ERK1 Associates with αvβ3 Integrin and Regulates Cell Spreading on Vitronectin*

Kinases that associate with integrins are likely to mediate the assembly/disassembly of cell:matrix junctions during cell migration. Here we show that ERK1 associates with αvβ3 integrin following the addition of platelet-derived growth factor to serum-starved Swiss or NIH 3T3 fibroblasts in an interaction that is mediated by the central region of the β3 integrin cytodomain. αvβ3·ERK1 association occurred prior to focal complex formation and was seen to initiate in small punctate complexes primarily in the peripheral regions of the plasma membrane. Expression of a dominant negative mutant of ERK1 (but not ERK2) significantly reduced the spreading of cells on vitronectin, whereas cell spreading on fibronectin was unaffected by inhibition of ERK1. In contrast, inhibition of ERK activation by PD98059 had no effect on the platelet-derived growth factor-regulated Rab4-dependent flux of αvβ3integrin from early endosomes to the plasma membrane, an event that is also necessary for cells to spread efficiently on vitronectin. We propose that αvβ3 integrin must recycle to the plasma membrane via the Rab4 pathway and recruit active ERK1 in order to function efficiently.

The integrins have received a great deal of attention in the literature over the last few years, not only for their ability to bind and model the extracellular matrix, but also due to their ability to activate a number of cell signaling cascades that influence a range of biological processes including cell growth, differentiation, migration, and apoptosis (1). The distal stretches of many of these signaling pathways are reasonably well defined. For instance, ligation of integrins is well established to both activate the MEK/ERK 1 signaling axis and to prolong the activation of the pathway in response to growth factors and thereby confer anchorage-dependent growth (2). However, precise descriptions of the upstream events in integrin signaling remain elusive.
Much of the evidence that kinases associate physically with integrins is controversial. The integrin-linked kinase was iden-tified as a ligand for the cytodomain of ␤ 1 integrin by yeast two-hybrid analysis, and engagement of ␤ 1 integrins has been shown to activate integrin-linked kinase (3). However, recent work in Drosophila suggests that the main function of integrinlinked kinase is that of an adaptor rather than a kinase (4). On the other hand pp125 FAK is well established to be activated by ligation of integrins and to be a key activator of numerous downstream signaling cascades (5), but it is unclear whether pp125 FAK associates physically with integrins. Recently, however, good data have been obtained indicating that the tyrosine kinase, Syk, associates with ␤ 3 integrin cytodomains, and activation of Syk by clustering of ␣ IIb ␤ 3 integrin is likely to be a key upstream event in the activation of platelets and locomotion of hemopoietic cells (6).
The transmembrane and extracellular domains of integrins also form so-called "lateral associations," and these are proposed to couple integrins to key signaling kinases. The ␣ 5 ␤ 1 , ␣ v ␤ 3 , and ␣ 1 ␤ 1 heterodimers associate with caveolin (7), most probably via their transmembrane stretches. There is no evidence that this interaction is direct, but it is likely to be important in linking integrins to the tyrosine kinase, Fyn, and the adaptor protein, Shc. The recruitment of Shc may be a critical link between integrins and the Ras/Raf/MEK/ERK signaling cascade and the regulation of anchorage-dependent growth (7). Integrins are now well established to form stable complexes with proteins of the transmembrane 4 superfamily (8). These tetraspannins have been shown to bind tyrosine kinases (9), phosphotidylinositol 4-kinase (10), and conventional protein kinase C (11), and the observed modulatory effects of integrintetraspannin complexes on adhesion-dependent signaling may involve these associations.
Prior to engagement with the extracellular matrix, integrins must be rendered competent to bind ligand via a process termed "inside-out" signaling (12). There are several ways in which the availability of ligand-competent integrin may be increased. Growth factors have recently been shown to regulate delivery of ␣ v ␤ 3 from endosomal compartments to the plasma membrane, and this process is necessary for efficient integrin function (13). Once upon the cell surface, conformational changes in an individual heterodimer may increase its affinity for monovalent ligand (12). This is well documented to be a key event in the activation of ␣ IIb ␤ 3 following treatment of platelets with thrombin. In addition to affinity modulation, the avidity of an integrin for a multivalent matrix ligand may be increased by regulating the clustering of active heterodimers (14). For example, clusters of ␣ v ␤ 3 integrin form at early stages during cell spreading and clearly prior to the assembly of focal complexes (15). Several growth factor-activated signaling pathways have been implicated in the activation of integrins prior to focal complex assembly. For instance, the small GTPase, Rab4 regulates delivery of ␣ v ␤ 3 to the plasma membrane (13); phosphatidylinositol 3-kinase (16) and H-Ras (17) are clearly in-volved in integrin affinity modulation; and the calciumdependent protease, calpain, has recently been shown to mediate integrin clustering (15). However, the overall picture is far from complete.
Clearly, any kinase or other signaling protein found to associate with an integrin shortly following cell activation but prior to its incorporation into focal complexes would be potentially of interest as a possible regulator of integrin function. To identify these proteins, we immunoprecipitated integrins from fibroblasts shortly following activation with platelet-derived growth factor (PDGF), and screened them for associated proteins that were rich in phosphotyrosine and phosphothreonine. We report that following PDGF addition, active ERK1 is an abundant component of ␣ v ␤ 3 integrin immunoprecipitates. The association of ERK with ␣ v ␤ 3 forms in plasma membrane complexes prior to delivery of integrin to focal complexes. Moreover, we find that association of active ERK1 with ␣ v ␤ 3 integrin is necessary for cells to spread effectively on vitronectin and thus may define a mechanism whereby a growth factor-activated signaling pathway can directly influence integrin function.
Cell Culture and Transfection-Swiss and NIH 3T3 mouse fibroblasts were grown in Dulbecco's modified Eagle's medium (DMEM) with 10% FCS and 100 units/ml penicillin, 100 g/ml streptomycin, and 0.25 g/ml amphotericin B at 37°C with 10% CO 2 . For transient transfection experiments, NIH 3T3 fibroblasts were grown to 50% confluence, fed with fresh DMEM containing 10% fetal calf serum, and transfected with integrin, ERK, and Rab constructs (integrin and Rab cDNAs were ligated into in pcDNA3 and are as described in Ref. 13; His-ERK1, HisERK1KϾR, HisERK2, and His-ERK2KϾR were in pCMV5 and are as described in Ref. 18) using Fugene 6 according to the manufacturer's instructions. The ratio of Fugene 6 to DNA was maintained at 3 l of Fugene 6:1 g of DNA. Immunoprecipitations, integrin recycling, and cell spreading assays were carried out 24 h post-transfection.
Expression and Purification of GST-Integrin Cytodomain Fusion Proteins-PCR-amplified DNA fragments corresponding to aa 728 -762, 728 -756, 728 -748, and 728 -741 of the human sequence of ␤ 3 integrin and to aa 764 -798 of the human sequence of ␤ 1 integrin were subcloned into the BamHI-EcoRI site of the pGEX-2TK vector. GST fusion proteins were expressed in Escherichia coli strain BL-21 and purified as described previously (19).
Immunoprecipitations and Pull-downs-Cells were grown to 90% confluence, serum-starved for 30 min, and treated with PDGF (10 ng/ml), epidermal growth factor (30 ng/ml), or lysophosphatidic acid (1 g/ml) where appropriate. Following this, cells were washed twice in ice-cold PBS and lysed in a buffer containing 200 mM NaCl, 75 mM Tris-HCl pH 7, 15 mM NaF, 1.5 mM Na 3 VO 4 , 7.5 mM EDTA, and 7.5 mM EGTA, 0.5% (v/v) Triton X-100, 0.25% (v/v) Igepal CA-630, 50 g/ml leupeptin, 50 g/ml aprotinin, and 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride (1.14 l/cm 2 culture area; giving a final concentration of 0.3% (v/v) nonionic detergent following dilution of the lysis buffer in the PBS wetting the cells) and scraped from the dish with a rubber policeman. Lysates were passed three times through a 27-gauge needle and clarified by centrifugation at 10,000 ϫ g for 10 min at 4°C. For immunoprecipitations, magnetic beads conjugated to sheep anti-mouse IgG were blocked in PBS containing 0.1% (w/v) BSA and then bound to anti-integrin or anti-His 6 monoclonal antibodies. For pull-downs, antimouse magnetic beads were bound to mouse anti-rabbit IgG, followed by rabbit anti-GST and finally GST or GST-integrin cytodomain fusion proteins as appropriate. Antibody and fusion protein-coated beads were incubated with lysates for 2 h at 4°C with constant rotation. Unbound proteins were removed by extensive washing in lysis buffer, and specifically associated proteins were eluted from the beads by boiling for 10 min in Laemmli sample buffer. Proteins were resolved by SDS-PAGE (8% gels under reducing conditions for ERKs, the hexa-His epitope, phosphotyrosine, and phosphothreonine; 6% gels under nonreducing conditions for integrins) and analyzed by Western blotting as described previously (20).
Integrin Recycling Assay-This was performed as described previously in (13). Cells were serum-starved for 30 min, transferred to ice, washed twice in cold PBS, and surface-labeled at 4°C with 0.2 mg/ml NHS-SS-biotin in PBS for 30 min. Labeled cells were transferred to serum-free DMEM at 22°C for 15 min to allow internalization of tracer into early endosomes. Cells were returned to ice and washed twice with ice-cold PBS, and biotin was removed from proteins remaining at the cell surface by incubation with a solution containing 20 mM sodium 2-mercaptoethanesulfonate (MesNa) in 50 mM Tris, pH 8.6, and 100 mM NaCl for 15 min at 4°C. The internalized fraction was then chased from the cells by returning them to 37°C in serum-free DMEM in the absence or presence of 10 ng/ml PDGF-BB for 10 min. Cells were returned to ice, and biotin was removed from recycled proteins by a second reduction with MesNa. MesNa was quenched by the addition of 20 mM iodoacetamide for 10 min, and the cells were lysed in 200 mM NaCl, 75 mM Tris, 15 mM NaF, 1.5 mM Na 3 VO 4 , 7.5 mM EDTA and 7.5 mM EGTA, 1.5% (v/v) Triton X-100, 0.75% (v/v) Igepal CA-630, 50 g/ml leupeptin, 50 g/ml aprotinin, and 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride. Lysates were passed three times through a 27-gauge needle and clarified by centrifugation at 10,000 ϫ g for 10 min. Supernatants were corrected to equivalent protein concentration, and levels of biotinylated integrin were determined by capture enzyme-linked immunosorbent assay. Maxisorb 96 well plates were coated overnight with 5 g/ml anti-␤ 3 integrin monoclonal antibodies in 0.05 M Na 2 CO 3 , pH 9.6, at 4°C and blocked in PBS containing 0.05% (v/v) Tween 20 (PBS-T) with 5% (w/v) BSA for 1 h at room temperature. Integrins were captured by overnight incubation of 50 l of cell lysate at 4°C. Unbound material was removed by extensive washing with PBS-T, and wells were incubated with streptavidin-conjugated horseradish peroxidase in PBS-T containing 1% (w/v) BSA for 1 h at 4°C. Following further washing, biotinylated integrins were detected by chromogenic reaction with ortho-phenylenediamine.
Immunofluorescence-Cells were plated onto glass coverslips and grown to 50 -70% confluence over 3 days. Cells were serum-starved for 30 min and treated with 10 ng/ml PDGF-BB for the indicated times prior to fixation in PBS containing 2% (w/v) paraformaldehyde for 20 min at room temperature. Nonspecific binding sites were blocked with PBS containing 10% (v/v) fetal calf serum (PBS-FCS) for 1 h, and cells were incubated with hamster anti-m␤ 3 monoclonal antibody at 5 g/ml in PBS-BSA at room temperature for 1 h. Following this, cells were permeabilized with 0.2% (v/v) Triton X-100 in PBS for 5 min and then reblocked in PBS-FCS. Integrin was detected by sequential application of a rabbit anti-hamster secondary antibody, followed by fluorescein isothiocyanate-conjugated goat anti-rabbit tertiary antibody. The cells were counterstained for ERKs using mouse anti-ERK1/2 monoclonal antibodies followed by a Texas Red-conjugated goat anti-mouse secondary antibody. Where appropriate, the actin cytoskeleton was counterstained with Texas Red-conjugated phalloidin in PBS for 10 min at room temperature. Cells were viewed on a Leica confocal laser-scanning microscope, and images were presented either as an extended focus projection (using the Leica "maxproj" algorithm) of a number of optical slices encompassing the depth of the cell (as for Figs. 4 and 6) or a single Z-section optical slice (as for Fig. 5).
Cell Adhesion and Spreading Assays-24-well tissue culture plates were coated overnight at 4°C with fibronectin (F-1141; Sigma) or vitronectin (V-8379; Sigma) at concentrations of 20 g/ml and then blocked with 2% (w/v) BSA. Cells were transfected with Rab4 or His-ERK constructs in conjunction with a ␤-galactosidase-expressing marker construct and, 24 h following transfection, were harvested by trypsinization and collected by centrifugation in the presence of 20 g/ml soyabean trypsin inhibitor. The cell suspensions were added immediately to ligand coated wells in serum-free DMEM containing 10 ng/ml PDGF-BB in the presence and absence of 12 M PD98059. Cells were allowed to attach for 60 min, and nonadherent cells were removed by washing six times with PBS. Attached cells were fixed for 1 min in 0.2% glutaraldehyde containing 5 mM EGTA, and ␤-galactosidase-expressing cells were visualized by incubation with 5 mM potassium ferricyanide and 1 mg/ml X-gal (5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside) overnight at 37°C. To obtain an index of cell spreading, the area of cells expressing ␤-galactosidase was determined by delineation of the cell envelope using the NIH Image software.

RESULTS
Active ERK1 Is Associated with ␣ v ␤ 3 Integrin-Swiss 3T3 fibroblasts were serum-starved for 30 min and then stimulated with 10 ng/ml PDGF for 10 min or allowed to remain quiescent. Cells were immediately cooled to 4°C and surface-labeled with NHS-SS-Biotin. Labeled cells were lysed in a buffer containing 0.5% (v/v) Triton X-100 and 0.25% (v/v) Igepal, and ␣ 5 ␤ 1 and ␣ v ␤ 3 integrin heterodimers were immunoprecipitated from lysates using monoclonal antibodies to either the mouse ␣ 5 or mouse ␤ 3 integrin chains, respectively. PDGF increased the surface expression of ␣ v ␤ 3 (but not ␣ 5 ␤ 1 ) by ϳ2-fold (Fig. 1A), and this is consistent with our previous work documenting the rapid stimulation of ␣ v ␤ 3 integrin recycling following growth factor addition (13).
To gain insight into key biochemical events that occur during the early stages of focal complex assembly, we investigated the complement of cell-signaling proteins that coimmunoprecipitate with integrins shortly following treatment of cells with growth factor. Therefore, integrin immunoprecipitates were analyzed by Western blotting with antibodies recognizing phosphotyrosine and phosphothreonine. In serum-starved cells, low levels of phosphotyrosine-or phosphothreonine-containing proteins coimmunoprecipitated with either ␣ 5 ␤ 1 or ␣ v ␤ 3 integrins (Fig. 1, B and C). However, following the addition of PDGF, a 44-kDa protein (p44) that was rich in both phosphotyrosine and phosphothreonine was particularly abundant in immunopre-FIG. 1. Active ERK1 is associated with endogenous mouse ␣ v ␤ 3 integrin in Swiss 3T3 fibroblasts. Swiss 3T3 fibroblasts were serum-starved for 30 min and incubated in the presence and absence of 10 ng/ml PDGF-BB (A-D, F), 30 ng/ml epidermal growth factor, or 1 g/ml lysophosphatidic acid (F) for 10 min or treated with 10 ng/ml PDGF-BB for the times indicated (E). Cells were surfacelabeled with 0.2 mg/ml NHS-SS-biotin for 30 min at 4°C and lysed in a buffer containing 0.5% (v/v) Triton X-100 and 0.25% (v/v) Igepal CA-630. Lysates were immunoprecipitated (IP) with monoclonal antibodies against mouse ␣ 5 (m␣5) and ␤ 3 (m␤ 3 ) integrins. Immobilized material was then analyzed by Western blotting with peroxidase-conjugated streptavidin (SA-HRP) (A), anti-phosphotyrosine (PY) (B), anti-phosphothreonine (PT) (C), anti-ERK1/2 (D-F), and anti-phospho-ERK1/2 (D). The positions of ␣ 5 , ␣ v , ␤ 1 , and ␤ 3 integrin chains, immunoglobulin heavy chain (IgG HC), and ERK1/2 are indicated.
cipitates of ␣ v ␤ 3 integrin (Fig. 1, B and C). p44 was only present at low levels in ␣ 5 ␤ 1 immunoprecipitates, and, moreover, there was no phosphotyrosine signal at 120 or 70 kDa, indicating that pp125 FAK and paxillin were not associated with either integrin. We were, however, able to coimmunoprecipitate small quantities of a phosphotyrosine-containing protein of 190 kDa with ␣ v ␤ 3 , and this is likely to represent association of the PDGF receptor with the integrin (21).
A previous report has documented the presence of the ERKs in newly forming focal complexes (22). Both p44 ERK1 and p42 ERK2 would be expected to be rich in phosphotyrosine and phosphothreonine following cell activation, so we tested for the presence of these kinases in the integrin immunoprecipitates. Western blotting with an antibody that recognizes both p44 ERK1 and p42 ERK2 revealed that p44 comprised ERK1, and this was associated with ␣ v ␤ 3 only following PDGF treatment (Fig. 1D). A small quantity of ERK1 was found to be associated with ␣ 5 ␤ 1 integrin; however, this was not increased by the addition of PDGF (Fig. 1D). Interestingly, only relatively small amounts of p42 ERK2 were associated with ␣ v ␤ 3 , indicating that recruitment of ERK1 was isoform-specific. ERK1 must be phosphorylated on both threonine 202 and tyrosine 204 to be active, and the presence of signals for phosphotyrosine and phosphothreonine in p44 implied that it was indeed active ERK1. This was confirmed by Western blotting using a phosphospecific antibody that recognized ERKs only when phosphorylated at both of these positions (Fig. 1D). The phosphotyrosine-containing protein at 60 kDa (p60; Fig. 1B) is likely to represent a Src family kinase. We are currently investigating the identity of p60 and the significance of its association with We investigated the time course over which the ERK1⅐␣ v ␤ 3 complex was established. Tyrosine phosphorylation of protein bands corresponding to the PDGF receptor (190 kDa), pp125 FAK (125 kDa), and paxillin (70 kDa) was maximal ϳ4 min after PDGF addition and subsided over the following 12 min (Fig. 1E, upper panel). Recruitment of ERK1 to ␣ v ␤ 3 integrin was slower than this and was maximal at 8 min following PDGF addition (Fig. 1E, lower panel). The association persisted for at least 16 min following PDGF addition (Fig. 1E) but abated somewhat over the following hour (data not shown). PDGF, lysophosphatidic acid, and epidermal growth factor were all able to induce substantial increases in tyrosine phosphorylation of a number of cellular proteins (notably of a band corresponding to pp125 FAK ) (Fig. 1F, upper panel). However, of the growth factors tested in the present study, only PDGF was able to elicit appreciable recruitment of ERK1 to ␣ v ␤ 3 (Fig. 1F,  lower panel). Increased tyrosine phosphorylation of a 36-kDa protein (marked p36 in Fig. 1, E and F, upper panels) was PDGF-specific and, moreover, occurred at a rate that paralleled recruitment of ERK1 to ␣ v ␤ 3 . Preliminary data indicate that p36 is likely to be annexin II (not shown), and the possible role of this protein in the assembly of integrin-containing complexes is discussed later.
To determine whether ERK1 was also able to associate with human ␣ v ␤ 3 and to confirm that its presence in mouse ␣ v ␤ 3 immunoprecipitates was not an artifactual characteristic of the antibody employed, human integrins were transfected into NIH 3T3 fibroblasts, and ␣ v ␤ 3 was immunoprecipitated with a monoclonal antibody that was specific for the human ␤ 3 integrin chain (h␤ 3 ). Surface labeling indicated that this antibody did not precipitate mouse ␣ v ␤ 3 ( Fig. 2A), and accordingly we were unable to detect any phosphotyrosine/phosphothreoninecontaining proteins or ERKs associated with anti-human ␤ 3 monoclonal antibody-coated beads when they were incubated with lysates prepared from mock-transfected cells even following PDGF treatment (Fig. 2, B-D). Following transfection of NIH 3T3 fibroblasts with the human ␣ v and ␤ 3 integrin chains, surface-labeled proteins corresponding to the ␣ v ␤ 3 heterodimer were prominent in the immunoprecipitates. These displayed increased surface expression following the addition of PDGF ( Fig. 2A), although Western blotting with an antibody recognizing the h␤ 3 chain revealed that the total quantity of integrin present in the immunoprecipitates was unaffected ( Fig. 2A). Moreover, in transfected cells, a profile of phosphotyrosine/ phosphothreonine-containing proteins, similar to that found associated with the mouse integrin, were coimmunoprecipitated with human ␣ v ␤ 3 following the addition of PDGF (Fig. 2, B and C), and Western blotting revealed that the 44-kDa band (p44) comprised active ERK1 (Fig. 2D).
To confirm the ability of ␣ v ␤ 3 integrin to recruit ERK1 in preference to ERK2 and to further test the specificity of this interaction, we performed immunoprecipitation studies on NIH 3T3 fibroblasts following transient expression of epitopetagged ERKs. Following the addition of PDGF, recruitment of His-ERK1 to ␣ v ␤ 3 immunoprecipitates was seen using a monoclonal anti-His 6 antibody to detect the epitope-tagged protein (Fig. 2E). Furthermore, Western blotting with anti-h␤ 3 indicated that ␣ v ␤ 3 integrin was precipitated by immunoisolation of His-ERK1 using magnetic beads conjugated to anti-His 6 (Fig. 2F). In contrast, following transient expression of His-ERK2, this epitope-tagged kinase was not detected in integrin immunoprecipitates (Fig. 2E); nor did ␤ 3 integrin co-precipitate with His-ERK2 (Fig. 2F). It is important to note that in these experiments cellular expression of His-ERK1 was somewhat greater than His-ERK2. This discrepancy was only observed when the ERKs were coexpressed with ␣ v ␤ 3 integrin (compare Fig. 2, E and F, with the expression levels of His-ERKs in Fig  9C, where no exogenous ␣ v ␤ 3 is expressed) and may indicate that overexpression of ␣ v ␤ 3 integrin can support increased cellular levels of ERK1. It is unlikely, however, that the lack of co-immunoprecipitation of His-ERK2 with ␣ v ␤ 3 (and vice versa) is due to this discrepancy, since we have performed experiments in which the expression of His-ERK2 was raised to exceed that of His-ERK1 (by increasing the quantity of cDNA employed for the transfection) and obtained similar results.
Recruitment of ERK to the ␤ 3 Integrin Cytodomain-A number of studies have indicated that the cytodomains of integrin ␤ subunits are responsible for recruiting signaling kinases and cytoskeletal proteins to the heterodimer (6,23). To test the involvement of the ␤ 3 cytodomain in ERK1 recruitment, we constructed a series of ␤ 3 integrin truncation mutants (Fig.  3A). These were aimed at sequentially deleting the membranedistal NITY motif including Ile 757 (⌬757), which has been im-plicated in the assembly of focal adhesions, the sequence intervening in the membrane-proximal NPXY and the NITY motifs (⌬749), and finally the removal of a large proportion of the integrin cytodomain (⌬728).
The three truncation mutants, ⌬757, ⌬749, and ⌬728, were all expressed at levels similar to that observed for full-length ␤ 3 , were immunoprecipitated efficiently with the anti-h␤ 3 monoclonal antibody (Fig. 3B), and formed heterodimers with the ␣ v subunit (not shown). ERK1 coimmunoprecipitated equally well with both full-length ␤ 3 integrin and ⌬757, indicating that the NITY motif and Ile 757 are not necessary for association with the kinase. However, removal of the 749 EAT-STFTN 759 sequence that intervenes in the NPXY and NITY motifs resulted in a substantial reduction in ERK1 recruitment (Fig. 3B).
To determine whether the ␤ 3 cytodomain was sufficient on its own to recruit active ERK, lysates from PDGF-stimulated NIH 3T3 fibroblasts were incubated with GST fusion proteins corresponding to the ␤ 3 integrin cytodomain and truncation mutants thereof (Fig. 3C). Interestingly GST-␤ 3 cytodomain was able to associate with both ERK1 and ERK2, whereas FIG. 3. The ␤ 3 integrin cytodomain recruits ERK. A and B, full-length ␤ 3 integrin and the ⌬757, ⌬749 and ⌬728 truncation mutants of ␤ 3 integrin shown in A were transiently expressed in NIH 3T3 fibroblasts together with the ␣ v integrin subunit. Transfected cells were serum-starved and challenged with 10 ng/ml PDGF-BB for 10 min. Following this, the monolayers were lysed and immunoprecipitated with monoclonal antibodies against human ␤ 3 integrin. Immobilized material was analyzed by Western blotting with anti-human ␤ 3 integrin (B; upper panel) and anti-ERK1/2 (B; lower panels). C and D, lysates from PDGF stimulated NIH 3T3 fibroblasts were incubated with magnetic beads conjugated to GST or the GST-integrin cytodomain fusion proteins indicated in C. Immobilized material was analyzed by Western blotting for ERK1/2 (D; upper panels), and the loading of the GST fusion proteins was confirmed by Western blotting for GST (D; lower panels).
GST-␤ 1 was less effective in this regard (Fig. 3D) and, consistent with the immunoprecipitation studies shown in Fig. 3B, association of ERKs was lost upon removal of the 749 EATST-FTN 759 sequence between the NPXY and NITY motifs (Fig.  3D). These results indicate that the cytodomain of the ␤ 3 integrin subunit is both necessary and sufficient to recruit an ERK-containing complex to ␣ v ␤ 3 integrin and that the central region of the cytodomain is involved in establishment of this association.
Recruitment of ␣ v ␤ 3 and ERK to Plasma Membrane Complexes-We have previously shown that ␣ v ␤ 3 integrin is incorporated into punctate plasma membrane complexes immediately following Rab4-dependent recycling, and these are subsequently redistributed into peripheral focal complexes (13). Given that ERK1 association with ␣ v ␤ 3 was established within 10 min of PDGF addition, we wished to determine whether the kinase was also recruited to plasma membrane complexes. Cells were serum-starved for 30 min and treated with PDGF for 10 min, and surface ␣ v ␤ 3 and ERKs visualized by confocal immunofluorescence microscopy. In serum-starved cells, immunoreactive ERKs were seen to focus in the perinuclear region, perhaps suggesting sequestration of the kinase upon an endomembrane compartment or even at the microtubule organizing center (Fig. 4C). Upon the addition of PDGF, ERKs rapidly redistributed such that they could now be seen in the nucleus and also dispersed into a punctate array across the cell surface (Fig. 4G). This resembled the distribution assumed by ␣ v ␤ 3 (Fig. 4E), and examination of the higher magnification confocal micrograph shown in Fig. 5 revealed a close colocalization of ERKs and ␣ v ␤ 3 integrin in these small punctate structures, which were particularly enriched toward the cell periphery (Fig. 5E). These complexes did not contain caveolin 1, so they are distinct from caveolin-containing membrane islands; nor did they contain paxillin or other markers of focal adhesions and complexes (data not shown). ␣ 5 ␤ 1 integrin was present in large deposits and a fibrillar distribution reminis- cent of the fibronectin network and was not seen to be distributed as a punctate array of plasma membrane complexes (Fig. 4M).
Following a longer (30-min) exposure to PDGF, ERKs were still visible in the nucleus, but the surface complexes were no longer prominent, and the kinase was incorporated into a fine array of focal complexes in the peripheral lamellae that paralleled the distribution of ␣ v ␤ 3 integrin (Fig 4, I-L). Taken together, these immunofluorescence and biochemical data suggest that shortly following growth factor addition, ERK1 and ␣ v ␤ 3 form a physical association within small punctate complexes in the plasma membrane. These then subsequently redistribute to form focal complexes in the peripheral lamellae.
Association of ERK1 with ␣ v ␤ 3 Integrin Requires the Activity of MEK-ERK1 is activated by phosphorylation on threonine 202 and tyrosine 204 by the dual specificity kinase, MEK1/2. We investigated whether treatment of cells with the MEK inhibitor, PD98059 (24), affected the association of ␣ v ␤ 3 with ERK1 and the recruitment of ERK1 to ␣ v ␤ 3 -containing complexes. Serum-starved cells were treated with 12 M PD98059 for 10 min, following which they were challenged with PDGF and lysed for immunoprecipitation or fixed for immunofluorescence. This concentration of PD98059 completely ablated association of ERK1 with immunoprecipitates of ␣ v ␤ 3 (Fig. 6A), and, in accordance with this, markedly reduced the appearance of ERK in integrin-rich plasma membrane complexes (Fig 6,  B-G). These data show that ERK1 must be phosphorylated by MEK in order to be recruited to ␣ v ␤ 3 integrin.
It is important to note that the concentration of PD98059 employed for these experiments (12 M), albeit sufficient to negate association of ERK1 with integrin, had only partial effects on the PDGF-induced recruitment of ERKs to the nucleus (Fig. 6F).
Association of ERK1 with ␣ v ␤ 3 Is Not Necessary for Integrin Recycling-We have previously shown that PDGF increases recycling of ␣ v ␤ 3 from early endosomes to the plasma membrane via a Rab4-dependent mechanism (13). To investigate the possibility that recruitment of ERK1 to ␣ v ␤ 3 is necessary for recycling, we studied the effect of PD98059 on ␣ v ␤ 3 recycling. Recycling of ␣ v ␤ 3 from early endosomes to the plasma membrane was assayed using the enzyme-linked immunosorbent assay-based method we have described previously (13), and PD98059 had no effect on the ability of PDGF to drive ␣ v ␤ 3 recycling from early endosomes (Fig. 7). These data are consistent with the images presented in Fig. 6, E-G, where PD98059 suppressed the colocalization of ␣ v ␤ 3 with ERK but did not affect surface expression of the integrin. It is interesting to note, however, that the integrin-containing complexes were smaller and more numerous in the presence of PD98059 (Fig. 6E), indicating that ERK may act to cluster ␣ v ␤ 3 .
Recycling of ␣ v ␤ 3 Is Not a Prerequisite for Association of ERK1 with ␣ v ␤ 3 -Having shown that delivery of ␣ v ␤ 3 to the plasma membrane was independent of ERK1 recruitment to the integrin, we wished to determine whether the recycling of ␣ v ␤ 3 integrin was a prerequisite for its association with ERK1. Recycling of ␣ v ␤ 3 from early endosomes to the plasma membrane was powerfully stimulated by PDGF, and, consistent with our previous studies (13), this component of integrin vesicular transport was completely ablated by expression of the dominant negative Rab4 construct, N121Irab4 (Fig. 8A). However, association of ERK1 with ␣ v ␤ 3 was unaffected by expression of N121Irab4 (Fig. 8B), despite the blockade of integrin recycling effected by this dominant negative construct.
Recycling of ␣ v ␤ 3 and Active ERK1 Are Required for Cell Spreading on Vitronectin-To investigate the role of ERK1 in ␣ v ␤ 3 integrin function, cells were allowed to spread on vitronectin, a good ligand for ␣ v ␤ 3 but not ␣ 5 ␤ 1 , in the presence and absence of 12 M PD98059. PD98059 inhibited cell spreading on vitronectin by ϳ40%, indicating a requirement for active ERK in this process (Fig. 9A). In contrast, PD98059 did not inhibit spreading on fibronectin. This matrix dependence of PD98059 action indicates that ERK activity is required for the function of a vitronectin-binding integrin and is consistent with our observation that ERK1 is found associated with ␣ v ␤ 3 and not ␣ 5 ␤ 1 . Inhibition of ␣ v ␤ 3 recycling by expression of dominant negative S22Nrab4, also inhibited cell spreading on vitronectin to a similar extent as PD98059 (Fig. 9A), indicating that the activities of Rab4 and ERK are both required for the function of Having demonstrated that recruitment of active ERKl to ␣ v ␤ 3 and the Rab4-dependent recycling of the integrin can be evoked independently from one another, we were interested in determining the effect of inhibiting both of these events simultaneously. The inhibitory effects of PD98059 and S22Nrab4 were not additive, indicating that they were affecting the function of the same integrin and that both efficient recycling and recruitment of active ERK1 to the integrin are necessary for ␣ v ␤ 3 function.
PD98059 opposes the activation of both ERK1 and ERK2. However, the observation that cellular ␣ v ␤ 3 integrin associates specifically with ERK1 in immunoprecipitates suggests the possibility of a special role for this kinase in regulating the assembly and remodeling of cell contacts with the vitronectin matrix. To test this, we employed dominant negative mutants of ERK1 and ERK2 (ERK1KϾR and ERK2KϾR, respectively), which have previously been shown to oppose ERK-induced c-fos expression and cell transformation in an isoform-specific fashion (18). Both of these dominant negative ERK mutants were expressed at similar levels in NIH 3T3 fibroblasts (Fig. 9C), but ERK1KϾR inhibited spreading onto vitronectin by ϳ60%, whereas ERK2KϾR was ineffective in this regard (Fig. 9B). Additionally, ERK1KϾR did not compromise spreading on fibronectin, indicating that ERK1 activity is particularly focused toward the function of ␣ v ␤ 3 and does not impinge on the function of other fibronectin-binding integrins such as ␣ 5 ␤ 1 .

DISCUSSION
Here we show that following addition of PDGF to serumstarved fibroblasts, a 44-kDa protein that is rich in phosphotyrosine and phosphothreonine coimmunoprecipitates with ␣ v ␤ 3 integrin. Western blotting with phosphospecific antibodies revealed that this protein was active ERK1. Experiments in which the C-terminal region of the ␤ 3 integrin subunit was truncated, and pull-downs with GST-␤ 3 fusion proteins show that the cytodomain of the integrin is both necessary and sufficient to recruit ERK and moreover that the 749 EATST-FTN 759 sequence interposing the NPXY and NITY motifs may be critical to this. Immediately following PDGF addition, ERK was seen to colocalize with ␣ v ␤ 3 in numerous small complexes at the plasma membrane, and only later did these redistribute to focal complexes in the peripheral lamellae. The association of ERK1 with ␣ v ␤ 3 was particularly sensitive to treatment of the cells with the MEK inhibitor, PD98059; however, this compound had no effect on the Rab4-dependent flux of integrin from early endosomes to the plasma membrane. Correspondingly, inhibition of Rab4 had no effect on recruitment of ERK1 to ␣ v ␤ 3 integrin, indicating that integrin recycling and the recruitment of active ERK1 are not interdependent. Expression   FIG. 6. PD98059 inhibits association of ERK with ␣ v ␤ 3 integrin. A, serumstarved Swiss 3T3 fibroblasts were incubated for 10 min in the absence or presence of 12 M PD98059 and then challenged with 10 ng/ml PDGF-BB or allowed to remain quiescent. Cells were lysed, and ␣ v ␤ 3 integrin was immunoprecipitated (IP) from the lysates with monoclonal antibodies against the mouse ␤ 3 integrin chain. Immobilized material was then analyzed by Western blotting with an antibody against ERK1/2. The migration positions of the ERKs1/2 are indicated. B-G, serum-starved Swiss 3T3 fibroblasts were incubated for 10 min in the absence (B-D) or presence (E-F) of 12 M PD98059, challenged with 10 ng/ml PDGF-BB, and then fixed in 2% paraformaldehyde. Surface ␣ v ␤ 3 integrin was visualized by indirect immunofluorescence (B and E; green). Following this, cells were detergent-permeabilized and counterstained for cellular ERK (C and F; red). Colocalization of the two fluorophores is shown in yellow (D and G). Bar, 10 m. 7. PDGF-stimulated recycling of ␣ v ␤ 3 does not require the activity of MEK. Serum-starved Swiss 3T3 fibroblasts were surfacelabeled with 0.2 mg/ml NHS-SS-Biotin for 30 min at 4°C, and internalization was allowed to proceed for 15 min at 22°C in the presence and absence of 12 M PD98059. Biotin was removed from receptors remaining at the cell surface by treatment with MesNa at 4°C, and cells were rewarmed to 37°C for 10 min in the absence or presence of 10 ng/ml PDGF-BB to allow recycling to the plasma membrane, followed by a second reduction with MesNa. Cells were lysed, and integrin biotinylation was determined by capture enzyme-linked immunosorbent assay using microtiter wells coated with anti-mouse ␤ 3 integrin monoclonal antibodies. The proportion of integrin recycled to the plasma membrane is expressed as a percentage of the pool of integrin labeled during the internalization period (values are mean Ϯ S.E. from three separate experiments). of a dominant negative mutant of ERK1 (but not ERK2) significantly reduced the spreading of cells onto vitronectin, whereas cell spreading on fibronectin was unaffected by inhibition of ERK1, consistent with a special role for this isoform of ERK in the regulation of ␣ v ␤ 3 (but not ␣ 5 ␤ 1 ) integrin function. PD98059 also reduced cell spreading on vitronectin, to the same extent as did dominant negative Rab4, and the effects of Rab4 and MEK inhibition were not additive. Taken together, these data indicate that ␣ v ␤ 3 must recycle to the plasma membrane via the Rab4 pathway and recruit ERK1 in order to function efficiently.

FIG.
Role of Integrins in ERK Translocation-In resting cells, ERK is retained in the cytoplasm in tight association with the microtubular cytoskeleton (25), and it is likely, therefore, that the perinuclear accumulation of ERK that we observe in serum-starved fibroblasts indicates association with the microtubule organizing center. Upon stimulation, ERK translocates from the cytoplasm to the nucleus, where it influences gene expression by phosphorylating transcription factors. This enhances expression of a number of early response genes, such as c-fos (26), and ultimately leads to the induction of cyclin D1 and progression through the G 1 phase of the cell cycle (27). The engagement of integrin is known to profoundly enhance ERK activation in response to growth factor addition, and this provides a rationale for the much studied phenomenon of anchoragedependent growth (28). Enhancement of ERK signaling is thought to be mediated by a diverse array of integrin-activated signaling pathways, most of which also lead to reorganization of the actin cytoskeleton. Indeed, a recent study has shown that integrin-mediated adhesion is necessary for efficient nuclear translocation of ERK via a mechanism that clearly requires an intact actin cytoskeleton (29). It is possible that association of ERK with the focal adhesion machinery may facilitate delivery of the kinase to the nucleus. Two aspects of our data, however, argue against this. First, ERK1 recruitment to ␣ v ␤ 3 is only fully established ϳ8 min following PDGF addition. However, the translocation of ERK to the nucleus is, if anything, faster than this, arguing against a sequence of events whereby ERK is obliged to associate with ␣ v ␤ 3 and passage through focal complexes in order to reach the nucleus. Second, the concentration of PD98059 employed in the present study was found to completely ablate association of ERK1 with ␣ v ␤ 3 but had no effect on nuclear accumulation of ERK. This implies that different pools of cytoplasmic ERK are destined for transport to the nucleus and the plasma membrane following growth factor addition, the activation of the former being less sensitive to treatment of cells with PD98059 than the latter.
Our data indicate that overexpression of h␣ v ␤ 3 integrin favors increased cellular expression levels of His-ERK1. This suggests that ␣ v ␤ 3 may play a role in the stabilization of the FIG. 8. Dominant negative Rab4 does not block association of ERK1 with ␣ v ␤ 3 integrin. NIH 3T3 fibroblasts were transfected with human ␣ v ␤ 3 integrin in combination with wild-type Rab4 (wtrab4) or N121Irab4 as indicated. A, transfected cells were serum-starved and surface-labeled, and integrin recycling was performed in the presence and absence of 10 ng/ml PDGF-BB as for Fig. 7. Biotinylated integrin was determined by capture enzyme-linked immunosorbent assay using microtiter wells coated with anti-human ␤ 3 monoclonal antibodies. Values are mean Ϯ S.E. from three separate experiments. B, transfected cells were serum-starved and then challenged with 10 ng/ml PDGF-BB or allowed to remain quiescent. Cells were lysed, and ␣ v ␤ 3 integrin was immunoprecipitated from the lysates with monoclonal antibodies against the human ␤ 3 integrin chain as for Fig. 2. Immobilized material was then analyzed by Western blotting with an antibody against ERK1/2. The migration positions of ERK1/2 are indicated. ERK1 protein, most likely by incorporation of the kinase into an integrin complex. The coordinated synthesis and degradation of many signaling proteins is likely to be integral to the establishment of anchorage-dependent growth, and it is possible that the ability of ␣ v ␤ 3 integrin to support increased cellular levels of ERK1 may contribute to this.
A Role for ERK at the Plasma Membrane-A number of recent studies have shown that ERK has an important role in the cytoplasm and that this is likely to be distinct from its activity in the nucleus. The sea star oocyte homologue of ERK1 directly phosphorylates myosin light chain kinase (30), and more recently activation of ERKs with a constitutively active MEK has been shown to enhance cell migration via phosphorylation of myosin light chain kinase (31). A more recent study has demonstrated that active ERK is recruited to focal adhesions and controls their assembly by virtue of its ability to phosphorylate and activate myosin light chain kinase (22). Thus, if phospho-ERK levels are lowered using U0126 (a more potent MEK inhibitor than PD98059), the assembly of focal complexes is inhibited, and consequently the ability of cells to spread on the extracellular matrix is compromised. We are able to confirm that ERK is indeed targeted to focal complexes and furthermore show that this is likely to be achieved by its association with an extracellular matrix receptor, ␣ v ␤ 3 integrin.
Recruitment of ERK is clearly mediated by the cytodomain of the ␤ 3 integrin subunit. The extreme C-terminal region of ␤ 3 is known to associate with both endonexin (32) and Syk (6), and a previous study has shown that Ile 757 is critical for targeting ␣ IIb ␤ 3 to focal adhesions (33). However, removal of the Cterminal NITY motif, including Ile 757 , has no effect on ERK recruitment to ␣ v ␤ 3 or to GST-␤ 3 cytodomain fusion proteins. On the other hand, our data indicate that the 749 EATSTFTN 759 sequence (immediately N-terminal to the NITY motif) is required for association with ERK. It has been known for some time that this portion of the ␤ 3 cytodomain is critical for integrin function. A serine to proline substitution in this region has been found in a patient with Glanzmann's thrombasthenia and renders the integrin refractory to inside-out activation (34), and, furthermore, substitution of the TST motif for AAA reduces ␤ 3 integrin function in cell attachment, spreading, and the initiation of tyrosine phosphorylation of pp125 FAK and paxillin (35). A recent report has documented a direct association of ERK with the cytodomain of ␤ 6 integrin (36). Synthetic peptides containing the central region of the ␤ 6 cytodomain were shown by these workers to bind directly to ERK. It is interesting to note that a TSTF motif in this region is conserved between ␤ 3 and ␤ 6 and is not present in other ␤-integrin cytodomains. Further experiments will be aimed at testing the ability of the TSTF motif in ␤ 3 to associate directly with ERKs.
Our data indicate a special relationship between ERK1 and ␣ v ␤ 3 , which is not shared by ERK2. First, ERK1 (and not ERK2) is recruited to immunoprecipitates of ␣ v ␤ 3 , and second His-ERK1KϾR opposes ␣ v ␤ 3 -mediated cell spreading, whereas His-ERK2KϾR is ineffective in this regard. From this, it would be tempting to speculate that this is due to an integrin-binding site that is present in ERK1 but not ERK2. However, it is clear that GST-␤ 3 cytodomain fusion proteins have the capacity to bind equally well to both ERK1 and ERK2. This may indicate that selectivity for ERK1 requires the presence of the ␣ v ␤ 3 heterodimer or alternatively may be influenced by accessory factors that are unable to function in the pull-down assays.
Our data indicate that the association of ␣ v ␤ 3 with ERK1 occurs rapidly following the addition of PDGF and that the resulting complex localizes to punctate clusters in the plasma membrane prior to its incorporation into focal complexes. Hith-erto, many studies have focused on the role of integrins in focal adhesions and complexes, and it is generally accepted that integrins are brought into close proximity with the various signaling molecules that mediate focal adhesion signaling, as a consequence of the activity of the Rho subfamily GTPases, such as RhoA and Rac (37). However, it is now becoming clear that certain pathways promote the association of integrins with other signaling components upstream of focal complex assembly. A recent study has highlighted a novel integrin complex, referred to as an integrin cluster, that forms upstream of Rac activation and focal complex assembly (15). These workers reported that integrin clusters differ from focal complexes in both their distribution and molecular composition. The ␣ v ␤ 3 ⅐ERK-containing complexes described in the present study also differ in their composition from focal complexes. For instance, they do not stain for established focal adhesion markers, such as vinculin and paxillin; nor do any of these proteins coimmunoprecipitate with ␣ v ␤ 3 following the addition of PDGF. It is interesting to speculate what kind of cellular structure these ␣ v ␤ 3 ⅐ERK-rich complexes may be. Labeling experiments with [ 32 P]orthophosphate have indicated that, even following extensive washing with nonionic detergent (0.5% (v/v) Triton X-100 and 0.25% (v/v) Igepal), labeled phospholipids are tightly associated with ␣ v ␤ 3 immunoprecipitates. Moreover, the quantity of coimmunoprecipitating phospholipid increased dramatically in response to the addition of PDGF, and this was opposed by PD98059 (data not shown). Reorganization of lipid rafts and other plasma membrane lipid subdomains has been reported to occur following activation of a number of signaling pathways (41). It is possible, therefore, that active ERK1 can act at the plasma membrane to induce clustering of integrin into large detergent-resistant raftlike membrane microdomains. Preliminary data indicate that p36 (Fig. 1, E and F) is likely to be annexin II (not shown). This Ca 2ϩ -, phospholipid-, and actin-binding protein is known to be a substrate for the PDGF receptor and Src tyrosine kinases (38), and more recently it has been shown to localize to plasma membrane lipid rafts (39) and participate in the reorganization of these domains during cell attachment (40). It will be interesting to determine whether annexin II is recruited to ␣ v ␤ 3 -rich membrane complexes and to investigate the possibility that it has a role in recruiting ERK1 to the plasma membrane.
Integrins have been reported to associate with many other types of transmembrane and other proteins, which in principle may coalesce to form an extensive network (42). Indeed, ␣ v ␤ 3 integrin is known to associate with CD47, or integrin-associated protein, in a plasma membrane lipid raft (43). Also, the integrin-associated tetraspannin, CD81, has been shown to localize to such a membrane domain (42), and it is interesting in this regard that we observe CD81 to colocalize with ␣ v ␤ 3 in puncta following PDGF treatment (data not shown).
Following expression of dominant negative Rab4, ␣ v ␤ 3 would be expected to accumulate in early endosomes. This construct, however, does not reduce the recruitment of ERK1 to ␣ v ␤ 3 immunoprecipitates and indicates that the association of active kinase to the integrin may be established either on the surface of endosomes or at the plasma membrane. Indeed, large integrin-tetraspannin complexes have been detected on intracellular vesicles (10), and this implies that associations made between integrins and other signaling molecules and membrane proteins may persist while the integrin engages in endoexocytic cycling. There is mounting evidence that the incorporation of membrane proteins into raftlike domains is a key sorting event in the secretory pathway (44). However, it is unlikely that recruitment of active ERK to ␣ v ␤ 3 at the endosome is necessary to direct its recycling, since the rate of delivery of early endo-somal ␣ v ␤ 3 to the plasma membrane was clearly unaffected by PD98059.
The observation that dominant negative ERK1 compromises cell spreading on vitronectin, but not on fibronectin, implicates the activity of this ERK in the activation of ␣ v ␤ 3 integrin. Many integrins, including ␣ v ␤ 3 , can assume different states with respect to ligand binding and engagement, and there are many examples of the transition between these states being controlled by signaling pathways within the cell, a phenomenon termed inside-out signaling (12). The affinity of an individual integrin heterodimer for its ligand may be increased, and changes in the lateral mobility and clustering of integrins can also affect the avidity of integrin binding to multivalent ligands. Regulation of either the affinity or avidity of an integrin for its ligand will have profound influence on the ability of cells to spread on the extracellular matrix. The ability of the platelet integrin ␣ IIb ␤ 3 to bind fibrinogen has been shown to be inhibited by dominant negative mutants of Raf-1 or MEK1 (45). This implicates the MEK/ERK signaling axis in inside-out signaling to ␤ 3 integrins, and this could by achieved by influencing either the affinity state or the clustering of integrin. We favor the explanation that the recruitment of ERK1 leads to integrin clustering, since we found that integrin-containing puncta were clearly smaller and more numerous in the presence of PD98059.
In summary, our studies have identified a novel association of active ERK1 kinase with ␣ v ␤ 3 integrin that is established in advance of the incorporation of the integrin into focal complexes. Formation of this complex was not necessary for the trafficking of ␣ v ␤ 3 through the Rab4-dependent recycling pathway; nor was the activity of Rab4 required for association of ␣ v ␤ 3 with ERK. However, the activity of ERK1 and Rab4 are clearly required for cells to spread on vitronectin. We suggest that recruitment of ERK1 to ␣ v ␤ 3 and the Rab4-dependent recycling pathway are parallel growth factor-activated events that are necessary for integrin function.