RPTPα is essential for NCAM-mediated p59^fyn activation and neurite elongation

The neural cell adhesion molecule (NCAM) is involved in several morphogenetic events, such as neuronal migration and differentiation, neurite outgrowth, and axon fasciculation. NCAM-induced morphogenetic effects depend on activation of Src family tyrosine kinases and, in particular, p59^fyn kinase (Schmid et al., 1999). NCAM-dependent neurite outgrowth is impaired in neurons from p59^fyn-deficient mice (Beggs et al., 1994) and is abolished by inhibitors of Src kinase family members (Crossin and Krushel, 2000; Kolkova et al., 2000; Cavallaro et al., 2001)_. The NCAM140 isoform has been observed in a complex with p59^fyn, whereas p59^fyn does not associate significantly with NCAM180 or glycosylphosphatidylinositol-linked NCAM120 (Beggs et al., 1997)_. However in oligodendrocytes, p59^fyn is also associated with NCAM120 in isolated lipid rafts (Kramer et al., 1999), whereas in tumor cells NCAM is also associated with pp60^c-src (Cavallaro et al., 2001), suggesting that additional molecular mechanisms may define NCAM's specificity of interactions with Src kinase family members. Several lines of evidence suggest that NCAM's association with lipid rafts is critical for p59^fyn activation. NCAM not only colocalizes with p59^fyn in lipid rafts (He and Meiri, 2002) but disruption of NCAM140 association with lipid rafts either by mutation of NCAM140 palmitoylation sites or by lipid raft destruction attenuates activation of the p59^fyn kinase pathway, completely blocking neurite outgrowth (Niethammer et al., 2002). However, in spite of compelling evidence that NCAM can activate Src family tyrosine kinases, the mechanism of this activation has remained unclear.

The activity of Src family tyrosine kinases is regulated by phosphorylation (Brown and Cooper, 1996; Thomas and Brugge, 1997; Bhandari et al., 1998; Hubbard, 1999; Petrone and Sap, 2000). The two best-characterized tyrosine phosphorylation sites in Src family tyrosine kinases perform opposing regulatory functions. The site within the enzyme's activation loop (Tyr-420 in p59^fyn) undergoes autophosphorylation, which is crucial for achieving full kinase activity. In contrast, phosphorylation of the COOH-terminal site (Tyr-531 in p59^fyn) inhibits kinase activity through intramolecular interaction between phosphorylated Tyr-531 and the SH2 domain in p59^fyn, which stabilizes a noncatalytic conformation.

A well known activator of Src family tyrosine kinases is the receptor protein tyrosine phosphatase RPTPα (Zheng et al., 1992, 2000; den Hertog et al., 1993; Su et al., 1996; Ponniah et al., 1999). It contains two cytoplasmic catalytic domains, D1 and D2, of which only D1 is significantly active in vitro and in vivo (Wang and Pallen, 1991; den Hertog et al., 1993; Wu et al., 1997; Harder et al., 1998). To activate Src family tyrosine kinase, constitutively phosphorylated pTyr789 at the COOH-terminal of RPTPα binds the SH2 domain of Src family tyrosine kinase that disrupts the intra-molecular association between the SH2 and SH1 domains of the kinase. This initial binding is followed by binding between the inhibitory COOH-terminal phosphorylation site of the Src family tyrosine kinase (pTyr531 in p59^fyn) and the D1 domain of RPTPα resulting in dephosphorylation of the inhibitory COOH-terminal phosphorylation sites in Src family tyrosine kinases (Zheng et al., 2000). These sites are hyperphosphorylated in cells lacking RPTPα, and kinase activity of pp60^c-src and p59^fyn in RPTPα-deficient mice is reduced (Ponniah et al., 1999). Like p59^fyn and NCAM, RPTPα is particularly abundant in the brain (Kaplan et al., 1990; Krueger et al., 1990), accumulates in growth cones (Helmke et al., 1998), and is involved in neural cell migration and neurite outgrowth (Su et al., 1996; Yang et al., 2002; Petrone et al., 2003).

Remarkably, a close homologue of RPTPα, CD45, associates with the membrane-cytoskeleton linker protein spectrin (Lokeshwar and Bourguignon, 1992; Iida et al., 1994), a binding partner of NCAM (Leshchyns'ka et al., 2003). Following this lead, we investigated the possibility that RPTPα is involved in NCAM-induced p59^fyn activation. We show that the intracellular domains of NCAM140 and RPTPα interact directly and that this interaction is enhanced by spectrin-mediated Ca²⁺-dependent cross-linking of NCAM and RPTPα. Levels of p59^fyn associated with NCAM correlate with the ability of NCAM-associated RPTPα to bind to p59^fyn, and the NCAM–p59^fyn complex is disrupted in RPTPα-deficient brains implicating RPTPα as linker molecule between NCAM and p59^fyn. RPTPα redistributes to lipid rafts in response to NCAM activation and RPTPα levels are reduced in lipid rafts from NCAM-deficient mice, suggesting that NCAM recruits RPTPα to lipid rafts to activate p59^fyn. Finally, NCAM-mediated p59^fyn activation is abolished in RPTPα-deficient neurons and NCAM-induced neurite outgrowth is blocked in RPTPα-deficient neurons or neurons transfected with dominant-negative RPTPα mutants, demonstrating that RPTPα is a major phosphatase involved in NCAM-mediated signaling.

Cross-linking of NCAM at the cell surface results in a rapid activation of p59^fyn kinase (Beggs et al., 1997; Niethammer et al., 2002) via an unknown mechanism. To analyze whether or not NCAM deficiency may affect the activation status of p59^fyn, we compared levels of activated p59^fyn characterized by dephosphorylation at Tyr-531 and phosphorylation at Tyr-420 in the brains of wild-type and NCAM-deficient mice. Whereas the level of p59^fyn protein was higher in brain homogenates of NCAM-deficient mice (Fig. 1 A), labeling with antibodies recognizing only p59^fyn dephosphorylated at Tyr-531 or with antibodies recognizing only p59^fyn phosphorylated at Tyr-420 was reduced in brain homogenates of NCAM-deficient mice (Fig. 1 B), indicating that activation of p59^fyn is inhibited in NCAM-deficient brains and suggesting that NCAM is involved in the regulation of p59^fyn function.

The intracellular domain of NCAM does not contain sequences known to induce p59^fyn activation. Thus, NCAM may form a complex with a protein, possibly a protein tyrosine phosphatase, to activate p59^fyn. One possible candidate is the RPTPα that dephosphorylates Tyr-531 of p59^fyn (Bhandari et al., 1998) and is highly enriched in neurons and growth cones (Helmke et al.,1998). Remarkably, in RPTPα-deficient cells, both dephosphorylation of the COOH-terminal tyrosine residue and autophosphorylation of the tyrosine residue within the activation loop of pp60^c-src is reduced (von Wichert et al., 2003), resembling the phenotype of NCAM-deficient mice. Furthermore, a close homologue of RPTPα, CD45, associates with the membrane-cytoskeleton linker protein spectrin (Lokeshwar and Bourguignon, 1992; Iida et al., 1994), a binding partner of NCAM (Leshchyns'ka et al., 2003). To investigate if NCAM interacts with RPTPα, we analyzed the distribution of both proteins in cultured hippocampal neurons. NCAM and RPTPα partially colocalized along neurites, and both proteins accumulated in growth cones where clusters of NCAM partially overlapped with accumulations of RPTPα (Fig. 2 A). To verify whether or not NCAM interacts with RPTPα, we induced clustering of NCAM at the cell surface of live hippocampal neurons by incubation with antibodies against NCAM. Clustering of NCAM enhanced overlap between NCAM and RPTPα localization (mean correlation between NCAM and RPTPα localization being 0.3 ± 0.05 and 0.6 ± 0.03 in neurons treated with nonspecific IgG or NCAM antibodies, respectively; Fig. 2 B), indicating that RPTPα partially redistributed to NCAM clusters and suggesting that NCAM and RPTPα form a complex. Because antibodies against RPTPα were directed against its intracellular domain, RPTPα contained in intracellular organelles could have been recognized as colocalizing with NCAM that associates with intracellular organelles of trans-Golgi network origin (Sytnyk et al., 2002). Thus, the redistribution of RPTPα to NCAM clusters may represent redistribution of intracellular carriers containing RPTPα. To analyze whether or not NCAM associates with RPTPα in the plasma membrane, we transfected neurons with RPTPα containing the HA tag in the extracellular domain and induced clustering of NCAM and HA-RPTPα with antibodies against NCAM and the HA tag. HA-RPTPα partially redistributed to NCAM clusters (Fig. 2 C), indicating that both proteins form a complex at the cell surface.

Finally, we examined the association between NCAM and RPTPα in the brain by coimmunoprecipitation. RPTPα coimmunoprecipitated with NCAM from brain homogenates (Fig. 2 D), confirming that NCAM associates with RPTPα. Interestingly, we found that the level of RPTPα was approximately two times higher in the brain of NCAM-deficient mice when compared with wild-type mice (Fig. 2 D), indicating a functional relationship between NCAM and RPTPα.

To identify the NCAM isoform interacting with RPTPα, we expressed NCAM120, NCAM140, and NCAM180 in CHO cells and analyzed their association with RPTPα by coimmunoprecipitation. CHO cells endogenously express RPTPα that was detected with RPTPα antibodies as a band with a molecular mass identical to RPTPα detected in brain homogenates (unpublished data). Although transfected CHO cells expressed NCAM120, NCAM140, and NCAM180 in similar amounts, RPTPα coimmunoprecipitated only with NCAM140 (Fig. 3 A). However, after prolonged exposure of the film we could also detect RPTPα in NCAM180 immunoprecipitates (unpublished data). RPTPα did not coimmunoprecipitate with NCAM120. We conclude that RPTPα associates predominantly with NCAM140 and to a lesser extent with NCAM180.

Inability of NCAM120, the GPI-linked NCAM isoform without the intracellular domain, to bind RPTPα suggested that the intracellular domain of NCAM is involved in the formation of a complex between NCAM and RPTPα. Furthermore, the extracellular domain of NCAM (NCAM-Fc) did not bind to RPTPα in brain lysates, confirming that the extracellular domain of NCAM does not bind to RPTPα (unpublished data). To verify that the NCAM intracellular domain interacts directly with the intracellular domain of RPTPα, we analyzed binding of the recombinant intracellular domain of RPTPα to the intracellular domain of NCAM180 or NCAM140 in a pull-down assay. For comparison, the intracellular domain of CHL1, another adhesion molecule of the immunoglobulin superfamily, was used. The intracellular domain of RPTPα bound to the intracellular domain of NCAM180 or NCAM140 but not to the intracellular domain of CHL1 (Fig. 3 B). Interaction between the intracellular domains of RPTPα and NCAM140 was severalfold stronger than between the intracellular domains of RPTPα and NCAM180 (Fig. 3 B). To confirm this finding, we examined the direct interaction between the intracellular domains of RPTPα and NCAM180 or NCAM140 by ELISA. Intracellular domain of RPTPα bound to the intracellular domains of NCAM180 or NCAM140 in a concentration-dependent manner, with the intracellular domain of NCAM140 binding with a higher affinity than the intracellular domain of NCAM180 (Fig. 3 C). No binding with the intracellular domain of CHL1 was observed (Fig. 3 C). We conclude that NCAM binds directly to RPTPα via the intracellular domain, with NCAM140 being the most potent RPTPα-binding NCAM isoform.

To identify the part of the intracellular domain of RPTPα responsible for the interaction with NCAM140, we coexpressed, in CHO cells, NCAM140 together with the wild-type form of RPTPα (wtRPTPα), RPTPα lacking the D2 domain (RPTPαΔD2), or catalytically inactive form of RPTPα containing a mutation within the D1 catalytic domain (RPTPαC433S) and analyzed binding of NCAM140 to these RPTPα mutants by coimmunoprecipitation. All transfected RPTPα constructs contained the HA tag to distinguish them from endogenous RPTPα. As seen for endogenous RPTPα, transfected wtRPTPα coimmunoprecipitated with NCAM140 (Fig. 4 A). Similar amounts of RPTPαC433S coimmunoprecipitated with NCAM140, whereas RPTPαΔD2 did not coimmunoprecipitate (Fig. 4 A), indicating that the D2 domain is required for the interaction between RPTPα and NCAM140.

Remarkably, among the major NCAM isoforms, only NCAM140 forms a complex with p59^fyn (Beggs et al., 1997) that we found to correlate with its ability to bind RPTPα (see the previous section). RPTPα directly interacts with p59^fyn (Bhandari et al., 1998). Accordingly, p59^fyn coimmunoprecipitated with wtRPTPα from transfected CHO cells (Fig. 4 A). Approximately the same amount of p59^fyn coimmunoprecipitated with RPTPαΔD2 (Fig. 4 A), indicating that this truncated construct also binds p59^fyn probably via the D1 domain. In accordance with previous reports, p59^fyn showed reduced ability to bind RPTPαC433S, a catalytically inactive mutant of RPTPα (Fig. 4 A; Zheng et al., 2000).

To analyze the role of RPTPα in NCAM140–p59^fyn complex formation, we coimmunoprecipitated p59^fyn with NCAM140 in the presence of RPTPα mutants. In CHO cells cotransfected with NCAM140 and wtRPTPα, p59^fyn coimmunoprecipitated with NCAM140 (Fig. 4 A). The amount of p59^fyn coimmunoprecipitated with NCAM140 was reduced in cells cotransfected with RPTPαC433S (Fig. 4 A), correlating with the reduced ability of this catalytically inactive RPTPα mutant to bind p59^fyn (see previous paragraph; Zheng et al., 2000). When NCAM140 was cotransfected with RPTPαΔD2, p59^fyn no longer coimmunoprecipitated with NCAM140 (Fig. 4 A). Because RPTPαΔD2 binds p59^fyn (Fig. 4 A), it is conceivable that this mutant, which does not bind NCAM140, competes with endogenous RPTPα for binding to p59^fyn and thus inhibits NCAM140–p59^fyn complex formation.

To extend this analysis to neurons, we transfected hippocampal neurons with GFP alone or cotransfected with GFP and RPTPαΔD2 or RPTPαC433S and analyzed the redistribution of p59^fyn to NCAM clusters after cross-linking NCAM with NCAM antibodies (Fig. 4 B). In neurons transfected with RPTPαΔD2 or RPTPαC433S, the level of p59^fyn in NCAM clusters was reduced by ∼30% when compared with GFP only transfected cells, suggesting that RPTPαΔD2 or RPTPαC433S inhibit NCAM–p59^fyn complex formation by competing with endogenous RPTPα. The combined observations indicate that NCAM140–p59^fyn complex formation correlates with the ability of NCAM140-associated RPTPα to bind to p59^fyn, implicating RPTPα as a linker between NCAM140 and p59^fyn.

To substantiate further our finding that RPTPα is a linker protein between NCAM and p59^fyn, we analyzed p59^fyn activation and association of p59^fyn with NCAM in RPTPα-deficient brains. As for NCAM-deficient brains, levels of p59^fyn dephosphorylated at Tyr-531 and levels of p59^fyn phosphorylated at Tyr-420 were reduced in brain homogenates of RPTPα-deficient mice (Fig. 5 A), further suggesting that RPTPα plays a role in NCAM-mediated p59^fyn activation in the brain. To analyze the role of RPTPα in the formation of the complex between NCAM and p59^fyn, we immunoprecipitated NCAM from wild-type and RPTPα-deficient brains and probed immunoprecipitates with antibodies against p59^fyn. Whereas p59^fyn coimmunoprecipitated with NCAM from wild-type brains, p59^fyn did not coimmunoprecipitate with NCAM from RPTPα-deficient brains (Fig. 5 B). Furthermore, when NCAM was clustered at the surface of wild-type and RPTPα-deficient cultured hippocampal neurons, levels of p59^fyn were significantly reduced in NCAM clusters in RPTPα-deficient neurons when compared with wild-type cells (Fig. 5 C), indicating that RPTPα is required for complex formation between NCAM and p59^fyn.

NCAM clustering at the cell surface induces rapid p59^fyn activation (Beggs et al., 1997). To analyze whether or not RPTPα is required for NCAM-induced p59^fyn activation, we treated live hippocampal neurons from wild-type and RPTPα-deficient mice with NCAM antibodies and analyzed levels of p59^fyn dephosphorylated at Tyr-531 along neurites of the stimulated neurons. Clustering of NCAM increased levels of Tyr-531–dephosphorylated p59^fyn along neurites of wild-type neurons by ∼60% (Fig. 5 D). However, NCAM-mediated p59^fyn activation was completely abolished in RPTPα-deficient neurons (Fig. 5 D), demonstrating that RPTPα is required for NCAM-mediated p59^fyn activation.

Coimmunoprecipitation experiments were performed either in the presence of Ca²⁺ or with 2 mM EDTA, a Ca²⁺-sequestering agent. Whereas RPTPα coimmunoprecipitated with NCAM from brain homogenates under both conditions, coimmunoprecipitated complexes were reduced by ∼60% in the presence of EDTA (Fig. 6 A), suggesting that Ca²⁺ promotes formation of the NCAM–RPTPα complex. These results are in accordance with findings of Zeng et al. (1999), who found that NCAM and RPTPα did not coimmunoprecipitate in the presence of EDTA. To analyze if the direct interaction between NCAM and RPTPα is Ca²⁺ dependent, we assayed binding of the intracellular domain of NCAM140 to the intracellular domain of RPTPα by ELISA in the presence or absence of Ca²⁺ (Fig. 6 B), showing that the direct interaction is Ca²⁺ independent and suggesting that additional binding partners of NCAM and/or RPTPα may enhance complex formation in a Ca²⁺-dependent manner. Spectrin, which directly interacts with the intracellular domain of NCAM (Leshchyns'ka et al., 2003) and contains a Ca²⁺ binding domain (De Matteis and Morrow, 2000), is one of the possible candidates. Indeed, RPTPα coimmunoprecipitated with spectrin from brain homogenates (Fig. 6 C). In the presence of 2 mM EDTA, RPTPα coimmunoprecipitating with spectrin was reduced by ∼80% (Fig. 6 C), whereas coimmunoprecipitation of NCAM with spectrin did not depend on Ca²⁺ (Fig. 6 C). We conclude that RPTPα directly interacts with NCAM in a Ca²⁺-independent manner. However, formation of the complex is enhanced by Ca²⁺-dependent cross-linking of NCAM140 and RPTPα via spectrin.

Whereas p59^fyn is mainly associated with lipid rafts (van't Hof and Resh, 1997; Niethammer et al., 2002; Filipp et al., 2003), only 4–8% of all RPTPα molecules were found in lipid rafts of brain (unpublished data). In hippocampal neurons extracted with cold 1% Triton X-100 to isolate lipid rafts (Niethammer et al., 2002; Leshchyns'ka et al., 2003), detergent-insoluble clusters of RPTPα only partially overlapped with the lipid raft marker ganglioside GM1 (Fig. 7 A), further confirming that RPTPα and p59^fyn are segregated at the subcellular level. Because activation of NCAM results in its redistribution to lipid rafts (Leshchyns'ka et al., 2003), it may also promote redistribution of NCAM-associated RPTPα to lipid rafts and thus activate raft-associated p59^fyn. To verify this hypothesis, we studied association of NCAM and RPTPα with lipid rafts in hippocampal neurons activated or not activated with NCAM-Fc or NCAM antibodies. In accordance with previous results (Leshchyns'ka et al., 2003), application of NCAM-Fc or NCAM antibodies increased GM1 levels in detergent-insoluble clusters of NCAM, indicating that NCAM redistributed to lipid rafts (Fig. 7, A–C). Application of NCAM-Fc or NCAM antibodies also increased the level of RPTPα in NCAM clusters, indicating that NCAM activation promoted NCAM–RPTPα complex formation (Fig. 7, A–C). Furthermore, NCAM activation also increased GM1 levels in detergent-insoluble clusters of RPTPα (Fig. 7 D), confirming that NCAM-associated RPTPα also redistributed to lipid rafts and suggesting that NCAM recruits RPTPα to lipid rafts. To further analyze this possibility, we compared levels of RPTPα in lipid rafts in brains of wild-type and NCAM-deficient mice. Indeed, RPTPα was reduced by ∼60% in lipid rafts isolated from NCAM-deficient brains (Fig. 7 E), confirming that NCAM plays a role in RPTPα targeting to lipid rafts. The levels of p59^fyn were increased in NCAM-deficient lipid rafts (100% and 124 ± 7.6% in wild-type and NCAM deficient rafts, respectively) probably reflecting increased levels of p59^fyn in NCAM-deficient brains. Levels of GM1 were not different in lipid rafts from wild-type and NCAM-deficient brains (100% and 103.3 ± 6.8% in wild-type and NCAM-deficient rafts, respectively), showing that lipid rafts were isolated with the same efficacy from wild-type and NCAM-deficient brains (Fig. 7 E).

NCAM activation increases intracellular Ca²⁺ concentrations via a FGFR-dependent mechanism (Walsh and Doherty, 1997; Kamiguchi and Lemmon, 2000; Juliano, 2002). This increase in intracellular Ca²⁺ may account for the enhanced association between NCAM and RPTPα after NCAM activation (Fig. 7, A–D) because of spectrin-mediated cross-linking of NCAM140 and RPTPα (Fig. 6). Interestingly, NCAM activation also induces redistribution of NCAM-associated spectrin to lipid rafts (Leshchyns'ka et al., 2003). To analyze the role of FGFR and Ca²⁺ in the recruitment of RPTPα to an NCAM complex, we estimated levels of RPTPα associated with NCAM following NCAM activation in control neurons and neurons incubated with BAPTA-AM, a membrane-permeable Ca²⁺ chelator (Williams et al., 1992; Cavallaro et al., 2001), or a specific FGFR inhibitor (Niethammer et al., 2002; Leshchyns'ka et al., 2003). Whereas NCAM activation increased levels of RPTPα and GM1 in NCAM clusters (Fig. 7 F), treatment with BAPTA-AM or FGFR inhibitor abolished recruitment of RPTPα to NCAM clusters in response to NCAM activation (Fig. 7 F). In accordance with previous findings (Leshchyns'ka et al., 2003), NCAM redistribution to lipid rafts was not affected by the FGFR inhibitor or BAPTA-AM (Fig. 7 F). BAPTA-AM or FGFR inhibitor did not affect the level of RPTPα associated with NCAM under nonactivated conditions (Fig. 7 F). We conclude that, whereas at resting conditions Ca²⁺ does not play a major role in the interaction between NCAM and RPTPα, NCAM-induced FGFR-dependent elevations of intracellular Ca²⁺ levels strengthen the interactions between NCAM and RPTPα in response to NCAM activation, most likely via spectrin (see the section Formation of the complex between RPTPα and NCAM is enhanced by Ca²⁺).

NCAM-induced neurite outgrowth depends on p59^fyn activation (Kolkova et al., 2000), suggesting that NCAM association with RPTPα may be involved. To analyze the role of protein tyrosine phosphatases in NCAM-induced neurite outgrowth, we incubated cultured hippocampal neurons with 100 μM vanadate, an inhibitor of these phosphatases (Helmke et al., 1998). NCAM-Fc–enhanced neurite outgrowth was abolished by vanadate, indicating that activation of protein tyrosine phosphatases is required for NCAM-mediated neurite outgrowth. Vanadate did not affect neurite outgrowth in nonstimulated neurons, indicating that vanadate does not lead to nonspecific impairments (Fig. 8 A). To directly assess the role of RPTPα in NCAM-induced neurite outgrowth, we transfected hippocampal neurons with the dominant-negative mutants of RPTPα. Both, RPTPαΔD2, which does not bind NCAM but associates with p59^fyn, and catalytically inactive RPTPαC433S, which associates with NCAM but binds p59^fyn with a lower efficiency than endogenous RPTPα, inhibited association of NCAM with p59^fyn by competing with endogenous RPTPα (Fig. 4). In neurons transfected with GFP only, stimulation with NCAM-Fc significantly enhanced neurite length when compared with control nonstimulated neurons (Fig. 8 B). However, neurons transfected with RPTPαΔD2 or RPTPαC433S remained unresponsive to NCAM-Fc stimulation (Fig. 8 B), indicating that RPTPα plays a major role in NCAM-induced neurite outgrowth. To confirm this finding, we analyzed NCAM-mediated neurite outgrowth in hippocampal neurons from RPTPα-deficient mice. Whereas NCAM-Fc enhanced neurite outgrowth in neurons from RPTPα wild-type littermates by ∼100%, NCAM-Fc–induced neurite outgrowth was completely abolished in RPTPα-deficient neurons (Fig. 8 C), further confirming that RPTPα is required for NCAM-mediated neurite outgrowth.

It is by now well established that in response to homophilic or heterophilic binding cell adhesion molecules of the immunoglobulin superfamily, such as NCAM, L1, or CHL1, activate Src family tyrosine kinases, and in particular pp60^c-src or p59^fyn, resulting in morphogenetic events, such as cell migration and neurite outgrowth. However, the mechanisms of Src family tyrosine kinase activation in these paradigms have remained unresolved. Here, we identify a cognate activator of p59^fyn, the receptor protein tyrosine phosphatase RPTPα, as a novel binding partner of NCAM. Activation of p59^fyn is reduced in NCAM-deficient mice and interaction between NCAM and p59^fyn is abolished in RPTPα-deficient brains. Interestingly, we found that the levels of p59^fyn and RPTPα are increased in NCAM-deficient brains, possibly reflecting a compensatory reaction to the decreased activity of these enzymes in the mutant and further indicating a tight functional relationship between NCAM, RPTPα, and p59^fyn. NCAM-induced neurite outgrowth is completely abrogated in RPTPα-deficient neurons or in neurons transfected with dominant-negative RPTPα mutants, indicating that RPTPα links NCAM to p59^fyn both physically and functionally.

Interactions between NCAM and RPTPα and NCAM–RPTPα–p59^fyn complex formation leading to neurite outgrowth are tightly regulated (Fig. 9). First, whereas direct interaction between NCAM and RPTPα is Ca²⁺ independent, NCAM–RPTPα complex formation is enhanced by Ca²⁺-dependent cross-linking via spectrin. Remarkably, NCAM activation results in an increase in intracellular Ca²⁺ concentration via influx through Ca²⁺ channels or release from intracellular stores, and may thus provide a positive feedback loop between NCAM activation and NCAM–RPTPα complex formation involving spectrin. RPTPα binding to spectrin may also elevate RPTPα enzymatic dephosphorylation activity (Lokeshwar and Bourguignon, 1992). Interestingly, NCAM activation also induces activation of PKC (Kolkova et al., 2000; Leshchyns'ka et al., 2003), which is known to phosphorylate RPTPα and stimulate its activity (den Hertog et al., 1995; Tracy et al., 1995; Zheng et al., 2002). Thus, a network of activated intracellular signaling molecules may underlie the induction and maintenance of NCAM-mediated neurite outgrowth. It is interesting in this respect that the NCAM140 isoform predominates in these interactions: it interacts more efficiently with p59^fyn via RPTPα and enhances neurite outgrowth more vigorously than NCAM180 (Niethammer et al., 2002). The structural dispositions of NCAM140 for this preference will remain to be established.

Additional regulation of RPTPα-mediated p59^fyn activation is achieved by segregation of RPTPα and p59^fyn to different subdomains in the plasma membrane (Fig. 9). Approximately 90% of all RPTPα molecules in the brain are located in a lipid raft-free environment and are thereby segregated from lipid raft-associated p59^fyn under nonstimulated conditions. Whereas segregation of receptor protein tyrosine phosphatases from their potential substrates due to targeting to different plasma membrane domains has been suggested as a general mechanism of the regulation of receptor protein tyrosine phosphatase function (Petrone and Sap, 2000), the mechanisms that target receptor protein tyrosine phosphatases to lipid rafts have remained unclear. We show that levels of raft-associated RPTPα in the NCAM-deficient brain are reduced, and NCAM redistribution to lipid rafts in response to NCAM activation also induces redistribution of RPTPα to lipid rafts via its NCAM association. The combined observations indicate that NCAM plays a role in recruiting NCAM-associated RPTPα to lipid rafts via NCAM palmitoylation (Niethammer et al., 2002; Fig. 9) or via NCAM interaction with GPI-anchored components of lipid rafts, such as the GPI-linked GDNF receptor (Paratcha et al., 2003). Investigations on the regulatory mechanisms underlying palmitoylation will be required to understand the subcellular compartment-specific distribution of the NCAM–RPTPα–p59^fyn complex. Furthermore, the localization of adhesion molecules and receptors will have to be elucidated in view of lipid rafts heterogeneities.

Besides NCAM, activation of L1 and CHL1, other cell adhesion molecules of the immunoglobulin superfamily, also results in the activation of Src family tyrosine kinases (Schmid et al., 2000; Buhusi et al., 2003). As for NCAM, the intracellular domains of L1 and CHL1 do not possess structural motif for protein tyrosine phosphatase activity, suggesting that yet unidentified protein tyrosine phosphatases may be involved. Identification of protein tyrosine phosphatases associated with other cell adhesion molecules of the immunoglobulin superfamily and conjunctions with RPTPα-activated integrins (Zeng et al., 2003) will be an important next step in the elucidation of the mechanisms that cell adhesion molecules use differentially to guide cell migration and neurite outgrowth in the developing nervous system.

Rabbit polyclonal antibodies against NCAM (Niethammer et al., 2002) were used in immunoprecipitation, immunoblotting, and immunocytochemical experiments; and rat mAbs H28 against mouse NCAM (Gennarini et al., 1984) were used in immunocytochemical experiments. Both antibodies recognize the extracellular domain of all NCAM isoforms. Hybridoma clone H28 was a gift of C. Goridis (Centre National de la Recherche Scientifique UMR 8542, Paris, France). Rabbit antibodies against RPTPα were a gift of C.J. Pallen (University of British Columbia, Vancouver, Canada) or were generated as described previously (den Hertog et al., 1994). Rabbit polyclonal antibodies against human erythrocyte spectrin, rabbit polyclonal antibodies against the HA tag, nonspecific rabbit immunoglobulins, and cholera toxin B subunit tagged with fluorescein to label GM1 were obtained from Sigma-Aldrich. Mouse mAbs against the HA tag (clone 12CA5) were obtained from Roche Diagnostics. Rabbit polyclonal antibodies and mouse mAbs against p59^fyn protein were purchased from Santa Cruz Biotechnology, Inc. Rabbit polyclonal antibodies against Tyr-527–dephosphorylated or Tyr-416–phosphorylated pp60^c-src kinase that cross-react with Tyr-531–dephosphorylated or Tyr-420–phosphorylated p59^fyn were obtained from Cell Signaling Technology. Secondary antibodies against rabbit, rat, and mouse Ig coupled to HRP, Cy2, Cy3, or Cy5 were obtained from Dianova.

To compare wild-type and NCAM-deficient mice, C57BL/6J mice and NCAM-deficient mice (Cremer et al., 1994) inbred for at least nine generations onto the C57BL/6J background were used. NCAM-deficient mice were a gift of H. Cremer (Developmental Biology Institute of Marseille, Marseille, France).To compare wild-type and RPTPα-deficient mice, RPTPα-positive and -negative littermates obtained from heterozygous breeding were used (see online supplemental material).

Coverslips were embedded in Aqua-Poly/Mount (Polysciences, Inc.). Images were acquired at RT using a confocal laser scanning microscope (model LSM510; Carl Zeiss MicroImaging, Inc.), LSM510 software (version 3; Carl Zeiss MicroImaging, Inc.), and oil Plan-Neofluar 40× objective (NA 1.3; Carl Zeiss MicroImaging, Inc.) at 3× digital zoom. Contrast and brightness of the images were further adjusted in Photo-Paint 9 (Corel Corporation).

Cells washed in PBS were incubated for 1 min in cold microtubule-stabilizing buffer (2 mM MgCl₂, 10 mM EGTA, and 60 mM Pipes, pH 7.0) and extracted 8 min on ice with 1% Triton X-100 in microtubule-stabilizing buffer as described previously (Ledesma et al., 1998). After washing with PBS, cells were fixed with cold 4% formaldehyde in PBS.

Colocalization quantification was performed as described previously (Leshchyns'ka et al., 2003). In brief, an NCAM cluster was defined as an accumulation of NCAM labeling with a mean intensity at least 30% higher than background. NCAM clusters were automatically outlined using the threshold function of the Scion Image software (Scion Corporation). Within the outlined areas the mean intensities of NCAM, RPTPα, p59^fyn, or GM1 labeling associated with NCAM cluster were measured. The same threshold was used for all groups. All experiments were performed two to three times with the same effect. Colocalization profiles were plotted using ImageJ software (National Institutes of Health).

Rat NCAM140 and NCAM180/pcDNA3 were a gift of P. Maness (University of North Carolina, Chapel Hill, NC). Rat NCAM120 (a gift of E. Bock, University of Copenhagen, Copenhagen, Denmark) was subcloned into the pcDNA3 vector (Invitrogen) by two EcoRI sites. The EGFP plasmid was purchased from CLONTECH Laboratories, Inc. cDNAs encoding intracellular domains of NCAM140 and NCAM180 were as described previously (Sytnyk et al., 2002; Leshchyns'ka et al., 2003). The plasmid encoding the intracellular domain of RPTPα was a gift of C.J. Pallen. Wild-type RPTPα, RPTPαC433S, and RPTPαΔD2 (containing RPTPα residues 1–516 [last 6 residues: KIYNKI]) were as described previously (den Hertog and Hunter, 1996; Blanchetot and den Hertog, 2000; Buist et al., 2000).

Details on cultures and transfection of hippocampal neurons and CHO cells, immunofluorescence labeling, ELISA and pull-down assay, coimmunoprecipitation, isolation of lipid enriched microdomains, gel electrophoresis, immunoblotting, and generation of RPTPα-deficient mice are given in online supplemental material.

We thank Achim Dahlmann and Eva Kronberg for genotyping and animal care, Drs. Patricia Maness and Elisabeth Bock for NCAM cDNAs, Dr. Catherine J. Pallen for antibodies against RPTPα and cDNA encoding the RPTPα intracellular domain, Dr. Harold Cremer for NCAM-deficient mice, and Dr. Christo Goridis for hybridoma clone H28.

This work was supported by Zonta Club, Hamburg-Alster (I. Leshchyns'ka), and Deutsche Forschungsgemeinschaft (I. Leshchyns'ka, V. Sytnyk, and M. Schachner).