Integrins enhance platelet-derived growth factor (PDGF)-dependent responses by altering the signal relay enzymes that are recruited to the PDGF beta receptor.

Since the extracellular matrix (ECM) can promote platelet-derived growth factor (PDGF)-dependent responses, we hypothesized that the ECM mediates this effect by preventing the PDGF beta receptor (betaPDGFR) from associating with the negative regulator, RasGAP (the GTPase-activating protein of Ras). We found that binding of RasGAP to the wild-type betaPDGFR was decreased; the activation of Ras and Erk was enhanced, and [3H]thymidine uptake was better in cells cultured on fibronectin than in cells cultured on polylysine. To investigate the mechanism by which culturing cells on fibronectin diminished the recruitment of RasGAP to the betaPDGFR, we focused on SHP-2 since it dephosphorylates the betaPDGFR at the phosphotyrosine required for binding of RasGAP. Culturing cells on fibronectin increased the amount of SHP-2 that associated with the betaPDGFR. Furthermore, cells expressing receptor mutants that failed to associate with SHP-2 were insensitive to fibronectin. The ECM enhances PDGF-dependent responses by increasing the association of SHP-2 with the betaPDGFR, which in turn decreases the time that RasGAP interacts with the receptor. Thus, fibronectin changes PDGF-dependent signaling and biological responses by altering the signal relay enzymes that are recruited to the receptor.

Since the extracellular matrix (ECM) can promote platelet-derived growth factor (PDGF)-dependent responses, we hypothesized that the ECM mediates this effect by preventing the PDGF ␤ receptor (␤PDGFR) from associating with the negative regulator, RasGAP (the GTPase-activating protein of Ras). We found that binding of RasGAP to the wild-type ␤PDGFR was decreased; the activation of Ras and Erk was enhanced, and [ 3 H]thymidine uptake was better in cells cultured on fibronectin than in cells cultured on polylysine. To investigate the mechanism by which culturing cells on fibronectin diminished the recruitment of RasGAP to the ␤PDGFR, we focused on SHP-2 since it dephosphorylates the ␤PDGFR at the phosphotyrosine required for binding of RasGAP. Culturing cells on fibronectin increased the amount of SHP-2 that associated with the ␤PDGFR. Furthermore, cells expressing receptor mutants that failed to associate with SHP-2 were insensitive to fibronectin. The ECM enhances PDGF-dependent responses by increasing the association of SHP-2 with the ␤PDGFR, which in turn decreases the time that Ras-GAP interacts with the receptor. Thus, fibronectin changes PDGF-dependent signaling and biological responses by altering the signal relay enzymes that are recruited to the receptor.
Upon exposure of cells to platelet-derived growth factor (PDGF), 1 the PDGF receptor (PDGFR) dimerizes and autophosphorylates on a number of intracellular tyrosine residues. One of the consequences of tyrosine phosphorylation of the PDGFR is the generation of binding sites for various SH2 domain-containing proteins. Some of the proteins that associate with the ␤PDGFR include Src family members, phosphatidylinositol 3-kinase (PI3K), the GTPase-activating protein of Ras (RasGAP), the tyrosine phosphatase (SHP-2), and phospholipase C␥1 (PLC␥) (reviewed in Ref. 1).
Evidence suggests that different signaling enzymes are re-quired for initiating different cellular responses. By using receptor mutants that bind only PI3K or only PLC␥, it has been demonstrated that PI3K and/or PLC␥ act as positive regulators of PDGF-dependent DNA synthesis (2,3), cellular transformation (4), vesicle trafficking (5), and cell migration (6,7). Microinjection of reagents that interfere with the action of the receptor-associated signal relay enzymes also prevents PDGFdependent entry into S phase (8 -10). In contrast, RasGAP is negative regulator of ␤PDGFR signaling. Analysis of ␤PDGFR mutants indicates that RasGAP prevents activation of PLC␥ (11) and PI3K (12). Furthermore, PDGF triggers activation of Ras better in cell lines that do not express RasGAP as compared with the corresponding control cell lines (13). A similar role for RasGAP has emerged in signaling initiated by other receptor tyrosine kinases. For instance, the RasGAP locus negatively regulates the Sevenless receptor tyrosine kinase during Drosophila eye development (14). Furthermore, the strength of Torso signaling is modulated by RasGAP and has dramatic consequences on terminal structure development in Drosophila (15,16). Hence, these studies indicate that RasGAP is a negative regulator of several different receptor tyrosine kinases. Given that RasGAP is a negative regulator of the ␤PDGFR, it is somewhat puzzling that the wild-type ␤PDGFR associates with RasGAP and is able to efficiently drive biochemical and biological responses. One explanation is that all activated ␤PDGFRs do not associate with RasGAP. This idea is supported by the observation that the receptor is not stoichiometrically phosphorylated at tyrosine 771, which is required for binding of RasGAP (17). This suggests that in an activated cell, the population of receptors are heterogeneous with respect to which sites are phosphorylated and consequently which signaling molecules are present in a receptor dimer. We hypothesize that changing the composition of signal relay enzymes that are recruited to the ␤PDGFR will alter the signal relay pathways that the receptor will activate.
Since changes in the extracellular environment alter the way in which cells respond to PDGF, this variable may influence which signal relay enzymes associate with the PDGFR. One extracellular component that could potentially play a role in changing the composition of signal relay enzymes that associate with the PDGFR are integrins. Integrins are heterodimeric receptors that mediate attachment to the extracellular matrix (ECM) and cell-cell adhesive interactions (reviewed in Ref. 18). Several lines of evidence suggest that integrins can modulate growth factor receptor function. Adherence to the extracellular matrix is critical for stimulation of cell proliferation by growth factors (19) and growth factor-stimulated Erk activation (20). Furthermore, association of integrins with activated insulin and PDGF receptors correlates with enhanced mitogenicity and chemotaxis (21,22).
Most of the data describing integrin-dependent modulation * This work was supported by National Institutes of Health Grant GM48339. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: Schepens Eye Research Institute, Harvard Medical School, 20 Staniford St., Boston, MA 02114. Tel.: 617-912-2517; Fax: 617-912-0128; E-mail: kazlauskas@ vision.eri.harvard.edu. 1 The abbreviations used are: PDGF, platelet-derived growth factor; PDGFR, platelet-derived growth factor receptor; ECM, extracellular matrix; DME, Dulbecco's modified Eagle's; PI3K, phosphatidylinositol 3-kinase; RasGAP, GTPase-activating protein of Ras; PLC␥, phospholipase C␥1; WT, wild type; GST, glutathione S-transferase; PAGE, polyacrylamide gel electrophoresis; BSA, bovine serum albumin; FBS, fetal bovine serum; PBS, phosphate-buffered saline. of growth factor receptor function have been obtained by investigating integrin-dependent signaling in cells that have been replated on ECM-coated dishes or beads. Although this commonly used system is an excellent approach to study signaling downstream of integrins, we were interested in how long term culturing of cells on ECMs alters growth factor signaling. Consequently, after replating cells on an ECM-coated dish, we waited until the integrin-dependent signaling subsided (12-14 h) and then stimulated the cells with PDGF. We found that recruitment of RasGAP to the wild-type ␤PDGFR was decreased in cells cultured on fibronectin; the duration of PDGFdependent Ras and Erk activation was prolonged, and the DNA synthesis response was enhanced. The decreased binding of RasGAP was tightly correlated with an increase in the recruitment of SHP-2 to the wild-type ␤PDGFR. In addition, mutation of the binding site on the ␤PDGFR for SHP-2 negated the effect of culturing cells on fibronectin. These findings suggest that long term engagement of integrins alters PDGF-dependent signaling by changing the amount of individual signaling molecules that associate with the wild-type ␤PDGFR, and the net effect is enhanced DNA synthesis. Finally, it appears that SHP-2 is one of the enzymes that acts as a liaison between the ECM and the ␤PDGFR.

Cell Lines
The Ph cell line was derived from mouse embryos homozygous for the Ph/Ph deletion that includes the ␣-PDGFR gene (23) and was kindly provided by Dan Bowen-Pope (University of Washington). These cells are mouse embryo fibroblasts that express endogenous ␤PDGFR at approximately 1 ϫ 10 5 ␤PDGFRs per cell and no ␣PDGFRs. Ph cells were maintained in Dulbecco's modified Eagle's (DME) medium supplemented with 5% calf serum, and 1 mg/ml G418 was added to cultures of cells expressing the introduced chimeric constructs. The F cells were obtained from embryos nullizygous for both the ␣ and ␤ PDGFRs and were kindly provided by Michelle Tallquist and Phillipe Soriano (Fred Hutchinson Cancer Research Center, Seattle). They were maintained in DME supplemented with 5% fetal bovine serum.

Construction of the PDGFRs
Chimeric PDGFRs-Construction of the wild-type chimeric PDGFR was described previously (4,24). The N 2 F771 chimera was constructed by subcloning a SacII/Sal (3Ј end of the ␤PDGFR up to and including the SacII site at position 1972) fragment of the F771 ␤PDGFR (17) into the SacII/Sal-cut wild-type chimeric receptor. The chimeras were subcloned into the pLXSN 2 retroviral vector (4,25); virus was generated using the 293T system (26), and the resulting virus was used to infect Ph cells. Receptor-expressing cells were selected in DME medium containing 5% calf serum and 1 mg/ml G418, and the cells were sorted so that the chimeric receptors were expressed at approximately the same number as the endogenous receptors, 1 ϫ 10 5 receptors per cell. Periodic assessment of the level of receptor expression by Western blot analysis indicated that the levels of expression were stable for at least 12 months.
␤PDGFR Constructs-The F1009 PDGFR cDNA (27) was subcloned into the pLXSHD (25), as a 4.2-kilobase pair EcoRI-SalI fragment. The resulting construct was introduced into 293GPG cells (28) using Lipo-fectAMINE. Virus was collected for 5 days, concentrated by centrifugation, and resuspended in TNE (50 mM Tris, pH 7.8; 130 mM NaCl, 1 mM EDTA). The resulting virus was used to infect the F cells, and mass populations of receptor-expressing cells were obtained by selecting the cells in DME supplemented with 5% FBS and 10 mM histidinol. The human wild-type ␤PDGFR was introduced into the amphotropic virus producing cell line PA317 as described previously (29). Virus from the PA317 cells was used to infect the F cells, and drug-resistant cells were grown out in medium containing 200 g/ml hygromycin. ␤PDGFR Western blot analysis of total cell lysates indicated that the WT-expressing cells had approximately 1.0 ϫ 10 5 receptors/cell.

Antibodies
PR292 is a mouse monoclonal antibody that recognizes an epitope in the extracellular domain of the ␣PDGFR and was purchased from Genzyme. 80.8 is a crude rabbit polyclonal antiserum raised against a glutathione S-transferase (GST) fusion protein including a portion of the first immunoglobulin domain (residues 52-94) and was used at a 1:1000 dilution. The ␤PDGFR was immunoprecipitated using crude polyclonal rabbit sera (30A) raised against glutathione S-transferase fusion proteins that included the entire carboxyl terminus of the human ␤PDGFR subunit (3). The ␤PDGFR (3), RasGAP (3), and the p85 subunit of PI3K (4) were blotted as described previously. For Ras Western blot analysis, a monoclonal pan-Ras antibody (Transduction Laboratories) was used at a 1:500 dilution. To blot for SHP-2, a monoclonal PTP1D antibody (Transduction Laboratories) was used at a 1:1000 dilution. For anti-phosphotyrosine Western blot analysis, a combination of PY20 (Transduction Laboratories) and 4G10 (Upstate Biotechnology, Inc.) was used, each at a 1:1000 dilution. The phospho-Erk antibody (New England Biolabs) is an affinity purified rabbit polyclonal antibody raised against a synthetic phosphotyrosine peptide corresponding to amino acid residues 196 -209 of human p44 Erk. It recognizes both tyrosine-phosphorylated and threonine/tyrosine doubly phosphorylated p42 and p44 Erk and was used at a 1:500 dilution. The pan-Erk antibody used for immunoprecipitation of Erk was a generous gift of Gen-Sheng Feng (Indiana University) and has previously been described (30). The pan-Erk antibody used for Western blot analysis was a monoclonal antibody available from Zymed Laboratories Inc. and was used at a 1:2500 dilution.

Ras Assay
The Ras assay is a modified version of the protocol previously described (31).
GST-Raf 50 -150 Isolation-A construct containing a GST fusion with amino acids 50 -150 of Raf was transformed into DH5␣ bacterial cells. A 20-ml starter culture of LB ϩ 100 g/ml carbenicillin was inoculated and grown overnight at 37°C. The starter culture was diluted 1:10 with LB ϩ carbenicillin and incubated at 37°C until the A 600 was 0.6 -0.8, at which time expression of the fusion protein was induced by addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 1 mM. The cultures were incubated at 37°C for an additional 2 h and centrifuged; the pellets were washed once with ice-cold H/S (32), and then the pellet was resuspended in HSE (20 mM HEPES, pH 7.5, 120 mM NaCl, 10% glycerol, 2 mM EDTA) ϩ 10 g/ml leupeptin and aprotinin, sonicated, centrifuged to remove the insoluble debris, and Nonidet P-40 added to a final concentration of 0.5%. The supernatant was incubated for 30 min at 4°C with 500 l of washed glutathione beads and washed eight times with HSE ϩ 0.5% Nonidet P-40 ϩ 10 g/ml leupeptin and aprotinin. The purified fusion proteins were left attached to the GST beads and used immediately to purify activated Ras.
Purification of Active Ras-Ph cells expressing the various receptor constructs were trypsinized and plated on polylysine (5 mg/ml) or fibronectin-coated plates (Falcon) in DME ϩ 2% calf serum. After the cells had attached and begun spreading (20 -30 min at 37°C), the cells were washed twice with PBS and placed in serum-free DME for 12 h. The cells were then exposed for 5-30 min to 50 ng/ml PDGF AA or buffer (2 mM acetic acid and 10 mg/ml BSA), washed 2ϫ in H/S (32), lysed, and then incubated with approximately 20 -30 g of GST-Raf fusion protein for 30 min at 4°C. The beads were washed three times in 600 l of lysis buffer; sample buffer (32) was added, and the proteins were resolved on a 15% SDS-PAGE gel. The proteins were transferred to Immobilon in a methanol-containing transfer buffer (150 mM glycine, 25 mM NaCl, 10% CH 3 OH) and subjected to a Ras Western blot except that Blotto without azide was used in place of Blotto.

Erk Activation
Parental Ph cells or Ph cells expressing a chimeric PDGFR were grown to 85-90% confluence, serum-starved in DME ϩ 0.1% calf serum for 18 -24 h, left resting (Ϫ) or stimulated with 50 ng/ml PDGF AA (Ph cells expressing a chimeric PDGFR) or 40 ng/ml PDGF BB (parental Ph cells) for 5-30 min at 37°C. The cells were washed with H/S (32), lysed in EB (32) without BSA, and centrifuged to remove the insoluble debris. The amount of protein in the lysates was determined by the BCA protein assay (Pierce), and equal amounts of protein were resolved on a 10% SDS-PAGE gel. The resolved proteins were transferred to Immobilon and probed with a phospho-Erk antibody. Activation of Erk was also monitored by assaying Erk activity in an in vitro kinase assay. Ph cells expressing the wild-type or N 2 F771 chimeric PDGFR were grown to 85-90% confluence, incubated in DME ϩ 0.1% calf serum overnight, and then stimulated with buffer or 50 ng/ml PDGF AA for 5-120 min at 37°C. The cells were washed twice with H/S ϩ 2 mM sodium orthovanadate (32) and lysed in EB (32), and Erk was immunoprecipitated using a pan-Erk antibody (3). Erk immunoprecipitates representing approximately 1.5 ϫ 10 5 cells were subjected to an in vitro kinase assay in the presence of 0.75 g of myelin basic protein (Sigma). The proteins were resolved on a 15% polyacrylamide gel, and the gel was stained to show the relative amounts of myelin basic protein present and then exposed to film. The phosphorylated bands were excised and counted in a scintillation counter.

Immunoprecipitation and Western Blot Analysis
For studies involving cells plated on plastic, Ph cells expressing the chimeric PDGFR were grown to 85-90% confluence, incubated in DME ϩ 0.1% calf serum for 18 -24, and stimulated with 50 ng/ml PDGF-AA for the times indicated. For studies involving cells plated on polylysine or an ECM, Ph cells were grown to 90% confluence, trypsinized briefly, and plated on 5 mg/ml polylysine (Sigma)-or fibronectin (Falcon)coated dishes. The cells were incubated for 30 min at 37°C, washed twice in PBS, and incubated in DME alone for 12 h at 37°C and 5% CO 2 . Under both plating conditions, the cells were then stimulated with buffer (10 mM acetic acid ϩ 2 mg/ml BSA) or 40 ng/ml PDGF BB for the times indicated, washed, and lysed in EB (32), and the ␤PDGFRs were immunoprecipitated using 30A (3). The immunoprecipitates were bound to formalin-fixed Staphylococcus aureus membranes and washed as described previously (33). The resulting gel was transferred to Immobilon and subjected to Western blot analysis. Proteins were detected using ECL (Amersham Pharmacia Biotech).

In Vitro Kinase Assay
The intrinsic tyrosine kinase activity of the PDGFR was analyzed using receptor immunoprecipitates from approximately 2 ϫ 10 5 cells as described previously (4).

[ 3 H]Thymidine Uptake
PDGF-stimulated [ 3 H]thymidine uptake was assayed as follows. Cells were plated at 3 ϫ 10 4 cells/ml in DME ϩ 2% FBS in 24-well dishes and incubated at 37°C for 1.0 h at which time they were washed 2ϫ in PBS and placed in DME containing 0.1% FBS. Cultures were incubated at 37°C and 5% CO 2 for 48 h at which time they were washed 2ϫ in PBS and incubated for an additional 24 h at 37°C in DME containing 2 mg/ml BSA. Buffer (10 mM acetic acid and 2 mg/ml BSA), PDGF BB (5 or 40 ng/ml), or 10% FBS was added; the cultures were incubated for 22 h and the media replaced by DME ϩ 5% FBS and 0.8 Ci of [ 3 H]thymidine per ml, and the incubation was prolonged for 4 h. The cells were then harvested as described previously (24).

RESULTS
We have previously found that artificially altering the composition of signaling molecules that are recruited to the ␤PDGFR profoundly changes the nature of the signaling cascades initiated by PDGF (3,11,12). Since extracellular changes such as cell density or ECM alter the ability of PDGF to initiate biological responses (21,22,34), we postulated that the basis of the altered responsiveness to PDGF is due to a change in signaling enzymes that associate with the receptor. In this study, we examined whether changing the composition of the signal enzymes which associated with the wild type ␤PDGFR was the mechanism by which integrins can modulate growth factor receptor function. Consequently, we focused on cell lines naturally expressing the wild-type ␤PDGFR, and we compared RasGAP co-immunoprecipitation with the ␤PDGFR when the cells were cultured on polylysine or fibronectin. Ph cells were grown to 90% confluence, trypsinized, resuspended in DME ϩ 2% calf serum, and then plated in dishes coated with polylysine or fibronectin. Once the cells had adhered (after 30 min), the cultures were washed twice with PBS and then incubated in serum-free DME at 37°C for 12-14 h. Pilot experiments indicated that Erk activity was elevated soon after plating and gradually returned to basal levels by 12-14 h (data not shown). The cells were either left resting or stimulated with 40 ng/ml PDGF BB for 5-30 min and lysed; the ␤PDGFR was immunoprecipitated, and the amount of co-immunoprecipitating Ras-GAP was assessed by Western blotting (Fig. 1A). In cells cultured on polylysine, the level of receptor-associated RasGAP was similar from 5 to 20 min, and less RasGAP was present in the 30-min sample. The decrease in the amount of co-precipitating RasGAP was seen earlier in cells cultured on fibronectin (at the 20-min time point), and by 30 min only a trace amount of RasGAP was present. Thus RasGAP associates with the receptor under both sets of conditions, but in the cells cultured on fibronectin, RasGAP binding is more transient than in the control cells.
Since binding of RasGAP depends on tyrosine phosphorylation of the ␤PDGFR, it is possible that the differences in Ras-GAP binding reflect an ECM-dependent alteration in the extent and/or the duration of receptor phosphorylation and/or kinase activity. To investigate this possibility, receptors were immunoprecipitated from cells that were treated as described in Fig. 1A, and the extent of receptor phosphorylation was examined by Western blot analysis (Fig. 1B). We found that tyrosine phosphorylation of the ␤PDGFR was similar in cells cultured on polylysine or fibronectin (Fig. 1B). The intrinsic kinase activity of the receptor was assessed by subjecting the receptor immunoprecipitates to an in vitro kinase assay in the presence of an exogenous substrate, GST-PLC␥. Altering the ECM did not detectably change the basal or PDGF-stimulated kinase activity of the receptor (Fig. 1C). Thus, while the overall extent of the ␤PDGFR phosphorylation or kinase activity was not altered by the ECM of cells, the duration that RasGAP stably associated with the ␤PDGFR decreased when cells are cultured on fibronectin.
Effect of ECM on Pathways Downstream of RasGAP-We next examined whether pathways downstream of RasGAP were affected by culturing cells on an ECM, and we focused on PDGF-dependent Ras and Erk activation. Ph cells were plated, starved, and stimulated as described above and lysed, and active GTP-bound Ras was recovered from the lysate with a GST fusion protein that includes the Ras binding domain of Raf. The samples were subjected to Western blot analysis with FIG. 1. Comparison of RasGAP binding. A, Ph cells were trypsinized, resuspended in DME ϩ 2% calf serum, and seeded onto dishes coated with fibronectin or polylysine. Thirty minutes after plating, both sets of cells had adhered and begun to spread. The cells were washed twice with PBS and then incubated in serum-free DME for 12 h. The cells were left resting (Ϫ) or stimulated with 40 ng/ml PDGF BB for the times indicated and then lysed, and the receptor was immunoprecipitated with an antibody against the tail of the ␤PDGFR. Receptor immunoprecipitates representing approximately 1.0 ϫ 10 6 cells were resolved by SDS-PAGE, transferred to Immobilon, and immunoblotted with the indicated antisera. In three experiments, similar amounts of RasGAP associated with the receptor at the earliest time points, whereas at the later time points (20 and 30 min) less RasGAP coimmunoprecipitated with receptor from cells cultured on fibronectin. B, the PDGFR immunoprecipitates shown in A were subjected to an antiphosphotyrosine Western blot analysis. C, the ␤PDGFR immunoprecipitates were subjected to an in vitro kinase assay in the presence of an exogenous substrate, GST-PLC␥; the proteins were resolved by SDS-PAGE, and the gel was subjected to autoradiography. a pan-Ras antibody; the signal was quantitated, and the fold activation was plotted as a function of time. Fig. 2A shows that PDGF-induced Ras activation comparably at the earliest time points under either culture conditions but that the levels of active Ras persisted longer in the cells cultured on fibronectin. Furthermore, the difference in Ras activation between the two culture conditions was most apparent at the times when Ras-GAP binding diminished. It is not obvious why receptors that bind RasGAP are able to activate Ras at the earliest time points (under either of the culture conditions), but this may be related to the timing of events that increase the level of active Ras. We also observed prolonged PDGF-dependent activation of Ras in cells cultured on fibronectin expressing a chimeric ␣/␤ PDGFR (data not shown).
To assess whether the ECM-dependent extension of Ras activation affected downstream signaling events, we examined PDGF-dependent Erk activation under different culture conditions. Ph cells were cultured, stimulated, and lysed as described in Fig. 1, and 30 g of Triton X-100 soluble cell lysate was subjected to Western blot analysis with antibodies that recognize phosphorylated p42 Erk and p44 Erk . Regardless of the culture conditions, the cells exhibited a PDGF-dependent in-crease in Erk phosphorylation that was greatest at the 5-min time point (Fig. 2B). For cells plated on polylysine, Erk activity returned to basal levels at approximately 30 min post-PDGF stimulation. In contrast, phospho-Erk persisted for at least 30 min in cells plated on fibronectin (Fig. 2B). These findings suggest that culturing cells on fibronectin decreases RasGAP recruitment to the receptor and regulates the duration of Ras and Erk activation.
RasGAP Regulation of Ras and Erk Using Receptor Mutants-The findings presented above showed that at times greater than 10 min post-PDGF, there was an inverse correlation between binding of RasGAP and the levels of active Ras. This suggests that at early times after exposure to PDGF, RasGAP plays little role in regulating the level of active Ras, and it is at the later time points when the presence of RasGAP results in a decrease in RasGTP levels. To investigate further the role of RasGAP in the kinetics of Ras activation, we compared the timing of Ras and Erk activation in cells expressing either WT or mutant ␤PDGFRs that do not bind RasGAP. Since Ph cells express the ␤PDGFR, we used ␣/␤ chimeric receptors, which can be selectively activated with PDGF AA (4). We constructed a mutant PDGFR chimera, N 2 F771, in which the tyrosine required for binding of RasGAP was mutated to phenylalanine. The wild-type (N 2 WT) and N 2 F771 chimeras were expressed in Ph cells to comparable levels (data not shown). As expected, the N 2 F771 receptor was not able to associate with RasGAP (data not shown). Additional characterization of the receptors indicated that they had comparable PDGF-stimulated kinase activity toward an exogenous substrate (data not shown). To compare the ability of N 2 WT and N 2 F771 chimera to activate Ras, cells were grown to 85-90% confluence, arrested by serum deprivation in DME ϩ 0.1% calf serum, and left resting or stimulated with PDGF AA for the times indicated. Active Ras was recovered with the GST-Raf fusion protein and quantitated as described in Fig. 2. At the 5-min time point, Ras was activated to a similar level in both cell types, indicating that the ability of RasGAP to associate with the receptor does not play a major role in the early accumulation of active Ras (Fig. 3A). At the later time points, the levels of active Ras diminished quickly in the N 2 WT receptor-expressing cells but persisted in the mutant receptor-expressing cells. Thus the ability to associate with RasGAP has the greatest effect on the levels of active Ras only at times greater than 5 min post-PDGF. We also measured Erk activation and found that like Ras activation, the differences between the two receptors became apparent at later time points (Fig. 3, B and C). In this series of studies, we obtained similar results using a phospho-Erk Western blot or monitoring kinase activity of Erk immunoprecipitates, strongly suggesting that either assay was an accurate measure of Erk activation. These studies indicate that binding of RasGAP to the ␤PDGFR does indeed alter PDGFdependent activation of Ras and that the effect is greatest after the initial accumulation of active Ras has occurred. Furthermore, preventing the association of RasGAP with the ␤PDGFR specifically at late time points post-PDGF appears to be the mechanism by which culturing cells on fibronectin prolongs the PDGF-dependent activation of Ras and Erk.
Molecular Basis for Decreased RasGAP Recruitment in Cells Cultured on Fibronectin-We next examined the molecular basis for the decreased recruitment of RasGAP to the ␤PDGFR in cells cultured on fibronectin. One possible explanation for the ECM effect is the action of protein tyrosine phosphatases, and a good candidate is SHP-2 which associates with the ␤PDGFR and selectively dephosphorylates phosphotyrosine 771 and 751 (35). To determine whether binding of SHP-2 to the ␤PDGFR was affected by the ECM of the cells, we com-

FIG. 2. Ras and Erk activation of cells plated on polylysine or fibronectin.
A, Ph cells were cultured and stimulated as described in Fig. 1. Active Ras was recovered from the total cell lysates representing 1.0 ϫ 10 7 cells by incubation with a GST fusion of Raf which contains the Ras binding domain. The recovered proteins were washed, resolved on an SDS-PAGE gel, transferred to Immobilon, and probed with a pan-Ras antibody. The bands were quantitated, and the data from three independent experiments were averaged and are plotted in A as fold activation Ϫ1. The arrows in A indicate the times at which RasGAP binding is reduced in PDGF-stimulated cells plated on fibronectin (Fig.  1). B, the cells were cultured, stimulated, and lysed as described in Fig.  1. Thirty g of clarified Triton X-100 soluble lysate was subjected to Western blot analysis with antibodies that recognize phospho-Erk (top panel). The bottom panel of each pair is a Western blot of the same samples and indicates the levels of protein (lysate standard). pared the amount of SHP-2 that co-precipitated with the ␤PDGFR in cells that had been cultured on either polylysine or fibronectin. We found that PDGF stimulated SHP-2 binding to the ␤PDGFR and that 2-3 times more SHP-2 was recovered in samples from cells that had been cultured on fibronectin (Fig.  4A). This effect appeared to be specific for SHP-2 since PDGFdependent binding of the p85 subunit of PI3K to the ␤PDGFR was comparable in cells cultured on polylysine or fibronectin (Fig. 4A). Despite the increased amount of SHP-2 that associated with the receptor at the earliest time points, a decrease in RasGAP binding was observed only at the later time points. It may be that the initial phase of PDGF-stimulated receptor kinase activity outweighs the initial phase of protein tyrosine phosphatase activity. Thus the results in Figs. 1A and 4 indicate that culturing cells on fibronectin selectively alters the proteins that are recruited to the ␤PDGFR. Furthermore, an increased association of SHP-2 correlates with the decreased binding of RasGAP. Consequently, we postulate that the ECM effect involves an increase in SHP-2 binding to the ␤PDGFR, which then decreases the time that RasGAP stays associated with the receptor, leading to prolonged Ras and Erk activity. Fig.  4 suggests that the ECM effects are mediated at least in part by SHP-2. If this is indeed the case, then cells expressing a ␤PDGFR that is unable to bind SHP-2 should be insensitive to the ECM. To test this idea, we compared the effect of ECM on PDGF-dependent Erk activation in cells expressing the F1009 ␤PDGFR. This receptor has a tyrosine to phenylalanine substitution at 1009 and does not bind SHP-2 effectively (27,36). This experiment could not be done in the Ph cells since they express the endogenous ␤PDGFR. Consequently, we chose F cells, which are nullizygous for both PDGFRs. The wild-type and F1009 ␤PDGFRs were each expressed to the level that is typically seen in fibroblasts (approximately 1 ϫ 10 5 receptors/ cell; data not shown). We first examined the ability of the WT and F1009 receptors to associate with SHP-2. Confluent, quiescent F cells expressing the WT or F1009 receptors were left resting or stimulated with 40 ng/ml PDGF BB for 5 min at 37°C. The cells were lysed, and the ␤PDGFR was immunoprecipitated with a receptor-specific antibody, and SHP-2 association was assayed by Western blot analysis. In response to the PDGF stimulation, the wild-type receptor bound SHP-2 in a PDGF-dependent manner, whereas SHP-2 was unable to efficiently associate with the mutant receptor (Fig. 5A, bottom  panel). Thus the F1009 ␤PDGFR was unable to bind SHP-2 when expressed in F cells, which is what we and others (3,27,36) have observed in several cell types.

Mutation of the SHP-2-binding Site Negates the Effect of Plating Cells on Fibronectin-The experiment presented in
We next examined if in this cell type, fibronectin has the same effect on RasGAP recruitment to the ␤PDGFR as in Ph cells. We immunoprecipitated the ␤PDGFR from F cells cultured on polylysine or fibronectin and examined RasGAP binding as described in Fig. 1A. Similar to what we observed in the Ph cells, RasGAP recruitment to the ␤PDGFR was decreased at 20 and 30 min post-PDGF stimulation in cells plated on fibronectin as compared with those cultured on polylysine (Fig.  5B). We then hypothesized that if the ECM effect requires SHP-2 binding to the ␤PDGFR, then cells expressing the F1009 receptor will be insensitive to the extracellular ECM. Consistent with our previous findings (36), we found that the F1009 receptor bound RasGAP slightly better than the WT ␤PDGFR. In addition, recruitment of RasGAP to the ␤PDGFR was similar in the F1009-expressing cells regardless of whether they were cultured on polylysine or fibronectin (Fig. 5C). Hence, mutation of the binding site for SHP-2 on the ␤PDGFR negates the effect of culturing cells on fibronectin. These findings FIG. 3. Receptors that cannot associate with RasGAP are better able to activate Ras and Erk. Ph cells expressing the N 2 WT or N 2 F771 receptors were grown to 95% confluence, starved overnight in DME ϩ 0.1% calf serum, and left resting (Ϫ) or stimulated with 25 (B and C) or 50 ng/ml (A) PDGF AA for the times indicated. A, Ras activation. Active Ras was recovered and detected as described in Fig. 2. B, Erk activation assessed by Western as described in Fig. 2. C, Erk activation as assessed by an in vitro kinase assay. The cells were washed and lysed, and Erk was immunoprecipitated. Erk immunoprecipitates representing approximately 4.0 ϫ 10 4 cells were subjected to an in vitro kinase assay in the presence of 0.75 g of myelin basic protein, and the proteins were resolved by SDS-PAGE, and the resulting gel was exposed to film. The extent of substrate phosphorylation was quantitated by excising the bands and counting them in a scintillation counter. The resulting data, calculated as fold activation Ϫ1, are plotted as a function of time.
strongly suggest that SHP-2 is a key regulator of changes in PDGFR signaling resulting from culturing cells on fibronectin.
Consequences of Decreased RasGAP Recruitment on PDGFdependent Biological Responses-We next determined if changes in PDGFR signaling had biological consequences. Previous reports have suggested that integrin engagement increased PDGF-dependent mitogenicity and chemotaxis (21,22). We tested the effect of ECM on PDGF-dependent responses in F cells. PDGF stimulated S phase entry in cells expressing the WT ␤PDGFR, and the response was modestly and reproducibly enhanced when cultured on fibronectin (Fig. 6A). At low and high doses of PDGF, the cells responded 61 and 67%, respectively, better on fibronectin than on polylysine. The ability of the F1009-expressing F cells to initiate DNA synthesis was also assessed. We found at low doses of PDGF, F1009 was less able than the WT-expressing cells to drive DNA synthesis, whereas at high doses of PDGF comparable responses are initiated (Fig. 6B). Importantly, culturing F1009-expressing cells on fibronectin did not improve the response, as it did with the WT ␤PDGFR. Collectively, these findings suggest that association of SHP-2 is required for cells to respond to the ECM and are consistent with the idea that SHP-2 is the liaison between integrins and ␤PDGFR signaling. DISCUSSION To test if culturing cells on ECMs alters PDGFR signaling, we compared the amount of signal relay enzymes that associate with the ␤PDGFR. Indeed, we found that the composition of the complex of receptor/signaling enzymes was sensitive to the ECM. RasGAP recruitment to the ␤PDGFR was decreased in cells cultured on fibronectin, and the net effect was an increase in the duration of Ras activation and pathways downstream of Ras, such as Erk. Furthermore, decreased RasGAP binding to the ␤PDGFR appeared to be a consequence of increased recruitment of SHP-2 to the ␤PDGFR. Finally cells expressing receptors that did not associate with SHP-2 were not affected by the ECM. Hence, integrins change signaling initiated by the ␤PDGFR by altering the ability of the receptor to recruit signal relay enzymes.
These studies strongly suggest that ECM modulates the wild-type receptor via SHP-2, which dephosphorylates the ␤PDGFR at phosphotyrosine 771 in the RasGAP-binding site.
Previous studies with mutant PDGFRs support the idea that this is one of the functions of SHP-2 (35) and are consistent with the findings of investigators studying Torso signal relay (15). Corkscrew (the Drosophila homolog of SHP-2) selectively dephosphorylates the Torso receptor at the tyrosine required for binding of D-RasGAP (15). Furthermore, dephosphorylation at the D-RasGAP-binding site modulates pathways downstream of GAP and the strength of Torso signaling (15,16). Hence, SHP-2 appears to play an important role in modulating receptor tyrosine kinase signaling via dephosphorylation of the tyrosine required for binding of RasGAP.
In addition to decreasing the duration of RasGAP binding to the ␤PDGFR, SHP-2 may also promote Ras activation via its ability to recruit Grb2 (37,38). Tyrosine phosphorylation of SHP-2 has been shown to enable SHP-2 to associate with Grb2. Furthermore, permitting ␤PDGFR mutants to associate with SHP-2 enhances activation of Ras (3). However, the timing of ECM-dependent enhancement of Ras activation correlated more strongly with a decrease in RasGAP binding than an increase in SHP-2 binding (Figs. 1 and 2). Thus it is possible that in the experiments described herein, SHP-2 acts to increase Ras activation by decreasing binding of RasGAP with the ␤PDGFR.
The findings that SHP-2 mediates the ECM effect initiate the question of how plating cells on fibronectin leads to increased association of SHP-2 with the ␤PDGFR. Integrin engagement has been shown to increase phosphorylation of the ␤PDGFR, in a PDGF-independent manner (39), and perhaps the receptor is phosphorylated better in cells plated on fibronectin. However, we did not detect any tyrosine phosphorylation of the receptor or association of SHP-2 in unstimulated cells, regardless of the ECM onto which the cells had been cultured (Fig. 1). Thus the difference appears to be a post-PDGF event and involves recruitment of SHP-2 to the ␤PDGFR.
It is possible that the ECM-dependent increase in association of SHP-2 with the ␤PDGFR is due to an enhancement in the availability of SHP-2 to the ␤PDGFR. Since the ␤PDGFR is localized to the plasma membrane and SHP-2 is cytosolic, SHP-2 must be recruited to the plasma membrane to bind to the PDGFR. Recent findings indicate that integrin aggregation induced localized membrane accumulation of SHP-2 (40), and integrin-induced activation of Src is a key step in these events (41). SHPS-1 (42)/SIRP (43) has recently been identified as a membrane glycoprotein that was tyrosine-phosphorylated and associated with SHP-2 in response to binding of integrins to ECM components, such as fibronectin and laminin (41). Thus culturing cells on an appropriate ECM increases tyrosine phosphorylation of SHPS-1, which in turn recruits SHP-2 to the plasma membrane. Such a change in the subcellular localization of SHP-2 could increase its availability to the ␤PDGFR and result in greater association with the ␤PDGFR once cells are exposed to PDGF.
The findings presented here raise a number of important questions about the nature of ␤PDGFR signaling. It is somewhat curious that whereas there is noticeably more SHP-2 associated with the ␤PDGFR at 5 min post-PDGF stimulation, the decrease in the amount of RasGAP bound to the ␤PDGFR is visible only at 30 min post-PDGF stimulation. This could arise because the steady state level of phosphorylation of tyrosine 771 (binding site for RasGAP) is a balance of phosphatase and kinase activity. At early time points (5 min) post-PDGF, SHP-2 could be dephosphorylating the phosphotyrosine at position 771, but the kinase activity of the receptor rephosphorylates it. At later time points (20 and 30 min), the kinase activity of the receptor begins to decrease, and the balance is shifted in favor of the phosphatase activity. The effect is that the ␤PDGFR kinase is no longer able to re-phosphorylate the Ras-GAP-binding site, and a decrease is observed in the amount of RasGAP that associates with the ␤PDGFR. Consistent with this hypothesis, we have found that the kinase activity of the ␤PDGFR begins to decrease 30 min post-PDGF stimulation (Fig. 1C).
Hence, during the first 30 min of PDGF stimulation, the nature of the signaling events are dynamic and can be influenced by the ECM of the cell. It is possible that multiple extracellular signals are integrated at these early times. Alternatively, we and others (31) 2 have detected signaling events that occur in mid to late G 1 , and therefore, it is possible that the ECM also influences these events. Identification of the times during G 1 progression that key signaling events occur, as well as the identity of the signaling enzymes involved, will facilitate future studies focused on how multiple environmental cues are coordinated and impact a cellular response.
FIG. 6. The affect of ECM on DNA synthesis in F cells expressing the WT or F1009 ␤PDGFR. F cells expressing wild-type (A) or F1009 (B) were plated at low cell density on dishes coated with polylysine (open boxes) or fibronectin (black boxes) and allowed to become quiescent. Buffer, 10% fetal bovine serum, or the indicated concentration of PDGF was added, and after 22 h the cells were pulsed with [ 3 H]thymidine and then harvested. Treatment with 10% FBS resulted in a 12-14-fold increase in DNA synthesis, which was consistently independent of the ECM. The data are expressed as a fold activation over the response to buffer. The experiment was performed three times, and this graph is representative of one of those three trials. Each condition was assayed in triplicate, and the data presented are the mean Ϯ S.D.