A lipid-regulated docking site on vinculin for protein kinase C.

During cell spreading, binding of actin-organizing proteins to acidic phospholipids and phosphorylation are important for localization and activity of these proteins at nascent cell-matrix adhesion sites. Here, we report on a transient interaction between the lipid-dependent protein kinase Calpha and vinculin, an early component of these sites, during spreading of HeLa cells on collagen. In vitro binding of protein kinase Calpha to vinculin tail was found dependent on free calcium and acidic phospholipids but independent of a functional kinase domain. The interaction was enhanced by conditions that favor the oligomerization of vinculin. Phosphorylation by protein kinase Calpha reached 1.5 mol of phosphate/mol of vinculin tail and required the C-terminal hydrophobic hairpin, a putative phosphatidylinositol 4,5-bisphosphate-binding site. Mass spectroscopy of peptides derived from in vitro phosphorylated vinculin tail identified phosphorylation of serines 1033 and 1045. Inhibition of C-terminal phospholipid binding at the vinculin tail by mutagenesis or deletion reduced the rate of phosphorylation to < or =50%. We suggest a possible mechanism whereby phospholipid-regulated conformational changes in vinculin may lead to exposure of a docking site for protein kinase Calpha and subsequent phosphorylation of vinculin and/or vinculin interaction partners, thereby affecting the formation of cell adhesion complexes.

During cell spreading, binding of actin-organizing proteins to acidic phospholipids and phosphorylation are important for localization and activity of these proteins at nascent cell-matrix adhesion sites. Here, we report on a transient interaction between the lipid-dependent protein kinase C␣ and vinculin, an early component of these sites, during spreading of HeLa cells on collagen. In vitro binding of protein kinase C␣ to vinculin tail was found dependent on free calcium and acidic phospholipids but independent of a functional kinase domain. The interaction was enhanced by conditions that favor the oligomerization of vinculin. Phosphorylation by protein kinase C␣ reached 1.5 mol of phosphate/mol of vinculin tail and required the C-terminal hydrophobic hairpin, a putative phosphatidylinositol 4,5-bisphosphate-binding site. Mass spectroscopy of peptides derived from in vitro phosphorylated vinculin tail identified phosphorylation of serines 1033 and 1045. Inhibition of C-terminal phospholipid binding at the vinculin tail by mutagenesis or deletion reduced the rate of phosphorylation to <50%. We suggest a possible mechanism whereby phospholipid-regulated conformational changes in vinculin may lead to exposure of a docking site for protein kinase C␣ and subsequent phosphorylation of vinculin and/or vinculin interaction partners, thereby affecting the formation of cell adhesion complexes.
Cell spreading is a complex process involving receptor-mediated adhesion, membrane protrusion, and the formation of discrete cell-matrix adhesion sites linked to a reorganization of the actin cytoskeleton (1)(2)(3). The modulation of proteins by protein kinase C (PKC) 1 -dependent phosphorylation and by phosphoinositide binding is critically involved in the regulation of these phenomena (4 -7). Both processes are part of signaling cascades but are not independent of each other and participate in other signaling pathways. PI4,5P 2 , the major phosphoinositide involved, is not only a precursor of the second messengers diacylglycerol, inositol 3,4,5-trisphosphate, and phosphatidylinositol 3,4,5-trisphosphate but is also critically involved in the organization of the actin cytoskeleton at the plasma membrane (8 -10). Its local concentration and accessibility control multiple aspects of the cytoskeleton-plasma membrane linkage via ERM proteins (11), assembly and disassembly of actin filaments (12)(13)(14), and of protein complexes bundling these filaments and connecting them to transmembrane receptors. Because of the heterogeneity and complexity of their protein composition, structure and composition of these complexes are only partially understood (2). Essential components include vinculin, talin, ␣-actinin, and filamin (15), which have all been proposed to be regulated in their conformation and ligand binding activity by PI4,5P 2 (16 -19).
Vinculin, an early and essential component of nascent cellmatrix adhesions (20), has been shown to associate with talin, ␣-actinin, paxillin, and members of the vasodilator-stimulated phosphoprotein/Ena and ponsin/ArgBP52/vinexin families (15). The function of vinculin in cell adhesions is not fully understood. By bringing multiple actin-organizing proteins in close proximity, vinculin may act as an adaptor protein in addition to its structural role in the binding and cross-linking of membrane-apposed actin filaments. The atomic structure of the entire vinculin molecule is still not known, but the overall structure as revealed by electron microscopy shows a tadpolelike molecule whose tail can be folded and attached to the more globular head portion (21). The vinculin tail, whose atomic structure has been resolved (22), consists of a bundle of five helices (H1-H5) in antiparallel orientation, with N-and Cterminal extensions emerging from the same side of the bundle. It contains two interaction sites for acidic phospholipids, one located at the surface of helices H2-H3 (23), and a second one comprising a C-terminal hydrophobic hairpin (22). In the closed vinculin conformation that is supposed to reflect the cytosolic moiety, binding sites for talin and ␣-actinin at the vinculin head, for F-actin at the vinculin tail, and for vasodilator-stimulated phosphoprotein and others at the proline-rich hinge between head and tail are not accessible (20, 24 -26). Both PI4,5P 2 binding to and phosphorylation of the tail are proposed to alter vinculin conformation and, thus, expose protein ligandbinding sites (15,16,27,28). Hence, PI4,5P 2 -regulated protein kinases seem good candidates to execute this change in vinculin conformation and activity, and PKC has been shown in vitro to accept the vinculin tail as a substrate (27,29).
PKC is a lipid-regulated serine/threonine kinase (30, 31) whose functional involvement in cell adhesion and spreading has been documented by pharmacological studies in numerous cellular systems (4,5,32,33). PKC isotypes ␣, ␦, and ⑀ have been implicated in the formation and maintenance of cellmatrix adhesion sites (34 -37). There are mainly three (classes of) proteins that have been identified as PKC binding partners in this location; the receptors for activated protein kinase C (RACKs), syndecan 4, and the transmembrane-4 superfamily (tetraspanins) (38 -40). All three bind to integrin ␤-chains, and the latter two are transmembrane proteins that function as integrin co-receptors (41,42). Binding partners and substrates of PKC directly involved in the actin organization of cell adhesion complexes, however, are less well defined (43). Although talin, vinculin, and filamin are considered potential PKC in vivo substrates (27,44,45), their role in PKC-dependent cell adhesion processes is still not clear.
In this report, we present evidence for a physical interaction of PKC␣ with vinculin during cell spreading. The docking of the enzyme depends on the C-terminal lipid-binding site of the vinculin tail, and PI4,5P 2 binding capacity correlates with phosphorylation of adjacent serine residues. These results suggest a possible mechanism whereby PI4,5P 2 -induced conformational changes in vinculin may result in PKC␣ binding and subsequent phosphorylation of vinculin and/or vinculin interaction partners as required for the formation of cell-matrix adhesion complexes.

EXPERIMENTAL PROCEDURES
Cloning, Mutagenesis, and Graphic Representation-Constructs used in this study were amplified according to standard PCR protocols, cloned in pCR-blunt (Invitrogen), and verified by DNA sequencing. Bovine PKC␣ cDNA (a gift of Dr. Peter J. Parker, Imperial Cancer Research Fund, London) was equipped either with an N-terminal His tag or a birch-profilin epitope tag (BiPro tag (46)) and cloned into pcDNA3 (Invitrogen). An ATP binding-deficient K368M mutant (47) was generated by site-directed mutagenesis according to the manufacturer's instructions using the QuikChange kit (Stratagene). Vinculin fragments encoding the tail domain (amino acids 858 -1066) were equipped with an N-terminal FLAG tag and cloned into pQE-30 (Qiagen) that contains a His tag. To remove potential phosphorylation sites at Ser-941/Thr-43, Ser-999, Thr-1050/Thr-55, and Thr-1062 single or double alanine mutants of Vt-(858 -1066) were generated, as described above. For the analysis of the lipid dependence of the PKC-vinculin interaction, mutants Vt-(R1060Q/K61Q), Vt-(T1062E), and the deletion fragment Vt-(858 -1052) were generated. Graphic views of Vt-(879 -1061) based on PDB entry 1QRK (22) were generated using SwissPDB viewer ((48) www.expasy.ch/spdbv) and POVray 3.1 (www.povray.org) on a Linux platform.
Protein Precipitation after in Situ Cross-linking-Protein complexes were obtained from HeLa cells that had been treated with the membrane-permeant cross-linker dithiobis(succinimidyl propionate) (DSP; Pierce) before lysis essentially as described (20). In brief, cells grown in Dulbecco's minimum Eagle's medium supplemented with 10% calf serum were transfected with His-or BiPro-tagged PKC␣ or vector control. After 24 h, cells were collected, and 3 ϫ 10 6 cells were plated on collagen-coated dishes for 10 -15 min or 16 h before DSP was added. After the addition of DSP, cells were kept in the incubator for another 15 min, excess cross-linker was quenched in 0.2 M glycine in phosphatebuffered saline, and cells were lysed in RIPA buffer (50 mM Tris, pH 7.2, 1% (v:v) Triton X-100, 0.25% deoxycholate, 1 mM EGTA, 150 mM NaCl, protease inhibitors). Cell debris, collected by a brief centrifugation, was boiled in 20 l of a 10% SDS solution to extract the maximal amount of cross-linked complexes and spun again. Supernatants of both centrifugation steps were pooled and adjusted with RIPA buffer to reduce SDS concentration below 0.1%. Precipitation was performed either using Ni-NTA-agarose (Qiagen) for His-tagged PKC␣ or a monoclonal antibody against the BiPro sequence (4A6 (46)) for BiPro-tagged PKC␣. Endogenous proteins were detected after Western blotting with the antibodies anti-vinculin (hVIN-1) and anti-actin (1A4) (Sigma). PKC␣ was probed with a monoclonal antibody (Upstate Biotechnology, Inc., Lake Placid, NY). Horseradish peroxidase-coupled secondary antibodies (Dianova, Hamburg, Germany) were used for detection by enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech).
Protein Expression and Purification-All recombinant Vt-(858 -1052/ 66) proteins bearing an N-terminal His tag and a FLAG tag in tandem were expressed in the Escherichia coli strain M15. Batch purification on Ni-NTA-agarose was performed according to the manufacturer's instructions (Qiagen). Purity of all proteins was checked by SDS-PAGE and immunoblots using a FLAG antibody (M2; Sigma). Protein concentration was determined by standard methods. Maltose-binding protein (MBP)-tagged Vt proteins (893-1066, 893-985, and 1016 -66) (a gift of Dr. Stefan Huttelmaier) were prepared as described in Huttelmaier et al. (49). Turkey gizzard ␣-actinin and vinculin were purified according to Feramisco and Burridge (50) using a Q-Sepharose (Amersham Biosciences, Inc.) instead of a DEAE column. Actin (a gift of Dr. Kathrin Schluter) was prepared from acetone powder of rabbit muscle as described in Spudich and Watt (51), with an additional gel filtration step.
Kinase Assays and Statistical Evaluation-Phosphorylation of Vt proteins was analyzed in a mixed micelle assay essentially as described (52). MALDI Analysis of Phosphopeptides and MS/MS Sequencing-Vt-(wild type) or Vt-(858 -1052) protein was phosphorylated by PKC for 1 h (see above), lipids were extracted (53), and pellets of phosphorylated protein were dissolved in 50 mM NH 4 HCO 3 buffer, pH 8. 1 g of protein was digested by 10 -50 ng of Glu-C or Lys-C endopeptidase (sequencing grade) according to the manufacturer's instructions (Promega). Mass spectra were obtained with a MALDI-time of flight Reflex II (Bruker-Daltonics, Bremen, Germany). For co-precipitation, analyte and matrix solutions (20 mg/ml ␣-cyano-4-hydroxy-cinnamonic acid (4CHCA) in 60% acetonitrile) were mixed in equal portions (0.5 l) on the MALDI target and dried. Spectra were externally calibrated using monoisotopic (MϩH) ϩ ions from two peptide standards. Approximately 200 shots were collected for each mass spectrum. MS/MS of phosphopeptides was performed by nano-electrospray ionization on a Q-time of flight mass spectrometer (Micro Mass, Manchester, UK).
Actin Filament Binding Assays-The interaction of recombinant proteins, Vt-(wild type), Vt-(R1060Q/K61Q), and Vt-(858 -1052), with actin filaments was analyzed in a sedimentation assay. All proteins used were centrifuged (100,000 ϫ g, 30 min) before to the sedimentation assays to remove protein aggregates. Pre-polymerized actin (3 M final concentration) was incubated in the absence or presence of recombinant Vt proteins at various molar ratios at 37°C for 2 h. Samples were subjected to centrifugation at high speed (100.000 ϫ g, 1 h) or low speed (12.000 ϫ g, 15 min). Pellets and supernatants were analyzed on Coomassie-stained gels.
Low shear viscometry was essentially performed as described (54). Recombinant Vt proteins (see above) were added to pre-polymerized actin (3 M final concentration) at various molar ratios. The solution was loaded in glass capillaries and incubated at 37°C for 2 h. The viscosity of the samples was determined relative to the time needed for a steel ball to pass the capillary filled only with an F-actin control.
Recombinant Vt proteins (see above) were added to pre-polymerized actin (3 M) at a molar ratio of 0.8 Vt:actin. After 2 h of incubation at 37°C, 10% (v/v) tetramethylrhodamine-labeled phalloidin was added from a stock solution (0.1 mg/ml in methanol), and incubation was continued for 1 h. Filaments were analyzed directly via fluorescence microscopy (Axiophot; Zeiss, Jena, Germany). Images were taken using a cooled CCD camera (Roper Scientific, Tucson, AZ) and the Meta-Morph software (Visitron, Puchheim, Germany).
PI4,5P 2 Binding Assays-Binding of the Vt C terminus to PI4,5P 2 micelles was tested as described (55). Briefly, Microlon ELISA plates (Greiner, Frickenhausen, Germany) were coated with Vt proteins (50 pmol each), blocked with 1% bovine serum albumin in phosphatebuffered saline, and incubated for 2 h with increasing amounts of PI4,5P 2 micelles (10 -500 pmol), prepared as described in Huttelmaier et al. (20). After intensive washing, bound PI4,5P 2 was detected using a PI4,5P 2 -specific antibody (Assay Design Inc., Ann Arbor, MI) and horseradish peroxidase-conjugated secondary antiserum (Dianova, Hamburg, Germany). A colorimetric reaction was developed using 2,2Јazinobis(3-ethylbenz-thiazoline-6-sulfonic acid) (ABTS) as substrate, and absorption A 410 was measured in an ELISA reader. To compensate for variations in protein adsorption and development between assays, we normalized the values using the highest absorption as reference. The value of each data point is reported as the mean Ϯ S.D. from three independent experiments.

PKC␣ and Vinculin Form Complexes in
Vivo-Because PKC effects on cell spreading have been observed during the early phase and flattening of cells (4, 5), we studied a possible association of vinculin with PKC during the formation of focal adhesion structures at early (compare with Huttelmaier et al. (20)) and late time points during spreading. HeLa cells transfected with PKC␣ equipped with either a His or a BiPro tag were seeded onto collagen and allowed to adhere and spread for either 10 -15 min or 16 h. To stabilize presumptive membraneassociated protein complexes, they were then treated with the membrane-permeant cross-linker DSP for a further 15 min. Subsequently, cells were lysed and rigorously extracted. PKC␣ was precipitated by either Ni-NTA-agarose or the BiPro-tag antibody, and precipitates were collected by centrifugation and subjected to SDS-PAGE and immunoblotting. As shown in Fig.  1, vinculin precipitated with His-tagged PKC␣ (Fig. 1A) and with BiPro-PKC␣ (Fig. 1B) when cells were harvested at the early time point after seeding. In contrast, such co-precipitation was not observed with lysates obtained after long term culture (Fig. 1B). Hence, these data support our hypothesis on a transient association between vinculin and PKC␣.
Interaction with Vinculin Depends on the PKC␣ Regulatory Domain-Early studies show that in vitro phosphorylation of vinculin by PKC requires acidic phospholipids (27,29) that bind to the vinculin tail (56). This requirement could derive from the lipid-regulated head-to-tail interaction in vinculin, masking potential PKC binding and/or phosphorylation site(s) that are also located in the vinculin tail. If the role of phospholipids in complex formation between PKC and vinculin would be confined to unmasking such sites, PKC should bind to the vinculin tail even in the absence of phospholipids. On the other hand, the regulatory domain of PKC␣ that consists of the C1 and C2 modules also binds phospholipids (57,58), and the C2 module requires a free Ca 2ϩ concentration of ϳ100 M. To analyze the conditions for the interaction between vinculin and PKC in more detail, we performed overlay experiments with purified smooth muscle vinculin or recombinant vinculin tail (Vt-(858 -1066)) and in vitro translated PKC␣ using PI4,5P 2 as an acidic phospholipid. As shown in Fig. 2, PKC binding to both vinculin and vinculin tail was observed only when these proteins were preloaded with PI4,5P 2 , indicating that the sole exposure of a PKC-binding site on the vinculin tail without PI4,5P 2 is insufficient for PKC recognition. Remarkably, the addition of the lipid to the PKC-containing solution used for the overlay experiments did not result in PKC binding (data not shown). Furthermore, even with PI4,5P 2 -preloaded vinculin proteins, the interaction was effectively blocked by addition of the Ca 2ϩ chelator EGTA (Fig. 2). These results strongly suggest a prominent role of the regulatory PKC-C2 module in the (protein-lipid)-protein interaction with vinculin. This was in contrast to smooth muscle ␣-actinin, which also binds to acidic phospholipids (18) and PKC␣, but the interaction was found independent of the free Ca 2ϩ concentration in this assay (Fig.  2). When an ATP binding-deficient PKC␣ mutant (47) was used that is characterized by a defectively folded kinase domain, the interaction with vinculin or vinculin tail was not reduced (Fig.  2). Taken together, these data allow for the conclusion that the contact between vinculin and PKC␣ is mediated by the PKC regulatory domain and not by the kinase domain.
Vinculin Tail Phosphorylation by PKC Depends on the C Terminus-We postulated that vinculin phosphorylation by PKC depends on the PI4,5P 2 -mediated interaction between the kinase and its substrate. The vinculin tail structure reveals two acidic lipid-interacting areas, with a basic ladder at the H2-H3 helical hairpin and a basic collar connected to the Cterminal hydrophobic hairpin (15,22) that both could support the lipid-regulated interaction with PKC. To clarify which of these sites might be involved, in vitro phosphorylation assays with recombinant PKC␣ and deletion fragments of the vinculin tail were performed. We used vinculin tail constructs tagged with MBP that had previously been shown to bind to actin like FIG. 1. Interaction of vinculin and PKC␣ is demonstrated by in vivo chemical cross-linking. PKC␣-transfected HeLa cells and vector-transfected controls were allowed to adhere on collagen for 10 -15 min or overnight. Subsequently, cells were treated with a membranepermeant cross-linker (DSP) for 15 min and lysed. PKC was precipitated from cleared supernatants. After reduction of DSP, precipitated proteins were analyzed in a Western blot. The membrane was probed sequentially with antisera specific to vinculin, PKC␣, or actin (as indicated). A, His-tagged PKC␣ was precipitated from lysates of spreading cells using Ni-NTA-agarose. P, precipitate. B, BiPro-tagged PKC␣ was precipitated using a BiPro-specific monoclonal antibody (4A6) (46). Cells were analyzed 15 min (15Ј) and 16 h after plating (immunoprecipitate (IP)). hc indicates the 50-kDa heavy chain of the BiPro-specific antibody. Note that co-precipitation of PKC␣ and vinculin was obtained by both experimental approaches in spreading cells, and actin was not part of the complex. the wild type vinculin tail and are thus probably in a native conformation (49). MBP itself was not phosphorylated by PKC. As shown in Fig. 3 and consistent with earlier results from our and other groups (27,29), a Vt-(893-1066) fragment comprising the five-helix bundle (H1-H5) and the C terminus was readily phosphorylated by PKC. A C-terminal deletion fragment, Vt-(893-985), was not phosphorylated (Fig. 3), although this protein contained the lipid binding domain in helices H2-H3 and several consensus motifs for PKC phosphorylation sites. In contrast, the short C-terminal construct Vt-(1016 -1066) comprising helix H5 and the C terminus, retained some phosphate incorporation (Fig. 3). This was ϳ20 -30% of the value obtained for the intact tail, as judged by densitometric analysis of the signal. This indicated that at least some of the phosphorylation sites are located at the very C terminus of the vinculin tail (amino acid residues 1016 -1066) and that this part also includes the main binding site for the kinase.
PKC Phosphorylates Serines 1033 and 1045 in Helix H5-PKC binding to vinculin during cell spreading is expected to have two major consequences, localization of the activated kinase to a specific site and phosphorylation of the binding partner(s). To further dissect the involvement of vinculin with respect to defining potential phosphorylation sites and analyzing the lipid binding properties of the C-terminal arm, a number of Vt mutants were generated by site-directed mutagenesis (Fig. 4A). Serine and threonine residues within a PKC consensus motif (59) were selected from accessible loop regions and the C terminus of the vinculin tail structure to generate phosphorylation site-deficient Vt proteins (Fig. 4, A and B). Mutant proteins included alanine substitutions of Ser-941/Thr-43 and Ser-999 from loop regions as well as Thr-1050/Thr-55 and Thr-1062 from the C terminus. However, phosphorylation assays with the corresponding recombinant Vt proteins and PKC␣ did not demonstrate a reduction of phosphate incorporation as compared with the wild type Vt protein (1.5 Ϯ 0.3 mol of phosphate/mol of Vt). This result renders it rather unlikely that any of the potential phosphorylation sites selected (Fig.  4B) contribute markedly to the overall phosphorylation. In contrast, when proteolytic peptides of phosphorylated vinculin tail were generated using Lys-C or Glu-C endopeptidase and subsequently analyzed by MALDI and electrospray ionization time-of-flight mass spectroscopy, peptides containing phosphorylated Ser-1033 and Ser-1045 were detected. The identity of these peptides and their phosphorylation sites were verified by MS/MS sequencing (Fig. 4B). Both sites are located in helix H5. This is consistent with our results on the phosphorylation of the MBP-Vt-(1016 -1066) polypeptide.

C-terminal Mutations in the Predicted PI4,5P 2 -binding Site Do Not
Interfere with Actin Binding-To estimate the importance of the lipid-binding site at the vinculin C terminus for PKC-vinculin interactions, we generated appropriate vinculin mutants. The selection of relevant mutants was based on the characterization of the PI4,5P 2 -binding site in syndecan 4 that is essential for PKC docking to this protein (60). PI4,5P 2 binding to syndecan 4 induces the oligomerization of the short cytoplasmic tail (61), which is also known for the vinculin tail (20). In addition, both syndecan 4 (amino acids 183-202) and the vinculin C terminus (1053-1066) do not contain the consensus motif described for PI4,5P 2 binding in actin binding proteins, (K/R)X 3-5 (K/R)X(K/R)(K/R) (62) (Fig. 4C). As shown in Fig. 4D, we constructed two point mutants with amino acid substitutions and one deletion mutant. The mutant Vt-(R1060Q/K61Q) was expected to display reduced or no binding of the negatively charged phosphoinositol ring (Fig. 4D), whereas the Vt-(T1062E) mutant should mimic the effect of a phosphorylation on Thr-1062. The deletion mutant Vt-(858 -1052) was generated to probe the importance of the vinculin C terminus for PKC binding.
As it had been shown previously that the binding of PI4,5P 2 and actin to the vinculin tail can mutually affect each other (16,27,63), we tested the actin binding of the presumed PI4,5P 2 binding mutants to exclude general effects on folding and ligand binding properties. The mutants Vt-(R1060Q/K61Q) and Vt-(858 -1052) were tested for their interaction with F-actin in cosedimentation and microscopic assays as well as in low shear viscometry (49, 54, 64). As shown in Fig. 5, actin binding and bundling was unaffected in standard high and low speed cen- trifugation assays (Fig. 5A, low speed data not shown), in actin-bundle formation of the co-polymers (Fig. 5B), and in low shear viscometry (data not shown). These results suggest that both mutant proteins are in their native conformation.
Deletion of the Presumed PI4,5P 2 -binding Region at the Vinculin C Terminus Results in Reduced Affinity for PI4,5P 2 -The Vt C-terminal arm seems critical for the binding to PI4,5P 2and phosphatidylserine-containing multilamellar vesicles. This was shown in cosedimentation assays using a Vt-(879 -1051) mutant (22). For our deletion mutant Vt-(858 -1052), we tested PI4,5P 2 -binding as compared with wild type Vt using an ELISA assay (Fig. 5C). Vt proteins (50 pmol each) were adsorbed to ELISA plates and incubated with 5-500 pmol of PI4,5P 2 vesicles in a total volume of 100 l. Lipid binding was monitored with a PI4,5P 2 -specific antibody as described earlier (55). As shown in Fig. 5C, the deletion mutant bound less lipid at lower PI4,5P 2 concentrations than wild type, suggesting that the affinity of this mutant protein for lipids is reduced. However, the overall capacity of lipid binding was comparable for both proteins when saturation levels of PI4,5P 2 were reached. -(858 -1066). Mutants were designed on the basis of the Vt structure that shows five amphipathic helices (H1-H5) forming an antiparallel bundle with short interconnecting stretches and arms at the N and C terminus (22). A, sequence of the wild type Vt-(858 -1066) expression construct with N-terminal epitope tags, His and FLAG, and positions of the helices (gray boxes). Positions of PKC phosphorylation site mutants are highlighted by numbers and red letters and positions of lipid-binding site mutants are highlighted by blue letters or a blue bar (deletion). B, ribbon representation of Vt- (879 -1061) showing the steric orientation of Ser/Thr sites with the PKC consensus motif that were mutated to alanine (black) and actual phosphorylation sites identified by mass spectroscopy after phosphorylation reaction (red). Masses of proteolytic peptides and sequences that were received are indicated (phosphoserine (S(P)). C, comparison of the vinculin C terminus with characterized PI4,5P 2 -binding sites of syndecan 4 (60) and of several actin-binding proteins, with the consensus sequence (K/R)X 3-5 (K/R)X(K/R)(K/R) (62). Peptides that are known to form homo-oligomers in the presence of PI4,5P 2

PKC Binding to Vt Mutants of the C-terminal PI4,5P 2 -bind-
ing Site Is Reduced-The mutants described and characterized above were used in phosphorylation assays to estimate relative binding affinities of PKC for the lipid-binding site of the Vt. Vt proteins were preincubated with PI4,5P 2 and analyzed for phosphorylation in a PKC-mixed micelle assay (52). There was no difference in the total amount of phosphate incorporated between these mutants and wild type Vt protein (1.5 Ϯ 0.3 mol of phosphate/mol of Vt protein). However, when the initial rate of phosphate incorporation (Ͻ0.4 mol phosphate/mol of Vt) was tested using a low enzyme to substrate ratio, a significant difference was observed. Variance analysis was performed to compare incorporation rates using data from four independent experiments. All substrate proteins showed a linear time dependence of phosphate incorporation (R 2 Ͼ 0.97). As shown in Fig. 6, Vt-(R1060Q/K61Q) and the deletion Vt-(858 -1052) exhibited only 50%, and Vt-(T1062E) exhibited 75% of wild type phosphorylation rate. Hence, deletion of or mutations in the presumptive lipid-binding motif at the vinculin C terminus affects the kinase activity, suggesting a possible mechanism for a regulated interaction between both proteins, PKC and vinculin, as a function of local PI4,5P 2 availability.

DISCUSSION
In this paper, we have analyzed the interaction of vinculin, a prominent component of focal contacts, and PKC, a Ser/Thr kinase that is known to be essential for cell adhesion and spreading of fibroblasts and other cell types (1). The interaction between PKC␣ and vinculin was monitored at an early time point (15 min) after seeding of HeLa cells, when adhesion has been completed, and the initial phase of spreading is induced. Engagement of integrins by a suitable substrate is accompanied by a rapid activation and translocation of PKC to the plasma membrane (4,65). To trap a presumedly labile and transient interaction of the kinase with vinculin at the membrane, we used a dual strategy. First, we expressed a PKC␣ construct to increase the number of kinase molecules available for complex formation with endogenous vinculin. Second, we used a cross-linking protocol to preserve membrane-associated complexes formed in situ that could be disintegrated during cell lysis. The usefulness of DSP cross-linking in stabilizing such complexes has previously been shown in several studies (20,66,67). The validity of both strategies is also demonstrated by the fact that the PKC-vinculin complexes identified were confined to spreading but were not seen with fully spread cells.
FIG. 5. F-actin binding and organization is not affected in lipid-binding site mutants, whereas binding to PI4,5P 2 is reduced. A, binding of the Vt mutants Vt-(R1060Q/K61Q) (Vt-(RK1060/61QQ)) and Vt-(858 -1052) to F-actin was compared with wild type Vt by high speed cosedimentation. Pairs of pellet (P) and supernatant (S) are shown for a range of Vt to actin ratios on Coomassie-stained gels. Note that both mutant proteins sediment with F-actin. B, F-actin bundling, as monitored by fluorescence microscopy. Actin filaments were polymerized in the presence of either Vt-(wild type), Vt-(R1060Q/K61Q), or Vt-(858 -1052) and stained with tetramethylrhodamine-labeled phalloidin. Note that the bundling capacity of all three Vt proteins is comparable. C, PI4,5P 2 binding of Vt-(858 -1052) (q) and Vt-(wild type) (OE) was compared in an ELISA assay. Protein (50 pmol each) was coated on plates and incubated with increasing amounts of PI4,5P 2 . Binding of the lipid was monitored by PI4,5P 2 -specific antiserum and expressed as mean values Ϯ S.D. for three independent experiments. Note that at low concentrations of PI4,5P 2 , Vt-(wild type) binds more efficiently than Vt-(858 -1052), whereas at high concentrations the same end point is reached.
Although PKC activation upon cell adhesion and spreading has been known for almost a decade, downstream events involving specific PKC-substrate interactions and the mechanism(s) underlying these time point-and site-specific interactions are poorly understood. The conformation and activation of both PKC and vinculin both depend on acidic phospholipids (31,68). Therefore, a mechanism controlling local availability of these lipids at the plasma membrane must be important for a regulated interaction. Local fluctuations of phospholipids may be regulated by GAP43-like proteins. These proteins accumulate at PI4,5P 2 -enriched rafts, where they co-distribute with PI4,5P 2 and promote its retention and clustering. Changes in actin dynamics are correlated with alterations in raft organization by GAP43-like proteins (7,10). The myristoylated alanine-rich protein kinase C substrate (MARCKS), a widely expressed member of this family, possibly acts as a key regulator in cell spreading. Sequestering of acidic phospholipids, phosphatidylserine, and PI4,5P 2 in membrane domains by MARCKS is abolished by PKC-dependent phosphorylation (69,70), resulting in a local release of these lipids. Furthermore, fibroblasts expressing a palmitoylated, permanently membrane-bound mutant of MARCKS do not spread. This loss of function was attributed to the fact that the PKC phosphorylationdependent, transient displacement of MARCKS from the membrane was blocked (6). The control by PKC of this "myristoylelectrostatic switch" (71) during cell spreading might form the basis for rapid and local acidic phospholipid-driven events. Concomitantly, active PKC recruited during spreading to the plasma membrane (4) can induce and modulate a phospholipid-dependent activation of focal contact proteins like vinculin.
In a dot-overlay assay, we analyzed the binding of PKC␣ to vinculin to better understand the characteristics of this (proteinlipid)-protein interaction. By comparison of vinculin and vinculin tail binding to in vitro translated PKC␣ or an inactive mutant, respectively, we demonstrated that the lipid-bound vinculin tail provides a Ca 2ϩ -dependent binding site for the PKC regulatory domain. In our assay, vinculin and its tail fragment were membrane-adsorbed after preincubation with PI4,5P 2 . With this protocol a Ca 2ϩ -dependent interaction with PKC was observed that resisted intensive washing with TBST. Earlier, Hyatt et al. (72) performed an overlay analysis with membrane-bound vinculin and a partially purified PKC fraction and observed a reversible interaction. The complex required the continuous presence of lipids and Ca 2ϩ throughout the washing and staining procedure or, alternatively, a fixation step after kinase addition. Fixation resulted in a residual binding of PKC␣ even in the absence of Ca 2ϩ . The apparent discrepancy between the results obtained by both procedures can be explained by the different mode of lipid application. The addition of the lipid to membrane-bound vinculin will loosen its intramolecular head-to-tail bonding and allow a lipid-mediated binding of the kinase. In contrast, pretreatment of soluble vinculin with PI4,5P 2 results in the formation of vinculin tail oligomers (20) that bind the lipid tightly, which probably reflects the in vivo situation (56,73). A similar effect of the state of oligomerization on the binding of PKC to a lipid-presenting partner was observed for syndecan 4. The cytoplasmic tail of syndecan 4 binds to and activates PKC␣ only in its lipid-bound, oligomeric conformation (61,74). Oligomerization of the synde- can 4 tail is inhibited by phosphorylation on Ser-183, which results in a reduction of PKC activity by 1 order of magnitude (74).
In a mixed micelle assay, we show that binding of PKC and subsequent phosphorylation require and are probably restricted to the C-terminal part of the vinculin tail (985-1066). This corroborates earlier reports that located the main PKC phosphorylation site(s) to the vinculin tail (27,29). Surprisingly, some of the sites in Vt with a PKC consensus motif, Ser-941 and Ser-999, predicted to be attractive sites for PKC phosphorylation (15) as well as the C-terminal threonine residues (Thr-1050, -55, -62) were not phosphorylated in our in vitro assay. In contrast, serines 1033 and 1045 in helix H5, adjacent to the PKC-binding site, were phosphorylated, and we have preliminary data indicating one to two further sites in helices H3 and H4. 2 In the compact conformation of the vinculin tail (879 -1066) that has been determined by x-ray structure analysis (22) the phosphorylation sites identified so far appear to be readily accessible for the kinase. Ser-1033 and -1045 both lie within the binding site for the vinculin head (1009 -1066) (24,27,75) and the C-terminal binding site for F-actin (1016 -1066) (49). Conversely, engagement of vinculin head or actinbinding sites on the Vt might modulate PKC-dependent phosphorylation. This concept is supported by an early study showing that talin, a ligand of the vinculin head domain, can increase PKC-dependent vinculin phosphorylation, whereas binding of vinculin to F-actin decreases phosphorylation (76).
Remarkably, the two acidic phospholipid-binding sites in vinculin tail, amino acids 916 -970 in helices H2, H3 (23) and amino acids 1053-1066, representing the hydrophobic finger at the C terminus (22), differed in their capacity to attract PKC. Although constructs including these sites individually, MBP-Vt-(893-985) and MBP-Vt-(1016 -1066), both perform homotypic interactions (49), bind to PI4,5P 2 (20), and contain PKC consensus phospho-rylation sites (Ref. 59 and this study), only the short Vt-(1016 -1066) peptide is accepted as a PKC substrate. This striking difference may result from the configuration by which the lipid moiety is bound to the respective peptide, thereby influencing the capacity of the PKC regulatory domain to bind the Vt-lipid complexes. This notion is supported by our kinetic analysis of the phosphorylation of Vt-(858 -1066) C-terminal lipid-binding site mutants. Mutants that were designed to reduce binding of the negatively charged head group of PI4,5P 2 show a reduced phosphorylation rate (25-50%), although the overall organization and activity was not affected, as judged by their capacity to interact with F-actin. Furthermore, the effect on the phosphorylation rates not simply reflects a reduced oligomerization of the vinculin tail mutants, since the deletion mutant Vt-(858 -1052) shows an increased capacity to oligomerize upon addition of PI4,5P 2 as compared with Vt-(wild type). 2 The syndecan 4/PKC␣ binding might again serve as a model. For this pair, a biophysical analysis suggests that PI4,5P 2 mediates their interaction, and PKC binding affinity mainly derives from its interaction with the inositol head group (60). The reduced initial rate of phosphorylation observed for lipid-binding site mutants of Vt suggests that PI4,5P 2 binding to the C terminus can control interaction kinetics with PKC␣. Given the low number of PKC molecules in cells as compared with vinculin, a lipid-induced rise in binding affinity may provide a mechanism for efficient phosphorylation of vinculin at the time point when focal contacts are being formed.
Although there are some reports on vinculin phosphorylation in cells (77,78), it has proven difficult to establish conditions of phosphorylation. There are, however, two reports analyzing specific activation events that provide evidence for a PKC-dependent phosphorylation and/or major subcellular redistribution of vinculin in adherent cells. During Ca 2ϩ -induced junctional sealing in Madin-Darby canine kidney cells, a PKC inhibitor-sensitive redistribution of vinculin was observed. Phosphorylation of Ser/Thr residues on vinculin immunoprecipitated from cell extracts was dependent on PKC activity (79). . I, upon cell-matrix adhesion PKC is translocated to and activated (*) at the membrane, where the kinase phosphorylates GAP43-like proteins, e.g. MARCKS, displacing them from PI4,5P 2 rafts. II, locally increased PI4,5P 2 attracts lipid-binding proteins, e.g. vinculin and induces conformational changes and/or oligomerization. III, lipid-bound oligomeric vinculin tail serves as a PKC-docking site, which leads to the phosphorylation and activation of vinculin and/or vinculin binding partners (see "Discussion," last paragraph).
Treatment of confluent epithelial cells (Int 407) with the inflammatory mediator leukotriene D4 resulted in a dissociation of vinculin from a complex containing ␣-catenin and an increased localization to focal contacts. These effects were mimicked by the PKC activator 12-O-tetradecanoylphorbol 13-acetate (TPA) and blocked by the specific inhibitor bisindolylmaleimide I (GF109203X) (80).
So far, the mode of regulation of vinculin and its function in modeling cell adhesion sites have escaped a comprehensive analysis. Our results indicate a lipid-dependent mechanism for the release of the inhibitory vinculin head-tail interaction by induction of membrane-associated PKC activity. The principles of this model are outlined in Fig. 7. Spreading-induced activation of PKC at the plasma membrane leads to the phosphorylation of GAP43 family proteins, like MARCKS. Phospho-MARCKS has a reduced binding affinity for acidic phospholipids and is released from PI4,5P 2 -enriched rafts. Cytosolic vinculin gains access to the surface of rafts, where PI4,5P 2 binding releases the head-tail interaction, thereby activating cryptic binding and homo-oligomerization sites in the molecule. Upon oligomerization of the vinculin tail, a docking site for PKC is provided. PKC phosphorylation of the vinculin tail and/or vinculin binding partners is expected to regulate the incorporation of vinculin in nascent cell adhesion complexes. Because of the large number of acidic phospholipid binding molecules involved in the regulation of the membrane-anchored actin cytoskeleton (2, 10), such a model may be expanded to other proteins.