3-Dimensional organization of the N-terminalvinculin head fragment

Jörg Winkler1), Brigitte M. JockuschCell Biology, Zoological Institute, Technical University of Braunschweig, Braunschweig/Germany

Received November 20, 2000Accepted November 21, 2000

Cytoskeleton ± vinculin ± electron microscopy ± electronspectroscopic imaging

The cytoskeleton-associated protein vinculin is composed of aglobular head and an elongated tail domain. The protein can becleaved by V8 protease treatment into two fragments withapparent molecular masses of 90 and 29/27 kDa, respectively.So far, no high-resolution data on the tertiary structure of the N-terminal 90-kDa fragment are available. We analyzed the 90-kDa fragment in detail, using electron spectroscopic imaging inconjunction with modelling experiments. The front viewprojection of this fragment appears roughly rhomboidal, with4 intensity maxima arranged at the vertices and a stain-filledregion in the center. Based on a detailed examination ofdifferent particle projections, a 3-dimensional model wasconstructed which appears as a flattened tetrahedron. Acomparison of the 90-kDa fragment with the intact proteinallows for a correlation between the subdomain organization ofthe vinculin head and the biochemically defined V8 proteasecleavage sites (aa 851 and 857).

Abbreviations. ESI Electron spectroscopic imaging. ± EFTEM Energy-filtered transmission electron microscopy. ± VASP Vasodilator-stimulatedphosphoprotein.

Introduction

Vinculin is a highly conserved 117-kDa structural proteinassociated with the cytoplasmic face of cell-cell and cell-matrixjunctions, where it contributes to the attachment of actinfilaments to the plasma membrane (Geiger, 1979; Otto, 1990;Jockusch et al., 1995). Within this junctional complex, vinculininteracts with a number of other proteins. The N-terminal 90-kDa head fragment contains binding sites for talin (Jones et al.,1989; Gilmore et al., 1992, Bass et al., 1999), a-actinin (Kroem-ker et al., 1994; McGregor et al., 1994) and a-catenin (Weiss

et al., 1998), whereas the 29/27-kDa C-terminal tail binds actin(Menkel et al., 1994; Hüttelmaier et al., 1997), paxillin (Woodet al., 1994) and acidic phospholipids (Isenberg, 1991; Fukamiet al., 1994; Johnson and Craig, 1995). Head and tail domainsare connected by a proline-rich region with inherent bindingsites for vasodilator-stimulated phosphoprotein (Brindle et al.,1996; Reinhard et al., 1996), ponsin (Mandai et al., 1999) andvinexin (Kioka et al., 1999). From several lines of evidence, ithas been proposed that the ligand-binding capability andassembly of vinculin into the adherens junctions is regulated bythe controlled exposure of discrete binding sites. This modelassumes a high degree of intramolecular flexibility between thehead and the tail. Previous data showing electron microscopicalimages of high resolution corroborates this assumption, pro-viding visual evidence for a highly flexible vinculin molecule, inwhich both, the head and tail domains are composed of globularsubdomains (Winkler et al., 1996). Crystal data revealed thatthe vinculin tail (aa 879 ± 1066) is composed of 5 amphipathichelices (Bakolitsa et al., 1999) which corresponds with struc-tural features formerly described by TEM (Winkler et al.,1996). However, detailed informations of the tertiary structureof the head fragment are missing, and its definite identificationwithin intact vinculin remains unknown. Hence, a detailedanalysis of the structural organization of the 90-kDa fragment,in particular of the topography of individual subdomains,should be valuable to understand the mechanism of ligandbinding. Here, we report on the fine structure of the isolated 90-kDa fragment. Images of high resolution and contrast wereobtained using energy-filtered transmission electron micros-copy (EFTEM). We show that the fragment, consisting of awell-ordered arrangement of subdomains, is organized in atertiary structure which is best described as a flattenedtetrahedron. These data allows for a definite correlationbetween the 90-kDa fragment with its inherent ligand-bindingsites and the intact vinculin polypeptide.

Materials and methods

Protein purificationTurkey gizzard vinculin was purified to homogeneity according to themethod of Feramisco and Burridge (1980), using low-ionic-strength

EJCB 201European Journal of Cell Biology 80, 201 ± 206 (2001, March) ´ � Urban & Fischer Verlag ´ Jenahttp://www.urbanfischer.de/journals/ejcb

0171-9335/01/80/03-201 $15.00/0

1) Dr. Jörg Winkler, Medizinische Universität zu Lübeck, Labor fürexperimentelle Ophthalmologie, Ratzeburger Allee 160, D-23538Lübeck/Germany, e-mail: jorgw@yahoo.com, FAX: � 494515004952.

extraction of minced tissue and ion exchange chromatography onDEAE cellulose (Whatman DE52, England). The 90-kDa N-terminalfragment was obtained by digestion with immobilized V8-protease

(Boehringer, Mannheim, Germany) as described by Johnson and Craig(1994). The digest was loaded onto a Q-Sepharose anion exchangercolumn (Pharmacia, Uppsala, Sweden) equilibrated with buffer B(20 mM Tris/HCl, pH 7.5, 20 mM NaCl, 0.1 mM EGTA, 0.2 mM DTE).Bound fragments were eluted with a NaCl gradient (20 to 300 mM) anddialyzed against buffer B. Purity of the 90-kDa vinculin fragment wasmonitored by SDS-PAGE (Laemmli, 1970) (Fig. 1). Protein concentrationwas determined according to Bradford (1976), using BSA as a standard.

Electron microscopyProtein samples used for electron microscopy were diluted with buffer Bto a protein concentration of 10 mg/ml. Negative staining of the proteinwas performed using the diffusion technique described by Valentineet al. (1968) and an aqueous solution of 4% (w/v) uranyl acetate. Thesamples were analyzed in a Carl Zeiss EM 902 equipped with an electronenergy-loss spectrometer. Each micrograph of a given area of particleswas imaged twice; (a) in the elastic bright-field, at an energy loss of 0 eV,and (b) in the ESI mode, using an energy loss of 115 eV, which is close tothe absorption edge of uranium (Bauer, 1988). Exposures were carriedout at � 50000 primary magnification with an energy resolution of20 eV and a 30-mm objective aperture.

Modelling experimentsThe structural organization of the 90-kDa fragment was modeled inplasticine. The model was formed according to the projections of

202 J. Winkler, B. M. Jockusch EJCB

Fig. 1. Coomassie Blue-stained profiles of purified turkey smoothmuscle vinculin with a molecular mass of 117 kDa (lane a), and its 90-kDa fragment generated by V8-protease cleavage (lane b), after SDS-PAGE on 12% gels.

Fig. 2. Field views of negatively stained 90-kDa vinculin fragments.(a) Elastic bright-field view of the sample taken at DE� 0 eV and (b)corresponding inelastic dark-field image taken at DE� 115 eV. Iden-tical molecules of both views are encircled. The heterogeneous shape ofthe molecules was classified into different groups designated by the

letters A ± D. The majority of particles appeared either as rhombo-idal (A) or triangular-shaped molecules (B). Particle projections withthree or two intensity maxima associated with one or two barely visiblecenters of mass are shown in (C). A minor quantity of particlesappeared as elongated fragments of undefined shape (D). Bar� 10 nm.

negatively stained fragments observed on electron micrographs. Forsimulation of staining effects, the model was immersed in a solution ofblack ink at different angular positions and staining depths (Birkenhä-ger et al., 1995).

Results and discussion

Ultrastructure of the 90-kDa head fragmentA survey electron micrograph of negatively stained 90-kDafragment is shown in Fig. 2. Images of the purified headfragment were taken in the elastic bright-field (Fig. 2a) and inthe inelastic dark-field mode, which significantly enhanced thecontrast ratio between the uranium salt layer and the proteinmolecules (Fig. 2b). On a background of ill-defined particles,projections with predominantly three to four intensity maximaare discernible, which are arranged in rhomboidal or triangularstructures (Fig. 2, encircled). Individual particle projections,obtained from images as depicted in Fig. 2b, were groupedaccording to their geometric appearance as shown in Figs. 3a ±d. Depending on the orientation of the molecules duringadsorption onto the carbon support film, and their individualstaining environment, the shape of the particles varied sub-stantially. Fig. 3a shows a gallery of fragments absorbed ontothe carbon film in such a way that front views resulted. Theoverall morphology of the particles is rhomboidal, with fourcenters of mass arranged at the vertices of the structures(arrows). The center of the particles appears light in the ESIdark-field mode, indicating a depression or a hole. The intensitymaxima often appeared merged with each other, rather thandisplaying four clearly separated centers of mass. Measure-ments of rhomboidal particle projections revealed an outerdiameter of 9.3 nm� 0.7 nm (n� 40). This value deviates byapproximately 10% from the values obtained for head domainsof intact vinculin (Winkler et al., 1996).

To better understand the spatial relationship between thefour intensity maxima, revealed within the isolated 90-kDafragment, we studied the morphology of particles projected indifferent orientations and under different staining conditions.In the gallery shown in Fig. 3b, the staining environmentchanges gradually from deep to shallower stain (Figs. 3b1±5).Under deep stain conditions, indicated by a bright background(Figs. 3b1,2), three intensity maxima of each fragment appearedcloser together displaying a ªY-shapedº structure (arrows). Thefourth center of mass was positioned between the two arms of

203Ultrastructure of the vinculin headEJCB

Fig. 3. Gallery of particle projections from a sample of purified 90-kDa fragments imaged in the uranium dark-field mode at DE� 115 eV.(a) Front-view projections of the fragment, with four intensity maxima,arranged in the vertices of a rhomboid (arrows) and with a protein-deficient central part. (b) Particles, projected in a staining environmentwhich changes from b1±5 from deep to shallow stain. The 2-dimensionalinterpretation of the subdomain arrangement is shown in a schematicdrawing (b6). (c, d) Particle projections which were interpreted to haveabsorbed in slightly different orientations onto the carbon support film.The molecules appear roughly as ªY-shapedº structures, associatedwith a fourth center of mass (arrow), which appears in (c1) less brightand smaller in diameter, in (c2) is barely visible, and finally disappearsin (c3). The particles shown in (c4) appear as three individual centers ofmass, oriented in a triangular arrangement (arrowheads). The rhom-boidal particles shown in (d) are composed of two intensity maximaoriented at the horizontal axis (arrowheads) and two barely visiblecenters of mass arranged at the vertical axis (arrows) of the molecules.Bar� 10 nm.

the ªYº (arrowhead). The particles shown in Figs. 3b3±5,projected under intermediate staining conditions revealedbasically the same structural arrangement as described inFigs. 3b1,2. The schematic drawing shown in Fig. 3b6 representsthe 2-dimensional interpretation of the particle structure asprojected in Figs. 3b1±5. In general, deep and intermediate stainconditions are more suitable to prevent artificial distortion andflattening of delicate structures than shallow staining; thus, weconclude that the particle projections shown in Figs. 3b1±5

represent molecules with a preserved natural morphology.The projections shown in Figs. 3c1 ± c4 display fragments

which were apparently absorbed under slightly differentorientations onto the carbon support film. In these images the3 subdomains comprising the ªYº appear of similar size, whilethe fourth center of mass (arrows) appears distinctly smaller(Fig. 3c1), is sometimes barely visible (Fig. 3c2) or not at alldiscernible (Figs. 3c3/4). This variable appearance of onesubdomain, as opposed to that of the three others, probablycorresponds to differences in stain immersion. Both, the regularprojection of the ªYº and the variable images of the fourthmass, argue against a planar rhomboidal particle morphologyand favour an overall configuration of a tetrahedron-likestructure. This structural interpretation is further supported byimages depicted in Fig. 3d. These particles are dominated bytwo larger circular intensity maxima (arrowheads) whichalternate with two smaller ones (arrows). The seeminglysmaller subunits might represent a different degree of immer-sion of these centers of mass in the negative stain due to theirlocation in different planes, thus, favouring again the image of anon-planar configuration of the fragment. Considering allparticle projections, the morphology of the isolated 90-kDahead fragment is best described as a distorted, flattenedtetrahedron, composed of 4 centers of mass arranged in twodifferent planes oriented orthogonal to each other.

Modelling studiesOn the basis of different projections of the isolated 90-kDavinculin fragment, as shown in Fig. 3, a plasticine model wasformed. As shown in Fig. 4a ± f, different orientations of thismodel confirm the conclusions drawn from the electronmicroscopic images. A combination of different projections of

204 J. Winkler, B. M. Jockusch EJCB

Fig. 4. 3-Dimensional model of the 90-kDa fragment. Both face-onviews of the model shown in (a) and (d) appear roughly rhomboidal.The two globular masses positioned at the horizontal axis and the twoglobules of the vertical axis are arranged in two different planes,oriented orthogonal to each other (c). The projections shown in (e) and

(f) were tilted in a way that in each case three globular masses arearranged in one plane revealing either a ªYº- (d) or ªvº-shaped (e)morphology, whereas the fourth globule is projected backwards,reaching into a deeper plane.

Fig. 5. Simulation of different molecule projections. The modelshown in Fig. 4 was immersed into a solution of dark ink, simulatingdifferent staining depths and rotational views of the molecules. Theprojections shown in (a1) and (b1) represent either shallow (ss) or deepstain (ds) conditions of the model, which is oriented as shown in Fig. 4d.The images shown in (b1/2) and (b3/4) display rotational views of themodel after tilting in angles of 108 or 208 in either direction around itshorizontal axis. Under deep-stain conditions the model projectionsshown in (b3) and (b4) revealed a reduced appearance composed of 3individual masses, oriented in a trigonal arrangement (c).

this model with different staining conditions, emphasizes thiscorrelation further. As seen in Fig. 5, tilting the model todifferent degrees in an ink solution of variable depth yieldsstrikingly different images. Starting with a rotational view asshown in Fig. 4d, the model was immersed in two differentdepths of ink, which simulate shallow (Fig. 5a1) or deep stain(Fig. 5a2) conditions. Immersed in a shallow layer of ink, themodel appears rhomboidal, which corresponds to projectionsof negatively stained fragments seen in Fig. 3a. Surrounded by adeeper layer of ink the model projected in an irregularmorphology, which is dominated by two larger globular masses,alternating with two smaller ones situated at the horizontal axisof the model (Fig. 5a2). The corresponding particle shape isshown in Fig. 3d. Figures 5b1 and 5b2 represent projections seenafter tilting the model in angles of 108 in both directions aroundits horizontal axis, i. e. towards the back (Fig. 5b1) or the front(Fig. 5b2), respectively. Three equally sized centers of mass areoriented in ªYº-shaped arrangements, in both cases associatedwith a fourth distinctly smaller mass. Such model morphologiessimulate projections of vinculin fragments shown in Figs. 3c1/2.Increasing the tilt angel to 208 in either direction (Figs. 5b3, b4)simulates molecule projections as shown in the electronmicrographs of Fig. 3c3. When more ink is added to sucharrangements as seen in Figs. 5b3 and 5b4, to simulate deepstain, both arrangements are reduced to three individualcenters of mass (Fig. 5c). A similar appearance of particleprojections was observed on microscopic images as shown inFig. 3c4. The conformity between the 2-dimensional particleprojections obtained by EFTEM with the 2-dimensionaltopography of the model, thus, support the proposed 3-dimensional interpretation of subdomain arrangement (Fig. 4).

Correlation with intact vinculinIn previous studies, the rotary-shadowed 90-kDa fragment ofsmooth muscle vinculin displayed a, roughly, circular shape(Milam, 1985), corresponding with the globular head domain ofintact vinculin (Molony and Burridge, 1985). Despite of visualinformation of low resolution, the 90-kDa fragment wasdescribed, a priori, as being the globular head component ofintact vinculin. Electron microscopic images of high resolutionrevealed a trilobar organization of the vinculin head, associatedwith further globular subdomains (Winkler et al., 1996). How-ever, since high-resolution images of the isolated 90-kDafragment were missing, a definitive structural correlation withthe head domain of intact vinculin was not possible. Here, wecould show that the head fragment actually contains fourcircular centers of mass, arranged as a distorted, flattenedtetrahedron. These data may be helpful to further analyse thepreliminary crystal data obtained for intact vinculin (Koganet al., 2000) and may contribute, as well, to a better under-standing of ligand binding regulated by intramolecular head-tail interaction. On the basis of former studies and consideringthe data presented here, we are able to correlate roughly theschematic organization of vinculin with the amino acidsequences of biochemically-defined ligand-binding sites(Fig. 6).

Acknowledgements. The expert technical assistance of Eva Saxinger(TU Braunschweig) and technical support of Dr. H. Lünsdorf (GBF,Braunschweig) is gratefully acknowledged. This work was financiallysupported by the Deutsche Forschungsgemeinschaft.

References

Bakolitsa, C., de Pereda, J. M., Bagshaw, C. R., Critchley, D. R.,Liddington, R. C. (1999): Crystal structure of the vinculin tailsuggests a pathway for activation. Cell 99, 603 ± 613.

Bass, M. D., Smith, B. J., Prigent, S. A., Critchley, D. R. (1999): Talincontains three similar vinculin-binding sites predicted to form anamphipathic helix. Biochem. J. 341, 257 ± 263.

Bauer, J. (1988): Electron Microscopy in Microbiology. In: F. Mayer(ed.): Methods Microbiol. Academic Press, London, New York, Vol.20, pp. 113 ± 146.

Birkenhäger, R., Hoppert, M., Deckers-Hebestreit, G., Mayer, F.,Altendorf, K. (1995): The F0 complex of the Escherichia coli ATPsynthase. Investigation by electron spectroscopic imaging andimmunoelectron microscopy. Eur. J. Biochem. 230, 58 ± 67.

Bradford, M. M. (1976): A rapid and sensitive method for thequantification of microgram quantities of protein utilizing theprinciple of protein-dye binding. Anal. Biochem. 72, 248 ± 254.

Brindle, N. P. J., Holt, M. R., Davies, J. E., Priece, C. J., Critchley, D. R.(1996): The focal-adhesion vasodilator-stimulated phosphoprotein(VASP) binds to the proline-rich domain in vinculin. Biochem. J. 318,753 ± 757.

Feramisco, J. R., Burridge, K. (1980): A rapid purification of a-actinin,and a 130000-dalton protein from smooth muscle. J. Biol. Chem. 255,1194 ± 1199.

Fukami, K., Endo, T., Imamura, M., Takenawa, T. (1994): a-Actinin andvinculin are PIP2-binding proteins involved in signaling by tyrosinekinase. J. Biol. Chem. 269, 1518 ± 1522.

Geiger, B. (1979): A 130 kDa protein from chicken gizzard: itslocalization at the termini of microfilament bundles in culturedchicken cells. Cell 18, 193 ± 205.

Gilmore, A. P., Jackson, P., Waites, G. T., Critchley, D. R. (1992): Furthercharacterisation of the talin-binding site in the cytoskeletal proteinvinculin. J. Cell Sci. 103, 719 ± 731.

205Ultrastructure of the vinculin headEJCB

Fig. 6. Correlation between the primary and the schematic tertiarystructure of vinculin and its protease derived fragments with the knownligand-binding sites. The 1066-amino-acid vinculin polypeptide iscomposed of a 90-kDa N-terminal head (aa 1 ± 850) and a 29/27 kDaC-terminal tail (aa 858 ± 1066), which is composed of 5 amphipathichelices (Bakolitsa et al., 1999). Both domains are connected by aproline-rich hinge region (aa 837 ± 878), which contains two V8protease cleavage sites (aa 851/aa 857).

Hazan, R. B., Kang, L., Roe, S., Borgen, P. I., Rimm, D. L. (1997):Vinculin is associated with the E-cadherin adhesion complex. J. Biol.Chem. 272, 32448 ± 32453.

Hüttelmaier, S., Bubeck, P., Rüdiger, M., Jockusch, B. M. (1997):Characterization of two F-actin-binding and oligomerization sites inthe cell-contact protein vinculin. Eur. J. Biochem. 247, 1136 ± 1142.

Isenberg, G. (1991): Actin-binding proteins-lipid interactions. J. MuscleRes. Cell Motil. 12, 136 ± 144.

Jockusch, B. M., Bubeck, P., Giehl, K., Kroemker, M., Moschner, J.,Rothkegel, M., Rüdiger, M., Schlüter, K., Stanke, G., Winkler, J.(1995): The molecular architecture of focal adhesions. Annu. Rev.Cell Biol. 11, 379 ± 416.

Johnson, R. P., Craig, S. W. (1994): An intramolecular associationbetween the head and tail domains of vinculin modulates talinbinding. J. Biol. Chem. 269, 12611 ± 12619.

Johnson, R. P., Craig, S. W. (1995): The carboxy-terminal tail domain ofvinculin contains a cryptic binding site for acidic phospholipids.Biochem. Biophys. Res. Commun. 210, 159 ± 164.

Jones, P., Jackson, P., Price, G. J., Patel, B., Ohanion, V., Lear, A. L.,Critchley, D. R. (1989): Identifiction of a talin binding site in thecytoskeletal protein vinculin. J. Cell Biol. 109, 2917 ± 2927.

Kioka, N., Sakata, S., Kawauchi, T., Amachi, T., Akiyama, S. K.,Okazaki, K., Yaen, C., Yamada, K. M., Aota, S. (1999): Vinexin: anovel vinculin-binding protein with multiple SH3 domains enhancesactin cytoskeletal organisation. J. Cell Biol. 144, 59 ± 69.

Kogan, O., Yarden, A., Gimona, M., Geiger, B., Safro, M. (2000):Prelimary crystallographic study of turkey gizzard vinculin. ActaCrystallogr. D Biol. Crystallogr. 56, 1055 ± 1057.

Kroemker, M., Rüdiger, A. H., Jockusch, B. M., Rüdiger, M. (1994):Intramolecular interactions in vinculin control a-actinin binding tothe vinculin head. FEBS Lett. 355, 259 ± 262.

Laemmli, U. K. (1970): Cleavage of structural proteins during theassembly of the head of bacteriophages T4. Nature 227, 680 ± 685.

Mandai, K., Nakanishi, H., Satoh, A., Takahashi, K., Satoh, K.,Nishioka, H., Mizoguchi. A., Takai, Y. (1999): Ponsin/SH3P12: An1-afadin- and vinculin-binding protein localized at cell-cell and cell-matrix adherens junctions. J. Cell Biol. 144, 1001 ± 1017.

McGregor, A., Blanchard, A. D., Rowe, A. J., Critchley, D. R. (1994):Identification of the vinculin-binding site in the cytoskeletal proteina-actinin. Biochem. J. 301, 225 ± 233.

Menkel, A. R., Kroemker, M., Bubeck, P., Ronsiek, M., Nikolai. G.,Jockusch, B. M. (1994): Characterization of an F-actin-bindingdomain in the cytoskeletal protein vinculin. J. Cell Biol. 126, 1231 ±1240.

Milam, L. M. (1985): Electron microscopy of rotary shadowed vinculinand vinculin complexes. J. Mol. Biol. 184, 543 ± 545.

Molony, L., Burridge, K. (1985): Molecular shape and self-association ofvinculin and metavinculin. J. Cell. Biochem. 29, 31 ± 36.

Otto, J. J. (1990): Vinculin. Cell Motil. Cytoskeleton 16, 1 ± 6.Reinhard, M., Rüdiger, M., Jockusch, B. M., Walter, U. (1996): VASP

interaction with vinculin: a recurring theme of interactions withproline-rich motifs. FEBS Lett. 399, 103 ± 107.

Valentine, R. C., Shapiro, B. M., Stadtman, E. R. (1968): Regulation ofglutamine synthetase. XII. Electron microscopy of the enzyme fromEscherichia coli. Biochemistry 7, 2143 ± 2152.

Weiss, E. E., Kroemker, M., Rüdiger, A. H., Jockusch, B. M., Rüdiger,M. (1998): Vinculin is part of the cadherin catenin junctionalcomplex: formation between alpha-catenin and vinculin. J. CellBiol. 141, 755 ± 764.

Winkler, J., Lünsdorf, H., Jockusch, B. M. (1996): The ultrastructure ofchicken gizzard vinculin as visualized by high-resolution electronmicroscopy. J. Struct. Biol. 116, 270 ± 277.

Wood, C. K., Turner, C. E., Jackson, P., Critchley, D. R. (1994):Characterisation of the paxillin-binding site and the c-terminal focaladhesion targeting sequence in vinculin. J. Cell Sci. 107, 709 ± 717.

206 J. Winkler, B. M. Jockusch EJCB