Association between Receptor Protein-tyrosine Phosphatase RPTPα and the Grb2 Adaptor

Receptor protein-tyrosine phosphatase RPTPα is found associated in vivo with the adaptor protein Grb2. Formation of this complex, which contains no detectable levels of Sos, is known to depend on a C-terminal phosphorylated tyrosine residue (Tyr798) in RPTPα and on the Src homology (SH) 2 domain in Grb2 (1, 2). We show here that association of Grb2 with RPTPα also involves a critical function for the C-terminal SH3 domain of Grb2. Furthermore, Grb2 SH3 binding peptides interfere with RPTPα-Grb2 association in vitro, and the RPTPα protein can dissociate the Grb2-Sos complex in vivo. These observations constitute a novel mode of Grb2 association and suggest a model in which association with a tyrosine-phosphorylated protein restricts the repertoire of SH3 binding proteins with which Grb2 can simultaneously interact. The function of the Tyr798 tyrosine phosphorylation/Grb2 binding site in RPTPα was studied further by expression of wild type or mutant RPTPα proteins in PC12 cells. In these cells, wild type RPTPα interferes with acidic fibroblast growth factor-induced neurite outgrowth; this effect requires both the catalytic activity and the Grb2 binding Tyr798 residue in RPTPα. In contrast, expression of catalytically active RPTPα containing a mutated tyrosine phosphorylation/Grb2 association site enhances neurite outgrowth. Our observations associate a functional effect with tyrosine phosphorylation of, and ensuing association of signaling proteins with, a receptor protein-tyrosine phosphatase and raise the possibility that RPTPα association may modulate Grb2 function and vice versa.

Phosphorylation of tyrosine residues in proteins constitutes a widespread regulatory modification, often serving to propagate intracellular signals involved in the generation of a specific cellular response to an extracellular stimulus (3)(4)(5). Controlled activation of this regulatory mechanism relies on the coordinated contribution of five classes of proteins: tyrosine kinases, their direct substrates, adaptor proteins, signaling molecules, and protein-tyrosine phosphatases (PTPases). 1 Ty-rosine phosphorylation can directly control the catalytic activity of a tyrosine kinase, such as the insulin receptor, or of a phosphorylated enzyme substrate, such as phospholipase C-␥. More often, however, the phosphorylation of specific tyrosine residues acts as a reversible switching mechanism governing the assembly of particular protein-protein complexes. Individual tyrosine autophosphorylation sites on activated receptor tyrosine kinases interact directly or indirectly with particular signaling molecules, and mutation of particular sites restricts the repertoire of signaling pathways being activated. It has been extensively documented that SH2 domains associate with phosphotyrosine in a context-specific manner (6). Other protein-protein associations dependent on the presence of phosphotyrosine can be mediated by the phosphotyrosine interaction/phosphotyrosine binding domain (7). As a consequence of such protein-protein associations, changes may occur in the catalytic activity or intracellular localization of proteins that are not necessarily phosphorylated themselves (see Refs. 3, 8, and 9 for reviews).
The adaptor protein Grb2 has become a paradigm for this type of interaction. This protein, consisting of one SH2 domain flanked by two SH3 domains, does not normally undergo tyrosine phosphorylation (10). Its structure allows Grb2 to serve as an adaptor between tyrosine-phosphorylated proteins (such as the EGF receptor, Syp, Shc, or insulin receptor substrate-1) and Sos, a guanine nucleotide exchange protein for Ras (Refs. [11][12][13][14][15][16][17]reviewed in Ref. 18). This role for Grb2 in Ras activation, thought to involve concentration of the Sos protein at a membrane location (19), is consistent with genetic data in lower organisms epistatically ordering Grb2 homologues between tyrosine kinases and Ras (20 -22). Recent data have drawn attention to the fact that the Grb2 protein, via its SH2 and SH3 domains, can interact with a large number of other proteins, such as Bcr, Bcr-abl, Fak, PI3 kinase, RPTP␣, Syp, p89, Slp-76, Cbl, dynamin, dystroglycan, synapsin I, Vav, and others (1,2,(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34). It is as yet not clear which of these interactions are physiologically relevant. Yet, the high number of Grb2 binding proteins raises a number of questions. First, what is the function of Grb2 association with such a diverse repertoire of proteins, including kinases, phosphatases, structural proteins, and others? More particularly, does the Grb2 protein, in all settings where its SH2 domain interacts with a tyrosinephosphorylated protein, act as an adaptor for Sos recruitment, or can the nature of the SH2 domain bound ligand determine the repertoire of proteins that can simultaneously interact with the SH3 domains of Grb2? Last, to what extent can the various Grb2 binding proteins compete for a shared pool of intracellular Grb2 protein?
Studies on the function of the hemopoietic CD45 antigen and its identification as a trans-membrane ("receptor") tyrosine phosphatase (R-PTPase), have drawn increased attention to the role of PTPases in signaling events. CD45 PTPase activity is required for signaling through the T-cell receptor; more particularly, in the absence of CD45 expression, ligation of the T-cell receptor fails to induce tyrosine phosphorylation of cellular proteins, suggesting that CD45 acts upstream of a tyrosine kinase. Evidence that CD45 controls the phosphorylation state of the negative regulatory tyrosine phosphorylation sites in the kinases Lck and Fyn (residues analogous to Tyr 527 in c-Src) suggests that CD45 acts as a necessary positive regulator of these Src family kinases (5,35).
In contrast to CD45, RPTP␣ is a widely expressed member of the R-PTPase family. This R-PTPase consists of a relatively small (130 residues) and highly glycosylated extracellular domain, linked to an intracellular domain containing two PTPase homology domains (36 -40). Overexpression of RPTP␣ can enhance the kinase activity of endogenous c-Src, concomitant with decreased c-Src phosphorylation levels at the Tyr 527 residue, suggesting that endogenous RPTP␣ may also contribute to the regulation of the phosphorylation level at Tyr 527 in c-Src, and thus the c-Src kinase activity (41,42).
In several cell types, endogenous RPTP␣ is constitutively phosphorylated on tyrosine. Interestingly, phosphorylation of RPTP␣ is required for association of RPTP␣ with the adaptor protein Grb2, with which RPTP␣ is found associated in vivo (1,2). Such tyrosine phosphorylation of PTPases and ensuing recruitment of signaling proteins is not an isolated observation. Analogous observations have been made for CD45 as well as for the SH2 domain containing PTPase Syp (e.g. Refs. 12,15,43,44). We here report on two questions raised by the observed tyrosine phosphorylation of RPTP␣ and its association with Grb2. First, we have investigated whether the Grb2 protein, when associated with tyrosine-phosphorylated RPTP␣, performs a similar adaptor function (i.e. Sos recruitment) as in the case of its association with tyrosine-phosphorylated EGF receptor or Shc. Second, using a strategy which has been highly successful in dissecting the signaling pathways controlled by activated tyrosine kinases, we have investigated the consequences of mutation of the phosphorylated and Grb2-binding tyrosine residue for RPTP␣ function in vivo.
Plasmid Construction and Mutagenesis-For point mutagenesis, the human RPTP␣ cDNA (38), cloned in the EcoRV site of pMJ30 (48), was mutagenized using the Transformer TM site-directed mutagenesis kit (Clontech) following the instructions of the manufacturer. Internal deletions were introduced in the human RPTP␣Y798F protein by restriction of the cDNA with XbaI and BspMI (⌬282-520) or XbaI and MfeI (⌬282-613), followed by Klenow treatment and religation, yielding in-frame fusions. The WT, RPTP␣CCSS, and RPTP␣Y798F mutant cDNAs were released by excision with XhoI and BglII and ligated between the XhoI and BamHI sites of pLXSHD (49).
In Vitro Binding Assays, Transient Expression Assays, and in Vivo Protein Association Assays-These tests were performed exactly as described previously (1). 36 h after calcium phosphate-mediated transfection of 293T cells, lysates were prepared in Triton lysis buffer (1% Triton X-100, 50 mM Hepes, pH 7.5, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 10% glycerol, 1 mM Na 3 VO 4 , 50 mM NaF, 10 mM Na 4 P 2 O 7 , 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin) and subjected to immunoprecipitation for 90 min at 4°C using protein A-Sepharose 4B (Zymed) or to in vitro binding reactions with GST fusion proteins previously immobilized on glutathione agarose (Sigma). In vitro binding reactions were for 90 min at 4°C, using 5 g of immobilized protein (except for the experiments in Fig. 4, where 15 g was used), followed by four washes with Triton lysis buffer. Immunoblots were developed using horseradish peroxidase-conjugated protein A (Kirkegaard & Perry) and an enhanced chemiluminescence detection reagent (Renaissance, DuPont NEN).
Generation of Stable Cell Lines-Retroviral constructs in pLXSHD (49) were cotransfected into 293T cells together with a -negative ecotropic helper construct (50), using a modified calcium phosphatemediated transfection protocol. The conditioned medium was collected after 24-h periods; virus titers, assayed by generation of histidinolresistant colonies in 3T3 cells, were approximately 10 4 to 10 5 colonyforming units/ml. Stable clones were generated by infection with an appropriately low virus titer for 3 h in the presence of polybrene, followed by selection for histidinol (Sigma) (4 mM) resistance starting 2 days after infection. Pools of clones were generated by infection of 2 ϫ 10 5 cells with 1 ml (10 4 to 10 5 colony-forming units/ml) of virus, followed by selection.
PC12 Cell Differentiation-10 5 PC12 cells were seeded per 3.5-cm plate and grown in complete medium (Dulbecco's modified Eagle's medium supplemented with 10% horse serum and 10% fetal calf serum) overnight. The cells were then transferred to Dulbecco's modified Eagle's medium containing 0.5% fetal calf serum and 0.5% horse serum supplemented with 50 ng/ml acidic fibroblast growth factor (aFGF) (a gift from Dr. I. Lax) and 50 g/ml heparin.

Association of Grb2 with Tyrosine-phosphorylated RPTP␣ Requires both the SH2 and C-terminal SH3 Domains of Grb2-
The RPTP␣ protein associates with Grb2 in vitro and in vivo. This interaction requires both tyrosine phosphorylation of RPTP␣, which occurs at residue Tyr 789 in mouse RPTP␣ (numbering of Ref. 36), corresponding to Tyr 798 of human RPTP␣ (numbering of Ref. 38), as well as a functional Grb2 SH2 domain (1,2). In preliminary studies, we observed that, by itself, an isolated Grb2 SH2 domain only weakly associates with phosphorylated RPTP␣. Similarly, we were unable to detect binding of a recombinant Grb2 SH2 domain protein with phosphorylated RPTP␣ by "far Western" analysis (data not shown). We therefore investigated in more detail the Grb2 domain requirements for the association of this adaptor protein with tyrosine-phosphorylated RPTP␣. To this purpose, we used a set of recombinant WT and mutant Grb2 proteins, which were generated as GST fusion proteins, immobilized on glutathione agarose beads and used as affinity matrices. The mutant proteins used were Grb2-FLVR, containing a point mutation (R86K), which disables interaction of the SH2 domain with phosphotyrosine (10, 11); Grb2-P49L and Grb2-G203R, point mutations interfering with the function of the N-and C-terminal SH3 domains of Grb2, respectively (11); and Grb2-⌬N and Grb2-⌬C, deletion mutants of the N-and C-terminal SH3 domains, respectively (46). The immobilized proteins were incubated in the presence of detergent lysates prepared from RPTP␣-expressing 293T cells, and binding of RPTP␣ protein was monitored by anti-RPTP␣ immunoblotting (Fig. 1A). Strikingly, deletion or point mutation of the C-terminal, but not the N-terminal SH3 domain of Grb2, almost completely abolished the ability of Grb2 to associate with RPTP␣. To establish the validity of these observations, we performed control experiments with two other proteins that interact with Grb2 in a better understood manner. First, we investigated the affinity of the HER for the same set of Grb2 proteins. Her14 cells (a 3T3 derivative) were stimulated with EGF for 5 min, and Triton lysates were incubated with the same Grb2 proteins used for the experiment in Fig. 1A. Consistent with previously reported data (10,14), it can be seen in Fig. 1B that tyrosine-phosphorylated EGF receptor binds equally efficiently to WT Grb2 protein as to Grb2 proteins in which the N-or C-terminal SH3 domains have been mutated or deleted. This proves that the Grb2-⌬C and Grb2-G203R proteins display an intact SH2 do-main function. We also monitored in vitro binding of the Sos protein to our battery of recombinant Grb2 proteins. This control experiment (Fig. 1C) showed that association of Sos with Grb2 involved a function for both the N-and C-terminal SH3 domains of Grb2, as reported previously (11). However, Sos binding was more sensitive to deletion of an entire SH3 domain than to the effect of point mutations P49L and G203R, suggesting again that the effect of the G203R mutation is unlikely to be due to gross misfolding of the Grb2 protein.
To extend these observations to the in vivo situation, we performed two types of transient coexpression experiments. First, we used an indirect assay to gauge RPTP␣-Grb2 association. Association with a matching SH2 domain protects tyrosine phosphorylation sites in proteins from phosphatase action; this effect has been used successfully to map binding sites for individual SH2 domains (14,51). In the case of a tyrosinephosphorylated tyrosine phosphatase, this protective effect may extend to both autodephosphorylation and dephosphorylation by other endogenous PTPases. We previously demonstrated that transient coexpression of Grb2, but not other SH2 domain-containing proteins, results in increased levels of tyrosine phosphorylation of RPTP␣ in vivo and that this effect is dependent on an intact Grb2 SH2 domain (although both the 110-and 130-kDa forms of RPTP␣ undergo tyrosine phosphorylation and associate with Grb2 in vitro and in vivo, RPTP␣ hyperphosphorylation induced by Grb2 overexpression is limited to the 130-kDa form of RPTP␣ (1); since the 110-kDa form is a precursor of the 130-kDa form (40), it may be differentially localized in the cell and thus less accessible to Grb2 or more susceptible to in vivo dephosphorylation). We extend this approach here to Grb2 proteins lacking the N-or C-terminal SH3 domain. As expected, coexpression of WT Grb2, but not Grb2-FLVR, resulted in the appearance of a novel 130-kDa tyrosinephosphorylated protein, corresponding to RPTP␣ (Fig. 2). While deletion of the N-terminal SH3 domain had no effect on the ability of Grb2 to enhance the phosphotyrosine content of RPTP␣, deletion of the C-terminal domain clearly interfered with this effect. This observation demonstrates that efficient protection of the RPTP␣ tyrosine phosphorylation site by association with the Grb2 SH2 domain also requires the C-terminal SH3 domain. Second, we used coimmunoprecipitation to determine the extent of association between RPTP␣ and Grb2 proteins lacking either the N-or C-terminal SH3 domain. Since the epitopes in RPTP␣ that are required for its interaction with Grb2 are unknown, and in order to avoid interference of the immunoprecipitating anti-RPTP␣ antiserum with association of RPTP␣ with the different Grb2 proteins, we used a chimeric RPTP␣ protein (HER-␣) in which the extracellular domain of RPTP␣ had been replaced by that of the HER. This approach allowed us to perform immunoprecipitations using an antibody to the EGF receptor ectodomain tag, which is unlikely to interfere with RPTP␣-Grb2 association. Coexpression of the HER-␣ protein with WT or mutant Grb2 proteins lacking one of the SH3 domains showed that deletion of the C-terminal SH3 domain completely abolished coimmunoprecipitation of Grb2 with HER-␣, whereas deletion of the N-terminal SH3 domain only partially abolished association (Fig. 3). We conclude from the above experiments that the RPTP␣-Grb2 complex combines SH2 and SH3 domain-mediated interactions.
The C-terminal SH3 Domain of Grb2 Engages in a Phosphotyrosine-independent Association with the Intracellular Domain of RPTP␣-A number of possible explanations could account for the observed requirement for the C-terminal SH3 domain of Grb2 for the interaction of this adaptor protein with tyrosine phosphatase RPTP␣. First, a conformational effect, caused by a mutation affecting the C-terminal SH3 domain of  (40). Grb2-FLVR is a Grb2 protein containing the R86K mutation in the FLVR motif of its SH2 domain. The R86K residue (part of the FLVR motif) is critical for the function of SH2 domains, and the R86K mutation is a classical mutation interfering with SH2 domain function (11). Grb2-P49L and Grb2-G203R are point mutations in the N-and Cterminal SH3 domains of Grb2, analogous to those found in the SH3 domains of the Caenorhabditis elegans homologue of Grb2, Sem-5. These mutations inactivate the respective SH3 domains, as revealed by their effects on binding to Sos (11). Grb2-⌬N has the N-terminal SH3 domain deleted, and Grb2-⌬C has the C-terminal SH3 domain deleted (46). B, control experiment monitoring the interaction of the mutant Grb2 proteins with phosphorylated EGF receptor. A similar experiment was performed as in Fig. 1A, except that the immobilized Grb2 proteins were incubated with lysates from EGF receptor expressing HER14 cells (stimulated with EGF before lysis), and the bound proteins were analyzed by immunoblotting with an antibody against the EGF receptor. C, association of the Sos protein with the various mutant Grb2 proteins. Untransfected 293T cell lysates were incubated with the various Grb2 fusion proteins, and retained proteins were analyzed by anti-Sos immunoblotting. In this experiment, a fusion protein consisting solely of the isolated SH2 domain of Grb2 was also included (Grb2-SH2).
Grb2, could indirectly affect the binding properties of the SH2 domain. Several studies suggest that the SH2 and SH3 domains appear to act as independent units (52)(53)(54). Recently, however, it was demonstrated that, in T cells, association of tyrosine-phosphorylated Shc with Grb2 enhances Grb2-Sos association, suggesting that associations via the Grb2 SH2 domain can affect SH3 domain function (55). Conversely, therefore, it is conceivable that mutation of the Grb2 SH3 domain negatively affects the affinity of its SH2 domain for tyrosinephosphorylated RPTP␣. However, such an effect would have to be specific, since C-SH3 deletion does not interfere with the ability of Grb2 to associate with the EGF receptor (Fig. 1B). Alternatively, the C-terminal SH3 domain of RPTP␣ could engage in a separate protein-protein association with an RPTP␣-associated SH3 domain binding site, thus contributing to an enhancement of the overall affinity between the two proteins. If this is the case, it should be possible to mutate this epitope and reduce the affinity of Grb2 for RPTP␣.
SH3 domains are protein modules that appear to participate in protein-protein interactions via their affinity for polyproline type II helices (9). Inspection of the RPTP␣ amino acid sequence for proline-rich sequences revealed the presence of a sequence, RKYPPLPVDK (residues 216 -225 in human RPTP␣; Ref. 38) in the juxtamembrane region with reasonable similarity to canonical SH3 domain-binding motifs (56). To test the hypothesis that this sequence in RPTP␣ contributes to the affinity of the protein for Grb2, the tandem proline residues 219 and 220 were mutated simultaneously to alanine. In in vivo and in vitro association assays, this mutation had no measurable effect on the association of Grb2 with RPTP␣, suggesting that the proline-rich motif in the RPTP␣ juxtamembrane domain is unlikely by itself to be required for Grb2-RPTP␣ association (data not shown).
We further explored the possibility that the requirement for the C-terminal SH3 domain of Grb2 reflects an independent protein-protein interaction by investigating directly whether Grb2 could bind to RPTP␣ in a manner independent of the Grb2 SH2 domain. For this purpose, in vitro binding assays were performed as for Fig. 1, except that we used a mutated RPTP␣ protein, RPTP␣Y798F, in which the tyrosine residue whose phosphorylation is required for Grb2 binding (Tyr 798 in human RPTP␣), had been mutated to phenylalanine. In addition, since we anticipated that an interaction mediated only by an SH3 domain might be of low affinity, these experiments were performed using larger amounts of recombinant Grb2 protein (Fig. 4A). As expected, under these conditions, binding of RPTP␣ that was not phosphorylated on tyrosine was not affected by mutational inactivation of the Grb2 SH2 domain. However, mutation or deletion of the C-terminal, but not the N-terminal, SH3 domain completely abolished retention of RPTP␣Y798F on the Grb2 affinity matrix.
We subsequently set out to further delineate the domain requirements for this SH3 domain-dependent component of RPTP␣-Grb2 association using deletion mutagenesis. Obviously, deletions in the intracellular domain of RPTP␣ will interfere with the catalytic activity of the protein and thus, by abolishing RPTP␣ autodephosphorylation, lead to increased RPTP␣ tyrosine phosphorylation, resulting in increased SH2 domain-dependent Grb2 binding (as documented previously in Ref. 2). To rule out this confounding factor, we introduced deletions within the context of the nonphosphorylated RPTP␣Y798F protein. It can be seen (Fig. 4B) that a deletion of the first PTPase domain of RPTP␣ (⌬282-520) abolishes association between unphosphorylated RPTP␣ and the Grb2 pro- The HER-␣ protein was used instead of WT RPTP␣ so as to allow us to use an immunoprecipitating antibody (against the HER ectodomain) that would not interfere with association of proteins with the intracellular domain of RPTP␣. This approach also rules out the possibility that different deletion mutants of Grb2 would sterically interfere to different extents with binding of antiserum to the intracellular domain of RPTP␣. The anti-RPTP␣ antiserum (number 210) used for immunoblotting is against an intracellular epitope in RPTP␣ (1), so that it also recognizes the HER-␣ protein. The Grb2 proteins all contained an HA tag at the N terminus, so that they could all be detected with the same antibody (46). In each lane, HER-␣ was transiently coexpressed in 293T cells, alone (Ϫ) or together with WT or mutant Grb2 proteins (Grb2-⌬N and Grb2-⌬C, lacking the N-and C-terminal SH3 domains, respectively). Left panels, anti-RPTP␣ (top) and anti-HA immunoblotting of total lysates of the transfected cells. Right panels, immunoprecipitation was performed with an antibody against the HER ectodomain, and the HER-␣ precipitates were analyzed by anti-RPTP␣ (top) or anti-HA tag immunoblotting (bottom).
tein. Taken together, the data in Fig. 4, A and B, demonstrate that the phosphotyrosine/SH2 domain-independent component of the association between RPTP␣ and Grb2 requires the Cterminal (but not N-terminal) SH3 domain of Grb2 as well as sequences within the first PTPase domain of RPTP␣. Thus, the SH3 domain-dependent component of the Grb2-RPTP␣ association shows a double element of specificity.
Several issues need to be considered when interpreting in vitro binding assays such as those illustrated by Figs. 1 and 4. First, since the RPTP␣ protein is supplied from a cell lysate, additional proteins in the lysate may be involved in formation of the RPTP␣-Grb2 complex. For instance, the Grb2 C-terminal SH3 domain could engage a third, unidentified, protein, which in turn could be required to stabilize the RPTP␣-Grb2 complex. Second, binding of the RPTP␣ protein in the lysate to the recombinant Grb2 matrix may be limited by an equilibrium between RPTP␣ and endogenous Grb2 present in the lysate. Additionally, other cellular proteins could compete with RPTP␣ for binding to the recombinant Grb2 used in the experiment. We therefore also investigated whether we could detect association of purified bacterially expressed, unphosphorylated RPTP␣ with purified Grb2 protein in a manner dependent on the C-terminal SH3 domain of Grb2. These experiments yielded inconsistent results (data not shown). This may point to a requirement for a cellular protein that modifies RPTP␣ or stabilizes its association with Grb2. Alternatively, it may simply reflect inappropriate folding of the bacterially expressed RPTP␣ protein.
In summary, our data are consistent with a model in which association between tyrosine-phosphorylated RPTP␣ and Grb2 involves two components. One is a canonical SH2-phosphotyrosine interaction. The second is a critical additional interaction that involves the Grb2 C-terminal SH3 domain. The latter association may involve direct contact between RPTP␣ and the C-terminal SH3 domain of Grb2 or could require participation by additional proteins in the RPTP␣-Grb2 complex.
Grb2 SH3 Domain Occupancy Interferes with Association between RPTP␣ and Grb2-The Grb2 protein functions as a signaling adaptor protein by mediating the formation of trimolecular complexes. Thus, Grb2 associates simultaneously, via its SH2 domain, with tyrosine-phosphorylated proteins (generally considered upstream in an information transfer pathway) such as the EGF receptor or Syp and, via its SH3 domains, with effector proteins such as Sos (Refs. 11, 12, 14, and 15; reviewed in Ref. 18). By analogy, Grb2 could also perform an adaptor function between tyrosine-phosphorylated RPTP␣ and a third protein such as Sos. We and others have thus far been unable to observe in vivo association between tyrosine-phosphorylated RPTP␣ and Sos (1,2). Arguably, this observation might simply reflect the insensitivity of our detection reagents or the low abundance of the complex. We therefore also performed in vitro binding studies, which, since they are more sensitive, might allow us to determine whether a trimolecular RPTP␣-Grb2-Sos complex can exist under forced conditions.
To address this issue, we first immobilized a recombinant protein corresponding to the carboxyl terminus of the Sos protein (residues 1132-1234). This domain was previously shown by several groups to contain high affinity binding sites for the SH3 domains of Grb2 (11,21,52). The immobilized protein was then incubated with lysates from 293T cells in which the RPTP␣ protein had been highly overexpressed, so as to further enhance the sensitivity of our assay (Fig. 5). Analysis of the Sos-bound proteins by anti-Grb2 immunoblotting showed that the recombinant Sos C terminus captured approximately 25% of the entire Grb2 protein population present in the lysates (slightly reduced upon RPTP␣ expression). Simultaneously, we FIG. 4. Unphosphorylated RPTP␣ binds Grb2 in a manner that depends on an intact Grb2 C-terminal SH3 domain and on the first PTPase domain in RPTP␣. A, an in vitro binding experiment was performed similar to the one described in Fig. 1A. However, in this case lysates were prepared from RPTP␣Y798F-expressing cells. The Tyr 798 phosphorylation site in this protein has been mutated, and the protein is not tyrosine-phosphorylated. Three times more immobilized Grb2 proteins were used than in Fig. 1, in order to raise the sensitivity of detection, since unphosphorylated RPTP␣Y798F protein has reduced affinity for Grb2 (1). We previously also showed that mutation of the SH2 domain of Grb2 also severely affects WT RPTP␣ binding (1). B, effect of deletion mutagenesis of the RPTP␣Y798F protein on its C-SH3dependent association with Grb2. Upper panel, deletions (white boxes in the bar diagram representations of RPTP␣) were introduced into the RPTP␣Y798F encoding cDNA, and the resulting proteins were analyzed for their ability to associate in vitro with immobilized WT Grb2 protein or Grb2-⌬C, by anti-RPTP␣ immunoblotting (antiserum 210). Lower panel, expression of the WT and mutant Grb2 proteins, as revealed by immunoblotting of total lysates with anti-RPTP␣ antiserum 210. The 210 antiserum used for immunodetection of RPTP␣ in both panels recognizes a peptide epitope (arrow) retained in the Y789F and ⌬282-520 proteins but not in the ⌬282-613 protein. Actual expression (and lack of binding) of the ⌬282-613 protein was detectable by immunoblotting with an antiserum (number 443) raised against the entire intracellular domain of RPTP␣ (data not shown). However, the latter antiserum (number 443) may recognize the deleted proteins with lesser affinity. Only the 210 antiserum allows unambiguous comparison of the expression levels of the WT, Y789F, and ⌬282-520 proteins; therefore, this serum was the one used for the analysis shown here. determined, by anti-RPTP␣ immunoblotting, whether any RPTP␣ protein associated, through Grb2, with the recombinant Sos C-terminal portion. Even after significant overexposure of the films used for chemiluminescence-based detection, no RPTP␣ protein was detectably retained under these conditions. We therefore conclude that a trimolecular RPTP␣-Grb2-Sos complex is unlikely to occur to any significant extent.
The absence of RPTP␣ protein from the population of Grb2 that is associated with the Sos C terminus in the above experiment prompted us to examine the possibility that SH3 domain-mediated association of Grb2 with a protein such as Sos would be incompatible with its simultaneous interaction with RPTP␣. For example, since the C-terminal SH3 domain of Grb2 appears to play a role in association with both RPTP␣ (see above) and Sos (11), steric hindrance, or direct competition for occupancy of the SH3 domain could prevent their simultaneous association. We therefore tested the possibility that occupancy of the peptide binding "groove" of the Grb2 SH3 domains, particularly the C-terminal SH3 domain, such as would occur by Sos binding to Grb2, interferes with RPTP␣ association. We addressed this issue by monitoring the effect of defined Grb2 SH3 domain-binding peptides on the association of RPTP␣ with recombinant Grb2. Lysates containing RPTP␣ were incubated with immobilized recombinant Grb2 proteins in the presence of either peptides that bind to the Grb2 SH3 domains or control peptides. Two Grb2 SH3 domain-binding peptides (one of which corresponds to a high affinity binding site in Sos) interfered with RPTP␣ retention on the Grb2 affinity matrix, whereas two irrelevant control peptides did not (Fig. 6A). Strikingly, a peptide (GC) known to bind mainly the C-terminal SH3 domain of Grb2 also interfered with RPTP␣-Grb2 association. This peptide also acted as an inhibitor of RPTP␣ association with Grb2 proteins in which the N-terminal SH3 domain had been disabled by point mutation or deletion (Fig. 6B). Such small peptides probably exert their effect solely by competition for the peptide binding "groove" of an SH3 domain, since they are known not to alter the affinity of interaction mediated by the SH2 domain (52,53), and they are unlikely to exert significant steric hindrance extending to other areas. We therefore conclude that the requirement of the C-terminal SH3 domain for RPTP␣ binding involves, at least in part, occupancy of the peptide binding area of this domain.
RPTP␣ Can Interfere with Association between Sos and Grb2 in Vivo-The above in vitro observations demonstrate that association of Grb2 with Sos prevents interaction of Grb2 with RPTP␣, possibly because of competition for Grb2 SH3 domain occupancy. Such apparent competition between Sos and RPTP␣ association with Grb2 suggests that RPTP␣ may actually be able to either decrease the formation of or dissociate the Grb2-Sos complex. To test this hypothesis, we reconstituted the Grb2-Sos complex by transient expression in 293T cells. We then monitored, by immunoprecipitation of the (HA epitopetagged) Sos protein followed by anti-Grb2 immunoblotting, the effect of increasing the amount of coexpressed RPTP␣ protein on the amount of Grb2 associated with Sos (Fig. 7). In such experiments, we observed that, at the highest expression levels, coexpression of RPTP␣ led to lower levels of Sos-Grb2 association, although the total amounts of Grb2 and Sos and the amount of immunoprecipitated Sos protein, remained unchanged. Similar results were obtained in cell lines stably expressing the RPTP␣ protein (see below; Fig. 9). While the level of transiently expressed RPTP␣ protein was approximately linear with respect to the amount of input DNA, the observation that displacement of Grb2 from Sos appeared to occur in a nonlinear manner (i.e. apparently above a threshold level of RPTP␣ expression) was reproducible. The reasons for this are unclear and could be manifold, one possibility being the formation of RPTP␣ oligomers (see "Discussion").

Blockage of aFGF-induced Neurite Outgrowth by RPTP␣ in PC12 Cells Requires both RPTP␣ Catalytic Activity and an
Intact Tyr 798 Tyrosine Phosphorylation/Grb2 Binding Site-Ectopic expression of receptor tyrosine kinases containing mutated tyrosine phosphorylation sites is a widely used strategy for understanding the function of individual phosphorylation sites (3). Whereas results obtained with this approach obviously cannot be automatically extended to the function of endogenous proteins, the use of such model systems has in many instances been illuminating. We therefore adopted a similar approach for RPTP␣. We chose to use PC12 cells as a recipient cell line, since these cells express only low levels of endogenous RPTP␣ protein (data not shown) and since tyrosine phosphorylation-mediated signaling pathways have been extensively studied in this cell line (57,58). Retroviral vectors were developed, encoding WT or mutant RPTP␣ proteins or, as a control, no protein ("empty" virus). The mutated RPTP␣ cDNAs encoded a protein in which the Tyr 798 phosphorylation/Grb2 association site was mutated to phenylalanine (RPTP␣Y798F) and a protein in which the catalytically important cysteine residues 442 and 732 in the first and second PTPase domains of human RPTP␣ (38) were mutated to serine (RPTP␣CCSS protein). Acting as phosphate acceptors in the reaction mechanism, these cysteine residues are critical for the catalytic activity of the PTPase, and their mutation is known to abolish catalytic activity (59,60). Retroviral stocks were generated and used to generate single PC12 clones (Fig. 8B) as well as pools of PC12 cells generated by infection en masse followed by selection. The latter approach allowed us to minimize stochastic clone-to-clone variation due to heterogeneity of the PC12 cell population or due to secondary events occurring as a result of RPTP␣ expression. Immunoprecipitation of RPTP␣ proteins followed by in vitro assay of PTPase activity toward an artificial substrate confirmed that the RPTP␣CCSS protein was catalytically inactive, while the RPTP␣Y798F retained in vitro catalytic activity comparable with that of WT RPTP␣ (data not shown), as reported previously (2). We also confirmed that, as expected, RPTP␣Y798F, in contrast to the WT protein, fails to associate with Grb2 in vivo (data not shown).
In order to assess the consequences of expression of the WT or mutant RPTP␣ proteins, we subsequently treated the infected PC12 clones and pools with aFGF in the presence of heparin and scored the neurite outgrowth response. After 2 days of growth factor treatment, the average neurite length was measured quantitatively by analysis of photographic fields containing 100 -250 cells each. The data obtained are reported in Table I, and illustrated in Fig. 8A. Expression of WT RPTP␣ protein led to a statistically significant reduction in average neurite outgrowth as compared with control cells infected with empty virus. Cell pools and clones infected with virus encoding the catalytically inactive RPTP␣CCSS protein behaved indistinguishably from control vector infected cells, demonstrating that the block imposed on neurite outgrowth by RPTP␣ is strictly dependent on catalytic activity. By contrast, cells infected with the RPTP␣Y798F expressing virus displayed a small but reproducible and statistically significant increase in neurite outgrowth as compared with control virus-infected cells (Table I).
We next determined the effect of expression of the WT or mutant RPTP␣ proteins on association between endogenous Sos and Grb2 in the infected PC12 cells. Similarly to the experiment illustrated by Fig. 7, the Sos protein was immunoprecipitated from clones of stably expressing PC12 cells, and the amount of associated Grb2 protein was assessed by immunoblot analysis of the washed immune complexes. This experiment revealed that expression of both WT RPTP␣ and RPTP␣CCSS, but not RPTP␣Y798F, interfered with association between endogenous Sos and Grb2 (Fig. 9). This observation is consistent with our observations of mutual incompatibility of RPTP␣ and Sos association with Grb2 in in vitro binding and transient coexpression experiments (Figs. 5, 6, and  7). It further shows that interference with Grb2-Sos interaction requires the Tyr 798 phosphorylation site, but not the catalytic  Table I. B, expression of WT RPTP␣, RPTP␣Y798F, and RPTP␣CCSS in PC12 cells. Total lysates from control virus-infected cells (i.e. the retroviral vector without a RPTP␣ cDNA insert) or infected cell clones expressing the various RPTP␣ mutants were analyzed by anti-RPTP␣ immunoblotting.
activity, of RPTP␣. Therefore, competition for binding to the Grb2 protein, resulting in decreased formation or increased dissociation of the Grb2-Sos complex, is the most likely mechanism of this effect. Finally, this experiment demonstrates that the ability of RPTP␣ mutants to interfere with RPTP␣-Grb2 association does not correlate with the ability to block neurite outgrowth in response to aFGF. In fact, blockage of neurite outgrowth by RPTP␣ requires both catalytic activity and the Tyr 798 tyrosine phosphorylation/Grb2 association site.

Mechanism of RPTP␣-Grb2
Association-The complexity of the PTPase family suggests that tyrosine dephosphorylation is subject to intricate regulation (reviewed in Refs. 60 and 61). Among the mechanisms that may control PTPase function are intracellular localization, serine/threonine phosphorylation, multimerization, and possibly association with extracellular ligands (Refs. 5, 47, and 62-64 and references therein). Maintenance of the specificity of intracellular information flow relies heavily on specific intermolecular interactions mediated by modular protein domains such as SH2, SH3, and phosphotyrosine binding/phosphotyrosine interaction domains (7)(8)(9). Is there evidence, as might reasonably be expected, that PTPases also participate in this cellular network of regulatory proteinprotein interactions? PTP1C and Syp are the only PTPases known to contain SH2 domains. These mediate the enzymes' association with tyrosine-phosphorylated proteins, with SH2 domain occupancy potentially modulating catalytic activity (65,66). Alternatively, PTPases may tie into the cellular signaling network by tyrosine phosphorylation of the PTPases themselves. For instance, tyrosine phosphorylation of Syp can recruit the Grb2-Sos complex, which may relate to a function for Syp in Ras signaling (12,15). RPTP␣ is phosphorylated on a C-terminal tyrosine and found in a complex with the Grb2 adaptor (1, 2). The use of PTPase inhibitors has also provided evidence for possible tyrosine phosphorylation of CD45, generating a docking site for the SH2 domain of Lck (43,44). Other R-PTPases such as RPTP, 2 and leukocyte common antigenrelated protein-related PTPases 3 can undergo tyrosine phosphorylation under certain circumstances.
The Tyr 798 phosphorylation site in RPTP␣ is required for association with Grb2 and occurs in a sequence context closely resembling the consensus binding site for the Grb2 SH2 domain (6). Increased Grb2 expression enhances phosphorylation of this site in vivo, an effect that requires a functional Grb2 SH2 domain (1). These observations are consistent with a direct interaction occurring between phosphorylated Tyr 798 and the Grb2 SH2 domain. Our present experiments demonstrate that RPTP␣-Grb2 association also strictly requires a functional C-terminal SH3 domain of Grb2. After submission of this manuscript, related observations were reported by others (67). This dual SH2/SH3 domain contribution to the association of Grb2 with RPTP␣ appears different from the situation for other tyrosine-phosphorylated proteins such as the EGF receptor (10,14) or Syp (12). In these cases, the Grb2 SH2 domain by itself was reported to be sufficient for association with its tyrosinephosphorylated partner.
Three interpretations are possible for the observed SH3 domain requirement. First, deletion or mutation of the C-terminal SH3 domain may cause a conformational change that specifically affects the affinity of the SH2 domain for phosphorylated RPTP␣. However, our experiments provide clear evidence for a phosphotyrosine-independent interaction between RPTP␣ and Grb2, which requires the C-terminal SH3 domain as well as the first PTPase domain of RPTP␣ (Fig. 4), suggesting other interpretations. Second, the SH3 domain requirement may reflect the interaction of this domain with a third (unidentified) protein, whose presence is required to stabilize the complex between RPTP␣ and Grb2. Third, the C-terminal Ϯ0.39 (n ϭ 133) (t ϭ 3.14)** a Percentage of neurite-bearing cells, estimated as the fraction of cells bearing neurites longer than 2 cell diameters, after 2 days of growth factor treatment.
b Average neurite length (restricted to cells bearing neurites longer than 1 cell diameter). Neurite length was measured on photographs of fields containing 100 -250 cells each, expressed as -fold average cell diameter, and averaged over the number of neurites (n).
c Standard deviation of neurite length and test of significance. Student's t values are given with respect to the values for control cells infected with an "empty" vector encoding no RPTP␣ protein. Probability values are 0.05 for t Ͼ 1.96, 0.01 for t Ͼ 2.58 (**), and 0.001 for t Ͼ 3.29 (***) (if a two-tailed test for differences only is used, since there is no a priori reason to assume that values will be larger or smaller than controls). SH3 domain may separately interact with an epitope in RPTP␣, and this interaction may enhance the overall affinity of association between both proteins, through cooperation with the SH2 domain-mediated binding. Such a bidentate interaction of the SH2 and C-terminal SH3 domains of Grb2 with RPTP␣ could involve either the same RPTP␣ molecule (in a 1:1 complex) or bridging between two different RPTP␣ molecules (leading to 2:1 or 2:2 RPTP␣-Grb2 complexes). Cooperativity between SH2 and SH3 domains is reminiscent of the instance of the Src protein. This kinase is thought to be kept in a repressed state through an intramolecular interaction involving its SH2 domain and a C-terminal phosphorylation site (Tyr 527 ). Mutational studies have shown that the SH2 and the adjacent SH3 domain both contribute to the maintenance of the intramolecular interaction and the repressed kinase activity (68). Affinity for left-handed polyproline helices is characteristic of SH3 domains (9,56). We have obtained no evidence for the involvement of a proline-rich site present in RPTP␣ as a potential ligand to the C-terminal Grb2 SH3 domain. However, SH3 domains can engage in other types of interaction, as exemplified by studies on the Vav and Lck SH3 domains (33,69).
Distinguishing between the two latter scenarios (stabilization by a third protein or direct association between the Cterminal SH3 domain and an epitope in RPTP␣) will require detailed mutational, biochemical, and biophysical analysis; yet, either would provide a reasonable explanation for the apparently mutually exclusive association of Sos and RPTP␣ with Grb2 (1, 2) (Figs. 5, 6, 7, and 9), due to occupancy of the peptide binding site of the Grb2 C-SH3 domain as an intrinsic consequence of Grb2-RPTP␣ association. Whereas the N-terminal SH3 domain is argued to be the most important for Grb2-Sos association, the C-terminal domain clearly contributes to the affinity between the two proteins (11,21,52) (Fig. 1). However, steric hindrance at other levels could also contribute to the lack of simultaneous association of RPTP␣ and Sos with Grb2.
RPTP␣ as a Regulator of Grb2 Function?-Since the Nterminal SH3 domain of the Grb2 adaptor is not required for association with RPTP␣, this domain could still be available for interaction with a third "effector" protein other than Sos. Thus, the particular mode of association of RPTP␣ with Grb2 would constrain the repertoire of proteins with which this adaptor, once associated with RPTP␣, can additionally interact, ensuring that Grb2 couples different effectors to RPTP␣ than it does to other tyrosine-phosphorylated proteins. This situation is reminiscent of that in T-cells, where Shc association with Grb2 can affect the affinity of Grb2 for Sos (55). Which, if any, effector proteins might be recruited by Grb2 into a tertiary complex with RPTP␣ obviously deserves further study. Such a role for Grb2 in RPTP␣ function might be independent of its role in the Ras signaling pathway. Genetic analysis of the Drosophila homologue of Grb2, Drk, points to a function for its C-terminal SH3 domain that is independent of sevenless signaling. Whereas a Drk protein that lacks the C-terminal SH3 domain can rescue the function of the sevenless pathway, this mutant does not rescue the lethality of homozygous loss-offunction mutations of Drk (21).
The observation that increased expression of RPTP␣ can interfere with Grb2-Sos association (Figs. 7 and 9) raises the issue of whether, under certain circumstances, endogenous RPTP␣ may control Grb2 availability or function. This potential of RPTP␣ to compete for the available Grb2 pool may be dependent on the expression and phosphorylation level of the protein. Different cell lines and tissues express different levels of RPTP␣ protein, and the RPTP␣ mRNA is transiently upregulated at the onset of neuronal differentiation (36,38,39,41). Under normal growth conditions, only approximately 20% of RPTP␣ in 3T3 cells is tyrosine-phosphorylated (2). The stimuli or processes that control tyrosine phosphorylation of RPTP␣ remain unknown, although autodephosphorylation may play a role (2). It is at least conceivable that, in certain instances, a local or temporal increase in RPTP␣ expression or phosphorylation may affect the degree of association between Grb2 and other proteins, and thus the function of Grb2-mediated signaling pathways.
A Potential Role for Tyrosine Phosphorylation and Grb2 Association in RPTP␣ Function-The above views of RPTP␣, as a substrate for tyrosine phosphorylation and assembly of a signaling complex, appear to be unrelated to its catalytic activity. It is worth pointing out in this context that nonreceptor kinases of the Src family perform important functions, related to their ability to engage in interaction with other proteins, which are independent of their kinase activity (70,71). Alternatively, tyrosine phosphorylation or Grb2 association may modulate the catalytic activity or substrate specificity of RPTP␣.
Various approaches can be taken to understand the effect of Tyr 798 phosphorylation and the ensuing Grb2 association on RPTP␣ function. A direct kinetic analysis of the effect of phosphorylation of Tyr 798 on the protein's catalytic activity presents serious methodological obstacles. It is technically difficult to prepare sufficient amounts of RPTP␣ stoichiometrically phosphorylated solely at the Tyr 798 residue. Moreover, RPTP␣ is capable of autodephosphorylation, further complicating a kinetic analysis (2). Finally, an in vitro analysis of the effect of tyrosine phosphorylation per se may lack relevance for the in vivo situation, since tyrosine-phosphorylated RPTP␣ protein is likely to exist largely as a component of a Grb2-containing complex. Therefore, we have taken an in vivo approach to start addressing the function of tyrosine phosphorylation of RPTP␣.
A major problem in the study of PTPases concerns the absence of suitable biological assays. Ectopic expression of RPTP␣ can result in tumorigenic conversion of embryonic fibroblasts and in permissivity for retinoic acid-induced neuronal differentiation of P19 embryonic carcinoma cells in the absence of cell aggregation (41,42). The latter phenotype is independent of the C-terminal tyrosine phosphorylation site in RPTP␣ (67). Both effects may relate to the ability of RPTP␣ to dephosphorylate the negative regulatory residue Tyr 527 in c-Src and thus activate c-Src kinase activity. Interestingly, direct expression of mutationally activated Src blocks neuronal differentiation of P19 cells (72,73), suggesting that other mechanisms may also be at work.
We show here that WT RPTP␣ expression inhibits aFGFinduced neurite outgrowth in PC12 cells and that this effect simultaneously requires the catalytic activity of, and the Grb2 binding site in, RPTP␣ (Fig. 8A, Table I). Consistent with our studies on the mechanism of RPTP␣-Grb2 association, RPTP␣ expression interferes with Grb2-Sos association in a manner that is dependent on the tyrosine phosphorylation site in RPTP␣ (Fig. 9). Yet, this observation does not explain the negative effect of RPTP␣ on aFGF-induced neurite outgrowth; the catalytically inactive RPTP␣CCSS protein, which remains phosphorylated (data not shown) and is more efficient in disrupting the Grb2-Sos complex than WT RPTP␣ (Fig. 9), fails to interfere with aFGF-induced neurite outgrowth. Thus, inhibition of neurite outgrowth by RPTP␣ does not correlate with Grb2 sequestration, and the effect of Grb2 binding by RPTP␣ manifests itself only within the context of a catalytically active protein. Presumably, therefore, the signaling pathways leading to neurite outgrowth in response to aFGF are functional at the low levels of Grb2-Sos association observed upon expression of RPTP␣CCSS.
How can catalytically active RPTP␣ both mediate blockage (in the presence of an intact Tyr 798 site) and stimulation (upon mutation of the Tyr 798 site) of neurite outgrowth in response to aFGF? One explanation is that RPTP␣ initiates a separate, stimulatory pathway leading to neurite outgrowth, which, in contrast to the aFGF-induced pathway, is highly dependent on Grb2 availability. For instance, the ability of RPTP␣ to activate the kinase activity of c-Src (41,42) might be related to the ability of RPTP␣Y798F to stimulate neurite outgrowth. Such a model will need to explain the abolition of the negative effect of catalytically active RPTP␣ on aFGF-induced neurite outgrowth by mutation of the Tyr 798 site as well as explaining the ligand dependence of the effect of RPTP␣Y798F. A complementary interpretation is that tyrosine phosphorylation of RPTP␣ and/or Grb2 association, exerts a direct effect on the catalytic function of RPTP␣ in vivo. This could occur through various testable mechanisms. First, tyrosine phosphorylation and subsequent assembly of a complex containing the Grb2 adaptor may modulate the activity or substrate specificity of one, or both, of the PTPase domains in RPTP␣, for instance by controlling substrate access. This seems particularly conceivable in light of our studies, which suggest that formation of the RPTP␣-Grb2 complex may entail a bidentate SH2/SH3 contact between Grb2 and the RPTP␣ intracellular domain or may involve assembly of a tertiary complex with another as yet unidentified protein (Fig. 4B). Second, Grb2 may function as an adaptor to recruit particular tyrosine-phosphorylated proteins as substrates for the PTPase. Only limited evidence exists concerning tyrosine phosphorylation of Sos or Grb2 itself (74), but examples of tyrosine-phosphorylated proteins that associate with the SH3 domains of Grb2 have been reported (e.g. Refs. 13, 27, 28, and 33). Third, Tyr 798 phosphorylation and/or Grb2 association may control the intracellular localization of RPTP␣. Either of these regulatory functions for tyrosine phosphorylation of a specific residue of an R-PTPase would constitute a striking analogy with the much more extensively documented role of tyrosine phosphorylation of receptor tyrosine kinases. Finally, in view of suggestions of RPTP␣ dimerization (40) and the fact that RPTP␣ can undergo trans-dephosphorylation (2), it is tempting to speculate that tyrosine phosphorylation/dephosphorylation of an R-PTPase such as RPTP␣ (and ensuing Grb2 association) may be controlled by the protein's oligomerization status, providing a potential mechanism for trans-membrane signaling. Examination of these hypotheses will require an extensive analysis of tyrosine phosphorylation events and the function of signaling pathways in cells expressing these and additional RPTP␣ mutants in response to various factors.