c-SRC mediates neurite outgrowth through recruitment of Crk to the scaffolding protein Sin/Efs without altering the kinetics of ERK activation.

SRC family kinases have been consistently and recurrently implicated in neurite extension events, yet the mechanism underlying their neuritogenic role has remained elusive. We report that epidermal growth factor (EGF) can be converted from a non-neuritogenic into a neuritogenic factor through moderate activation of endogenous SRC by receptor-protein-tyrosine phosphatase alpha (a physiological SRC activator). We show that such a qualitative change in the response to EGF is not accompanied by changes in the extent or kinetics of ERK induction in response to this factor. Instead, the pathway involved relies on increased tyrosine phosphorylation of, and recruitment of Crk to, the SRC substrate Sin/Efs. The latter is a scaffolding protein structurally similar to the SRC substrate Cas, tyrosine phosphorylation of which is critical for migration in fibroblasts and epithelial cells. Expression of a dominant negative version of Sin interfered with receptor-protein-tyrosine phosphatase alpha/EGF- as well as fibroblast growth factor-induced neurite outgrowth. These observations uncouple neuritogenic signaling in PC12 cells from sustained activation of ERK kinases and for the first time identify an effector of SRC function in neurite extension.

Most analyses that use reductionist systems such as PC12 cells to identify pathways governing neuronal process formation have focused on the Ras/ERK 1 pathway, the requirement of which is well established. The neuritogenic effect of constitutively active MEK has been taken to indicate that ERK activation is also sufficient for neuritogenesis. Yet, physiological ERK activation, e.g. in response to EGF, often does not engender neurite extension. An attempt to resolve this paradox has been the postulate that the kinetic pattern of ERK signaling after a stimulus is what dictates the nature of the ensuing response (1). In this view, "sustained" activation of ERK (hours), such as typically seen for FGF or NGF, would be the signal that specifies a neuritogenic outcome as opposed to the "transient" (Ͻ1 h) ERK induction associated with non-neuritogenic factors such as EGF. However, such a correlation does not necessarily imply sufficiency or necessity.
Diverse and independent approaches have exposed the necessity of postulating pathways to process formation other than ERK. Recurrently, tyrosine kinases of the SRC family (SFKs) have been proposed as candidates for such "parallel pathways." A mutation in Trk retains ERK induction but abolishes neurite outgrowth by NGF (2). Analysis of platelet-derived growth factor receptor mutants showed that sustained ERK activation per se is insufficient but that neuritogenesis requires additional signals possibly involving SRC (3). v-SRC induces neurite outgrowth in PC12 cells (4), whereas c-SRC inhibition can block outgrowth (5,6). This crucial role of SFKs in neurite extension is not restricted to PC12 cells but is equally clearly encountered in conditionally immortalized (7) and primary neurons (8,9). SFKs associate with adhesion molecules that facilitate process elongation (10,11) and are enriched in nerve growth cones, where they interact with the cytoskeleton in a protein-tyrosine phosphatase (PTP)-dependent manner (12).
In fibroblasts and epithelial cells, SRC is intimately implicated in phosphorylation of focal adhesion proteins, turnover of these structures, and cell motility (13,14). v-SRC can activate or inhibit ERK kinases, depending on the cell type and stage of transformation (15,16). Its action as a transforming oncogene can be separated from its ability to activate Ras and ERK (17,18). One of the major SRC substrates, Cas, in complex with the adaptor Crk, is a key mediator of cell migration in fibroblasts and epithelial cells (19 -23). By contrast, in neuronal cells, the identity of signaling steps downstream of c-SRC has remained obscure. One reason for this lack of progress has been the previous extensive reliance on mutationally activated v-SRC. Given the drastic degree of kinase activation and deregulation of the latter, this approach has not been informative regarding the function of endogenous c-SRC. Indeed, the signaling pathways downstream of oncogenic and cellular SRC proteins were recently shown to differ substantially; for instance, c-SRCmediated activation of certain promoters relies exclusively on Rap1, whereas transforming SRC alleles also signal through Ras (24). Approaches using oncogenic SRC alleles are particularly compromised by the experimental difficulty of separating the biological effect of mutationally activated v-SRC from its ability to deregulate ERK kinases (15,16).
SFKs are regulated by a conformational mechanism that is controlled by phosphorylation. Intramolecular interactions between the SH2 domain and a C-terminal tyrosine phosphorylation site (Tyr-527 in chicken SRC) in combination with SH3mediated interactions stabilize a kinase-inactive conformation (25). In consequence, wild type SRC family kinases are reversibly activated in situ by ligands to their SH2 and SH3 domains (24,26,27) or by PTPs that mediate Tyr(P)-527 dephosphorylation (28). Extensive evidence identifies receptor-PTP␣ (RPTP␣) as one such physiological Tyr(P)-527 phosphatase. This PTP, which is particularly abundant in neural tissue, associates physically with SRC and Fyn (26,29,30), and its overexpression activates these kinases (26, 29 -32); conversely, loss of RPTP␣ leads to dose-dependent reductions in SRC and Fyn kinase activities and generates integrin signaling deficits similar to SRC Ϫ/Ϫ cells (33,34). RPTP␣ itself undergoes tyrosine phosphorylation at a residue (Tyr-798) in its C terminus. This modification leads to Grb2 recruitment (35,36) and alters RPTP␣ function (36), SRC-activating ability (26), and localization (37). In the present study, we have exploited the role of RPTP␣ as a physiological SRC activator to identify signaling events downstream of SRC in neuronal cells.

MATERIALS AND METHODS
Cell Culture-PC12 cells were cultured in Dulbecco's modified Eagle's medium plus 10% fetal calf serum and 10% horse serum. Retrovirus production and infection were as described (36). For neuritogenesis, 5 ϫ 10 4 cells were seeded per 35-mm plate, grown overnight, and starved (0.5% fetal calf serum plus 0.5% horse serum) for 16 -20 h. To this medium was then added 50 ng/ml acidic FGF plus 5 g/ml heparin or 100 ng/ml EGF.
Data Analysis-After 2 days of stimulation, neurite length was measured on photographed fields containing 100 -250 cells. Data were expressed in two ways; first, as total neurite length averaged over cell number (bar diagrams; y axis ϭ cell diameters); second, as percent neurite-bearing cells (percent cells bearing at least one neurite larger than two cell diameters). The level of statistical significance was assessed by a two-sided t test (unequal variance). In bar diagrams, error bars always indicate 95% confidence intervals. For numerical data (% neurite-bearing cells), the extent of a 95% confidence interval is indicated by the number between brackets. All key conclusions were reconfirmed on independent clones and/or pools of clones selected en masse.
Antibodies and Plasmids-Anti-RPTP␣ (36), and anti-Sin (38) sera have been described. Anti-SRC was from Calbiochem. For the immunoprecipitation/in vitro ERK assay and immunoblotting, anti-ERK-1 C-16 and anti-ERK2 C-14 (Santa Cruz) were used, respectively; antiphospho-ERK antibody was from New England Biolabs. Anti-phosphotyrosine antibody 72 was described (36); 4G10 was from Upstate Biotechnology. Anti-Cas was a gift of T. Parsons and A. Bouton (University of Virginia). Anti-CrkL C-20 was from Santa Cruz. Anti-Nck sera were provided by E. Skolnik (New York University) or from Santa Cruz (C-19).
RPTP␣ constructs were described (36). The Sin deletion SinSD, lacking residues 101-256 (38), was generated using the Exsite kit (Stratagene). Two retroviral vectors were used: pLXSHD (36), and pBabeI-EG (D. Unutmaz, NYU). The latter encodes an long terminal repeat-driven di-cistronic transcript consisting of the transduced cDNA and green fluorescent protein (3Ј to an internal ribosomal entry site). cDNA inserts were ligated between XhoI and BamHI of pLXSHD or into the BamHI site of pBabeI-EG.
Immunofluorescence-Cells on collagen-coated coverslips were fixed (4% paraformaldehyde/phosphate-buffered saline, 20 min) and permeabilized (0.1% Triton X-100, phosphate-buffered saline, 20 min). They were blocked in phosphate-buffered saline plus 10% fetal calf serum for 30 min and incubated with primary antibody for 1 h followed by 10 g/ml secondary antibody and examined under a Zeiss photomicroscope.
ERK1 assays involved precipitation from Triton lysate and incubation in 25 l of buffer (10 mM Tris, pH 7.3, 10 mM MgCl 2 ) with 10 Ci of [␥-32 P]ATP plus 12.5 g of myelin basic protein for 25 min at 30°C. Reactions were stopped by SDS-PAGE sample buffer, and gel autoradiographs were quantitated using phosphorimaging.

EGF Induces Neurite Outgrowth in c-SRC-overexpressing
PC12 Cells-v-SRC induces growth factor-independent neuritogenesis in PC12 cells (4), but how this may relate to the physiological function of c-SRC is poorly understood. We used retroviral infection to generate a pool of PC12 cells overexpressing c-SRC (Fig. 1A). This conferred upon EGF (normally a solely mitogenic factor (1)) an ability to induce neurite formation equal to that of a bona fide neuritogenic factor (FGF) in control cells (Fig. 1C). This suggests that an SRC-dependent function can contribute to converting a non-neuritogenic factor (EGF) into a neuritogenic one. Increases in tyrosine phosphorylation after c-SRC overexpression were largely limited to four proteins of 130, 90, 70, and 60 kDa (the latter being SRC itself) (Fig. 1B).
RPTP␣ Localizes to Cell-Cell Contact Zones and Tips of Spontaneous Spikes and Activates Endogenous SRC-To ask whether endogenous levels of c-SRC could similarly contribute to conversion of EGF into a neuritogenic factor, we relied on the known c-SRC-activating function of RPTP␣ (26,29,(31)(32)(33)(34). We previously reported generation of PC12 lines expressing wt or mutant RPTP␣ (36). The mutants used were RPTP␣CCSS, catalytically inactive due to mutation of the active site cysteine residues in either PTP domain (Cys-442, Cys-732) to serine, and RPTP␣Y798F, in which Tyr-798, the site of tyrosine phosphorylation in RPTP␣, was mutated to phenylalanine (resulting in loss of Grb2 recruitment but normal in vitro catalytic activity (35)). Immunofluorescence revealed RPTP␣ was membrane-localized and concentrated in cell-cell contact zones and at the tips of spontaneous spikes ( Fig. 2A); no significant differences were seen in localization of wt versus mutant proteins (not shown).
FIG. 1. SRC expression allows neurite outgrowth in response to EGF. PC12 cells were infected with an SRC-encoding or control retrovirus, and lysates were analyzed by anti-SRC (A) or anti-phosphotyrosine (B) immunoblotting (IB). C, neurite formation in the absence or presence of EGF or FGF. Data are expressed as averaged neurite length/cell (bar diagram; y axis ϭ cell diameters) or as % neuritebearing cells (% n, numbers below diagram ϭ % of cells bearing a neurite larger than two cell diameters). 95% confidence intervals are indicated by error bars (in the bar diagram indicating neurite lengths) or by the numbers between brackets (next to numbers below indicating % neurite bearing cells). Statistical significance is tested with respect to unstimulated control PC12 cells or to unstimulated c-SRC expressing cells (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).
wt RPTP␣ and RPTP␣Y798F, but not RPTP␣CCSS, moderately increased c-SRC activity (by 2-fold), with the level of c-SRC protein remaining unchanged (Fig. 2B). This increase in specific activity likely reflects the well documented ability of RPTP␣ to reduce the phosphorylation level of the Tyr-527 residue in c-SRC (33,34). Contrasting with the situation for c-SRC-overexpressing cells, we observed no constitutive effects of RPTP␣ on cell morphology or on neurite extension in the absence of added growth factors (data not shown and Fig. 3B).
RPTP␣ Stimulates EGF-induced Neurite Outgrowth in a Manner That Requires Both SRC and ERK-As reported previously (36), the presence of wt RPTP␣, but not catalytically inactive RPTP␣CCSS, impairs the ability of FGF (and NGF; data not shown) to cause neurite outgrowth. Mutation of the Tyr-798 phosphorylation site in RPTP␣ not only abolishes, but inverts this effect of wt RPTP␣, since RPTP␣Y798F behaves instead as a net potentiator of FGF-induced outgrowth (36) (Fig. 3A, right panel).
We here report that, unexpectedly, the effect of wt RPTP␣ on responsiveness to EGF was the opposite of that on FGF responsiveness; that is, wt RPTP␣ converted EGF into a promoter of neurite outgrowth (Fig. 3A, left panel). Furthermore, RPTP␣Y798F promoted neurite outgrowth in response to EGF even more powerfully than wt RPTP␣ (Fig. 3A, left panel). In contrast, RPTP␣CCSS did not potentiate EGF responsiveness; this catalytically inactive protein even somewhat reduced the formation of short spikes that occurs in response to EGF. It is conceivable that this weak inhibitory effect of RPTP␣CCSS reflects dominant negative-like interference with the function of endogenous RPTP␣ or of a related RPTP. However, to what extent C-to-S mutant PTPs can indeed be relied on to be specific and true dominant-negatives is as yet still unclear; hence, interpretation of the significance of this observation is probably premature. Assessment of mRNA levels by Northern blotting for the metalloprotease transin, a late marker for NGF-induced neuronal differentiation of PC12 cells (39,40), revealed that the EGF-induced neurite extension correlated fully with expression of this marker. EGF treatment lead to transin induction in cells expressing wt RPTP␣ or RPTP␣Y798F but not RPTP␣CCSS; moreover, this effect was stronger in RPTP␣Y798F than in wt RPTP␣-expressing cells (supplemental data included in the on-line version of manuscript). This indicates that EGF induced not merely morphological effects but also changes in gene expression.
The ability of RPTP␣ to confer outgrowth-promoting activity on the previously non-neuritogenic EGF is reminiscent of c-SRC overexpression, although less drastic, since the effect of RPTP␣ remained fully EGF-dependent ( Fig. 1C versus Fig. 3, A and B, and data not shown). Pharmacological SRC inhibition using PP1 abolished the neuritogenic effect of RPTP␣ and RPTPY␣798F on stimulation with EGF, whereas cell morphology and viability remained unaffected (Fig. 3B), indicating that the stimulatory effect of RPTP␣ on EGF-induced neuritogenesis depends on the activity of a SFK. At the same time, EGF-induced neurite outgrowth still continued to be dependent on ERK activity, as shown by application of PD98059 (a selective MEK1 inhibitor) (Fig. 3B). We did observe a tendency of EGF-stimulated RPTP␣Y798F-expressing cells to still form "stumps" in the presence of 10 M PD98059. However, Ͻ2% of cells formed protrusions longer than 2 cell diameters, and any Lysates were subjected to anti-SRC immunoprecipitation (IP), and precipitates were split in two. One-half was subjected to in vitro SRC kinase assay using enolase as exogenous substrate (top); the other half was analyzed by anti-SRC immunoblotting (IB).

FIG. 3. RPTP␣ causes SRC-and MEK-dependent neurite outgrowth in response to EGF, and this effect is potentiated by mutation of the Tyr-798 phosphorylation site in RPTP␣.
PC12 clones (control or expressing wt RPTP␣, RPTP␣CCSS (catalytically inactive), or RPTP␣Y798F (lacking the Tyr-798 phosphorylation site in RPTP␣)) were exposed to EGF or FGF for 48 h in the absence or presence of the MEK inhibitor PD98059 (10 M) or the SRC inhibitor PP1 (1 M). The inhibitory effects of wt RPTP␣ and stimulatory effect of RPTP␣Y798F on FGF-induced neuritogenesis were described previously (36) and are included here as controls. A, neurite outgrowth data measured over 100 -250 cells. Data and 95% confidence intervals are expressed as described under "Materials and Methods" and in the legend of Fig. 1 (histogram ϭ average neurite length; numbers ϭ fraction of neurite-bearing cells). The level of statistical significance was determined by a two-sided t test with respect to stimulated vector (V; control) cells (i.e. left bar in each graph) (*, p Ͻ 0.05; ***, p Ͻ 0.001). B, photographs of representative fields after EGF treatment in the absence or presence of the indicated inhibitors.
protrusions formed were Ͻ1 cell diameter. The number of these protrusions could be reduced significantly by raising the concentration of PD98059 to 25 M (data not shown). Hence, these protrusions cannot be referred to as neurites, and their true biological significance is questionable at best.

Conversion of EGF into an Outgrowth-promoting Factor by RPTP␣ Is Not Accompanied by Alteration in ERK Kinetics-A
widely cited model traces back the divergent effects of EGF versus FGF on normal PC12 cells to differences in their kinetics of ERK activation. It has been proposed that the relatively transient ERK activation induced by EGF is insufficient for neuritogenesis, which would require the more sustained (Ͼ2 h) activation of ERK that is observed for NGF or FGF. This hypothesis was prompted by observations that experimental induction of neurite formation by overexpression of many signaling molecules is invariably accompanied by a shift in ERK activation kinetics from a transient to a sustained mode (1). Hence, we wished to determine whether the ability of wt RPTP␣ and RPTP␣Y798F to convert EGF into a neuritogenic factor could similarly be accounted for by an alteration, from transient to sustained, in the kinetics of EGF-induced ERK activation.
We observed that wt RPTP␣ inhibited ERK activation in response to FGF at both the 5-min and 6-h time points, with wt RPTP␣ more potent at reducing ERK activation than RPTP␣CCSS and RPTP␣Y798F (Fig. 4A), suggesting that optimal inhibition required both catalytic activity and Grb2 binding. Although reduced ERK activation might explain the ability of wt RPTP␣ to inhibit FGF-induced neurite formation (Fig.  3A), this correlation broke down for RPTP␣Y798F, which reduced ERK activation by FGF (Fig. 4A and data not shown) but enhanced neurite extension induced by this factor (Fig. 3A).
The situation in the case of EGF was analyzed in extensive detail. As expected from the literature (1), in control cells, EGFand FGF-induced ERK activities were comparable at the 5-min time point but differed significantly at later time points, with FGF-induced activation sustained longer (Fig. 4, A-C). Strikingly, however, neither wt RPTP␣ nor RPTP␣Y798F (which both potentiate EGF-induced neurite outgrowth; see Fig. 3) potentiated the extent of ERK induction by EGF. In the short term (5 min), expression of wt or mutant RPTP␣ actually tended to reduce ERK activation (Fig. 4B, left panel). More importantly, none of the wt or mutant RPTP␣ proteins significantly altered EGF-induced ERK activation at the late time point (6 h) (Fig. 4B, right panel). As expected from the literature (1), in the control (vector) cells at late time points, EGFinduced ERK activation was significantly lower than FGFinduced activation (Fig. 4B, right panel, and C). Similar conclusions regarding the effect of RPTP␣ on ERK activation were reached by in vitro kinase assay (Fig. 4B) and by immunoblotting with phospho-specific antibodies for the activated state of ERK 1 and 2 (Fig. 4C).
We conclude that wt RPTP␣ and RPTP␣Y798F counteract the induction of ERK by FGF but do not alter the extent or kinetics of ERK activation in response to EGF. This breakdown of the correlation between neurite outgrowth and the extent of sustained ERK induction leaves the ability of wt RPTP␣ and RPTP␣Y798F to induce a neuritogenic response to EGF unexplained.
RPTP␣ in a SRC-dependent Manner Elevates Tyrosine Phosphorylation of the Docking Proteins Cas and Sin-In a search for alternative explanations for conversion of EGF into an outgrowth-promoting factor, we assessed how RPTP␣ affected cellular tyrosine phosphorylation. We found that wt RPTP␣ and RPTP␣Y798F (but not RPTP␣CCSS) specifically elevated tyrosine phosphorylation of a 90-kDa protein and also caused a more modest increase in a 130-kDa species (Fig. 5A). This pattern is similar to that observed after an increase in c-SRC expression (Fig. 1B). EGF or FGF did not affect the phosphotyrosine content of these 90-and 130-kDa proteins, and wt or mutant RPTP␣ did not alter the overall patterns of tyrosine phosphorylation induced in response to EGF or FGF (not shown).
In trying to characterize these proteins, we found (data not shown) that RPTP␣ had not affected phosphorylation of cortactin and Cbl (both SRC substrates) or of FRS-2 (implicated in FGF-dependent PC12 differentiation). However, we successfully identified the proteins whose tyrosine phosphorylation level was elevated by RPTP␣ as Sin (90 kDa) and Cas (130 kDa) FIG. 4. RPTP␣ decreases ERK activation in response to acidic FGF but does not affect the extent and kinetics of ERK activation in response to EGF. A and B, relative ERK activity at various time points as determined by in vitro ERK kinase assay. A representative autoradiograph is shown in A. ERK activities in response to FGF relative to the vector control were 0.4 (wt), 0.5 (CCSS), and 0.6 (YF) at 5 min and 0.4 (wt), 0.6 (CCSS), and 0.6 (YF) at 6 h. Quantitation and statistical analysis of 3 independent experiments is shown in B. Error bars denote 95% confidence intervals. Statistical significance was tested by a two-sided t test with respect to EGF-stimulated empty vector-infected (control) cells (i.e. left column in each graph) (*, p Ͻ 0.05). C, lysates from clones exposed to EGF for the indicated times were analyzed by immunoblotting with an antibody against activated (phosphorylated) ERK1/2. For each time point, vector-infected (V, control) cells were also stimulated with acidic FGF as a positive control. MBP, myelin basic protein. (Fig. 5, B and D), 2 related docking proteins and known SRC substrates (21,22). Quantitative immunodepletion experiments demonstrated that these 2 proteins (or associated proteins of similar sizes) accounted for the bulk of the RPTP␣induced increase in tyrosine phosphorylation at 90 and 130 kDa (Fig. 5B).
Pharmacological SRC inhibition was used to determine whether the RPTP␣-induced increase in tyrosine phosphorylation of Sin required SRC activity. As shown in Fig. 5D, the SRC inhibitor PP1, along with its ability to inhibit neurite outgrowth described above (Fig. 3B), also reduced tyrosine phosphorylation of Sin to undetectable levels, indicating that RPTP␣-induced phosphorylation of Sin is fully SRC-dependent.
RPTP␣ Induces Complex Formation of Sin and Cas with the Adaptors Crk and Nck-In fibroblasts, Cas associates with, and is a substrate, an effector, and an activator of c-SRC (22), controlling organization of the actin cytoskeleton, cell migration, and activation of JNK kinases (21,41,42). Sin (Efs), whose expression is more restricted than Cas, was isolated in independent screens as a ligand for the SH3 domains of both SRC (38) and Fyn (43). A third mammalian member of this family, HEF-1, was isolated on the basis of its ability to induce pseudohyphal growth in yeast (21). All three share a common domain structure consisting of an SH3 domain, a "substrate" domain containing tyrosine phosphorylated SH2 binding sites, a proline-rich region, and a C-terminal domain with SH3 and SH2 consensus binding sites for SFKs (Fig. 6A) (21). Their association with cytoplasmic tyrosine kinases is followed by phosphorylation of the substrate region and ensuing recruitment of adaptors and other signaling proteins, leading to assembly of higher order complexes with the potential for interactions between associated effectors (21). RPTP␣ enhanced such recruitment of the adaptor CrkL to Sin and Cas, as shown by analysis of anti-CrkL immune precipitations from control and wt RPTP␣-expressing cells by anti-Cas or anti-Sin immunoblotting (Fig. 5C). Similar results were obtained for CrkII (which is less abundant in PC12 cells than CrkL (44); not shown) and for Nck (Fig. 5C). Consistent with the ability of the SRC inhibitor PP1 to antagonize the increase in tyrosine phosphorylation of Sin that is induced by RPTP␣ (Fig. 5D), PP1 also reversed the increase in RPTP␣-induced complex formation between Sin or Cas and CrkL (Fig. 5E).
Sin-Crk Coupling Is a Necessary Mediator of RPTP␣/SRCdependent Neuritogenesis-To test the functional relevance of increased phosphorylation of Sin as caused by RPTP␣-induced activation of SRC, we designed a mutant, SinSD, which lacks the substrate region that encompasses the tyrosine residues whose phosphorylation recruits Crk and Nck (Fig. 6A). By retaining the binding sites for the SH2 and SH3 domains of SRC and Fyn, we expected SinSD to compete with endogenous Sin for association with SRC and, thus, to antagonize assembly of productive complexes around endogenous Sin. An analogous mutant of Cas has been successful in elucidating the contributions of Cas to cell migration (20) and JNK activation in fibroblasts (42).
Parental PC12 cells and wt RPTP␣ or RPTP␣Y798F expressors were infected with a control or SinSD-expressing retrovi- rus. As expected, SinSD successfully and specifically interfered with tyrosine phosphorylation of endogenous Sin (Fig. 6B). No effect of SinSD was noted on the growth rate or morphology or unstimulated cells (not shown). However, SinSD interfered with the neuritogenic response of wt RPTP␣-and RPTP␣Y798Fexpressing cells to EGF (Fig. 6C) and of parental PC12 cells to FGF (Fig. 6D). This indicates that endogenous Sin is a necessary effector for the ability of RPTP␣ to convert EGF into a neuritogenic factor and that normal endogenous Sin function is necessary for optimal FGF-induced outgrowth. We also observed that Sin overexpression potentiated the RPTP␣ effect on EGF-induced neuritogenesis (Ref. 45; data not shown), further consistent with Sin being a limiting effector downstream from RPTP␣-activated c-SRC.
A caveat to the above (and many) dominant negative approaches is that SinSD may compete for access to c-SRC, not only with endogenous Sin but also with other SRC substrates. Hence, whereas the result in Fig. 6C actually constitutes powerful additional evidence for the involvement of SRC (or Fyn) in RPTP␣-induced neurite outgrowth in response to EGF, the effect of SinSD might still merely reflect displacement of another, more critical but unidentified, SRC substrate. To alleviate this concern, we asked whether Crk, as a downstream effector of Sin, would be similarly important for EGF/RPTP␣-FIG. 6. RPTP␣/SRC-dependent neurite extension requires tyrosine phosphorylation of Sin and Crk. A, schematic diagram of the Cas, Sin, and SinSD proteins. SR, substrate region containing phosphorylatable tyrosines that constitute binding sites for adaptors; PL, proline-rich SRC-SH3 binding sites; Y, tyrosine motif constituting a SRC-SH2 binding site; C, C-terminal region of homology between Cas and Sin (21). B, expression of SinSD interferes with tyrosine phosphorylation of endogenous Sin. Parental cells and cells expressing wt RPTP␣ or RPTP␣Y798F were infected with control (vector (V)) or SinSD-expressing virus. Total lysates (TL) were analyzed by immunoblotting (IB) with anti-RPTP␣ (upper) and with anti-Sin (second panel from top). Anti-Sin immune precipitates (IP) were analyzed by immunoblotting with anti-phosphotyrosine (third panel from top) and with anti-Sin (bottom panel). C, SinSD blocks neuritogenesis of RPTP␣-expressing cells in response to EGF. Pools of cells expressing wt RPTP␣ or RPTP␣Y798F in the absence of presence of SinSD were exposed to EGF, and average neurite length (expressed in cell diameters) was measured after 48 h. Error bars denote 95% confidence intervals; Statistical significance for SinSD-infected RPTP␣ (WT) and SinSD-infected RPTP␣ (YF) cells was tested with respect to vector-infected RPTP␣ (WT) cells and vector-infected RPTP␣ (YF) cells, respectively (***, p Ͻ 0.001). D, FGF-induced neurite outgrowth in parental PC12 cells (Uninf.) or cells infected with control empty vector or a SinSD-expressing retrovirus. E, schematic representation of dominant-negative CrkR38K, in which a lysine residue critical for SH2 domain function was mutated to arginine. F, expression of CrkR38K. Pools of cells expressing wt RPTP␣ (WT) or RPTP␣Y798F (YF) were infected with empty (vector (V)) or CrkIIR38K-encoding retrovirus, and expression of RPTP␣ and CrkIIR38K was monitored by immunoblotting. G, CrkR38K blocks neuritogenesis of RPTP␣-expressing cells in response to EGF. Cells expressing wt RPTP␣ (left) or RPTP␣Y798F (right) in the presence or absence of CrkIIR38K were exposed to EGF for 48 h, and average neurite length was measured (**, p Ͻ 0.01). Average neurite length in control (non-Crk)-superinfected RPTP␣-expressing cells was somewhat lower than in Fig. 3 due to reduced RPTP␣ expression as a consequence of superinfection (data not shown). induced neuritogenesis. We introduced a CrkII mutant, CrkR38K, into wt RPTP␣-or RPTP␣Y798F-expressing cells (Fig. 6F); this mutant adaptor contains a point mutation in its SH2 domain that abolishes its ability to associate with tyrosine-phosphorylated docking proteins (Fig. 6E). It can be expected to interact with the full range of effectors available to endogenous CrkII but to sequester them in a complex incapable of associating with endogenous docking proteins for Crk such as Cas and Sin. The experiment revealed clear interference of CrkR38K with RPTP␣/SRC-dependent neurite outgrowth (Fig. 6G). DISCUSSION Correlative evidence originally gave rise to the hypothesis that the qualitative distinction between a non-neuritogenic versus a neuritogenic response is dictated by quantitative differences in ERK signaling (1). We observed here conversion of a non-neuritogenic (EGF) into a neuritogenic factor without alteration in the extent or kinetics of ERK induction. Moreover, in response to FGF, wt RPTP␣ and RPTP␣Y798F inhibited ERK activation, yet the latter elicited neurite extension. Our data thus demonstrate that the kinetic pattern of ERK activation does not constitute a necessary determinant of growth factor specificity. Potentially related observations have been made using other approaches, e.g. studying the synergism between NGF and IL-6 in a particular variant PC12 line (46), or studying the effect of stable transfection of GTPase-deficient G protein subunits (47) (however, in the latter case, the effect was constitutive, not showing the growth factor dependence observed here). In the present study, we identify SRC-mediated assembly of Sin-Crk complexes as the "alternative" pathway that contributes to formation of neuritic processes.
The latter conclusion is based on 1) the well established function of RPTP␣ in SRC activation (26,29,(31)(32)(33)(34) (Fig. 2B); 2) the similar effects of c-SRC (Fig. 1C) and RPTP␣ (Fig. 3) on EGF-induced neuritogenesis; 3) the fact that pharmacological SRC inhibition abolished the ability of RPTP␣ to alter the nature of the response to EGF (Fig. 3B); 4) the similarity between the patterns of tyrosine phosphorylation induced by RPTP␣ and c-SRC (Figs. 1B and 5A); 5) the observed SRC-dependent increase in phosphorylation and Crk binding of Sin and Cas, two known c-SRC substrates (Fig. 5B); and 6) the ability of both the dominant-negative SRC substrate SinSD and a dominant-negative version of the associated adaptor CRK to inhibit neuritogenesis (Fig. 6, C and F). Although each argument is by itself subject to alternative interpretations, taken together they strongly implicate the cascade SRC Ͼ Sin Ͼ Crk as a signaling cassette capable of contributing to neurite outgrowth. Notwithstanding the wealth of previous evidence implicating SRC in neurite formation, the identity of its downstream effectors had remained surprisingly enigmatic; to our knowledge, the present demonstration of the role of Sin-Crk complexes constitutes the first elucidation of the mechanism of action of SFKs in this process. We consistently noted that expression of dominant negative versions of Sin or Crk reduced RPTP␣/EGF-or FGF-induced neurite outgrowth but did not altogether abolish it (Fig. 6, C, D, and G). Further study will determine whether this reflects an only partial dominant negative effect (for instance because of an inability to achieve insufficient expression levels of SinSD or Crk-R38K) or, more interestingly, reflects the existence and function of alternative effectors (other than Sin) downstream of SRC that function in neurite outgrowth.
Docking of Crk to Cas, a close homolog of Sin, is a crucial signaling step in fibroblast migration, acting in an ERK-independent manner (19 -21, 23), and PTPs that dephosphorylate Cas also severely affect cell migration (28). Given the mecha-nistic analogies between migration and neurite extension, the contribution of related scaffolding proteins and effector pathways to the latter would thus not seem implausible, yet had thus far not been recognized.
The pathway outlined above needs to be qualified in two ways. First, it is unclear to what extent the effects observed occur through activation of SRC versus Fyn. Although RPTP␣ regulates both kinases (30,33,34), the role of Fyn in PC12 cells is a totally unexplored issue that may require further study; however, because Sin is a common substrate for both (38,43), it may constitute a point of convergence. Second, we have not been able to dissect with reasonable certainty the respective contributions of Cas versus Sin. Given their similarity, antibody cross-reaction, common binding partners, and ability to heterodimerize (21,48), resolution of this issue will have to rely on reagents as yet to be developed, such as cell lines that are null for either.
Strikingly, wt RPTP␣ potentiated the neuritogenic capacities of EGF but inhibited FGF-dependent neuritogenesis (36). The latter function likely results from a separate ability of RPTP␣ to impair FGF-induced ERK activation, which offsets the neuritogenesis-stimulating effect of activating the SRC-Sin-Crk pathway. Two arguments suggest that this inhibitory effect of RPTP␣ on FGF signaling involves a non-SRC substrate for RPTP␣ that is important only for FGF/NGF-mediated outgrowth. First, c-SRC overexpression mimicked the stimulating effect of RPTP␣ on EGF responsiveness but not its inhibitory effect on FGF responsiveness (Fig. 1C versus Fig.  3A). Hence, the effect of RPTP␣ on FGF signaling must diverge from that on EGF responsiveness upstream of SRC. Second, the two functions could be mutationally separated; the ability of RPTP␣ to inhibit FGF-induced outgrowth is dependent on its Tyr-798 phosphorylation site, but its outgrowth-eliciting ability in response to EGF is not. Thus, abolishing the ability of RPTP␣ to be phosphorylated (the Y798F mutation) converts RPTP␣ from an inhibitor to a net stimulator of FGF-induced neurite outgrowth (most likely due to an additive effect of enhanced Sin-Crk signaling) and further potentiates the ability of RPTP␣ to induce neurite outgrowth in response to EGF. We suggest that phosphorylation of Tyr-798 in RPTP␣ is necessary for the ability of RPTP␣ to dephosphorylate a substrate in the pathway from the FGF receptor to ERK activation. Such modulation of the substrate specificity of RPTP␣ by Tyr-798 phosphorylation could occur in various ways, e.g. displacement (as described for the RPTP␣-Tyr-798/SRC-SH2 interaction in fibroblasts (26); but see below), RPTP␣-bound Grb2 acting as an adaptor mediating recruitment of a specific substrate to RPTP␣, or altered RPTP␣ intracellular localization (37).
Zheng et al. (31), and den Hertog et al. (32) reported 3-6-fold and 4 -6-fold increases in SRC kinase activity following RPTP␣ overexpression. The comparatively more modest (2-fold; Fig. 2) effect of RPTP␣ on c-SRC activity seen here in PC12 cells may reflect a more modest level of RPTP␣ overexpression in the present case or, alternatively, may be an intrinsic feature of PC12 cells. De-phosphorylation of Tyr-527 activates c-SRC 10 -20-fold (49). However, increases of this magnitude in total cellular SRC activity are never observed. Rather, increases in overall cellular SRC activity that are of physiological significance tend to be much more modest, in all probability reflecting activation of only a distinct subpopulation of the total content of cellular SRC. EGF and NGF were reported to increase tyrosine phosphorylation of c-SRC by a few fold (50). Determining to what extent the RPTP␣-activated and growth factor-activated SRC pools may be related or overlapping should be an area of detailed and careful further analysis.
Although mutation of Tyr-798 abolished the ability of RPTP␣ to inhibit FGF-dependent outgrowth, we observed it to have only minor effects on the ability of RPTP␣ to activate SRC. Thus, wt RPTP␣ and RPTP␣Y798F both enhanced in vitro SRC kinase activity (Fig. 2B), in vivo phosphorylation of the SRC substrate Sin (Fig. 5), and EGF-dependent neurite outgrowth itself (Fig. 3). We have tried to quantitate the effect of the Y798F mutation precisely enough so as to be able to document small effects of it on SRC activation or on tyrosine phosphorylation of Sin. These experiments (not shown) suggested that the slightly lower SRC activation by RPTP␣Y798F as compared with wt RPTP␣ (Figs. 2B, and 5A) (24)), rendering the contribution of tyrosine-phosphorylated Tyr-798 in RPTP␣ less rate-limiting. Signaling downstream of SRC, Cas, and Sin is thought to rely on the small GTPase Rap1 (24). Although Crk-mediated Rap1 activation is itself implicated in sustained ERK activation (44), our data argue for a contribution of Crk separate from this role. A worthwhile avenue for further dissection of the mechanism of control of neurite outgrowth by Sin-Crk coupling may involve the contribution of Rac-like small G proteins. The controlling role of Cas-Crk complexes in fibroblast motility has been shown to depend on this pathway (20). DOCK180, recruited by Crk, activates Rac1 (51), which could thus exert localized effects on the actin cytoskeleton. Rac and Cdc42 stimulate filopodia and lamellipodia in the growth cone of neuronal cells, and a balance between Rho-like and other small G protein activities may control neurite outgrowth (52). The viral v-Crk oncogene has been shown to control Rho and to modulate axonal growth in vitro and in vivo (53,54).
SRC activity and Cas phosphorylation are induced in response to cell-cell contact (55) and membrane depolarization (56). We suggest that neuritogenic signaling by the SRC Ͼ Sin-Cas Ͼ Crk cassette may be particularly relevant for stimuli that rely heavily on SFKs for promoting neurite outgrowth, such as cell-cell adhesion molecules (8,9). Indeed, RPTP␣ localizes to areas of cell-cell contact (Fig. 2), and several cell-cell adhesion molecules interact with PTPs in neuronal cells (28,57); particularly, glycosylphosphatidylinositol-linked contactin associates in cis with RPTP␣ and the SFK Fyn (11). We have thus far not observed clear effects of selected cell-cell adhesion or extracellular matrix molecules or growth factors on the total level of Sin phosphorylation (data not shown); very likely this issue will require a detailed one-by-one analysis of phosphorylation sites in Sin. At any rate, a role for SRC Ͼ Sin-Cas Ͼ Crk signaling in specifying neurite outgrowth may help solve two paradoxes associated with the concept of sustained ERK activation as necessary and sufficient for neurite outgrowth (1). First, given the complex multiplicity of neural cell types, "process extension" would not be expected to be a unitary event, making reliance on a single pathway unlikely (3,7). Second, the ability of cells to elaborate multiple processes with different and plastic properties demands control mechanisms that can act locally. A role for SRC, whose localization and activity are tightly regulated, in assembly of localized signaling complexes around scaffolding proteins such as Sin or Cas may be relevant in this respect.