Activation of c-Src and Fyn Kinases by Protein-tyrosine Phosphatase RPTPα Is Substrate-specific and Compatible with Lipid Raft Localization*

Src family tyrosine kinases (SFKs) function in multiple signaling pathways, raising the question of how appropriate regulation and substrate choice are achieved. SFK activity is modulated by several protein-tyrosine phosphatases, among which RPTPα and SHP2 are the best established. We studied how RPTPα affects substrate specificity and regulation of c-Src and Fyn in response to epidermal growth factor and platelet-derived growth factor. We find that RPTPα, in a growth factor-specific manner, directs the specificity of these kinases toward a specific subset of SFK substrates, particularly the focal adhesion protein Paxillin and the lipid raft scaffolding protein Cbp/PAG. A significant fraction of RPTPα is present in lipid rafts, where its targets Fyn and Cbp/PAG reside, and growth factor-mediated SFK activation within this compartment is strictly dependent on RPTPα. Forced concentration of RPTPα into lipid rafts is compatible with activation of Fyn. Finally, RPTPα-induced phosphorylation of Paxillin and Cbp/PAG induces recruitment of the SFK inhibitory kinase Csk, indicative of negative feedback loops limiting SFK activation by RPTPα. Our findings indicate that individual SFK-controlling PTPs play important and specific roles in dictating SFK substrate specificity and regulatory mechanism.

Src family kinases (SFKs), 3 for which c-Src is the prototype, play crucial and partly overlapping roles in proliferation, differ-entiation, adhesion, motility, and survival (1,2). The ubiquity of their involvement in signaling pathways (downstream of integrins, growth factor, antigen, and G protein-coupled receptors among others), and the multitude of substrates they can phosphorylate, generate a need for mechanisms that confer stimulus-and context-appropriate specificity. SFK activity is regulated by intramolecular interactions that depend on an equilibrium between tyrosine phosphorylation and dephosphorylation. In the unstimulated state, catalytic activity is constrained by intramolecular interactions, such as engagement of the SH2 domain by a phosphorylated C-terminal tyrosine (generically referred to here as Tyr 527 after the chicken c-Src sequence). Disruption of these interactions, e.g. through engagement of SFK SH2 and SH3 domains by third proteins and by dephosphorylation of Tyr(P) 527 results in SFK activation. Activation is characterized by autophosphorylation of a conserved tyrosine in the activation loop, generically referred to as Tyr 416 (after the chicken c-Src sequence).
Two kinases are firmly implicated in phosphorylation of Tyr 527 , C-terminal Src kinase Csk, and Ctk; expression of the former is ubiquitous, whereas the latter is more restricted (e.g. to neuronal and hemopoietic tissue, and breast cancer cell lines). In hemopoietic cells, the inhibitory effect of Csk on SFK activity is strengthened further through dephosphorylation of Tyr 416 by the Csk-associated PTP PEP (3). By contrast, an increasing number of protein-tyrosine phosphatases (PTPs) are being implicated in controlling Tyr 527 phosphorylation and thus SFK activity. Such SFK-activating PTPs include both nonreceptor (e.g. PTP1B, SHP1, and SHP2) and receptor PTPs (RPTPs, such as CD45, RPTP␣, RPTP⑀, and LAR) (4,5). PTP regulation of SFK function is rendered even more complex by the variety of underlying mechanisms. SHP2 reduces SFK Tyr 527 phosphorylation indirectly, by dephosphorylating Cbp/ PAG (an adaptor protein) and thus controlling Csk recruitment into the vicinity of SFKs, leading to net SFK activation (6). CD45 likely dephosphorylates Tyr(P) 527 of SFKs directly in hemopoietic cells but also acts on their conserved Tyr(P) 416 , thus combining positive and negative functions (7). We and others have shown that another PTP expressed more ubiquitously than CD45, RPTP␣, activates c-Src and Fyn by reducing phosphorylation of Tyr(P) 527 ; this activation is important for adhesionrelated signaling in fibroblasts (8,9). Physical interaction between RPTP␣ and c-Src (10) suggests that RPTP␣ dephosphorylates Tyr(P) 527 directly, but the alternative possibility of Csk regulation by RPTP␣ has never been investigated. In unstimulated thymocytes, RPTP␣ suppresses Fyn by reducing not only phosphorylation of Fyn Tyr 527 but also Tyr 416 (11).
The involvement of multiple PTPs in SFK regulation generates a confusing picture, with two questions standing out. First, the multiplicity of co-expressed SFK-activating PTPs (e.g. fibroblasts express RPTP␣, PTP⑀, SHP2, and PTP1B) raises the question of their respective individual contributions. SFKs can phosphorylate a very large number of substrates, such as focal adhesion proteins (focal adhesion kinase, Paxillin, and p130 cas ), cytoskeletal proteins (cortactin, catenins, and caveolin), enzymes (phosphatidylinositol 3-kinase, phospholipase C␥, p190 RhoGAP, and Syk/ZAP), adaptors (Shc, Cbl, and DOK), and receptor proteins. Do these various PTPs all modulate SFK activity globally, or is their role distinct where these various SFK substrates are concerned? Second, how are the potential activating and inhibitory effects of PTPs on SFKs (through dephosphorylation of Tyr(P) 527 and Tyr(P) 416 , respectively) on SFKs coordinated? SFKs are typically concentrated in specialized domains such as supramolecular activation clusters, focal adhesions, or lipid rafts. However, little is known about the intracellular distribution of PTPs and their dynamics in response to extracellular stimuli. It has been postulated that net activation of an SFK by a PTP requires a measure of physical segregation between both, to specifically favor dephosphorylation of Tyr(P) 527 , while allowing SFK-activating autophosphorylation on Tyr 416 to occur in an unimpeded manner (12)(13)(14). Conversely, in this model, co-concentration in the same compartment of an SFK with its controlling PTP would favor net downregulation of SFK activity. This hypothesis has received experimental support in the case of CD45 regulation of lipid microdomain-associated SFK activity and T-cell receptor signaling (15), but its wider relevance remains unknown.
Here we address these questions in the case of RPTP␣ during growth factor stimulation of fibroblasts. We find that RPTP␣ enhances the basal and growth factor (EGF and PDGF)-induced kinase activities of c-Src and Fyn but differentially affects tyrosine phosphorylation of individual SFK substrates in a growth factor-specific manner. We demonstrate that one quarter of cellular RPTP␣ is found in lipid rafts. SFK activation by growth factors in this compartment is absolutely dependent on RPTP␣ and occurs concurrently with additional recruitment of RPTP␣. Moreover, manipulative overconcentration of RPTP␣ into lipid rafts also preserves its ability to activate the lipid raftassociated SFK Fyn. Our findings suggest a model in which individual PTPs specifically shepherd SFK activity toward distinct local SFK pools and SFK substrate repertoires. They do not support a role for physical segregation between this PTP and its targets for achieving net activation of lipid raft-localized SFKs.

EXPERIMENTAL PROCEDURES
Reagents-Enolase from rabbit muscle was purchased from Sigma. Recombinant human EGF was from Chemicon, and human PDGF-BB was from Peprotech. Protein A-and G-Sepharose were purchased from Amersham Biosciences and Upstate Biotechnology, Inc., respectively.
Retroviral Constructs-pBabe-WT-RPTP␣, a retroviral vector expressing human WT RPTP␣, was a gift from Leslie Holzinger and Bahija Jallal (Sugen, Inc.). Lipid raft-targeted RPTP␣ (F-PTP␣) was constructed as follows. A 5Ј primer was designed containing a BamHI site, a Kozak sequence, a start codon, and the first 16 N-terminal amino acids of Fyn fused to the first 6 amino acids of RPTP␣ intracellular region (GCGGGATCCA-CCATGGGCTGTGTGCAATGTAAGGATAAAGAAGCA-GCGAAACTGACGAGAGGTTTAAGAAATACAAGC). A 3Ј primer that began 1420 bases from RPTP␣ start codon (GAC-ATACTGCATATCGGTTTGCACCATCTG) was also designed. The PCR product generated from these two primers using pBabe-WT-RPTP␣ as template was digested by BamHI and used to replace the corresponding fragment in pBabe-WT-RPTP␣. A retroviral vector expressing mouse Cbp/PAG (pBabe-Cbp/PAG) was also constructed as follows. An expressed sequence tag clone of mouse PAG (IMAGp998N023527Q1) cloned in pT7T3D-PacI was purchased from ImaGenes GmbH (Berlin, Germany). In a first step, the fragment BamHI-Cbp-EcoRI was cloned into pBabe. In parallel, the Cbp/PAG N terminus was amplified by PCR using the expressed sequence tag clone as template and ATTGCCGGCAC-CATGGGCCCTGCAGGAAGCGTACTGAGCAGT and CTC-ACAGCCAGGGCTGCAGACACG as 5Ј and 3Ј primers. After digestion by Nae1 and BamHI, the PCR product was inserted into pBabe containing the fragment BamHI-Cbp-EcoRI.
Preparation of Embryonic Mouse Fibroblasts and Thymocytes-The RPTP␣-deficient (Ptpra Ϫ/Ϫ ) mice have been previously described (8), and the allele was maintained on a C57Bl6 background. To establish littermate-matched WT and Ptpra Ϫ/Ϫ mouse primary fibroblast cultures, heterozygote mice were mated. At embryonic day 13.5, embryos were extracted from pregnant females, placed in cold PBS, and processed and genotyped individually. Internal organs were discarded, and the head was used for genotyping. The remaining tissue was dissociated by three passages through a 16-gauge needle in 5 ml of 0.25% Trypsin-EDTA solution (Invitrogen), transferred into a sterile Erlenmeyer containing glass beads and a magnetic stirrer, and digested further for 5 min at 37°C. 5 ml of DMEM (Invitrogen) supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin were added to the mixture. Cells from each embryo were seeded in two 10-cm dishes and cultured at 37°C. Cells from multiple embryos of the same genotype were pooled. For thymocyte preparations, thymi from 9-week-old WT or Ptpra Ϫ/Ϫ mice were dissected and smashed between two sterile frosted glass slides in 5 ml of DMEM containing 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin to obtain cell suspensions.
Cell Cultures, Infections, and Stimulations-RPTP␣-deficient spontaneously immortalized mouse fibroblasts have been previously described (8), and were maintained in DMEM with 10% fetal bovine serum, 100 units/ml penicillin and 100 g/ml streptomycin. These cells were infected either with pBabe, pBabe-WT-RPTP␣, pBabe-F-PTP␣, or pBabe-CS-RPTP␣ to express WT, lipid raft-targeted, or catalytically inactive RPTP␣, respectively. Infected cells were selected as pools and cultured with 4 g/ml puromycin (Sigma). Before stimulation, the cells were starved for 24 h in DMEM with 0.5% fetal bovine serum and stimulated for the times indicated with growth factors (50 ng/ml EGF or 25 ng/ml PDGF). When necessary, the cells were pretreated with 10 M PP1 (Calbiochem) or SU6656 (provided by Dr. S. Courtneidge, Sugen Inc.) for 1 h at 37°C before growth factor exposure.
In Vitro Src and Fyn Kinase Assays-After stimulation, the cells were washed two times with cold PBS and lysed on ice with radioimmune precipitation assay buffer (50 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EGTA, 1% Triton X-100, 1% sodium deoxycholate, 0,1% SDS, 1 mM Na 3 VO 4 , 50 mM NaF, 10 g/ml leupeptin, 10 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride). Equal amounts of proteins from cleared lysates were used to immunoprecipitate c-Src or Fyn. Immune complexes were washed four times with cold radioimmune precipitation assay buffer followed by two washes with cold TBS (20 mM Tris, pH 7.5, 150 mM NaCl) and reacted in 50 l of kinase buffer containing 10 Ci of [␥-32 P]ATP and 12.5 g of aciddenatured enolase at 30°C for 10 min. After electrophoresis and transfer onto polyvinylidene difluoride membranes, membranes were incubated for 1 h at 55°C in 1 M KOH. Phosphorylation of enolase was quantified with a phosphorimaging device (Fujifilm).
Lipid Raft Isolations-After stimulation, the cells were washed two times with cold PBS and lysed on ice in 25 mM Hepes, pH 7.1, 150 mM NaCl, 1% Triton, 1 mM Na 3 VO 4 , 50 mM NaF, 10 g/ml leupeptin, 10 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Equal amounts of proteins were adjusted to 2 ml with lysis buffer, placed in the bottom of an ultracentrifuge tube (344060; Beckman Coulter), and mixed with an equal volume of 80% sucrose (w/v) in lysis buffer. The samples were overlaid with 6 ml of 35% sucrose and 2 ml of 15% sucrose in lysis buffer and centrifuged at 35,000 rpm in an SW40Ti rotor for 17-20 h at 4°C. The gradients were separated into 12 fractions of 1 ml from the top. Equal aliquots of each fraction were precipitated with an equal volume of acetone for 15 min on ice and centrifuged at 13,000 rpm for 15 min. at 4°C. The pellets were resuspended in Laemmli buffer and resolved by SDS-PAGE for immunoblotting.

RPTP␣ Is Required for Basal and Growth Factor-induced SFK Activity in Fibroblasts and Is Particularly Important for SFK
Stimulation by EGF-We and others have previously reported that RPTP␣ activates SFKs during integrin-mediated responses in fibroblasts (8,9,17). Because SFKs are also activated during growth factor stimulation, we investigated the possible role of RPTP␣ in this process by measuring basal and growth factorinduced c-Src and Fyn kinase activity in a spontaneously immortalized Ptpra Ϫ/Ϫ fibroblast cell line (8) rescued with either WT RPTP␣ or a vector alone (Fig. 1A).
It has been reported that the SFK epitope against which antibodies are directed can affect the ability to detect changes in in vitro SFK kinase activity; for instance, antibodies against the C terminus detect the largest increases in SFK activity (18). However, we feared that interpretations of experiments using such antisera could be biased by differential recognition of phosphorylated versus nonphosphorylated proteins. For our analysis, we used anti-Src and -Fyn antibodies directed against the N-terminal domains of the kinases. We found that RPTP␣ significantly increased both basal and growth factor-induced activity of c-Src (Fig. 1B) and Fyn (Fig. 1C).
There were significant differences in the effect of RPTP␣ on activation of these kinases by PDGF versus EGF. Most particularly, RPTP␣ was required for the ability of EGF, but not PDGF, to activate Fyn (Fig. 1C). For c-Src (Fig. 1B), there was a similar tendency toward an obligate requirement for RPTP␣ in EGFinduced c-Src activation, although significance was at the borderline (p ϭ 0.06). Under our conditions, we did not observe significant activation of c-Src by PDGF, neither in the absence nor in the presence of RPTP␣; this discrepancy from findings by others (18) may reflect the cell type used (spontaneously immortalized mouse fibroblasts) and/or the use of an antibody against the c-Src N terminus.
To follow up on our findings using an independent approach, we performed additional experiments on primary mouse fibroblasts derived in parallel from WT and Ptpra Ϫ/Ϫ embryos. Furthermore, as an alternative way of investigating SFK kinase activity, we assessed phosphorylation of SFK Tyr 416 , which is autophosphorylated along with SFK activation (1). A phosphospecific antibody reacting with this site mainly detects a single 60-kDa species, most likely c-Src, although because of the lack of gel resolution, this may also incorporate a component of the 59-kDa Fyn protein (Fig. 1D). Consistent with previous studies, in the absence of growth factor stimulation, basal Tyr 416 phosphorylation of SFKs was lower in RPTP␣-deficient than in WT fibroblasts. In response to EGF, Tyr 416 phosphorylation was increased in the presence of RPTP␣ but not in its absence. To determine whether the RPTP␣ effect on SFK activation was dependent on PTP catalytic activity, we also examined the effect of expression of a catalytically inactive mutant of RPTP␣ on SFK Tyr 416 phosphorylation. Expression of this mutant protein, in which the catalytic cysteine residues in both PTP DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 domains have been mutated to serine, left the levels of basal and EGF-induced SFK activation comparable with empty vectorinfected control cells, suggesting that the effect of RPTP␣ requires the catalytic activity of the protein (supplemental data Fig. S2). In response to PDGF, SFK activation occurred both in the presence and absence of RPTP␣ (Fig. 1D) (p values for induced versus uninduced in RPTP␣-deficient cells: 0.226 (n ϭ 6) for EGF and 0.022 (n ϭ 7) for PDGF). Taken together, these data show an excellent correspondence between reconstituted immortalized cells and primary fibroblasts using two different assays for SFK function: the in vitro kinase assays and the Tyr 416 phosphorylation levels. In both settings, RPTP␣ played a more important role in EGF-induced than in PDGF-induced SFK activation, indicating that its function in SFK activation is stimulus-specific.

RPTP␣ and Substrate Specificity of Src Family Kinases
RPTP␣ Ablation Differentially Affects Tyrosine Phosphorylation of SFK Substrates-Next, we assessed the role of RPTP␣ for phosphorylation of various individual SFK substrates in vivo, by comparing WT and Ptpra Ϫ/Ϫ primary fibroblasts. We and others already reported that the absence of RPTP␣ compromises phosphorylation of various substrates involved in integrin signaling (8,9). Here, we focused on a series of signaling intermediates whose tyrosine phosphorylation in response to growth factors is known to be mediated by SFKs, i.e. STAT-3, Vav2, Cbl, Paxillin, and Cbp/PAG (19 -21).
As expected, we found that both EGF and PDGF enhanced phosphorylation of all five proteins in WT cells. Strikingly, tyrosine phosphorylation of STAT3 was entirely independent of RPTP␣ at all time points ( Fig. 2A and data not shown). We confirmed that, consistent with the literature (19), tyrosine phosphorylation of STAT3 was indeed SFK-dependent, because it could be entirely abolished by treatment with two established SFK inhibitors, PP1 and SU6656 ( Fig. 2A).
RPTP␣ ablation had only very minor effects on phosphorylation of Vav2 or Cbl. Following acute (10 min) stimulation with either EGF or PDGF, phosphorylation of these two proteins was indistinguishable between WT and RPTP␣-deficient cells. In the basal state, a slight reduction in Vav2 phosphorylation was observed in the absence of RPTP␣; this reduction reappeared at the late time points for EGF (60 min) but not for PDGF (60 or 120 min) (Fig.  2B). Cbl phosphorylation showed no differences between genotypes at the basal state or after acute (10 min) stimulation but a more transient phosphorylation in response to EGF in RPTP␣-deficient cells (Fig. 2C). By contrast, the absence of RPTP␣ led to a significant reduction in tyrosine phosphorylation of the SFK substrate and focal adhesion protein Paxillin. Growth factor stimulation only marginally altered the level of Paxillin tyrosine phosphorylation in both cell types, and its tyrosine phosphorylation in RPTP␣deficient cells essentially was similarly reduced at all time points (Fig. 2D). Of note, the absence of RPTP␣ did not affect the growth factor-induced mobility shift of Paxillin, which is thought to reflect serine/threonine phosphorylation (22).
In contrast to Paxillin, tyrosine phosphorylation of the lipid raft Fyn substrate Cbp/PAG was clearly increased by growth factors in our hands, in WT as well as in RPTP␣deficient cells, and this increase persisted for at least 60 min. The absence of RPTP␣ lowered the level of Cbp/PAG phosphorylation to the same extent under all conditions and thus did not affect growth factor inducibility of Cbp/PAG phosphorylation per se (Fig. 2E). Taken together, this analysis shows that the impaired SFK activity seen in absence of RPTP␣ ( Fig. 1) does not affect tyrosine phosphorylation of SFK substrates indiscriminately; rather, RPTP␣ is of selective importance for the control of phosphorylation of a specific subset of SFK targets.
RPTP␣ Enhances Paxillin Phosphorylation on Tyr 31 and Tyr 118 and Recruitment of Csk to Paxillin-Paxillin is a scaffolding protein phosphorylated by SFKs on several tyrosines, among which Tyr 31 and Tyr 118 have been the best characterized (23). Investigating how phosphorylation of these individual sites was dependent on SFK activation by RPTP␣, we found that both Tyr 31 and Tyr 118 were hypophosphorylated in Ptpra Ϫ/Ϫ cells (Fig. 3A).
Phosphorylated Paxillin Tyr 31 and Tyr 118 constitute, among others, binding sites for Csk (23), itself a key negative regulator of SFK activity via phosphorylation of SFK C-terminal tyrosine. Accordingly, we determined whether Csk recruitment on Paxillin was affected in the absence of RPTP␣, by immunoprecipitating Csk from cell lysates and immunoblotting the membrane with an anti-Paxillin antibody. Under basal conditions, the amount of Paxillin bound to Csk was lower in RPTP␣-deficient fibroblasts than in their WT counterparts. This difference persisted after growth factor stimulation (Fig. 3B). Taken together, our results indicate that enhancement of Paxillin tyrosine phosphorylation via RPTP␣-activated SFKs causes Csk recruitment, resulting in a feedback loop that may limit SFK activation by RPTP␣. Comparison of tyrosine phosphorylation of individual SFK substrates in WT and RPTP␣-deficient fibroblasts following growth factor stimulation. Serum-starved WT or RPTP␣-deficient embryonic fibroblasts were treated with either EGF or PDGF for the indicated times. Quantifications of band intensities normalized to 1 for unstimulated WT cells are given under each panel. A, lysates resolved by SDS-PAGE were immunoblotted with an anti-phospho-STAT3 antibody. The blots were stripped and reprobed with an anti-STAT3 antibody. Alternatively, cells were pretreated with PP1 and SU6656 (SU) for 1 h at 37°C and stimulated with EGF (E) for 10 min. No differences were observed between WT and RPTP␣-deficient cells at 30 and 60 min after stimulation (data not shown). B, Vav2 was immunoprecipitated (IP) from cell lysates and probed for phosphotyrosine content. The amount of Vav2 in the immunoprecipitates was controlled for by immunoblotting with anti-Vav2 (a representative of two experiments is shown). C, the same as B, but with Cbl. D, immunoprecipitation of Paxillin blotted with anti-phosphotyrosine and anti-Paxillin (one representative of each experiment is shown of two for EGF and three for PDGF stimulation). E, Cbp/PAG immunoprecipitates revealed by anti-phosphotyrosine and anti-Cbp/PAG antibodies (a representative of two experiments is shown for each growth factor).

RPTP␣ Is Localized in Fibroblast Lipid Rafts and Is Strictly Required for Activation of Raft-associated SFKs by Growth
Factors-Both Fyn and its substrate, the scaffolding protein Cbp/PAG, are characteristically and highly enriched in the "lipid raft" compartment of the plasma membrane (24,25). Colocalization with substrates is a well established mechanism for conferring specificity upon phosphorylation and dephosphorylation reactions. Overexpressed RPTP␣ has been observed in plasma membrane and in an intracellular vesicular compartment (26). Mutant versions of RPTP␣ have been observed in focal adhesion plaques (27). Our own immunofluorescence analysis of WT RPTP␣ protein in fibroblasts reveals a diffuse membrane stain (data not shown). Whether and how localization affects RPTP target specificity is currently a matter of uncertainty and even contention. It has been suggested that the ability of RPTP␣ and CD45 to activate SFKs requires physical separation between these PTPs on the one hand and their target SFKs and the SFK substrates on the other. Co-localization of these PTPs with their substrates could result in net negative consequences, because of loss of specificity for the Tyr 527 site over the Tyr 416 site and over the SFK substrate sites. This hypothesis has been invoked to explain the inhibitory effect of CD45 in macrophage adhesion sites, e.g. on Hck and Lyn (13), and more recently the inhibitory effect of RPTP␣ on Fyn in thymocyte rafts (11). Because the RPTP␣ targets Fyn and Cbp/ PAG are significantly concentrated in membrane lipid rafts (28), we investigated whether or not RPTP␣ also localized to these structures.
We prepared detergent-resistant membranes, which contain the membrane raft fraction, from primary WT fibroblasts by ultracentrifugation-mediated fractionation on sucrose gradients (29). As shown in Fig. 4A, lipid rafts of WT fibroblasts were present in fractions 1-3, as indicated by the markers Caveolin-1, Fyn, and Cbp/PAG. The enrichment of Cbp/PAG in fibroblast rafts is absolute, and the majority of Fyn is also present in the raft fraction. By contrast, Paxillin was completely excluded from lipid raft fractions, and Csk was present both outside and inside lipid rafts, as previously reported (25). We found a substantial fraction (24%) of RPTP␣ inside of fibroblast lipid rafts (Fig. 4A). In addition, growth factor stimulation led to a modest but reproducible recruitment of additional RPTP␣ to lipid rafts (Fig.  4B, representative of three independent experiments).
The recruitment of RPTP␣ to lipid rafts in response to EGF and PDGF (Fig. 4B) occurs concurrently with activation of the total cellular pools of the SFKs c-Src and Fyn (Fig.  1). To investigate specifically how RPTP␣ affects the pools of SFK present within lipid rafts (predominantly Fyn (24,25)), we subjected the isolated raft fractions to immunoblotting with an anti-SFK Tyr(P) 416 antibody. We found that, in the unstimulated state, SFK phosphorylation in lipid rafts on Tyr 416 is impaired in absence of RPTP␣ (Fig.  4C). Moreover, stimulation with either growth factor activated the SFK pool inside lipid rafts in a manner that was absolutely dependent on the presence of RPTP␣. This contrasts with the situation for the total cellular SFK pool, which remains PDGFresponsive in the absence of RPTP␣ (Fig. 1D, right panels).
Subjecting raft fractions to anti-phosphotyrosine blotting, we observed that the absence of RPTP␣ decreased phosphorylation of a protein of approximately 75 kDa that co-migrates with Cbp/PAG (Fig. 4C, arrows). However, because of technical limitations (sensitivity of immunoprecipitation after sucrose gradient centrifugation), we were not able to definitely identify this protein as Cbp/PAG. Taken together, these results indicate that, in fibroblasts, a substantial fraction of RPTP␣ is present in lipid rafts, where it is strictly required for activation of SFK activity by growth factors and for phosphorylation of a major lipid raft protein likely to be Cbp/PAG.
Lipid Raft-targeted RPTP␣ Is a Fyn Activator in Fibroblasts-It has been proposed, based on studies in thymocytes, that colocalization of RPTP␣ with Fyn in the detergent-resistant lipid raft compartment would result in dephosphorylation of both the Tyr 416 -and Tyr 527 -homologous sites in Fyn, with the net consequence of repressing Fyn activity (11). This observation contrasts with our results on primary fibroblasts (Fig. 4). Direct evaluation of the relevance of RPTP␣ recruitment to lipid rafts FIGURE 3. RPTP␣ enhances the tyrosine phosphorylation of Paxillin and its association with Csk. Serumstarved WT or RPTP␣-deficient embryonic fibroblasts were treated with either EGF or PDGF for the indicated times. A, cell lysates resolved by SDS-PAGE were immunoblotted with phosphotyrosine-specific Paxillin antibodies. After stripping, the blots were probed with an anti-Paxillin antibody. B, Csk was immunoprecipitated from cell lysates, and the amount of Paxillin co-immunoprecipitated was visualized by immunoblotting with an anti-Paxillin antibody. The blots were stripped and reprobed with an anti-phosphotyrosine antibody to visualize Paxillin phosphorylation and with an anti-Csk antibody.
for the choice between activation and inhibition of SFKs in this compartment requires experimental manipulation of the ratio of RPTP␣ inside and outside lipid rafts. Such a study has never been performed.
We engineered a lipid raft membrane-targeted RPTP␣ protein (F-PTP␣) by fusing the 16 N-terminal amino acids of Fyn (responsible for targeting Fyn to lipid rafts (30)) to the intracellular catalytic domain of RPTP␣. We expected this Fyn-RPTP␣ fusion to faithfully recruit the RPTP␣ catalytic domain to the same membrane compartment as the Fyn protein. The con-struct, or its WT RPTP␣ counterpart as control, was then stably reintroduced into an immortalized RPTP␣-deficient fibroblast cell line. In this context, we generated two pools of fibroblasts, obtained using different multiplicity of infection (Fig. 5A). The WT RPTP␣ protein assumed an approximately equal distribution between raft and nonraft fractions (Fig. 5A), largely consistent with our results on primary cells in Fig. 4A. F-PTP␣ displayed a lower molecular weight (because it lacks the ectodomain) and, as expected, was highly enriched into the lipid raft fraction in both the low and high expressing pools. F-PTP␣ Low expresses F-PTP␣ in lipid raft fractions at a level comparable with that of WT RPTP␣ in this compartment, whereas the nonraft level of the protein remains much lower than that of its WT counterpart. By contrast, the level of F-PTP␣ High protein level in lipid rafts is higher than that normally seen for WT RPTP␣ inside rafts, with the amount of F-PTP␣ in nonraft fractions now being roughly comparable with that in cells expressing the WT protein. As an important control, anti-Fyn immunoblotting shows that expression of the F-PTP␣ protein does not result in competition and subsequent exclusion of endogenous Fyn protein from lipid rafts ( Fig. 5A and data not shown).
We used these rescued pools to examine the effect on Fyn kinase activity of enhancing the ratio between RPTP␣ within and outside of lipid rafts. Fig. 5B shows that, consistent with our observations in Fig. 1C and with previously reported experiments (8,9), expression of WT RPTP␣ significantly increased Fyn activity. Importantly, expression of either low or high levels of lipid raft-targeted F-PTP␣ similarly increased the kinase activity of Fyn in total cell lysate (Fig. 5B). Lipid raft-associated SFK activity and phosphorylation status were also assessed by Western blotting using phosphospecific antibodies against Tyr(P) 416 and Tyr(P) 527 . We found that both WT RPTP␣ and F-PTP␣ reduced Tyr 527 phosphorylation, while increasing Tyr 416 phosphorylation of the SFK pool in the lipid raft fraction (supplemental data Fig. S1). These observations indicate that even considerable enrichment of RPTP␣ into fibroblast lipid FIGURE 4. A fraction of RPTP␣ is localized in lipid rafts where it controls SFK activity and tyrosine phosphorylation. A, WT embryonic mouse fibroblasts were subjected to sucrose density gradient centrifugation. Fractions were removed from the top of the tube and probed for the indicated proteins. B and C, serum-starved WT or RPTP␣-deficient embryonic mouse fibroblasts were treated with either EGF or PDGF for the indicated times, lysed in 1% Triton X-100-containing buffer, and subjected to sucrose density gradient centrifugation. Fractions 1-3 containing lipid rafts were pooled and resolved by SDS-PAGE. B, membranes were blotted with an anti-RPTP␣ antibody, stripped, and reprobed with an anti-Caveolin 1 antibody. C, blots were probed with anti-Tyr(P) 416 SFK, anti-Fyn, anti-phosphotyrosine, anti-Cbp/PAG, and anti-Caveolin-1 antibodies with stripping in between.
rafts results in specific dephosphorylation of the inhibitory Tyr(P) 527 site, concomitantly with increased phosphorylation at the activating Tyr 416 site, of SFKs inside this compartment.
The Lipid Raft Composition of SFK-regulating Proteins Is Different between Fibroblasts and Thymocytes-The above experiments indicate that large differences in the relative distribution of RPTP␣ inside and outside of lipid rafts are compatible with Fyn activation and that differential recruitment of RPTP␣ to rafts is not likely to underlie the opposing effects of this signaling protein on Fyn activity in thymocytes (inhibitory (11)) versus fibroblasts (stimulatory). Addressing this issue further, we also directly compared the protein content of fibroblast and thymocyte lipid rafts, focusing not only on RPTP␣ and Fyn but also on other proteins regulating Fyn activity, i.e. Cbp/PAG, SHP2, and Csk. Immunoblots with anti-Fyn and anti-RPTP␣ antibodies revealed very similar levels and distributions of either protein among fibroblast and thymocyte lipid rafts (Fig. 5C, top two panels). However, we found thymocyte rafts to contain vastly elevated levels of Cbp/PAG but lower levels of SHP2 than fibroblasts (Fig. 5C); these differences in protein levels were more pronounced in the lipid raft fractions than in total cell lysates. The high level of Cbp/PAG in thymocyte rafts (as compared with fibroblasts) is accompanied by the prominent presence of an abundant tyrosine phosphorylated protein of approximately 75 kDa (the size of Cbp/ PAG; Fig. 5C, arrow) and higher Csk recruitment to this compartment. Thus, thymocyte rafts contain higher protein levels of two proteins (Cbp/PAG and Csk) implicated in negative regulation of Fyn and lower levels of a candidate positive regulator of Fyn (SHP2) than fibroblasts.

DISCUSSION
It is well established that the Src family kinases c-Src and Fyn are subject to regulation by the widely expressed receptor protein-tyrosine phosphatase RPTP␣. Many overexpression and knock-out studies have revealed a net activating effect of this PTP on c-Src and Fyn kinase activity, e.g. in cell lines (31,32), primary fibroblasts (8), primary neurons (33), and tissues; one study reveals that RPTP␣ also acts as a negative regulator of the SFK Fyn in thymocytes (11). Yet the precise functional outcome of RPTP␣-mediated SFK modulation remains poorly understood. Animal studies show only a very partial overlap between the Ptpra Ϫ/Ϫ and the Src Ϫ/Ϫ or Fyn Ϫ/Ϫ phenotypes. Documentation of the role of RPTP␣-mediated SFK regulation in intracellular phosphorylation and signaling responses is limited. An increase in phosphorylation of the SFK substrates focal adhesion kinase and p130 cas during integrin-mediated adhesion has been reported (8,34,35), as well as enhancement of SFK-mediated phosphorylation of N-methyl-D-aspartic acid receptor subunits (36,37). The relevance of RPTP␣ for control of the multiple relevant SFK substrates that are responsive to growth factors has never been explored.
We find that RPTP␣ is more important for stimulation of Fyn kinase activity (Fig. 1C), and for phosphorylation of the active loop tyrosine Tyr 416 in SFKs (Fig. 1D), by EGF than by PDGF. SFKs can be activated by release of the SH2/Tyr(P) 527 interaction. This can occur by engagement of a competing external SH2 domain ligand (such as activated receptor), or by PTPmediated dephosphorylation of Tyr(P) 527 . The simplest explanation for the greater importance of RPTP␣ in EGF-as opposed to PDGF-induced SFK activation might reside in a lower in vivo affinity of the SFK (Fyn) SH2 domain for phosphorylated EGFreceptor than for phosphorylated PDGF-receptor. Our study provides the starting point for testing this hypothesis.
Compared with assessment of SFK kinase activity or Tyr 416 phosphorylation (Fig. 1), our analysis of individual SFK substrates (Fig. 2) reveals a much more nuanced view of RPTP␣dependent SFK modulation. Thus, SFK-mediated tyrosine phosphorylation of STAT3 in response to growth factors remains entirely independent of RPTP␣ at all time points. That of Vav2 and Cbl becomes weakly RPTP␣-dependent during the late stages of EGF stimulation only. By contrast, Paxillin and Cbp/PAG phosphorylation is reduced in Ptpra Ϫ/Ϫ cells at all time points examined.
It is instructive to compare these functions of RPTP␣ with those of another PTP that is also a well established regulator of SFKs, SHP-2. During integrin stimulation, SHP2, much like RPTP␣, activates SFK activity, and is required for full phosphorylation of focal adhesion kinase, p130 cas , and Paxillin, and for normal spreading on fibronectin (38). Contrasting with RPTP␣, however, in response to short-term (5 min) growth factor stimulation, this PTP is clearly required for adequate phosphorylation of Vav2, as well as multiple other substrates (6). In even more contrast with RPTP␣ following growth factor stimulation, Shp2 deletion increases phosphorylation of Cbp/PAG and Paxillin, whereas we here reveal that phosphorylation of these same SFK substrates is decreased in RPTP␣-deficient cells. Thus, 2 co-expressed PTPs are both capable of stimulating the in vitro kinase of SFKs, yet exert different effects on regulated phosphorylation of the repertoire of SFK substrates.
In addition to SHP2 and RPTP␣, already several other PTPs (among others CD45, PTP1B, PTP⑀, and LAR) have also been shown to participate in positive regulation of SFKs, all apparently dephosphorylating the inhibitory "Tyr 527 " site in the latter. It is clear from biological studies that these PTPs are notredundant, and in fact act in highly specific manners (e.g. (39)). This multiplicity of SFK-regulating PTPs contrasts with a mere 2 known Tyr 527 -phosphorylating kinases (Csk and Ctk). We suggest that the multiple SFK-activating PTPs each direct SFK activity toward distinct although partially overlapping subsets of SFK substrates. In this way, PTPs could fine-tune the specificity of signaling by SFKs, which are involved in the response to a bewildering range of stimuli, and capable of phosphorylating a large number of substrates. Conceivably, the existence of distinct SFK-activating PTPs may confer additional specificity to SFK regulation by mediating differential coupling to upstream stimuli; for instance, SHP2 activity is triggered via engagement of its N-terminal SH2 domain by specific ligands; RPTP␣ can be activated through phosphorylation by upstream kinases (40), and by cross-linking of associated NCAM (33).
The ability of specific SFK-activating PTPs coupling to specific substrate sets depends on two requirements. First, the existence of mechanisms that couple individual PTPs to SFKmediated phosphorylation of specific substrate sets. This could occur through activation of different subpopulations of a particular SFK, depending on the subcellular localization of the PTP; for example, plasma membrane RPTP␣ may access different SFK subpopulations, and thus SFK substrates, than endosomal PTP1B. Similarly, physical anchoring may occur between an SFK-activating PTP, an SFK, and a substrate; for instance, in neurons, PSD95 links RPTP␣ together with SFKs and the NR2 subunit of the N-methyl-D-aspartic acid receptor (41). A second requirement of a model involving physical proximity is that the substrate specificity of the PTP for the Tyr 527 over the Tyr 416 site (or any phosphorylation sites in the affected SFK substrates themselves) be sufficiently high to allow net SFK activation and an increase in substrate phosphorylation.
In this context, we focused our subsequent studies on stimulation of the SFK Fyn and its substrate Cbp/PAG by RPTP␣. Cbp/PAG can be dephosphorylated by SHP2, resulting in reduced Csk recruitment, concomitant SFK activation, and activation of the pool of Ras located on intracellular membranes (6). By contrast, we here show that RPTP␣ increases phosphorylation of Cbp/PAG, presumably through Fyn activation as an intermediary, because this is the major SFK responsible for tyrosine phosphorylation of Cbp/PAG (24,25). This observation clearly supports the thesis that RPTP␣ activates Fyn by direct dephosphorylation of the C-terminal (Tyr 527 ) site of the latter rather than by reducing Csk function. Furthermore, the positive effect of RPTP␣ on tyrosine phosphorylation of Cbp/PAG and Paxillin, both proteins that upon phosphorylation recruit the SFK inhibitory kinase Csk, suggests the existence of negative feedback loops that limit RPTP␣-mediated activation of Fyn in lipid rafts and of Src in focal adhesions through ensuing counterbalancing recruitment of Csk to these locales.
We used the case of Fyn and Cbp/PAG, both predominantly found localized within detergent-resistant lipid "rafts" (24,25) to determine whether Fyn activation and Cbp phosphorylation are compatible with close vicinity of RPTP␣ to these targets, a requirement if differently localized PTPs are to be capable of specifically activating subsets of SFK populations and their substrates. We find that a significant fraction of RPTP␣ is found within lipid rafts (Fig. 4A), that activation of raft-localized Fyn by growth factors is accompanied by recruitment of additional RPTP␣ to this compartment (Fig. 4B), and that activation of the pool of SFKs present within lipid rafts by EGF as well as PDGF is strictly dependent on RPTP␣ (Fig. 4). The latter observation contrasts with the only partial dependence on RPTP␣ of growth factor activation of the total cellular SFK pool (Fig. 1D); the fact that total cellular SFK activity remains inducible by PDGF in the absence of RPTP␣ (Fig. 1D) suggests that other PTPs can substitute for RPTP␣ in other locations, but not in the detergent-resistant membrane compartment. Finally, we show that tilting the balance toward the increased relative recruitment of RPTP␣ to Fyn-containing lipid rafts and even significant overrecruitment of RPTP␣ to the latter (by fusing the RPTP␣ catalytic domain to the raft-targeting N terminus of Fyn itself) maintains the ability for net Fyn activation (Fig. 5, A and B).
Our results in fibroblasts contrast with the situation in thymocytes, where RPTP␣ is a net inhibitor of Fyn, presumably through relaxation of the specificity for the inhibitory C-terminal (Tyr 527 ) site over the activating Tyr 416 phosphorylation site in the activation loop of Fyn (11). Future elucidation of the reasons for this discrepancy may provide additional insights into the cross-talk between RPTP␣ and SFKs. Our data (Fig. 5C) make differences in the relative abundance of Fyn and RPTP␣ in lipid rafts an unlikely explanation. Changing the composition of fibroblast lipid rafts to mimic that of thymocyte rafts by overexpression of Cbp/PAG did not affect the capability of RPTP␣ to activate c-Src or Fyn (data not shown); however, interpretation of this experiment was complicated by failure to achieve additional tyrosine phosphorylation of the protein, perhaps because of tight control by a Cbp/PAG phosphatases(s). Other factors, such as the contribution of yet additional proteins in lipid rafts or of an alternatively spliced form of Fyn in thymocytes, will need to be explored to understand what determines whether RPTP␣ will be a net activator or inhibitor of SFK activity.
SFK activity is frequently deregulated in human cancers, promoting both tumor invasiveness and metastasis (42). Osteoporosis and stroke are examples of additional pathologies where pharmacological modulation of SFK activity may be relevant (43,44). Although SFK inhibitors are currently under development, side effects and toxicity constitute a concern given the many physiological processes in which SFKs participate (45). Our observation that an SFK-activating PTP such as RPTP␣ directs and fine-tunes SFK activity toward specific substrate sets suggests the usefulness of PTP inhibition as a therapeutic strategy.