo and helps state exists. define specific protein-actin interactions as well as any polymor- The solution NMR structure of V1, also known as villin 14T, re- phisms in the crosslinked filaments. Using multivariate data vealed an overall fold that is very similar to that of gelsolin G1 as analysis (MDA) of subvolumes extracted from the tomograms, predicted by the high sequence identity (Markus et al., 1994, we show that the villin-actin contacts are homogeneous and 1997). Several structures exist for the villin head piece domain different from contacts for other actin-binding proteins. Without (Frank et al., 2004; McKnight et al., 1997; Meng et al., 2005; Var- calcium to activate it, villin can crosslink F-actin without disrupt- dar et al., 1999; Vermeulen et al., 2004). Cysteine scanning mu- ing the filaments. We find that the filament arrangement in tagenesis of this domain has provided comprehensive data for F-actin:villin 2D arrays differs from F-actin:fimbrin 2D arrays, the surface distribution of basic residues and their presentation suggesting that 3D actin bundles in microvilli involve arriving at to actin (Doering and Matsudaira, 1996). A recent NMR structure a compromise structure or bonding rule. We also explore the of a V6-headpiece construct demonstrated that the 40 amino novelty of these villin-actin contacts in the context of villin’s acid connecting linker is unfolded and disordered (Smirnov homology with gelsolin. et al., 2007). Villin and fimbrin, another actin crosslinking protein, are found RESULTS together in the actin core of finger-like projections known as microvilli in intestinal epithelium and kidney proximal tubules. Each Image Alignment and Volume Classification microvillus contains 20+ polar actin filaments (Tilney and Moose- In the presence of EGTA, villin forms regularly spaced crosslinks ker, 1971). In the normal microvillus, actin filaments are separated along F-actin in 2D paracrystalline arrays R1 mm wide formed on laterally by 125 A˚ (DeRosier et al., 1977). Microvilli form in villin- lipid monolayers. We collected six different single-axis tilt series deficient mice but they lack the typical discrete F-actin bundles and computed six tomograms. Subvolumes were combined within the core (Ferrary et al., 1999; Pinson et al., 1998), suggest- after initial alignment on the actin filaments. Resolution was trun- ing that whereas the loss of villin can be partially compensated by cated to 38 A˚ , consistent with the position of the first node of the fimbrin, the finer structure of the microvillus may be specific to contrast transfer function. The villin crosslinks, which occur once villin’s functions. Moreover, villin’s ability to induce microvillus per actin crossover, appear as long, extended ‘‘squiggles’’ in the formation is a function of its crosslinking ability, not its nucleating raw tomogram (Figure 1A; see Figure S1A available online), mea- ability (Friederich et al., 1999). It is not known how both fimbrin suring between 126 and 140 A˚ in length. The measured volume of and villin cooperate in the microvillus core or in vitro to bundle 130.2 3 103 A˚ 3 for the villin crosslink indicates that 95 kD villin F-actin. Each crosslinking protein by itself may produce its own binds as a monomer. Actin filaments are predominantly sepa- preferred F-actin bundle geometry, whereas in combination, rated by 126 A˚ (Table 1), a separation that is close to the value a single compromise bundle geometry may occur. of 120 A˚ found for arrays of actin and fimbrin (Volkmann

Table 1. F-Actin:Villin Raft Parameters ˚ ˚ Tomogram Actin Symmetry Interfilament Spacing (A) Filament Shift (A) strans Filament Rotation srot Ideal Rotation Filaments/Repeat 1 80/37 125.3 16.3a 2.2 5.7c 1.1 5.4 5 2 186/86 129.5 16.9a 2.4 10.0c 2.0 9.0 3 3 80/37 125.3 16.9a 1.2 5.5c 1.2 5.4 5 4 13/6 133.4 17.5b - 0.0 - 0.0 1 5 186/86 112.5 27.0b - 0.0 - 0.0 1 6 54/25 129.7 17.1b - 0.0 - 0.0 1 Average 125.9 18.6 s 7.1 4.1 a Calculated from Equation 13 of Sukow and DeRosier (1998) using the idealized filament rotation in column 8. b Calculated from the reciprocal lattice vectors. c Calculated from Equation 13 of Sukow and DeRosier (1998) using an average filament translation of 17.3 A˚ (average from tomograms 4 and 6). et al., 2001) and nearly identical to values obtained for microvilli crosslinks during our initial rounds of alignment. Once the subvo- (DeRosier et al., 1977; Matsudaira et al., 1983). lumes were sufficiently aligned to the global average, further Actin filaments in the arrays varied in their helical symmetries classification of the villin crosslink was done based on the vari- and their rotational orientations. Computed Fourier transforms ance calculated in the region of the villin density. The resulting of well-ordered regions within each of the six tomograms (e.g., class averages (Figure 2) show that, in some cases, potential Figure 1B; Figure S1B) were analyzed according to Sukow and villin crosslinking sites are unoccupied, some sites show partial DeRosier (1998). Predominately, each neighboring filament is villin density, whereas other variations can be explained by villin axially shifted 16.9 ± 0.4 A˚ up, from left to right (Table 1). The shift flexibility. The partial villin density could be attributed either to was confirmed from measurements made on two global aver- negative staining effects or proteolysis because villin, like gelso- ages, one in negative stain (Figure 1C) and one in ice (Figure 1D). lin, is susceptible to cleavage of its long interdomain linkers Of the well-ordered arrays analyzed, three showed no rotation (Bazari et al., 1988). between adjacent actin filaments, two showed a rotation of 5.4, and one showed a rotation of 10. Actin filament symme- An F-Actin:Villin Crosslink Model tries varied from 13/6 (166.15 rotation between actin subunits) By rigid-body fitting, we docked the atomic model for a 13/6 actin to 54/25 (166.67). One array in Table 1 (#5) showed a 17 A˚ filament into our F-actin volume in the tomogram. A quasi-atomic smaller interfilament spacing and a 27 A˚ axial offset, values model for chicken villin was created (see Experimental Proce- more than 3s different from the averages of the other five. dures) based on the known structures of villin domains V1 and The axial rotational differences between adjacent filaments the head piece and homology models from gelsolin for domains were also determined by rotational crosscorrelation in 1 steps. V2–V6. We then placed V1, V2, V3, and the head piece as individ- Although less accurate than the raft analysis described above ual domains into the villin volume, and placed V4–V6 as a group because the missing wedge may affect the alignment, it mea- (Figure 3). In the class averages there appear to be at least three sures the entire set of crosslinks. Histograms of the rotation contacts of villin with the actin filament, with a fourth possible. angles show predominately three groups centered on 0,5, The two contacts at the top of the density are far enough apart and 10 (Figure S2), consistent with the raft analysis. The most to accommodate two gelsolin domains (Figures 3A and 4A). favored rotation by far is 0, whereas there is little difference in These are tentatively labeled V1 and V2, because at this resolu- the frequency of occurrence for 5–6 and 9–11. The average tion it is impossible to distinguish them. Because of the connect- axial shift obtained from this analysis was 13.5 ± 3.7 A˚ , in agree- ing linker, it is also not possible to determine with certainty the ment with that determined from the array analysis. Thus, although orientation of V1 and V2. The small tail region fits the villin head the axial shift seemed fixed at 17 A˚ , variations of up to 10 in piece structure, but is too small for a gelsolin-like domain (Fig- filament rotation appeared to be tolerated. ure 4D). Likewise, the head piece is too small for the V1-V2 den- The global average of the aligned subvolumes shows clearly sity (Figure 4C). The head piece crown of positive charge, located the shape of the villin and its location relative to the actin subunits on the C-terminal helix, was placed facing the actin filament (Fig- (Figure 1C). The 17 A˚ axial filament shift brings the right-side ac- ures 3 and 4B) according to the fitting of the dematin head piece tin subunit closer to the left-side subunit. Thus, the actin subunits on F-actin (Chen et al., 2003). The 40 amino acid unfolded linker that interact with villin are offset axially only 27 À 17 = 10 A˚ . Three between V6 and the head piece may explain the thin and variable contacts are seen between villin and actin, two across the top crook of density we see leading from the villin core to the F-actin and one on the left side at the bottom. These contacts are also interacting site (Smirnov et al., 2007). This same weak density seen in the projection of frozen-hydrated arrays (Figure 1D). corresponding to the N terminus of the head piece was observed The two projections are virtually identical, showing that the struc- by Chen et al. in an EM map of the dematin head piece-decorated tures of negatively stained arrays are the same as those in F-actin (Chen et al., 2003). The dematin head piece has 47% frozen-hydrated arrays. sequence identity with the villin head piece (Rana et al., 1993). Using MDA (Winkler and Taylor, 1999), we separately charac- Perhaps the most curious aspect of the villin contacts with terized the heterogeneity in the actin filaments and the villin F-actin is that the location on actin is different from that of other

03 1 4 5 7 62 Figure 2. Class Averages Based on the Variance within the Villin Crosslink Labels across the top are in order of relatedness from the classiﬁcation. Numbers at the bottom are the number of class members. (A) Projection images. Class average 1 shows A 412 503 453 637 831 533 542 296 incomplete ‘‘half-villin’’ density. Average 0 has no villin crosslink. Class average 2 illustrates a fraction of crosslinks that are highly variable at one end. (B) Same averages as above, but shown as volumes. All averages in (B) are contoured to the same threshold.

B actin-binding proteins. The arrangement of V1 and V2 within the actin as determined separately without steric clash. Recent density indicates an interaction with actin subdomains 1 and/or results on the structure of F-actin decorated with fimbrin ABDs 4, unlike those typical for gelsolin-like domains (Burtnick et al., indicate that the actin-bound ABD1 is poorly ordered, suggest- 2004; Choe et al., 2002). G1 binds in a hydrophobic pocket be- ing flexibility or polymorphism in one of its CH domains’ interac- tween actin subdomains 1 and 3 (Figure 4; Figure S3). However, tion with actin (Galkin et al., 2008). binding of G1 is characteristic of end-capping proteins. The For the second model (Figure 6B), we remodeled fimbrin as actin-binding domain (ABD) of a-actinin binds to two adjacent described above as well as villin and kept the filaments in axial actin protomers by contacting subdomains 1 and 2 on one actin register with no offset. A villin crosslinker must bind both actin protomer and subdomain 1 of the next protomer up the filament. filaments, and this could be accomplished using no axial offset This corresponds to actin residues 86–117 of the lower protomer by eliminating the V2-actin interaction. V2 still remains close to and residues 350–375 of the upper protomer (Fabbrizio et al., actin, but any residual interaction must involve different surfaces 1993; Lebart et al., 1990, 1993; McGough et al., 1994; Mimura of both V2 and actin. The head piece linkage to V6 also required and Asano, 1987). The gelsolin G2 domain has a similar actin remodeling, but it is known to be flexible. interaction (McGough et al., 1998). Although the loops on the ends of the actin-binding helices in V1 and V2 are in close prox- DISCUSSION imity to actin residues 358–365 (Figure 5), the rest of the protein density is oriented back and away from the a-actinin/G2 binding Villin is a compact 84 A˚ molecule in solution but extends to about region. 123 A˚ on the binding of calcium (Hesterberg and Weber, 1983b). Although the exact binding site of villin’s head piece on F-actin Our map is well fit by an extended villin molecule and bears is unknown, competition-binding assays have indicated that the a striking similarity to a recent structure obtained by small-angle villin head piece can displace the binding of the actin depolyme- X-ray scattering of Ca2+-activated gelsolin (Ashish et al., 2007). rizing factor, which suggests that it binds to actin in the same However, our specimens are formed in low [Ca2+] and with actin. hydrophobic binding pocket as G1 and other capping proteins Although we do not know whether the villin conformation in the (Pope et al., 1994). However, the position of the villin head piece crosslink is the same as that in an actin-free calcium medium, density in our reconstruction is not in this hydrophobic pocket. the two forms of these highly similar molecules, one compact, Instead, it is positioned near two a helices, one from actin subdo- the other extended, reinforce the hypothesis that villin contains main 3 (residues 308–320) and one from subdomain 4 (residues a number of flexible hinges. 222–236). The head piece localization is invariant among the class averages. Villin Forms an Unusual F-Actin Crosslink The class averages of F-actin:villin were surprisingly similar (Fig- Compatibility with the F-Actin:Fimbrin Crosslink Model ure 4), consistent with constancy of villin-actin contacts. Poly- The F-actin:fimbrin 2D arrays and the F-actin:villin 2D arrays morphism in the arrays is largely due to differences in rotations have different geometries: the former has either a 0 or 55 A˚ offset, of adjacent actin filaments around the filament axis. The glancing and the latter have a 17 A˚ offset, which raises the question of contacts of villin at the outermost edge of the actin protomers what types of changes would be required to place the villin (Figures 3, 4A, and 5) suggest that each bond contributes only model into the F-actin:fimbrin crosslink model. We explored weakly to the total binding. Multiplicity of weak interaction sites, this in two ways (Figure 6). For both models, we built a modified in fact at least three such sites, cooperatively stabilizes the F-actin:fimbrin crosslink by combining the coordinates of the crosslinking in the presence of the high F-actin concentration in F-actin:fimbrin crosslink from Volkmann et al. (2001) with recent the microvillus. It would also give villin a dynamic interaction with data for the interaction of fimbrin’s ABD2 with F-actin from the actin bundle, allowing it to rapidly convert between different Galkin et al. (2008). The first model (Figure 6A) was built using bound configurations. the 17 A˚ filament offset with a modified fimbrin crosslink. Despite Unlike gelsolin, villin possesses an F-actin crosslinking activ- the offset, the two fimbrin coordinate sets could be placed on ity, which was largely thought to reside in its unique head piece.

Figure 3. Model of Villin Crosslinking F- Actin (A) Placement of the villin homology model atomic coordinates within the averaged villin density. (B) Schematic of villin domain topology with the approximate location of linkers shown.

minimum, two contacts by villin are required for crosslinking. Our modeling (Figure 6B) suggests that a minimal villin crosslink could adapt to the axially aligned filament geometry dictated by fimbrin by eliminating the V2-actin bond and remodeling the link between the villin head piece and V6 to accommodate the 17 A˚ shift. Although plausible, the cross- AABlink would probably be less energetically favorable compared to the three-bond However, a head piece by itself cannot crosslink filaments, un- crosslink seen in the F-actin:villin rafts. Another possibility is less villin crosslinks as a dimer, which our data show is not the that once localized in the densely packed microvillus, villin could case. We suspect that the head piece is essential for holding play a passive role in bundle geometry by simply binding to the villin on F-actin and that crosslinking occurs via V1-V2. In the side of a single actin filament using the V2 and head piece bind- quasi-atomic model, end loops on V1 and V2, both of which con- ing sites. However, side binding would be less effective for tain arginine and lysine residues, are placed on an acidic a helix targeting to the microvillus than crosslinking, which is a function of actin subdomains 1 and 4, forming a predominately electro- of bundle geometry. Either way, the ability of villin to determine/ static association. The buried surface is small, suggesting a dictate the physical geometry of the bundle is reduced because weak and dynamic interaction. It is not known whether a V1-V3 one interaction is lost. villin fragment could crosslink actin filaments on its own, as sug- Both villin and fimbrin crosslinks produce exceptionally high gested by one of our class averages, but optimal placement of local actin concentrations, negating the need for strong actin af- the actin filaments produced by full-length villin crosslinks may finity for bundling. Cooperativity among multiple binding sites be an important factor in the present case. would stabilize the bundle. With weak interactions, the off rates for any one of the villin-actin bonds would be high, promoting Comparison with F-Actin:Fimbrin Arrays the ability to change conformation quickly if conditions change. The organization of filaments in F-actin:fimbrin arrays differs The high local actin concentration would produce high on rates from that found in F-actin:villin arrays. Filament separation in that keep villin concentrated within the bundle. If fimbrin is the F-actin:fimbrin arrays is 120 A˚ whereas it is 126 A˚ in F-actin:vil- dominant influence on the bundle structure, possibly villin would lin arrays, identical to that found in microvilli. Fimbrin also formed have only two binding sites, making it highly dynamic in vivo. two types of arrays, one with a 0 A˚ and the other with a 55 A˚ axial Multiple weak interactions could serve as a positioning/locating offset between neighboring filaments. Both fimbrin arrays, on mechanism for villin on the filaments, facilitating concentration of average, have no relative rotation between actin filaments. villin into the bundle, even if the rate might be slow. In high [Ca2+], F-actin:villin arrays have a consistent 17 A˚ axial shift between both fimbrin and villin would lose their crosslinking capability, so neighboring actin filaments but a variable 0–10 filament rota- that the bundle would collapse rapidly with no active crosslinkers tion. Our results show that the fimbrin and villin binding sites in and an active severing protein embedded within it. We envision the bundles do not overlap, so that coexistence in the same a scenario of quick conversion from bundling to severing without bundle is possible. In other respects, some accommodation is large-scale diffusion of villin, confining actin filament cleavage to necessary for an integrated bundle geometry. a localized area. It would appear that for villin and fimbrin to coexist in the same microvillar bundle requires either villin, fimbrin, or both to Implications for In Vivo Actin Bundles alter their mode of interaction with actin to accommodate the dif- Theoretically, one way to form a hexagonal bundle of actin ference in preferred actin filament offsets, with perhaps the filaments using a single crosslinker is for the actin filaments to linkers between domains performing a key role. It is somewhat all be arranged in axial register (DeRosier and Tilney, 1982). surprising that fimbrin, with shorter linkers between its ABDs, This allows a single quasi-equivalent bonding rule to define all demonstrates two different crosslink geometries whereas villin, the crosslinks, even for actin filaments with different helical pitch. possessing long linkers between domains, demonstrates a single However, there are published examples of images and diffrac- geometry. tion patterns of actin bundles possessing well-oriented actin fil- The model shown in Figure 6A suggests that the 17 A˚ offset aments with crossovers having regular offsets (DeRosier et al., seems to be compatible for forming a fimbrin crosslink. At a 1977; DeRosier and Tilney, 1982; Tilney et al., 1980). Indeed,

Figure 4. Close-Up of Villin-Actin Interac- tions (A) A top-down view of the model in Figure 3. The N terminus of the long V2 helix, residues 151–158, points between two actin subunits. On the left- hand filament, which is interacting with villin A C domain V2 (red), the actin subdomain 1 (S.D. 1) helices, residues 112–126 and 358–365 of one subunit and on the next subunit up, actin subdomain 4 (S.D. 4) helix residues 222–236, all of which are involved in villin interaction, are shown in red. Likewise, V1 (blue) is positioned between the same regions of actin on the right-hand filament. (B) Position of the villin head piece. The head piece helix residues 62–76, which include the ‘‘crown of positive charge,’’ are positioned toward actin and in the vicinity of helix residues 308–320 on actin subdomain 3 (S.D. 3). The disordered linker can be readily accommodated within the linker density. BD (C) Replacement of V2 by the head piece. The head piece is too small for the V2 density. (D) Replacement of the head piece by V1. The gelsolin-like domain is too large for this density. Thus, the head piece placement determines the overall orientation of the villin crosslink model. one report (DeRosier and Tilney, 1982) observed an axial offset reduce this amount. Assuming the helix is regular, our results of 18 A˚ in the actin bundles in stereocilia, in which fimbrin is show that villin crosslinks form with azimuthal variations be- the dominant crosslinker. However, these offsets are generally tween actin filaments of up to 10. On this criterion, villin appears attributed to bending of the actin bundle (Tilney et al., 1983) to be well suited for a crosslinking role in a hexagonal bundle. In rather than to a general feature of the bundle organization. other ways, villin appears to be poorly suited to hexagonal The formation of hexagonal bundles of actin filaments with bundle formation. a regular twist requires that the crosslinkers tolerate 4–5 of At a minimum, two actin contacts are required for villin to azimuthal variation in the orientation of their actin binding sites crosslink F-actin, one on V1 and the other on the head piece. (DeRosier and Tilney, 1982). Local variations in the F-actin helical However, our reconstructions show three: an additional contact twist (Egelman et al., 1982) or even actin protomer flexibility may is formed by V2. We think that the rigid selection of a 17 A˚ offset is determined by the V1-V2 arrangement because the density connecting the head piece with V6 is very thin, suggesting flexibility in this location, consistent with the disorder observed in solution (Smirnov et al., 2007). Flexibility between the head piece and V6 would make the head piece an effective crosslinker under a vari- ety of conditions, but poor with respect to enforcing a particular geometry. In addition, one class average having the same 17 A˚ offset as the others appeared to contain a villin molecule trun- cated after V3 and therefore had no villin head piece (#3 in Fig- ure 2) but still retained the V1-V2 arrangement. We suspect that a flexible head piece would allow villin to accommodate a filament arrangement, specifically a filament alignment between crosslinked actins dictated by the fimbrin crosslinks. Unlike the other gelsolin-like domains, V1 and V2 appear to be relatively inflexibly joined; otherwise, there should be more variation in AB the actin filament offsets. In the class average used for quasi- atomic model building, the density connecting head piece and Figure 5. Villin and Gelsolin Binding Sites Do Not Overlap V6 is narrow but would accommodate the linker in a folded con- Villin density map and atomic model are shown next to gelsolin crystal struc- formation (although our structure here is entirely hypothetical), tures (colored copper) bound to actin. suggesting that crosslinking may induce a folded state of this (A) N terminus of gelsolin bound to actin as in the crystal structure. G2 binds via otherwise mobile segment. However, other classes show even its long a helix, the ‘‘actin binding helix,’’ between actin subdomains 1 and 3; less linker density, consistent with high flexibility. G3 does not make contact with actin. G1 bound in this manner effectively caps The apparent incompatibility between F-actin:villin bundles the barbed end of the filament (PDB ID code 1RGI). (B) C terminus of gelsolin bound to actin. G4 makes contact with actin subdo- and F-actin:fimbrin bundles may be explicable in two ways. Villin main 3. G5 and G6 are not bound to actin; however, the long a helix of G6 is in is not necessary for formation of microvilli, but its absence leads close proximity to helices of actin subdomains 3 and 4 (PDB ID code 1H1V). to poorly ordered bundles (Pinson et al., 1998). Villin appears

Figure 6. Fimbrin and the F-Actin:Villin Crosslink Model To demonstrate compatibility between these two co-crosslinkers, we placed the atomic coordinates of fimbrin ABD2 (orange) on the left-hand actin filament and ABD1 and the calmodulin-like domain (violet) on the closest right-hand actin monomer. It is not known exactly how these domains interact with one another in a crosslink, but all are accommodated within the context of our model without a steric clash. (A) Model built with the two actin filaments offset by 17 A˚ . (B) Model built with the two actin filaments in axial register. In this model, the interaction between V2 and actin must be broken and the head piece linker rebuilt.

AB early in microvilli formation and is followed by fimbrin (Ezzell Despite the high sequence identity between villin and gelsolin et al., 1989). If villin appears early, it might favor formation of in the six main domains, the F-actin-villin interactions observed crosslinks with 17 A˚ offsets. An offset of 17 A˚ is closer to the here are uncharacteristic when compared to the known interac- aligned crosslink produced by fimbrin, rather than the alternative tions of gelsolin with F-actin or G-actin. However, our work per- 55 A˚ offset that fimbrin can form. Our modeling suggests that tains to a capability of villin for interacting with actin in low [Ca2+] with a 17 A˚ offset, fimbrin can form a crosslink, which we spec- that gelsolin completely lacks. Villin’s actin crosslinking activity is ulate would anneal/adjust to one with no offset as long as the present in low [Ca2+], under which conditions its severing and villin bonds are forming and dissociating rapidly, which our struc- capping activity are minimal. Essentially, villin crosslinks F-actin ture suggests is indeed the case. Villin would then be the agent while in its enzymatically ‘‘inactive’’ state. Gelsolin is also enzy- favoring an aligned bundle. Its absence would force a disordered matically inactive in low [Ca2+] and forms a folded conformation bundle, as fimbrin by itself would form crosslinks with different that blocks its surfaces that participate in actin binding, severing, offsets. Villin, having multiple weak and variable interactions and capping. Our reconstructions show that the villin ‘‘inactive’’ with actin, would be retained in the aligned bundle, where it state bears no resemblance to the gelsolin inactive state. In low performs an important role for remodeling and disassembly of [Ca2+], the homologous actin interaction sites of gelsolin are not the microvilli (Glenney et al., 1980). placed appropriately for F-actin crosslinking as they are in villin. Villin’s multiple weak interactions with actin would also be It is therefore not surprising that the F-actin:villin crosslinks in low important for the mechanical characteristics of the bundle. Villin [Ca2+] cannot be predicted by gelsolin-actin interactions in high and fimbrin occur in approximately equal quantities in the micro- [Ca2+]. Interactions with F-actin in high [Ca2+], under which con- villus (Matsudaira and Burgess, 1979), and our results show that ditions both villin and gelsolin cap and sever F-actin, could be their interaction sites do not compete with each other. If villin closely parallel, but that remains to be shown. It also remains crosslinked actin as strongly as fimbrin, it would have important to be determined whether villin forms a gelsolin-like inactive con- consequences for the mechanical characteristics of the bundle. formation in the absence of F-actin, which would imply that actin Fimbrin by itself has crosslinking characteristics that are ideal to induces a conformational change that results in crosslinking. accommodate bending in the bundle (Claessens et al., 2006). Vil- That the more than 6500 villin crosslinks were so homogeneous lin, as a weak actin binder, would have dynamic bonds with rapid in their structure suggests that what we have observed is not an on and off rates that may have little effect on the mechanical exceptional case for actin interaction in this protein, but rather properties of the bundle while permitting villin to occupy the a legitimate mode of interaction for gelsolin-like domains. This bundle in high concentration. is a further illustration that inferring protein-protein interactions based on homology alone may be misleading unless the func- Modes of F-Actin Binding among Homologous Domains tional roles are also homologous. It is commonly assumed that homologous actin binding proteins In summary, we have used electron tomography of 2D interact similarly with actin, although this is not always the case, F-actin:villin arrays and subvolume classification to obtain as shown for the widely studied ABDs composed of calponin what to our knowledge is the first 3D image of villin in a crosslink- homology domains which include those of fimbrin, a-actinin, ing role. The structure of the villin crosslinks and the interactions utrophin, dystrophin, and calponin itself (Galkin et al., 2006, of the individual villin domains with actin could not have been 2008). Our research identifies a novel interaction between actin predicted from the known gelsolin-actin structures. The specific and villin that is not predicted by actin-bound structures of villin’s geometry of the 2D arrays determined by the villin crosslink is closest homolog gelsolin and bears on the question of similarity different from that found in F-actin:fimbrin arrays. However, villin of actin interactions by homologous proteins. possesses some rather long interdomain linkers, which may

facilitate accommodation of villin into actin bundles whose struc- the structures for villin 14T (PDB ID code 2VIK), gelsolin domains G2-G3 in an ture is determined by fimbrin crosslinks. Our modeling suggests open conformation (PDB ID code 1RGI), the gelsolin G3-G4 linker through G6 that villin has sufficient flexibility to bind or crosslink F-actin in in the closed conformation (PDB ID code 1D06), and the villin head piece (PDB ID code 1YU5). The disordered sequence linking V6 to the head piece was cooperation with fimbrin in the microvillus, thereby stabilizing modeled based on other flexible linkers from the above structures. Manipula- the structure and, at the same time, embedding within it the tion of the PDB files and energy minimization was performed using the MMTSB apparatus for rapid disassembly. Tool Set (Feig et al., 2004). Coordinates were checked with PROCHECK (Laskowski et al., 1993). Models of villin crosslinks were created in Chimera EXPERIMENTAL PROCEDURES (Pettersen et al., 2004).

Protein Purification and Array Formation ACCESSION NUMBERS Villin was purified from chicken intestines (Burgess et al., 1987). Actin was prepared from rabbit muscle acetone powder (Pardee and Spudich, 1982), The tomograms, the global average, and the eight major class averages have followed by a Superose-12 column. Fresh G-actin was prepared by dialysis been deposited in the European Bioinformatics Institute under accession overnight against 2 mM Tris-Cl, 0.2 mM Na ATP, 0.02% b-mercaptoethanol, 2 numbers EMD-1491, 1496, 1497, 1537–1539. 0.2 mM CaCl2, 0.01% NaN3 (pH 8.0). The protein was clarified by high-speed centrifugation prior to sample setup. Arrays were grown on positively charged lipid monolayers (Taylor and Taylor, 1994). A 3:7 volume ratio of DLPC:DDMA SUPPLEMENTAL DATA lipids (Avanti Polar Lipids) at 1 mg/ml in chloroform was layered over a solution containing 0.35 mM a-actin and 0.14–0.17 mM villin in phosphate-buffered Supplemental Data include three figures and can be found with this article on- polymerizing buffer without CaCl2 and with 1 mM EGTA (pH 6.5). The actin line at http://www.cell.com/structure/supplemental/S0969-2126(08)00422-X/. was polymerized in the presence of villin at 4C. ACKNOWLEDGMENTS EM Data Collection The 2D arrays were recovered using 200–300-mesh copper grids with a retic- We thank Hanspeter Winkler for his guidance, Niels Volkmann and Dorit Ha- ulated carbon support film (Kubalek et al., 1991). Samples were stained with nein for providing their actin:fimbrin crosslink coordinates, Edward Egelman 2% uranyl acetate, dried, and stabilized by a thin layer of evaporated carbon. for the fimbrin ABD2 coordinates, Greta Ouyang for the preparation of villin Data were collected on a Philips CM300-FEG. Six tomograms were recorded protein and the 2D F-actin:villin class average in ice, and David Burgess for at 24k magnification with 1.83 postmagnification at the CCD camera. The teaching us how to make villin. This research was supported by NIH grant sample was tilted using a 2 Saxton scheme (Saxton and Baumeister, 1982) U54 GM64346 to the Cell Migration Consortium. and recorded on a 2k 3 2k CCD camera (TVIPS; GMBH, Gauting, Germany). Received: June 13, 2008 Image Processing Revised: September 13, 2008 Tilt series were aligned using marker-free alignment (Winkler and Taylor, 2006). Accepted: September 30, 2008 Repeat coordinates centered on each actin crossover from each tomogram Published: December 9, 2008 were manually picked using the EMAN Boxer utility (Ludtke et al., 1999). Raw repeats (6566) were extracted as 64 3 144 3 40 voxel boxes. Images REFERENCES were then normalized by subtracting the mean and scaling the standard devi- ation to 1. Repeats from each tomogram were aligned in 3D to the two cross- Ashish, Paine, M.S., Perryman, P.B., Yang, L., Yin, H.L., and Krueger, J.K. linked actin filaments using in-house software. After the best alignments were (2007). Global structure changes associated with Ca2+ activation of full-length attained, as judged by classification of the actin filaments, the repeats from human plasma gelsolin. J. Biol. Chem. 282, 25884–25892. each tomogram were compiled into a single data set and aligned once more to the global average. Bazari, W.L., Matsudaira, P., Wallek, M., Smeal, T., Jakes, R., and Ahmed, Y. (1988). Villin sequence and peptide map identify six homologous domains. Classification Proc. Natl. Acad. Sci. USA 85, 4986–4990. Aligned subvolumes were classified using MDA with a 3D mask derived from Bretscher, A., and Weber, K. (1980). Villin is a major protein of the microvillus the variance map of the global average. Classifications were done using be- cytoskeleton which binds both G and F actin in a calcium-dependent manner. tween 8 and 32 factors, but ones made with the top 16 gave the best results. Cell 20, 839–847. The number of factors (eigenimages) used in the classification was determined Burgess, D.R., Broschat, K.O., and Hayden, J.M. (1987). Tropomyosin distin- by inspecting each factor and from a plot of the variance (eigenvalues) as guishes between the two actin-binding sites of villin and affects actin-binding a function of factor number. properties of other brush border proteins. J. Cell Biol. 104, 29–40.

Analysis of 2D Array Structure Burtnick, L.D., Koepf, E.K., Grimes, J., Jones, E.Y., Stuart, D.I., McLaughlin, Regions of homogeneous crosslinks were boxed from each tomogram, pro- P.J., and Robinson, R.C. (1997). The crystal structure of plasma gelsolin: impli- jected, and their Fourier transforms calculated after padding into a square cations for actin severing, capping, and nucleation. Cell 90, 661–670. array. The layer line and row line spacing was determined from the transforms Burtnick, L.D., Urosev, D., Irobi, E., Narayan, K., and Robinson, R.C. (2004). and the reciprocal lattice drawn. Vertical row lines were placed over the three Structure of the N-terminal half of gelsolin bound to actin: roles in severing, inner layer lines, 0, 1, and 2, and over the outer layer lines, 5, 6, and 7. When the apoptosis and FAF. EMBO J. 23, 2713–2722. two sets of row lines intersect, there is no rotation between neighboring Chen, J.Z., Furst, J., Chapman, M.S., and Grigorieff, N. (2003). Low-resolution filaments and when they do not, there is a rotation about the filament axis, structure refinement in electron microscopy. J. Struct. Biol. 144, 144–151. the amount of which can be determined, provided that the axial translation Choe, H., Burtnick, L.D., Mejillano, M., Yin, H.L., Robinson, R.C., and Choe, S. is known. This analysis also determined the helical symmetry. (2002). The calcium activation of gelsolin: insights from the 3 A˚ structure of the G4–G6/actin complex. J. Mol. Biol. 324, 691–702. Modeling The Protein Data Bank (PDB) coordinates of an actin monomer were used to Claessens, M.M., Bathe, M., Frey, E., and Bausch, A.R. (2006). Actin-binding generate an atomic model of F-actin with 13/6 helical symmetry (PDB ID proteins sensitively mediate F-actin bundle stiffness. Nat. Mater. 5, 748–753. code 1ATN). Homology models for chicken villin (Swissprot accession number DeRosier, D.J., and Tilney, L.G. (1982). How actin filaments pack into bundles. P02640) were created using SWISS-MODEL (Schwede et al., 2003) based on Cold Spring Harb. Symp. Quant. Biol. 46, 525–540.

DeRosier, D., Mandelkow, E., and Silliman, A. (1977). Structure of actin- Ludtke, S.J., Baldwin, P.R., and Chiu, W. (1999). EMAN: semiautomated soft- containing filaments from two types of non-muscle cells. J. Mol. Biol. 113, ware for high-resolution single-particle reconstructions. J. Struct. Biol. 128, 679–695. 82–97. Doering, D.S., and Matsudaira, P. (1996). Cysteine scanning mutagenesis at Markus, M.A., Nakayama, T., Matsudaira, P., and Wagner, G. (1994). Solution 40 of 76 positions in villin headpiece maps the F-actin binding site and struc- structure of villin 14T, a domain conserved among actin-severing proteins. tural features of the domain. Biochemistry 35, 12677–12685. Protein Sci. 3, 70–81. Egelman, E.H., Francis, N., and DeRosier, D.J. (1982). F-actin is a helix with Markus, M.A., Matsudaira, P., and Wagner, G. (1997). Refined structure of villin a random variable twist. Nature 298, 131–135. 14T and a detailed comparison with other actin-severing domains. Protein Sci. Ezzell, R.M., Chafel, M.M., and Matsudaira, P.T. (1989). Differential localization 6, 1197–1209. of villin and fimbrin during development of the mouse visceral endoderm and Matsudaira, P.T., and Burgess, D.R. (1979). Identification and organization of intestinal epithelium. Development 106, 407–419. the components in the isolated microvillus cytoskeleton. J. Cell Biol. 83, Fabbrizio, E., Bonet-Kerrache, A., Leger, J.J., and Mornet, D. (1993). Actin- 667–673. dystrophin interface. Biochemistry 32, 10457–10463. Matsudaira, P., Mandelkow, E., Renner, W., Hesterberg, L.K., and Weber, K. Feig, M., Karanicolas, J., and Brooks, C.L., III. (2004). MMTSB Tool Set: (1983). Role of fimbrin and villin in determining the interfilament distances of enhanced sampling and multiscale modeling methods for applications in actin bundles. Nature 301, 209–214. structural biology. J. Mol. Graph. Model. 22, 377–395. McGough, A., Way, M., and DeRosier, D. (1994). Determination of the a-acti- Ferrary, E., Cohen-Tannoudji, M., Pehau-Arnaudet, G., Lapillonne, A., Athman, nin-binding site on actin filaments by cryoelectron microscopy and image R., Ruiz, T., Boulouha, L., El Marjou, F., Doye, A., Fontaine, J.J., et al. (1999). In analysis. J. Cell Biol. 126, 433–443. vivo, villin is required for Ca(2+)-dependent F-actin disruption in intestinal McGough, A., Chiu, W., and Way, M. (1998). Determination of the gelsolin brush borders. J. Cell Biol. 146, 819–830. binding site on F-actin: implications for severing and capping. Biophys. J. Finidori, J., Friederich, E., Kwiatkowski, D.J., and Louvard, D. (1992). In vivo 74, 764–772. analysis of functional domains from villin and gelsolin. J. Cell Biol. 116, McKnight, C.J., Matsudaira, P.T., and Kim, P.S. (1997). NMR structure of the 1145–1155. 35-residue villin headpiece subdomain. Nat. Struct. Biol. 4, 180–184. Frank, B.S., Vardar, D., Chishti, A.H., and McKnight, C.J. (2004). The NMR McLaughlin, P.J., Gooch, J.T., Mannherz, H.G., and Weeds, A.G. (1993). structure of dematin headpiece reveals a dynamic loop that is conformation- Structure of gelsolin segment 1-actin complex and the mechanism of filament ally altered upon phosphorylation at a distal site. J. Biol. Chem. 279, 7909– severing. Nature 364, 685–692. 7916. Meng, J., Vardar, D., Wang, Y., Guo, H.C., Head, J.F., and McKnight, C.J. Friederich, E., Vancompernolle, K., Louvard, D., and Vandekerckhove, J. (2005). High-resolution crystal structures of villin headpiece and mutants (1999). Villin function in the organization of the actin cytoskeleton. Correlation with reduced F-actin binding activity. Biochemistry 44, 11963–11973. of in vivo effects to its biochemical activities in vitro. J. Biol. Chem. 274, 26751–26760. Mimura, N., and Asano, A. (1987). Further characterization of a conserved actin-binding 27-kDa fragment of actinogelin and a-actinins and mapping of Galkin, V.E., Orlova, A., Fattoum, A., Walsh, M.P., and Egelman, E.H. (2006). their binding sites on the actin molecule by chemical cross-linking. J. Biol. The CH-domain of calponin does not determine the modes of calponin binding Chem. 262, 4717–4723. to F-actin. J. Mol. Biol. 359, 478–485. Pardee, J.D., and Spudich, J.A. (1982). Purification of muscle actin. Methods Galkin, V.E., Orlova, A., Cherepanova, O., Lebart, M.C., and Egelman, E.H. Enzymol. 85(Pt B), 164–181. (2008). High-resolution cryo-EM structure of the F-actin-fimbrin/plastin ABD2 complex. Proc. Natl. Acad. Sci. USA 105, 1494–1498. Pettersen, E.F., Goddard, T.D., Huang, C.C., Couch, G.S., Greenblatt, D.M., Meng, E.C., and Ferrin, T.E. (2004). UCSF Chimera—a visualization system Glenney, J.R., Jr., Bretscher, A., and Weber, K. (1980). Calcium control of the for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612. intestinal microvillus cytoskeleton: its implications for the regulation of micro- filament organizations. Proc. Natl. Acad. Sci. USA 77, 6458–6462. Pinson, K.I., Dunbar, L., Samuelson, L., and Gumucio, D.L. (1998). Targeted Glenney, J.R., Jr., Geisler, N., Kaulfus, P., and Weber, K. (1981). Demonstra- disruption of the mouse villin gene does not impair the morphogenesis of tion of at least two different actin-binding sites in villin, a calcium-regulated microvilli. Dev. Dyn. 211, 109–121. modulator of F-actin organization. J. Biol. Chem. 256, 8156–8161. Pope, B., Way, M., Matsudaira, P.T., and Weeds, A. (1994). Characterisation of Hesterberg, L.K., and Weber, K. (1983a). Demonstration of three distinct the F-actin binding domains of villin: classification of F-actin binding proteins calcium-binding sites in villin, a modulator of actin assembly. J. Biol. Chem. into two groups according to their binding sites on actin. FEBS Lett. 338, 258, 365–369. 58–62. Hesterberg, L.K., and Weber, K. (1983b). Ligand-induced conformational Rana, A.P., Ruff, P., Maalouf, G.J., Speicher, D.W., and Chishti, A.H. (1993). changes in villin, a calcium-controlled actin-modulating protein. J. Biol. Cloning of human erythroid dematin reveals another member of the villin Chem. 258, 359–364. family. Proc. Natl. Acad. Sci. USA 90, 6651–6655. Kubalek, E.W., Kornberg, R.D., and Darst, S.A. (1991). Improved transfer of Saxton, W.O., and Baumeister, W. (1982). The correlation averaging of a two-dimensional crystals from the air/water interface to specimen support regularly arranged bacterial cell envelope protein. J. Microsc. 127, 127–138. grids for high-resolution analysis by electron microscopy. Ultramicroscopy Schwede, T., Kopp, J., Guex, N., and Peitsch, M.C. (2003). SWISS-MODEL: an 35, 295–304. automated protein homology-modeling server. Nucleic Acids Res. 31, 3381– Kumar, N., and Khurana, S. (2004). Identification of a functional switch for actin 3385. severing by cytoskeletal proteins. J. Biol. Chem. 279, 24915–24918. Smirnov, S.L., Isern, N.G., Jiang, Z.G., Hoyt, D.W., and McKnight, C.J. (2007). Laskowski, R.A., MacArthur, M.W., Moss, D.S., and Thornton, J.M. (1993). The isolated sixth gelsolin repeat and headpiece domain of villin bundle F-actin PROCHECK: a program to check the stereochemical quality of protein struc- in the presence of calcium and are linked by a 40-residue unstructured tures. J. Appl. Crystallogr. 26, 283–291. sequence. Biochemistry 46, 7488–7496. Lebart, M.C., Mejean, C., Boyer, M., Roustan, C., and Benyamin, Y. (1990). Sukow, C., and DeRosier, D. (1998). How to analyze electron micrographs of Localization of a new a-actinin binding site in the COOH-terminal part of actin rafts of actin filaments crosslinked by actin-binding proteins. J. Mol. Biol. sequence. Biochem. Biophys. Res. Commun. 173, 120–126. 284, 1039–1050. Lebart, M.C., Mejean, C., Roustan, C., and Benyamin, Y. (1993). Further Taylor, K.A., and Taylor, D.W. (1994). Formation of two-dimensional com- characterization of the a-actinin-actin interface and comparison with filamin- plexes of F-actin and crosslinking proteins on lipid monolayers: demonstration binding sites on actin. J. Biol. Chem. 268, 5642–5648. of unipolar a-actinin-F-actin crosslinking. Biophys. J. 67, 1976–1983.

Tilney, L.G., and Mooseker, M. (1971). Actin in the brush-border of epithelial Volkmann, N., DeRosier, D., Matsudaira, P., and Hanein, D. (2001). An atomic cells of the chicken intestine. Proc. Natl. Acad. Sci. USA 68, 2611–2615. model of actin filaments cross-linked by fimbrin and its implications for bundle Tilney, L.G., Derosier, D.J., and Mulroy, M.J. (1980). The organization of actin assembly and function. J. Cell Biol. 153, 947–956. filaments in the stereocilia of cochlear hair cells. J. Cell Biol. 86, 244–259. Way, M., Pope, B., and Weeds, A.G. (1992). Evidence for functional homology Tilney, L.G., Egelman, E.H., DeRosier, D.J., and Saunder, J.C. (1983). Actin in the F-actin binding domains of gelsolin and a-actinin: implications for the filaments, stereocilia, and hair cells of the bird cochlea. II. Packing of actin requirements of severing and capping. J. Cell Biol. 119, 835–842. filaments in the stereocilia and in the cuticular plate and what happens to Winkler, H., and Taylor, K.A. (1999). Multivariate statistical analysis of three- the organization when the stereocilia are bent. J. Cell Biol. 96, 822–834. dimensional cross-bridge motifs in insect flight muscle. Ultramicroscopy 77, Vardar, D., Buckley, D.A., Frank, B.S., and McKnight, C.J. (1999). NMR 141–152. structure of an F-actin-binding ‘‘headpiece’’ motif from villin. J. Mol. Biol. 294, 1299–1310. Winkler, H., and Taylor, K.A. (2006). Accurate marker-free alignment with simultaneous geometry determination and reconstruction of tilt series in Vermeulen, W., Vanhaesebrouck, P., Van Troys, M., Verschueren, M., Fant, F., electron tomography. Ultramicroscopy 106, 240–254. Goethals, M., Ampe, C., Martins, J.C., and Borremans, F.A. (2004). Solution structures of the C-terminal headpiece subdomains of human villin and advil- Yin, H.L., and Stossel, T.P. (1979). Control of cytoplasmic actin gel-sol trans- lin, evaluation of headpiece F-actin-binding requirements. Protein Sci. 13, formation by gelsolin, a calcium-dependent regulatory protein. Nature 281, 1276–1287. 583–586.