Essential Tyrosine Residues for Interaction of the Non-receptor Protein-tyrosine Phosphatase PTP1B with N-cadherin*

Expression of a dominant-negative, catalytically inactive form of the nonreceptor protein-tyrosine phosphatase PTP1B in L-cells constitutively expressing N-cad-herin results in loss of N-cadherin-mediated cell-cell adhesion. PTP1B interacts directly with the cytoplasmic domain of N-cadherin, and this association is regulated by phosphorylation of tyrosine residues in PTP1B. The following three tyrosine residues in PTP1B are potential substrates for tyrosine kinases: Tyr-66, Tyr-152, and Tyr-153. To determine the tyrosine residue(s) that are crucial for the cadherin-PTP1B interaction we used site-directed mutagenesis to create catalytically inactive PTP1B constructs bearing additional single, double, or triple mutations in which tyrosine was substituted by phenylalanine. Mutation Y152F eliminates binding to N-cadherin in vitro , whereas mutations Y66F and Y153F do not. Overexpression of the catalytically inactive PTP1B with the Y152F mutation in L-cells constitutively expressing N-cadherin has no effect on N-cadherin-me-diated adhesion, and immunoprecipitation reveals that the mutant Y152F PTP1B does not associate with N-cadherin in situ . Furthermore, among cells overexpress-ing the Y152F mutant endogenous PTP1B associates with N-cadherin and is

Members of the cadherin family of cell-cell adhesion molecules are key players in morphogenetic processes, and regulation of cadherin function, as opposed to transcription and translation, is thought to be responsible for many of the rapid changes that occur during development. Classic cadherins are characterized by a highly conserved intracellular domain that interacts with the actin-containing cytoskeleton, an interaction essential for function. This interaction is mediated by ␣and ␤-catenin (1-4); ␤-catenin associates directly with a 20-amino acid domain near the carboxyl terminus of cadherin (5,6) and with ␣-catenin, which, in turn, interacts with actin, either directly (7) or indirectly, through ␣-actinin (8). ␤-catenin not only performs a bridging role between cadherin and actin, but free ␤-catenin can be translocated to the nucleus where it regulates transcription of cadherin and other gene products (9,10). Thus, the regulation of free ␤-catenin is of critical importance, and, consequently, the interaction of ␤-catenin with cadherin has multiple ramifications on cellular function (11,12).
Regulation of the interaction of ␤-catenin with N-cadherin is mediated by the phosphorylation of tyrosine residues on ␤-catenin (13,14). In embryonic chick neural retina cells, hyperphosphorylation of ␤-catenin is correlated with loss of its association with N-cadherin and loss of cadherin function (13,14). Enhanced phosphorylation of ␤-catenin has also been correlated with loss of E-cadherin function (15)(16)(17)(18)(19). These data suggest that tyrosine kinases and/or phosphatases must play a critical role in maintaining ␤-catenin association with cadherin and/or its ability to mediate the cytoskeletal linkage. We have reported that the nonreceptor protein-tyrosine phosphatase PTP1B binds to the cytoplasmic domain of N-cadherin and regulates its function by dephosphorylating ␤-catenin (13,14). Furthermore, transfection of mouse L-cells constitutively expressing N-cadherin with a catalytically inactive PTP1B (substitution of cysteine 215 for serine) abolishes the ability of these cells to form N-cadherin-mediated adhesions. The mutant PTP1B associates with N-cadherin displacing endogenous PTP1B, resulting in dissociation of the cadherin-actin connection and accumulation of cadherin-free tyrosine-phosphorylated ␤-catenin (14). PTP1B is targeted to many distinct cellular locations based on specific residues or domains in the molecule. The largest single pool is localized to the cytoplasmic face of the endoplasmic reticulum through a carboxyl-terminal domain (20). PTP1B also interacts with the insulin receptor and the EGF receptor and is phosphorylated on tyrosine residues in response to receptor stimulation (21)(22)(23). We have also reported that PTP1B is physically and functionally associated with focal adhesion complexes (24). This association may depend on binding to p130 cas through a proline-rich site (25). Binding of PTP1B to N-cadherin requires that PTP1B itself be phosphorylated on tyrosine residues (13,14). In this study we show that the in vitro and in situ interaction between PTP1B and Ncadherin depends on phosphorylation of tyrosine residue 152.

EXPERIMENTAL PROCEDURES
Antibodies-Monoclonal mouse anti-PTP1B antibody was purchased from Calbiochem. Anti-N-cadherin antibodies were NCD-2, a rat monoclonal specific to chick N-cadherin (grown in our laboratory from a culture provided by M. Takeichi, Kyoto University, Kyoto, Japan), and polyclonal anti-pan-cadherin (Sigma). Monoclonal rabbit anti-phosphotyrosine antibody (PY20) was from Transduction Laboratories (Lexington, KY). Anti-HA antibody was from Babco, Richmond, CA). HRP 1conjugated anti-mouse and anti-rat secondary antibodies were from Organon Teknika Co. (Durham, NC). Goat-HRP anti-rabbit antibody and fluorescein isothiocyanate-conjugated anti-rat IgG were from Jackson Immunoresearch Laboratories, Inc. (West Grove, PA). Antibodies * This work was supported in part by Grant EY12132 from the National Institutes of Health (to J. L. and J. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Site-directed Mutagenesis-All mutant forms of PTP1B were generated using recombinant PCR. For bacterial expression in pGEX-KG (Amersham Pharmacia Biotech), we added a SmaI and an XhoI restriction site at the 5Ј and 3Ј ends, respectively. The oligonucleotide primers were as follows: forward primer, 5Ј-TCCCCCGGGGGACATGGAGAT-CGAGAAGGAGTTCC-3Ј; reverse primer, 5Ј-CCGCTCGAGCGGCCAT-CAATGAAAACATACCCTG-3Ј. The underlined bases indicate the start and stop codon. For expression in eukaryotic cells, the forward primer included a KpnI restriction site and an HA tag at the 5Ј end, and the reverse primer contained an XhoI restriction site at the 3Ј end to facilitate cloning into the pcDNA3.1(ϩ)zeo mammalian expression vector (Invitrogen, Carlsbad, CA). The oligonucleotide primers used were as follows: 5Ј primer with a KpnI restriction site, 5Ј-GGGGTACCGCCAC-CATGGCATACCCATACGATGTTCCAGATTACGCTGAGATCGAGA-AGGAGTTCCA-3Ј; 3Ј primer with an XhoI restriction site, 5Ј-CCGCT-CGAGCGGCCATCAATGAAAACATACCCTG-3Ј. The underlined bases indicate the start and stop codon.
Preparation of GST Fusion Proteins-PTP1B cDNA constructs were subcloned in pGEX-KG as SmaI/XhoI fragments. The resulting plasmids were transformed into Epicurian coli TKB1 cells (Stratagene, La Jolla, CA) that constitutively express a tyrosine kinase. Cultures were induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside and allowed to express GST-PTP1B fusion proteins for 3 h. Induced cultures were harvested by centrifugation at 3,000 ϫ g for 10 min, and the bacterial pellets were stored at Ϫ70°C until ready for use. The frozen bacterial pellets were resuspended in B-PER bacterial protein extraction reagent (Pierce) containing 1% protease inhibitor mixture (Sigma) and 1 mM sodium orthovanadate (Sigma). The suspended cultures were incubated for 15 min at room temperature with gentle shaking. Soluble proteins were separated from insoluble residue by centrifugation at 27,000 ϫ g for 15 min and stored at Ϫ70°C for future use. Expression of GST-PTP1B was confirmed by SDS-PAGE and Western blot.
The cDNA fragment corresponding to the cytoplasmic domain of N-cadherin (cyt-N-cad) was generated by PCR and subcloned as a SmaI/XbaI fragment into pGEX-KG. The oligonucleotide primers used were as follows (the underlined bases are nucleotides corresponding to the 5Ј or 3Ј end of the cytoplasmic sequence of N-cadherin cDNA): forward, 5Ј-TCCCCCGGGGGACTTCGTAGTATGGATGAAGCG-3Ј; reverse, 5Ј-GCTCTAGAGCGTCAGTCACTCAGTCATCACCTCCACC-3Ј. The GST fusion protein of cyt-N-cad was prepared as described above and purified using glutathione-Sepharose 4B according to the manufacturer's instructions (Amersham Pharmacia Biotech). The purified GST fusion protein was confirmed by SDS-PAGE and Western blot.
In Vitro Binding Assay-Purified GST-cyt-N-cad was biotinylated using EZ-Link Sulfo-NHS-LC-Biotin (Pierce), and biotinylation was confirmed by immunoblot using streptavidin-HRP. Biotinylated cyt-Ncad (30 g/well in PBS) was applied to a streptavidin-coated 96-well plate (Roche Molecular Biochemicals). The plate was incubated for 1 h at room temperature and washed three times with PBS, blocked with 2% BSA (Sigma) in PBS for 1 h at room temperature, and washed again with PBS. Aliquots of GST-PTP1B mutants (50 g/well in PBS) were added to the wells, and the plate was incubated for 1 h at room temperature. After several washes in PBS, anti-PTP1B antibody (in 0.5% BSA, PBS) was added to the wells, followed by a 1-h incubation at room temperature and three washes with TBST (50 mM Tris, 150 mM NaCl, 0.2% Tween 20). Polyclonal anti-mouse HRP antibody (in 0.5% BSA, TBST) was then added, and the plate was incubated for 1 h at room temperature and washed three times with TBST. O-Phenylenediamine dihydrochloride (Sigma) was used as substrate, and absorbance was measured at 492 nm.
Immunoprecipitation and Immunoblotting-Cells were washed with ice-cold PBS and incubated for 30 min on ice with lysis buffer (1% Nonidet P-40 and protease inhibitor mixture (Sigma) in PBS). Cells were harvested by scraping, and the cell lysate was centrifuged at 15,000 ϫ g for 10 min. Aliquots containing equivalent amounts of protein were incubated overnight at 4°C with 1 l of rabbit anti-HA tag antibody (1 mg/ml). 10 l of goat anti-rabbit IgG conjugated to magnetic beads were then added to the supernatant, and the mixture was incubated for 1 h at 4°C with mixing. The magnetic beads were collected using a magnetic stand, washed one time with lysis buffer and three times with PBS, dissolved in SDS sample buffer, separated by SDS-PAGE, and transferred to PVDF membranes. The membranes were immunoblotted with anti-PTP1B, anti-HA, and anti-N-cadherin antibodies as described (14).
To analyze the precipitation of endogenous PTP1B with N-cadherin, anti-N-cadherin antibody NCD-2 was covalently linked to protein Gagarose beads (Pierce) and incubated with neutral detergent extracts of cells prepared as described above. Bound protein was eluted, fractionated by SDS-PAGE, transferred to PVDF membranes, and immunoblotted with the appropriate antibodies and developed as described.
Adhesion Assays-96-well plates coated with protein L (Pierce) were incubated with anti-N-cadherin antibody NCD-2 (20 g/ml in PBS; 50 l/well) overnight at 4°C. The wells were washed three times with PBS and blocked with 1% BSA for 1 h at room temperature. Cells in semiconfluent monolayers were washed in serum-free medium and incubated overnight in methionine-free Dulbecco's modified Eagle's medium containing 1 Ci/ml 3 H-methionine (PerkinElmer Life Sciences). The cells were then washed twice in HBSGKCa (20 mM HEPES, 150 mM NaCl, 3 mM KCl, 2 mM glucose, 1 mM CaCl 2 ), released from the plate with a 0.002% trypsin solution prepared in the same buffer, washed, and resuspended in the same buffer containing 0.1% BSA, 10 g/ml DNase, and 0.4 mM AESBF (Calbiochem). Approximately 4 ϫ 10 4 cells were added to each well. The plate was incubated for 45 min at 37°C and washed 4 times with HBSGKCa. The cells remaining on the wells were solubilized in 0.5% SDS, and radioactivity was determined by liquid scintillation.

Tyrosine Residues 66, 152, and 153 in PTP1B Are Targets for
Phosphorylation-The amino acid sequence of chick PTP1B has eleven tyrosine residues; however, only three of those fit the consensus substrate site for most protein-tyrosine kinases (26). To determine the residues essential for interaction between N-cadherin and PTP1B we used the catalytically inactive C215S PTP1B mutant to create point mutations substituting phenylalanine for tyrosine residues 66, 152, and 153. This substitution is the most conservative, maintaining the structure and size of the amino acid, but eliminating the phosphorylation site. A diagram of all the constructs is shown in Fig. 1. The mutated PTP1B cDNAs were subcloned into pGEX-KG and expressed as GST fusion proteins in the bacterial strain TKB, which expresses a tyrosine kinase with broad specificity, able to phosphorylate a variety of proteins. The GST fusion proteins were analyzed for reactivity with anti-PTP1B and anti-phosphotyrosine antibodies ( Fig. 2A). All PTP1B fusion proteins migrate as multiple bands on SDS-PAGE, with apparent molecular masses between ϳ60 and 76 kDa ( Fig. 2A), reflecting the added masses of GST (ϳ26 kDa) and PTP1B (ϳ50 kDa). The multiple bands do not appear to reflect differential PTP1B Binding to N-Cadherin phosphorylation, as immunoblotting with an anti-phosphotyrosine antibody reveals only two major bands. The triple mutant, Y66F/Y152F/Y153F, does not show any reactivity with anti-phosphotyrosine antibody, demonstrating that these tyrosine residues are indeed the only substrate sites for Src-like tyrosine kinases. The wild-type enzyme also shows minimal tyrosine phosphorylation as compared with the C215S mutants because of its phosphotyrosine phosphatase activity.
Tyr-152 Is the Crucial Residue for PTP1B Binding to the Cytoplasmic Domain of N-cadherin in Vitro-To determine the tyrosine residue(s) critical for the interaction of PTP1B with N-cadherin, we analyzed the ability of the various GST-PTP1B mutants to bind to the cytoplasmic domain of N-cadherin in vitro. cyt-N-cad was prepared as a GST fusion protein, purified on glutathione-conjugated Sepharose 4B, and covalently labeled with biotin on lysine residues (Fig. 2B, bottom). The labeled cyt-N-cad was further purified to eliminate free biotin and bound to neutravidin-coated 96-well plates. The amount of bound biotin-cyt-N-cad was determined by enzyme-linked immunosorbent assay using an antibody to the carboxyl terminus of N-cadherin. Wells coated with saturating amounts of Ncadherin or BSA were then incubated with the various GST-PTP1B fusions, as well as with GST only, as a control. After washing and blocking the wells with BSA, the amount of PTP1B bound was determined using anti-PTP1B antibody, which recognizes all the PTP1B mutants equally well (see Fig.  2A), followed by an HRP-conjugated secondary antibody. Optimal binding of PTP1B to immobilized N-cadherin depends on phosphorylated tyrosine residues. Fusion proteins lacking phosphorylated tyrosine residues, the C215S triple mutant (Y66F/Y152F/Y153F), and the wild-type bind minimally, showing only about 25% that of the C215S mutant with no substituted tyrosine residues (Fig. 3A). Among the C215S mutants bearing one Tyr 3 Phe substitution, only the Y152F shows a significant reduction in binding, suggesting that residue 152 is

PTP1B Binding to N-Cadherin
the most critical determinant of PTP1B binding to N-cadherin in vitro. In agreement with this, the C215S double mutants containing a 152 mutation (Y66F/Y152F and Y152F/Y153F) also show reduced binding, whereas the C215S Y66F/Y153F double mutant binds as well as the unsubstituted C215S (Fig.  3). These results are true over a wide concentration range (Fig.  3B); concentrations of C215S PTP1B that show saturation binding still fail to show binding of the Y152F mutant. It is interesting to note that the Y66F mutant actually facilitates binding.
Tyr-152 Is Essential for PTP1B Interaction with N-cadherin-To determine the interaction of the PTP1B mutants with N-cadherin in cells, the several PTP1B cDNA constructs were subcloned into the pcDNA3.1(ϩ)zeo vector and trans-fected into LN-cells (14). A 9-amino acid sequence coding for the hemagglutinin sequence was added to the amino terminus of the PTP1B sequence to facilitate detection of the transfected enzyme. Stable cell clones were established by culturing in the presence of Zeocin and Geneticin (for stable N-cadherin expression). Cells were grown to near confluency, lysed with nonionic detergent in the presence of tyrosine phosphatase inhibitors, and immunoprecipitated with anti-HA antibody. Immunoprecipitated material was fractionated by SDS-PAGE and transferred to PVDF membranes, and the membranes were probed with anti-N-cadherin antibody (NCD-2) and anti-PTP1B antibody (Fig. 4A). In agreement with what we observed in the in vitro binding assays, the Y152F mutation alone is enough to eliminate binding to N-cadherin (Fig. 4A). Furthermore, all combinations of mutant tyrosine residues that include Tyr-152 behave identically (not shown), whereas mutation at tyrosine residues 66 and 153 alone (Fig. 4A) or in combination (not shown) have no effect on binding of PTP1B to N-cadherin.
As in embryonic chick retina cells (13), endogenous PTP1B is associated with N-cadherin in control LN-cells (transfected with vector alone) and is phosphorylated on tyrosine residues (Fig. 4B). Expression of the dominant-negative C215S mutant PTP1B in LN-cells prevents the association of endogenous PTP1B with N-cadherin (see Fig. 4B and Ref. 14). In contrast, expression of PTP1B carrying both the C215S and the Y152F mutations does not alter the association of endogenous PTP1B with N-cadherin. Thus tyrosine 152 is critical for in situ binding and displacement of endogenous PTP1B from cadherin.
The Y152F Mutation Reverses the C215S Dominant-Negative Effect on N-cadherin-mediated Adhesion-The catalytically inactive C215S PTP1B mutant acts as a dominant-negative when introduced into LN-cells, inhibiting N-cadherin-mediated cell interaction (14). By introducing a mutation that eliminates binding to N-cadherin in the C215S PTP1B, the dominantnegative effect should be abolished; this is indeed the case (Fig. 5). N-cadherin-mediated cell adhesion is abolished in the C215S mutants but restored in the C215S mutants that also have a Y152F mutation. In comparison, mutations in tyrosine residues 66 and 153 alone or in combination have no effect (Fig.  5). This effect on N-cadherin-mediated adhesion is reflected in the cells phenotype; LN-cells grow in clusters of tightly adherent cells because of expression of N-cadherin ( Fig. 6A; see also Ref. 14). In the dominant-negative C215S mutant this phenotype is lost because of inactivation of N-cadherin (compare Fig.  6, A and B) but recovered in the C215S mutant bearing the Y152F mutation (Fig. 6C). In contrast, mutation of either tyrosine 66 or 153 has little or no effect on the dominant-negative phenotype.

DISCUSSION
Our laboratory has demonstrated that PTP1B interacts directly with N-cadherin and that phosphorylation of PTP1B on tyrosine residues is necessary for this association (13,14). We now identify tyrosine residue 152 in PTP1B as the critical residue for PTP1B-N-cadherin interaction. PTP1B mutants that have tyrosine 152 replaced by phenylalanine do not interact with N-cadherin in in vitro binding assays. Moreover, in L-cells expressing N-cadherin and HA-tagged PTP1B carrying the Y152F and C215S double mutation, HA-PTP1B does not coimmunoprecipitate with N-cadherin, indicating a lack of association between the two molecules in situ. This is also reflected in the loss of the dominant-negative effect on adhesion of the C215S mutation on N-cadherin function. Furthermore, in LN-cells expressing the Y152F mutation endogenous PTP1B is associated with N-cadherin, and it is tyrosinephosphorylated.
The multiple intracellular roles played by PTP1B require interactions with many different intracellular partners. The needed binding specificity appears to be achieved by compartmentalization or by targeting mediated by specific domains. The carboxyl terminus of PTP1B directs its localization to the cytoplasmic face of the endoplasmic reticulum, thus restricting the number of potential interactors (20). In platelets and activated T-cells, proteolytic cleavage in the ER targeting domain results in translocation of PTP1B to the cytoskeletal/membrane fraction (27)(28)(29). This cleavage is dependent on integrin engagement, resulting in increased Ca 2ϩ levels and, consequently, activation of calpain. We also find that PTP1B associated with N-cadherin in vivo migrates faster on SDS-PAGE than the intact ϳ50-kDa enzyme, suggesting cleavage (13,14). The N-cadherin-associated PTP1B represents a small fraction of the total and colocalizes with N-cadherin in sites of cell-cell contacts and at the tips of growing neurites (14,30). Elimination of the ER localization signal does not alter the interaction of PTP1B with N-cadherin, suggesting that targeting of PTP1B to the N-cadherin complex does not depend on prior targeting to the ER. Furthermore, targeting to specific plasma membrane locations does not appear to depend on cleavage of the ER targeting sequence, as the PTP1B associated with focal adhesion complexes (24) and the insulin receptor (22) have an apparent molecular mass of ϳ50 kDa.
Phosphorylation on tyrosine residues is important for targeting of PTP1B to at least two of its interacting partners. As we demonstrate here, phosphorylation of tyrosine 152 is critical for binding to N-cadherin. Additionally, interaction of PTP1B with the insulin receptor results in phosphorylation of tyrosine residues 66 and 152/153. Phosphorylation of these residues further promotes binding to the receptor. Tyrosine 66 is the major target for phosphorylation of PTP1B by the insulin receptor, creating a site essential for downstream signaling (22). In contrast, tyrosine phosphorylation on PTP1B does not appear to play a role in the binding of PTP1B to p130 cas (25). This interaction, which probably mediates targeting of PTP1B to the integrin complex, is mediated by a proline-rich, SH3-binding domain in PTP1B (25). These differences highlight the fact that even though PTP1B is a ubiquitous enzyme, it plays a pivotal role in regulating many cellular functions through specific protein-protein interactions.