Regulation of Human Endothelial Cell Focal Adhesion Sites and Migration by cGMP-dependent Protein Kinase I*

cGMP-dependent protein kinase type I (cGK I), a major constituent of the atrial natriuretic peptide (ANP)/nitric oxide/cGMP signal transduction pathway, phosphorylates the vasodilator-stimulated phosphoprotein (VASP), a member of the Ena/VASP family of proteins involved in regulation of the actin cytoskeleton. Here we demonstrate that stimulation of human umbilical vein endothelial cells (HUVECs) by both ANP and 8-(4-chlorophenylthio)guanosine 3′:5′-monophosphate (8-pCPT-cGMP) activates transfected cGK I and causes detachment of VASP and its known binding partner (zyxin) from focal adhesions in >60% of cells after 30 min. The ANP effects, but not the 8-pCPT-cGMP effects, reversed after 3 h of treatment. In contrast, a catalytically inactive cGK Iβ mutant (cGK Iβ-K405A) was incapable of mediating these effects. VASP mutated (Ser/Thr to Ala) at all three of its established phosphorylation sites (vesicular stomatitis virus-tagged VASP-AAA mutant) was not phosphorylated by cGK I and was resistant to detaching from HUVEC focal adhesions in response to 8-pCPT-cGMP. Furthermore, activation of cGK I, but not of mutant cGK Iβ-K405A, caused a 1.5–2-fold inhibition of HUVEC migration, a dynamic process highly dependent on focal adhesion formation and disassembly. These results indicate that cGK I phosphorylation of VASP results in loss of VASP and zyxin from focal adhesions, a response that could contribute to cGK alteration of cytoskeleton-regulated processes such as cell migration.

cGMP-dependent protein kinase type I (cGK I), a major constituent of the atrial natriuretic peptide (ANP)/ nitric oxide/cGMP signal transduction pathway, phosphorylates the vasodilator-stimulated phosphoprotein (VASP), a member of the Ena/VASP family of proteins involved in regulation of the actin cytoskeleton. Here we demonstrate that stimulation of human umbilical vein endothelial cells (HUVECs) by both ANP and 8-(4-chlorophenylthio)guanosine 3:5-monophosphate (8-pCPT-cGMP) activates transfected cGK I and causes detachment of VASP and its known binding partner (zyxin) from focal adhesions in >60% of cells after 30 min. The ANP effects, but not the 8-pCPT-cGMP effects, reversed after 3 h of treatment. In contrast, a catalytically inactive cGK I␤ mutant (cGK I␤-K405A) was incapable of mediating these effects. VASP mutated (Ser/Thr to Ala) at all three of its established phosphorylation sites (vesicular stomatitis virus-tagged VASP-AAA mutant) was not phosphorylated by cGK I and was resistant to detaching from HUVEC focal adhesions in response to 8-pCPT-cGMP. Furthermore, activation of cGK I, but not of mutant cGK I␤-K405A, caused a 1.5-2-fold inhibition of HUVEC migration, a dynamic process highly dependent on focal adhesion formation and disassembly. These results indicate that cGK I phosphorylation of VASP results in loss of VASP and zyxin from focal adhesions, a response that could contribute to cGK alteration of cytoskeleton-regulated processes such as cell migration.
Nitric oxide and atrial natriuretic peptide (ANP) 1 activate soluble or particulate guanylyl cyclases, respectively. The cGMP produced by these cyclases has multiple effectors within the cell, including cGMP-gated channels, cGMP-regulated phosphodiesterases, and cGMP-dependent protein kinases (cGKs). cGK I (including ␣and ␤-splice variants) and cGK II have been identified in mammalian cells. The cGK I studied here is a major mediator of the nitric oxide/cGMP signal transduction pathway in several tissues (1); and recently, the functional role of cGK I in relaxation of smooth muscle, inhibition of Ca 2ϩ transients, and inhibition of platelet adhesion and aggregation was consolidated by data from cGK I knockout mice (2,3). Furthermore, cGK I is present in only certain endothelial cells, and only these respond to cGMP with a reduction in thrombininduced calcium transients and paracellular permeability (4,5).
The relationship between cGK I substrate phosphorylation and cellular function is still an emerging field and is the subject of this investigation. One extensively studied substrate of cGK I is the vasodilator-stimulated phosphoprotein (VASP), originally discovered in human platelets (6), but a substrate of cGK I in many cells, including endothelial cells (1,5). In platelets, both cAMP-and cGMP-dependent phosphorylation of VASP closely correlate with inhibition of agonist-induced aggregation and in particular with inhibition of fibrinogen receptor (integrin ␣ IIb ␤ 3 ) activation (7). In knockout mice lacking endogenous VASP, cGMP-mediated inhibition of platelet aggregation was reduced, and agonist-induced platelet activation was enhanced (8,9). Similarly, knockout mice lacking endogenous cGK I also displayed defective cGMP-mediated inhibition of platelet aggregation (3). cGMP-dependent phosphorylation of VASP (only Ser 157 was measurable at that time), but not cAMPdependent phosphorylation of VASP, disappeared from the platelets of cGK I knockout mice (3) as well as from cGK I-deficient platelets from chronic myelocytic leukemia patients (10) and from endothelial cells that lost cGK I in culture (5), clear evidence that cGMP-dependent phosphorylation of VASP is mediated by cGK and not by activation of cAK. In contrast, cAMP-dependent phosphorylation of VASP is mediated by cAK.
In cells studied so far, VASP has been shown to be localized at highly dynamic membrane regions, focal adhesions, and cell-cell contacts and in a periodic pattern along stress fibers (11,12). Recently, VASP-related proteins (now designated as members of the Ena/VASP protein family), which share some structural and functional characteristics with VASP, were identified (13,14). Demonstration of VASP interaction with other proteins present at several intracellular sites suggested a role for VASP in cytoskeletal function. The N-terminal Ena/ VASP homology 1 (EVH1) domain of the Ena/VASP protein family was shown to interact with a proline-rich FP 4 -containing motif present in the focal adhesion proteins zyxin and vinculin (15) and also present in ActA, a surface protein of the intracellular pathogen Listeria monocytogenes to which VASP is recruited to facilitate F-actin assembly essential for Listeria motility (15)(16)(17)(18)(19)(20)(21). The central VASP domain containing four copies of a GP 5 motif binds profilin, an actin regulatory protein that has been shown to promote F-actin assembly (22). In addition, the VASP C-terminal EVH2 domain was shown to bind F-actin and to facilitate actin bundling (23,24), to influence VASP localization to focal adhesions (25), and also to be responsible for tetramerization of VASP (23), which may stabilize interactions of the VASP EVH1 domain with the focal adhesion proteins vinculin and zyxin (26).
VASP is phosphorylated by cAMP-and cGMP-dependent protein kinases on Ser 157 , located just distal to the EVH1 domain in the central proline-rich domain of VASP, and on Ser 239 and Thr 278 , located in the VASP C-terminal EVH2 domain (23,25,27). So far, only phosphorylation of Ser 157 and Ser 239 has been demonstrated to be of major significance in intact cells (27), and phosphorylation of these sites can be distinguished using specific antibodies (28). The phosphorylation site containing Ser 157 is preferred by cAK, whereas the site containing Ser 239 is preferred by cGK in vitro and in intact cells; however, both kinases phosphorylate all three sites (27,28). Information is, however, scarce concerning the influence of cGMP-mediated VASP phosphorylation on the described VASP interactions with other proteins or on cytoskeletal function. Furthermore, studies on many cell types, including endothelial cells, are hampered by the rapid decline in cGK levels in cell culturing and passaging (5,29). Therefore, in this work, we present a system for studying the effect of VASP phosphorylation, using endothelial cells lacking endogenous cGK I, in the absence and presence of expressed cGK I. cGK I phosphorylation of VASP promoted detachment of VASP (but not a triple phosphorylation site mutant of VASP) and zyxin from focal adhesions, thus altering the composition and possible functional integrity of these structures. In support of this, cGK I also inhibited collagen-, fibronectin-, and fibrinogen-mediated endothelial cell migration, which is a highly dynamic process dependent on focal adhesion formation and destruction.

EXPERIMENTAL PROCEDURES
Cells-Endothelial cells were isolated from human umbilical veins and cultured as described previously (30). HUVECs from only passages 1-3 were used for experiments. Primary human dermal fibroblasts (passages 7-10) and the human embryonic kidney cell line (HEK 293) were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc., Eggenstein, Germany) supplemented with 10% fetal calf serum and 1% antibiotic solution. All cells were kept at 37°C with 5% CO 2 .
Vectors and Constructs-The recombinant adenoviral vector Ad5-cGK I␤ containing human cGK I␤ cDNA (33,34) was constructed as described previously (35). The catalytically inactive cGK I␤-K405A mutant was newly constructed by oligonucleotide-directed mutagenesis (Transformer site-directed mutagenesis kit, CLONTECH, Palo Alto, CA) of cGK I␤ cDNA. The sequence of the mutagenic primer was 5Ј-GTT TCT TGA GAA TCg cCA TTG CAA ACG TTT TG-3Ј, and the mutated region was checked by sequencing. The recombinant virus Ad5-cGK I␤-K405A was developed by cloning human mutant cGK I␤ into the adenoviral transfer plasmid pZS2 using the multiple cloning site between the cytomegalovirus promotor and the bovine growth hormone 3Ј-untranslated region. The resulting recombinant plasmid pZS2/cGK I␤-K405A also contained 500 base pairs of 5Ј-terminal sequence of wild-type adenovirus type 5 upstream of cGK I␤. This plasmid was linearized with XbaI and ligated with the long XbaI fragment of RR5 DNA, an Ad5 mutant that carries a unique XbaI site and a deletion of the E1 region ranging from nucleotides 445 to 3333 of the wild-type adenovirus type 5 genome. Subsequently, the ligation product was transfected into E1a-transformed HEK 293 cells; and after 10 -14 days, recombinant virus was recovered and plaque-purified as described previously (36).
Plasmid expression vectors were constructed using pcDNA3 (Invitrogen, DeSchelp, The Netherlands) and human VASP N-terminally tagged with an epitope of VSV glycoprotein G. The vectors for wild-type VASP and for three mutants carrying single amino acid substitutions at each phosphorylation site (S157A, S239A, and T278A) were described previously (28). Now, double (AAT) and triple (AAA) phosphorylation mutants of VSV-VASP were generated by combining one or more singly mutated regions using unique restriction sites.
Expression Experiments and Western Blot Analysis-HUVECs or human fibroblasts seeded on six-well plates were infected with Ad5 vectors at concentrations of 10 9 to 10 11 particles/ml in serum-free medium for 2 h at 37°C. Subsequently, the virus-containing medium was replaced by fresh serum-containing medium. Alternatively, plasmid transfections of HUVECs were performed using Superfect (QIAGEN Inc., Hilden, Germany) according to the instructions of the manufacturer. Later, 16 -24 h after either infection or transfection, cells were washed once with serum-free medium and incubated for 5-30 min with 100 nM ANP (Sigma), 100 M 8-pCPT-cGMP (BioLog, Bremen, Germany), or 3 M forskolin (Sigma) in serum-free medium containing 0.1% bovine serum albumin at 37°C. Subsequently, cells were briefly washed with PBS and either lysed in hot SDS sample buffer for later immunoblot analysis or fixed for immunofluorescence as described below. For immunoblotting, samples were separated by SDS-polyacrylamide gel electrophoresis using 9% gels and subsequently analyzed on Western nitrocellulose blots. Nitrocellulose sheets were blocked overnight in blocking medium (PBS, 0.3% Triton X-100, 0.05% Tween 20, 1% hemoglobin, and 0.01% NaN 3 ); incubated for 1 h at room temperature with anti-cGK I antiserum A10 (diluted 1:200), monoclonal anti-VASP antibody IE273 (diluted 1:1000), monoclonal anti-phospho-Ser 239 VASP antibody 16C2 (1 g/ml), or monoclonal anti-phospho-Ser 157 VASP antibody 5C6 (2 g/ml); and then subsequently incubated for 1 h with 3.7 kBq/ml 125 I-labeled protein A or 7.4 kBq/ml 125 I-labeled sheep antimouse antibody (both from Amersham Pharmacia Biotech, Freiburg, Germany) followed by autoradiography (see Figs. 1A and 2) or with horseradish peroxidase-coupled anti-mouse or anti-rabbit secondary antibodies followed by ECL detection (Amersham Pharmacia Biotech) (see Figs. 1B, 4, and 10A). The amount of cGK expressed ( Fig. 1A) (data not shown) was determined by cutting the radioactively labeled bands out of the nitrocellulose sheet and counting them in a scintillation counter. Various concentrations of purified recombinant cGK were used as a protein standard.
VASP Immunoprecipitation and in Vitro Phosphorylation-Plasmid vectors for VSV-tagged VASP and VASP phosphorylation site mutants (Ser/Thr to Ala) were transfected into HEK 293 cells using calcium phosphate. Two days after transfection, cells were harvested and solubilized in buffer containing 20 mM Tris-HCl (pH 7.4), 75 mM NaCl, 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS, 10 mM EDTA, 10 mM sodium pyrophosphate, and 20 units/ml aprotinin. The expressed proteins were immunoprecipitated using an anti-VSV antibody and rat anti-mouse IgG1 coupled to magnetic beads (Dynal, Oslo, Norway). Immune complex phosphorylation was performed with the precipitates in buffer containing 10 mM HEPES (pH 7.4), 5 mM MgCl 2 , 1 mM dithioerythritol, 0.2 mM EDTA, 4 M ATP, 1 Ci of [␥-32 P]ATP, and either 50 ng of cAK catalytic subunit purified as described (37) or 50 ng of purified cGK I (38) activated with 20 M cGMP. Samples were phosphorylated for 30 min at 30°C and then stopped and analyzed using Western blotting as described above. Expressed VSV-VASP was detected using the monoclonal anti-VSV antibody, a horseradish peroxidase-coupled anti-mouse secondary antibody, and ECL analysis. After washing ECL labeling from the nitrocellulose, 32 P incorporation into VASP was subsequently detected by autoradiography.
Indirect Immunofluorescence-HUVECs, grown on glass coverslips previously coated with 1% gelatin for 45 min and then fixed for 15 min with 0.5% glutaraldehyde and washed five times with M199 medium, were treated with cGMP-elevating agents as described below, washed twice with PBS, fixed with 4% paraformaldehyde for 15 min on ice, washed twice with PBS, permeabilized by 0.2% Triton X-100 in PBS for 10 min at room temperature, and then incubated with primary antibodies (see below) for 1 h at 37°C and with secondary antibodies for 1 h at 37°C. The coverslips were mounted in Moviol 4-88 (Hoechst, Frankfurt, Germany)-containing mounting medium with 1% N-propyl gallate, and cells were examined with a Leitz Aristoplan microscope. Photographs were taken with Tri-X-Pan 400 or TMAX 400 film (both from Eastman Kodak Co.). 2 U. Walter and H. Hoschuetsky, manuscript in preparation.
In HUVECs transfected with plasmid vectors for cGK I and VSVtagged wild-type or mutant VASP, cGK I was stained using polyclonal anti-cGK I antibody A10 diluted 1:1000 and fluorescein isothiocyanatelabeled anti-rabbit secondary antibody diluted 1:30; expressed VSV-VASP was stained using monoclonal anti-VSV antibody P5D4 diluted 1:10,000 and Texas Red-labeled anti-mouse secondary antibody diluted 1:50. The percentage of cGK I-expressing HUVECs (50 -100 cells analyzed per experiment) staining positive for VSV-VASP (or the VSVtagged triple phospho-VASP mutant) at focal adhesions after cGK I activation was determined.
Migration Assay-Haptotactic cell migration assays were performed using modified Boyden chambers (Transwell filters, 8-m pore size, Costar Corp., Cambridge, MA). The lower surface of the filter membranes was coated with 3 g/ml human fibronectin, 10 g/ml collagen I, or 10 g/ml fibrinogen (all from Sigma). The upper surface of the membrane and the remaining free binding sites on the lower surface were blocked with 0.5% bovine serum albumin. Subconfluent HUVECs were trypsinized and resuspended in M199 medium containing 25 mM HEPES, 1 mM MgCl 2 , and 0.5% bovine serum albumin, and then 30,000 cells were seeded per filter. After 5 h of incubation at 37°C with or without 100 M 8-pCPT-cGMP or 500 nM ANP, cells on the upper surface of the membrane were removed with a cotton swab. Migrated cells on the lower membrane surface were stained with 1% (w/v) crystal violet (Sigma) in 2% (v/v) ethanol, washed with H 2 O, and dried. The stain was eluted with 10% (v/v) acetic acid, and its absorbance was measured at 600 nm.

RESULTS
Functional Expression of cGK I␤ in HUVECs-A replicationdeficient adenoviral vector containing the cDNA of human cGK I␤ was constructed and used to evaluate the function of cGK I␤ in HUVECs lacking endogenous cGK I. Infection of HUVECs with different amounts of vector (10 9 to 10 11 particles/ml) resulted in the expression of 0.2-2.0 g of cGK I␤/mg of cell protein as assessed by immunoblotting ( Fig. 1A) (data not shown), which is in the range of the endogenous cGK I concentration in other types of human endothelial cells (5). In subsequent experiments, adenoviral vectors were usually used at concentrations of 10 10 particles/ml or less. At 10 10 particles/ml (Fig. 1A), a well known cGK I degradation product (65 kDa) was observed; and at higher levels of expression, additional smaller (most likely breakdown) products were detected. Expressed cGK I␤ was demonstrated to be functionally active by treatment of HUVECs with a cell membrane-permeant, hydrolysis-resistant, cGK-selective cGMP analog, 8-pCPT-cGMP, which activates the kinase. In response to 8-pCPT-cGMP, cGK I phosphorylated its well characterized substrate, VASP, as demonstrated by Western blot analysis (Fig. 1B). An antibody against total VASP (IE273) detected a shift in the electrophoretically determined size of VASP, from an apparent size of 46 kDa in unstimulated cells to 50 kDa in cGMP-stimulated, cGK I-expressing cells (Fig. 1B, first panel, sixth lane), a shift resulting from phosphorylation of VASP Ser 157 . This was confirmed using a newly developed antibody (5C6) specific for Ser 157 in its phosphorylated state (Fig. 1B, second panel). VASP phosphorylation on Ser 157 was also observed with uninfected HUVECs stimulated with 3 M forskolin, an activator of adenylate cyclase and thus cAK ( ther with a newly developed, catalytically inactive cGK I␤ mutant (cGK I␤-K405A) (Fig. 1B, third and fourth lanes; and see Fig. 4, first through third lanes) or with a luciferase control vector or in uninfected HUVECs (data not shown). These results clearly demonstrate that cGMP-dependent phosphorylation of VASP does not occur in the absence of cGK and therefore is not mediated by cAK. Similar results were obtained from in vitro kinase assays performed with lysates from cells treated as shown in Fig. 1B, i.e. cGMP-activable kinase activity was present only in cGK I␤-infected cells (data not shown). Polymerase chain reaction analysis also detected no cGK I mRNA in pure HUVECs uncontaminated by fibroblasts, which do contain cGK I (data not shown).
Mutant cGK I␤-K405A was not only itself catalytically inactive, but was also capable of inhibiting (Fig. 2B, third through sixth lanes) wild-type cGK I activity (second lane), as demonstrated in passage 8 human dermal fibroblasts containing endogenous cGK I (cGK I observed upon longer film exposure than that shown in Fig. 2A). VASP phosphorylation by 8-pCPT-cGMP-stimulated endogenous cGK I was completely inhibited by 5 ϫ 10 10 particles of mutant cGK I␤-containing adenoviral vector/ml, which produced a severalfold overexpression of the mutant kinase compared with the endogenous kinase. The mutated Lys 405 is the homolog of the essential Lys 72 in the ATP-binding site of cAK that is conserved in all protein kinases (39). The mutant's inhibitory effect on wild-type cGK I function could be abolished by raising the concentration of 8-pCPT-cGMP from 100 to 500 M (data not shown), suggesting that the inactive mutant competed with wild-type cGK I for the cGMP analog. In contrast, VASP phosphorylation in fibroblasts induced by 10 M prostaglandin E 1 , an agonist of endogenous cAK, was not inhibited by mutant cGK I␤ (data not shown).
Early Detachment of VASP from Focal Adhesions after cGK I␤ Activation in HUVECs-In cultured HUVECs, VASP is localized at focal adhesions, microfilaments, and cell-cell contacts (Fig. 3, A, C, and E) (5). In our present studies, VASP localization was studied by indirect immunofluorescence in HUVECs infected with the adenoviral vector for cGK I (Fig. 3, C-F). Cells were infected with 10 9 particles of viral vector for cGK I␤/ml and 24 h later were treated with 100 M 8-pCPT-cGMP for 30 min to activate the kinase, the procedure that produced maximal VASP phosphorylation (blot in Fig. 4). Subsequently, cells were immediately fixed for immunofluorescence. Double labeling with antibodies against cGK I and VASP demonstrated that VASP was no longer detectable at focal adhesions (Fig. 3C) in any cGK I-expressing HUVECs (bright cells in Fig. 3D) in nonconfluent monolayers. Instead, a diffuse VASP signal was observed in the cytosol, and some VASP staining was still detected along stress fibers, although it did not resemble the normal periodic pattern. Whereas complete loss of VASP from focal adhesions occurred at 30 min, a reduction in VASP at focal adhesions was observed already after 5 min of 30 M 8-pCPT-cGMP treatment, the earliest time point and lowest cGMP analog concentration at which maximal phosphorylation of VASP was detectable by immunoblotting (data not shown). This early loss appeared as a marked general reduction in focal adhesion labeling intensity, which progressed at later times to complete loss of labeling. The specificity of the effect of cGK I␤ expression and activation on VASP localization was demonstrated by the lack of any such response to 8-pCPT-cGMP in uninfected cells (Fig. 3A), in cells at the periphery in Fig. 3C corresponding to cells not expressing detectable cGK I␤ (Fig.  3D), or in cells infected with a control (luciferase) adenoviral vector (data not shown) or with an adenoviral vector for the catalytically inactive cGK I␤-K405A mutant (see Fig. 5A and 6A). Although cGK I activation caused a loss of VASP from focal adhesions in nonconfluent monolayers of HUVECs, it did not alter the predominant VASP localization at cell-cell contacts in confluent monolayers (Fig. 3E).
Loss of Zyxin and Vinculin from Focal Adhesions after cGK I␤ Activation in HUVECs-To investigate cGK-mediated rearrangement of focal adhesion proteins in more detail, we analyzed the localization of VASP as well as that of the established VASP-binding proteins zyxin and vinculin at different times of cGK activation. cGK was activated either directly with 8-pCPT-cGMP or indirectly with ANP, which elevates intracellular cGMP levels via the ANP receptor guanylyl cyclase. The phosphorylation of VASP in HUVECs expressing cGK I, incubated for 5 min to 3 h with 100 nM ANP or 100 M 8-pCPT-cGMP, is shown in Fig. 4. More important, no cGMP-dependent VASP phosphorylation was observed in HUVECs expressing the catalytically inactive cGK I␤-K405A mutant. In contrast to 8-pCPT-cGMP (hydrolysis-resistant) induction of constant maximal VASP phosphorylation for 3 h, ANP-induced VASP phosphorylation declined at 3 h. In parallel experiments, HU-VECs were double-labeled using antibodies against either VASP and vinculin or zyxin and vinculin for immunofluorescence analysis. The transfection efficiency was adjusted so that nearly all cells expressed cGK. After 30 min of cGK activation by ANP or 8-pCPT-cGMP, VASP (Figs. 5E and 6E) as well as zyxin (Figs. 7E and 8E) were detached from focal adhesions (quantitation in Fig. 9). Vinculin staining per focal adhesion was diminished after 30 min of cGK activation by either 8-pCPT-cGMP or ANP (Figs. 5F, 6F, 7F, and 8F), i.e. the size of vinculin-containing focal adhesions often appeared reduced, although the total number of cells exhibiting vinculin staining was not greatly changed at 30 min (Fig. 9). However, 3 h of 8-pCPT-cGMP treatment caused depletion of all three proteins (VASP, zyxin, and vinculin) from focal adhesions (Figs. 6, G-H;  8, G-H; and 9) and, at the same time, a general loss of fluorescein isothiocyanate-phalloidin staining of the microfilament system (data not shown). In contrast, 3 h of ANP treatment of HUVECs resulted in some VASP and zyxin relocalization to focal adhesions, and the size of vinculin-staining focal adhesions increased again (Figs. 5, G-H; 7, G-H; and 9). This recovery corresponds to the decline in ANP-stimulated cGK-dependent VASP phosphorylation (Fig. 4, upper panel). Thus, ANP caused a time-dependent and reversible depletion of primarily VASP and zyxin from focal adhesions. The effects of both 8-pCPT-cGMP and ANP were dependent on cGK since HUVECs transfected with the inactive cGK I␤-K405A mutant did not display any change in VASP phosphorylation (Figs. 1B and 4) or in VASP, zyxin, or vinculin localization in focal adhesions (Figs. 5-8, A and B).

Dependence of VASP Intracellular Redistribution on VASP
Phosphorylation by cGK I-The role of cGK I-dependent VASP phosphorylation in the observed redistribution of VASP was examined using VASP phosphorylation mutants in which Ser 157 , Ser 239 , and Thr 278 were replaced by Ala, either singly or by double mutation of both serines (AAT) or by triple mutation of all three phosphorylation sites (AAA). To distinguish both transfected wild-type and mutant VASPs from endogenous VASP present in HUVECs, the former were expressed as VSV epitope-tagged fusion proteins and detected using an antibody directed against VSV in both Western blotting and immunofluorescence.
Efficacious VASP phosphorylation site mutation was demonstrated by expressing wild-type or mutant VASP in HEK 293 cells, immunoprecipitating them, and determining their ability to be phosphorylated by purified cGK I or cAK in vitro using [␥-32 P]ATP. The anti-VSV antibody detected essentially equal amounts of all VSV-VASP forms on Western blots of the phosphorylated samples (Fig. 10A); however, phosphate incorporation (Fig. 10B) into the double mutant (AAT, Ser 157 and Ser 239 both changed to Ala) was severely impaired, and no significant phosphate incorporation into the triple mutant protein (AAA) was observed.
Cotransfection of plasmids of cGK I and VSV-tagged wildtype VASP into HUVECs, followed by 30 min of 8-pCPT-cGMP stimulation of kinase, led to detachment of VASP from focal adhesions (Fig. 11C). In contrast to the previously described experiments using adenoviral vectors, in which virtually all cGK I-expressing cells demonstrated cGMP-dependent VASP loss from focal adhesions under the same conditions used here, experiments using plasmid vectors demonstrated that only ϳ43% of cGK I-expressing cells lost VSV-tagged wild-type VASP from focal adhesions. However, a similar pattern of VASP relocalization was observed as in the adenoviral vector experiments, i.e. a diffuse VASP staining in the cytosol sometimes appearing as micro-aggregates, with residual stress fiber staining devoid of periodic character. Mutant VASP could localize to focal adhesions (Fig. 11E), indicating that this basic property was not disturbed by the mutations. However, after stimulation of cotransfected cGK I, the triple phosphorylation mutant of VASP was more resistant to detachment from focal adhesions (lost from only 11 Ϯ 7% of cGK I-expressing cells) than was wild-type VASP (lost from 43 Ϯ 9% of cells) (Fig. 11, G and C, respectively; percentages in text represent means Ϯ S.E. of four independent experiments in which 50 -100 cells were analyzed per experiment). The detachment of the double mutant of VASP from focal adhesions was not observably different from that of wild-type VASP in response to 8-pCPT-cGMP (data not shown).
Inhibition of HUVEC Migration by cGK I␤-The functional consequence of cGK I activation on HUVEC migration was examined. In a haptotactic migration assay, HUVEC migration on fibronectin (Fig. 12A) as well as on collagen and fibrinogen matrices (Fig. 12B) was inhibited by expression and activation of cGK I␤, but not the catalytically inactive cGK I␤-K405A mutant. This was the case when the expressed kinase was activated either directly with 100 M 8-pCPT-cGMP or indirectly with 500 nM ANP. The inhibitory effect of ANP in the 5-h migration assay was slightly weaker than that of 8-pCPT-cGMP, perhaps reflecting the observed decline in ANP activation of cGK after 3 h (Fig. 4). Our studies demonstrate for the first time in endothelial cells that cGK, and not cAK, mediates cGMP-dependent inhibition of migration since the catalytically inactive cGK I␤-K405A mutant did not inhibit migration. Collagen and fibrinogen are known to activate the migration of HUVECs via integrins ␣ 2 ␤ 1 and ␣ 5 ␤ 1 , respectively (40,41). Fibronectin binds to integrins ␣ 5 ␤ 1 , ␣ v ␤ 3 , and ␣ 4 ␤ 1 on endothelial cells (42). Thus, cGK I inhibition of haptotactic cell migration was independent of the integrins involved.

DISCUSSION
Expression and activation of cGK I␤ (but not a catalytically inactive cGK I␤-K405A mutant) in primary cultures of HU-VECs resulted in depletion of several cytoskeletal proteins from focal adhesions, beginning with VASP and zyxin within 30 min and later vinculin, followed by microfilament disruption, thus causing a destabilization of motility structures that may contribute to the observed cGK I-dependent inhibition of HU-VEC migration. Of the cytoskeletal proteins affected, only VASP is a known cGK I substrate, suggesting that perhaps VASP phosphorylation leads to the subsequent chain of events. In the past, few effects of VASP phosphorylation have been reported. The localization of the Ser 157 phosphorylation site in the central domain of VASP, in close proximity to the binding site for profilin, suggested an influence of VASP phosphorylation on VASP/profilin interaction; however, in vitro binding of VASP to matrix-bound profilin was not grossly altered by VASP phosphorylation (22). Recently, only VASP phosphorylation on Ser 157 was analyzed and suggested to enhance VASP affinity for F-actin in vitro (20); however, the contribution of the other VASP phosphorylation sites was not investigated.
In this study, all three VASP phosphorylation sites were investigated in intact cells using a combination of site-specific antibodies and site-specific mutation. In particular, VSVtagged VASP mutants that could be expressed and distinguished from endogenous VASP were used to demonstrate that direct VASP phosphorylation was required for VASP detachment from focal adhesions since the VSV-tagged VASP-AAA triple phosphorylation mutant was highly resistant to 8-pCPT-cGMP-dependent detachment from focal adhesions in comparison with wild-type VASP. However, in the absence of 8-pCPT-cGMP, all phosphorylation site mutants of VASP were targeted to focal adhesions and showed a periodic distribution along stress fibers, suggesting that the mutations did not induce any major changes in the tertiary structure of VASP. VASP detachment from focal adhesions was inhibited only if all three phosphorylation sites were mutated, suggesting that all three sites are important for regulation of VASP interaction with cytoskeletal components. Ser 157 is located just distal to the VASP EVH1 domain, and Ser 239 and Thr 278 are located in the VASP EVH2 domain (12,23). Specific functions ascribed to these domains are consistent with our results that point mutations of phosphorylation sites close to and within these domains affect VASP localization at focal adhesions. VASP truncation and other experiments have shown that both the EVH1 and EVH2 domains are involved in VASP localization at focal adhesions and that the EVH2 domain affects the function and binding properties of the EVH1 domain (discussed in Ref. 23). Addi- tional data clearly demonstrated that both EVH1 and EVH2 domains of Ena/VASP family members are essential but alone not sufficient for the overall function of these proteins (43). Furthermore, during the review process for our manuscript, another publication appeared demonstrating that expression of the EVH2 domain in keratinocytes inhibited VASP function in adhesion zipper and epithelial sheet formation; and in keratinocytes of transgenic mice overexpressing the EVH2 domain, abnormalities in intercellular adhesion were observed (44). Thus, it is highly conceivable that Ser 278 phosphorylation, in combination with Ser 157 and Ser 239 phosphorylation, could affect important VASP functions, ultimately resulting in a major effect like VASP displacement from focal adhesions.
The exact role of each individual VASP phosphorylation site in either VASP loss from or recovery to focal adhesions is not clear. It was noted, however, that after 3 h of ANP treatment, VASP localization at focal adhesions began to recover concomitant with a relative loss of Ser 157 phosphorylation. This was not observed after 3 h of 8-pCPT-cGMP treatment, most likely due to the greater metabolic stability of this cGMP analog in comparison with unmodified cGMP formed in response to ANP. It is possible that incomplete VASP phosphorylation is insufficient for initiating VASP removal from focal adhesions and perhaps sufficient for its recovery. All available data support the interpretation that all three VASP phosphorylation sites are critical for the interaction of VASP with focal adhesion components. This complexity is perhaps not surprising since VASP interacts via its EVH1 and EVH2 domains with at least two distinct components present in focal adhesions, FP 4 motif-containing proteins (e.g. zyxin and vinculin) and F-actin, as discussed below.
In comparison with VASP, VASP-related proteins show conservation of only Ser 157 and Ser 239 (Mena), only Ser 157 (Evl), or no sites (Ena), suggesting that VASP homologs may display different regulation in response to phosphorylation. Phosphorylation of these sites has not been extensively studied, but a cAK-induced shift in Mena migration in SDS-polyacrylamide gel electrophoresis has been shown (14), and our 16C2 antibody detects phosphorylation of Mena, probably at Ser 376 homologous to VASP Ser 239 . 3 Interestingly, tyrosine phosphorylation of Ena on other sites by Abl tyrosine kinase attenuates Ena association with profilin or SH3 domain-containing proteins, and mutation of these phosphorylation sites partially impairs the in vivo function of Ena, indicating that phosphorylation is required for optimal Ena function (45). Phosphorylation state has also been shown to regulate focal adhesion localization of other proteins. Paxillin dephosphorylation on tyrosine resulted in its loss from focal adhesions in response to cAMP (46), whereas serine and threonine phosphorylation has been shown to accelerate fibronectin-induced localization of paxillin to focal adhesions (47). In contrast, serine phosphorylation of talin in response to interleukin-1␤ decreased talin's focal adhesion localization (48), and phorbol ester stimulation of protein kinase C increased talin serine phosphorylation and talin removal from focal adhesions (49). Thus, cGK I regulation of focal adhesion constituents exemplifies "inside-out" regulation of adhesion sites by yet a new signal transduction pathway. Our results in intact cells provide a physiological correlate to the previous demonstrations of in vitro interactions between VASP and either zyxin or vinculin (26,32). These interactions were demonstrated using one solid phase (nitrocellulose, column, or microtiter plate)-bound partner to bind another from solution. More recently, VASP and vinculin have also been shown to co-immunoprecipitate after in situ cross-linking experiments (50). The direct binding partner that attaches VASP to focal adhesions is unclear, although vinculin and zyxin are likely candidates since they have been shown to directly interact, via their FP 4 motifs, with the N-terminal EVH1 domain of proteins of the Ena/VASP protein family (15,26,32), and injection of peptides containing the FP 4 motif into human fibroblasts and HeLa cells caused a depletion of VASP from focal adhesions, resulting in a diffuse cytosolic distribution of VASP (15,17). Although the role of the Ena/VASP family EVH2 domain is not entirely clear, it does appear to regulate tetramerization, which is important for EVH1 domain interactions with other proteins, including ones present at focal adhesions (23). Detachment of VASP and zyxin from focal adhesions may also have additional cellular consequences since zyxin has been shown to shuttle between focal contacts and the nucleus in fibroblasts (51).
Prolonged effects of 8-pCPT-cGMP treatment were also observed after VASP phosphorylation and disappearance from focal adhesions. A progressive reduction of the actin microfilament system and vinculin at focal adhesions was observed, which was complete after 3 h of 8-pCPT-cGMP treatment. VASP influence on the microfilament system may be related to the role shown for VASP in enhancing actin assembly required for Listeria motility (15)(16)(17)(18)(19)(20)(21). FP 4 domain-containing proteins were also implicated in bacterial motility since motility was blocked by peptides containing the FP 4 motif (52). In mammalian aortic smooth muscle and endothelial cells, disassembly of focal adhesions induced by thrombospondin and tenascin has recently been suggested to be mediated by cGK I, although GMP-dependent protein kinase I alone had no effect (53). In our studies, cGK I stimulation resulted in loss of proteins from focal adhesions, and this was not mimicked by a catalytically inactive cGK I␤-K405A mutant. This mutant is not only itself inactive, but could also compete with wild-type cGK I for the cGMP analog. Although it was not clear whether mutant and wild-type cGKs I could form heterodimers, this did not seem to be the case since a large ratio of mutant to wild-type cGK I was required for inhibition of wild-type cGK I, and inhibition could be overcome by increasing the cGMP analog concentration. Furthermore, a heterodimer holoenyzme might still display activity since even cGK I monomers demonstrate catalytic activity (54).
cGK I effects on focal adhesion integrity may also underlie the observation that cGK I and VASP have been implicated in inhibition of fibrinogen binding to its integrin receptor, ␣ IIb ␤ 3 . (7-9). The time course and concentration of cGMP that caused VASP phosphorylation were very similar to those that inhibited fibrinogen binding. However, VASP phosphorylation may be only one of the effects of cGK I on the cytoskeletal system. cGK I in platelets is known to inhibit phospholipase C and myosin light chain (MLC) phosphorylation (10,55), which are suggested to be involved in tension-and stress fiber-mediated focal adhesion assembly (56,57). Inhibition of MLC phosphorylation is unlikely to be mediated by cGK I phosphorylation of MLC kinase since this did not affect MLC kinase activity (58). Alternatively, a role for cGK in MLC phosphatase activation has been discussed (59 -61). Another potential target of cGK could be the small GTP-binding protein RhoA, which has been shown to mediate the thrombin-induced increase in MLC phosphorylation in endothelial cells (62). RhoA can be inhibited by phosphorylation through cAK, which correlates with an inhibition of lymphocyte motility by cAMP (63).
The regulation of focal adhesions is intimately involved in the process of cell migration, which is a dynamic process involving polarized formation and disassembly of focal adhesions at opposite cell ends (64,65). Lack of focal adhesions and stress fibers in Chinese hamster ovary cells expressing a mutant integrin resulted in decreased migration in haptotactic assays like that used here, although the cells were motile in a random migration assay (65). However, others have reported that decreasing vinculin and ␣-actinin via antisense RNA increased 3T3 cell motility in random migration assays (66,67). In our experiments on haptotaxis, 8-pCPT-cGMP-dependent vinculin loss was observed at the time of inhibition of HUVEC migration. The cGK I-dependent reorganization of focal adhesion proteins and ultimate destabilization of the microfilament system that we observed may facilitate cGK I-dependent inhibition of cell migration. This inhibition was independent of which extracellular matrix or integrins were involved. In agreement with our observations, Ikeda et al. (68) demonstrated that FIG. 12. Inhibition of HUVEC migration by cGK I␤ expression from adenoviral vectors. HUVECs were uninfected (control) or infected with equal amounts (5 ϫ 10 9 particles/ml) of Ad5-cGK I␤ or the catalytically inactive Ad5-cGK I␤-K405A mutant. One day later, HU-VECs were analyzed in migration assays using modified Boyden chambers. Shown is the migration in response to the following extracellular matrix proteins: A, fibronectin; B, collagen (left) and fibrinogen (right). In the presence of either 100 M 8-pCPT-cGMP (black bars) or 500 nM ANP (hatched bars) to activate cGK I␤, HUVEC migration was ϳ1.5-2-fold less than that observed for untreated cells (white bars). In contrast, the migration of HUVECs expressing mutant cGK I␤-K405A was not significantly inhibited in response to 8-pCPT-cGMP. Migration of control uninfected cells (A) in the absence of 8-pCPT-cGMP or ANP (open bar) was designated as 100% for normalizing data for cell migration in response to fibronectin (similar control data for collagen and fibrinogen are not shown). Each bar represents the mean Ϯ S.E. of three individual experiments. migration of rat aorta endothelial cells was inhibited by 8-Br-cGMP as well as by natriuretic peptides that elevate cGMP. However, focal adhesions and the presence of endogenous cGK I were not investigated. The migration of smooth muscle cells stably transfected with cGK I was also inhibited by cGMP analogs (69). Furthermore, overexpression of endothelial nitricoxide synthase in balloon-injured carotid arteries resulted in impaired endothelial regeneration (70), consistent with possible nitric oxide (or cGMP)-dependent inhibition of endothelial cell proliferation and migration.
Collectively, our data demonstrate that cGMP has effects on VASP phosphorylation and localization and on cell migration that are mediated by cGK and not cAK. Use of a combination of VASP phosphorylation site-specific antibodies and mutations demonstrated the importance of cGK I-dependent VASP phosphorylation in inhibiting VASP localization at focal adhesions. Continued studies are necessary to more closely define the farreaching consequences of cGK I-dependent inhibition of focal adhesion integrity and cellular processes that depend on them.