Intramolecular Interactions between the Juxtamembrane Domain and Phosphatase Domains of Receptor Protein-tyrosine Phosphatase RPTP m REGULATION OF CATALYTIC ACTIVITY*

RPTP m is a receptor-like protein-tyrosine phosphatase (RPTP) whose ectodomain mediates homotypic cell-cell interactions. The intracellular part of RPTP m con-tains a relatively long juxtamembrane domain (158 amino acids; aa) and two conserved phosphatase domains (C1 and C2). The membrane-proximal C1 domain is responsible for the catalytic activity of RPTP m , whereas the membrane-distal C2 domain serves an un-known function. The regulation of RPTP activity remains poorly understood, although dimerization has been proposed as a general mechanism of inactivation. Using the yeast two-hybrid system, we find that the C1 domain binds to an N-terminal noncatalytic region in RPTP m , termed JM (aa 803–955), consisting of a large part of the juxtamembrane domain (120 aa) and a small part of the C1 domain (33 aa). When co-expressed in COS cells, the JM polypeptide binds to both the C1 and the C2 domain. Strikingly, the isolated JM polypeptide fails to interact with either full-length RPTP m or with trun-cated versions of RPTP m that contain the JM region, consistent with the JM-C1 and JM-C2 interactions being intramolecular rather than intermolecular. Further-more, we find that large part of the juxtamembrane domain (aa 814–922) is essential for C1 to be catalytically nonradioactive protein-tyrosine phosphatase assay kit (Roche Molecu- lar Biochemicals) according to the manufacturer’s instructions. Signals on Western blots were detected by chemiluminescence (ECL, Amer- sham Pharmacia Biotech).

Protein-tyrosine phosphatases (PTPs) 1 play important roles in signal transduction pathways regulated by tyrosine phosphorylation. Members of the superfamily of PTPs use the same catalytic mechanism and are broadly classified into transmembrane or receptor-like PTPs (RPTPs) and intracellular, nonreceptor PTPs (reviewed in Refs. 1 and 2). Members of the RPTP subfamily are type I membrane proteins consisting of a variable ectodomain, a single membrane-spanning region, and in most cases, two conserved intracellular phosphatase domains. The RPTPs are further classified according to the structure of their ectodomains (reviewed in Refs. 3 and 4). The large variety in ectodomain structure suggests the existence of an equal number of putative ligands, yet in most cases the corresponding ligands have not been identified.
RPTP is the prototype member of a subfamily of RPTPs that mediate homophilic cell-cell interactions via their ectodomains and, hence, are thought to play a role in cell adhesionmediated processes (5)(6)(7)(8). The ectodomain of RPTP shows similarities with that of cell-cell adhesion molecules and consists of an N-terminal "MAM" domain, which is critical for mediating cell-cell adhesion (9), followed by an Ig-like domain and four fibronectin type III repeats (10). Its intracellular part consists of a juxtamembrane domain of 158 amino acids (aa), which is relatively long compared with that in other RPTPs, and two tandem phosphatase domains referred to as C1 and C2. As in most other RPTPs, the membrane-proximal C1 domain of RPTP is catalytically active, whereas the membranedistal C2 domain shows no activity, at least in vitro (11). The C2 domains of most RPTPs have been proposed to play a regulatory role (12), but how it might contribute to RPTP activity is not known.
One major unresolved question is how ligand binding may influence the catalytic activity of RPTPs to affect signal transduction events. A recently proposed model involves dimerization, as inferred from the crystal structure of RPTP␣ (13). This model suggests that ligand binding induces the formation of a symmetrical dimer in which the catalytic site of one molecule is blocked by specific interactions with a helix-turn-helix segment (termed the "wedge") in the juxtamembrane domain of the other (13). There is no wedge-like region present directly upstream of the C2 domain, suggesting a fundamental difference between the C1 and C2 domains. Based on these structural studies, dimerization has been proposed to be a universal mechanism of inactivation of RPTPs (reviewed in Ref. 14). Consistent with this, earlier studies had already indicated that the leukocyte-specific RPTP CD45 can form homodimers (15) and that artificial induction of CD45 dimerization may lead to loss of function (16). Using a epidermal growth factor receptor-CD45 chimera, a part of CD45 homologous to the inhibitory helix-turn-helix wedge in RPTP␣ was recently shown to inhibit CD45 function after ligation by epidermal growth factor (17), in support of the dimerization model. On the other hand, however, the crystal structure of RPTP does not reveal such intermolecular interactions between a wedge region and the C1 domain (18). It seems that the catalytic site of RPTPC1 is unhindered and adopts an open conformation similar to what is observed in the cytosolic PTP, PTP1B (19). It was suggested that the RPTP dimer may be the consequence of crystallization, because dimers were not found in solution. Furthermore, some residues important for the proposed dimerization mechanism are less conserved in RPTP (18,20), suggesting that RPTP may not be regulated by dimerization (reviewed in Refs. 14 and 20).
Here we present evidence for a new type of interdomain interaction involved in the regulation of RPTP activity. In a search for potential binding partners of the C1 domain using the yeast two-hybrid system, we isolated a cDNA clone encoding part of RPTP itself, consisting of a large part of the juxtamembrane domain and a small part of the C1 domain. We show that this "JM" segment can interact with both the C1 and C2 domain and present evidence suggesting that this interaction is intramolecular rather than intermolecular. We further show that the juxtamembrane domain is essential for catalytic activity of the C1 domain. Based on these findings, we propose a model in which the juxtamembrane domain may contribute to the regulation of RPTP activity.

EXPERIMENTAL PROCEDURES
Cells, Transfections, and Antibodies-COS-7 cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with antibiotics and 8% fetal calf serum. Transient transfections of COS-7 cells were performed by the DEAE-dextran method as described in Ref. 21. Antibodies against the HA tag (12CA5) and Myctag (9E10) were obtained from hybridoma supernatants. Biotinylated anti-HA antibody and anti-FLAG tag monoclonal antibody M2 were purchased from Roche Molecular Biochemicals and Eastman Kodak Co., respectively. Monoclonal antibody 3D7 directed against the extracellular domain of RPTP has been described previously (22).
Yeast Two-hybrid Library Screen-For use as a bait in the twohybrid screen, the first catalytic domain of RPTP (RPTPC1) was polymerase chain reaction-amplified using primers 5Ј-TATGTCGA-CAACAGAATGAAGAACAGATACG and 5Ј-CCGGAATTCCTCTTTA-ATCTG. RPTPC1 was fused to the Gal4 DNA binding domain by SalI-EcoRI subcloning into pMD4 (23) containing a trp1 marker for selection. pMD4-RPTPC1 was co-transfected into the lacZ and his3 containing yeast strain Y190, together with a pVP16-based (24) human testis cDNA library (kindly provided by R. Bernards) that carries the leu2 marker. Yeast transformants expressing the reporter genes were selected on medium lacking histidine and supplemented with 25 mM 3-amino-1,2,4-triazole. Positive colonies were identified by ␤-galactosidase filter assays.
Protein Analysis and Phosphatase Assays-Cells were washed once with ice-cold phosphate-buffered saline and lysed on ice in 1 ml (per 10 cm plate) of Nonidet P-40 lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1.5 mM EDTA, 10% glycerol, 1% Nonidet P-40) supplemented with 5 g/ml leupeptin, 2.5 g/ml aprotinin, and PefablocSC (Roche Molecular Biochemicals). After centrifugation, 200 l of supernatant (for immunoblot analysis) or 1 ml of supernatant (for phosphatase assays) was incubated for 2 h with protein A-Sepharose beads (Amersham Pharmacia Biotech) precoupled to specific antibodies. Immunoprecipitates were washed three times with lysis buffer and analyzed by Western blotting or assayed for phosphatase activity. For expression controls, 10 l of total lysate was analyzed by Western blotting. Tyrosine phosphatase activity of immunoprecipitates was measured using a nonradioactive protein-tyrosine phosphatase assay kit (Roche Molecular Biochemicals) according to the manufacturer's instructions. Signals on Western blots were detected by chemiluminescence (ECL, Amersham Pharmacia Biotech).

C1
Interdomain Interaction in the Yeast Two-hybrid System-In an attempt to identify proteins that interact with the C1 domain of RPTP, we used this domain as bait (RPTPC1, aa 923-1190) in a yeast two-hybrid screen of a human testis cDNA library. Two positive colonies were identified that contained identical testis-derived cDNA clones, termed clone 1 and clone 2. Strikingly, both clones encode a membrane-proximal region of RPTP (aa 803-955) consisting of a large part of the juxtamembrane domain (120 aa) and a small part of the C1 domain (33 aa) (Fig. 1A), which we refer to as either cl.2 or the JM region. As shown in Fig. 1B, co-expression of RPTPC1 and cl.2/JM in yeast results in the activation of the lacZ reporter gene. These results strongly indicate that the C1 domain undergoes either intermolecular or intramolecular interaction with the JM region.
The Membrane-proximal JM Region Interacts with Both Catalytic Domains of RPTP in COS Cells: Evidence for Intramolecular Interactions-To confirm the observed C1-JM interdomain interaction in mammalian cells, HA-tagged clone 2 (HAcl.2) encoding the JM polypeptide was transiently expressed together with epitope-tagged C1 (FLAG-RPTPC1) in COS cells ( Fig. 2A). When both proteins were co-expressed, C1 was co-precipitated with the anti-HA monoclonal antibody (not shown), whereas HA-cl.2 was co-precipitated with anti-FLAG monoclonal antibody. Thus, the C1-JM interdomain interaction occurs in both yeast and mammalian cells. Given the sequence similarities between the C1 and C2 domains, we tested whether the C2 domain might also interact with JM. As shown in Fig. 2A, this is indeed the case; when HA-cl.2 was co-expressed with the isolated C2 domain (FLAG-RPTPC2; aa 1191-1452) or with both C1 and C2 domains in tandem (FLAG-RPTPC1C2; aa 923-1452), HA-cl.2 was co-precipitated with anti-FLAG monoclonal antibody. It thus appears that the JM polypeptide does not discriminate between the C1 and C2 domain for binding in COS cells. It is of note that the RPTPC1C2 construct, containing both phosphatase domains, yields a stronger binding signal than either C1 or C2 ( Fig. 2A,  last lane), as one would expect if each catalytic domain binds one HA-cl.2 molecule (JM polypeptide).
We next examined whether the nature of the JM-C1 and JM-C2 interactions is intramolecular or intermolecular. To this end, COS cells were transfected with various epitope-tagged RPTP constructs and then subjected to immunoprecipitation and blotting assays. Strikingly, whereas HA-cl.2 (JM polypeptide) co-precipitates with the individual phosphatase domains as well as the tandem C1C2 domain ( Fig. 2A), HA-cl.2 fails to interact with longer versions of RPTP: either a Myc-tagged polypeptide consisting of a large part of the juxtamembrane domain and the C1 domain (JC1, aa 814 -1190) (not shown) or full-length RPTP (Fig. 2C). Furthermore, we find that HAtagged JC1 does not co-precipitate with Myc-tagged JC1 (Fig.  2B) nor with full-length RPTP (Fig. 2C). The results of the co-immunoprecipitation analysis are summarized in Fig. 2D. From these findings we conclude that the observed JM-C1/C2 interactions do not occur between different RPTP molecules.
Thus, our results can only be explained by the JM-C1/C2 interactions being intramolecular rather than intermolecular.
Mutational Analysis: Effects of Point Mutations on the JM-C1/C2 Interaction-To determine which residues are involved in the interaction between the JM and C1 domains, we transfected RPTP constructs in which the following critical residues were mutated: cysteine 1095 to a serine (C1095S) and glutamate 896 to an arginine (E896R). The conserved cysteine 1095 is essential for catalytic activity of the C1 domain of RPTP. Mutation of cysteine 1095 to a serine (C1095S) was shown to completely abolish phosphatase activity (11). Glutamate 896 is analogous to aspartate 228 of RPTP␣. In the RPTP␣ dimer, this residue is located in the N-terminal wedge that inserts into the catalytic pocket of the C1 domain of the juxta-posed RPTP␣ molecule and thereby may block its activity (13). Glutamate 896 of RPTP is also analogous to glutamate 624 in CD45; mutation of this residue was shown to abolish the inhibitory effect on T-cell receptor signaling caused by CD45 dimerization (17). We find, however, that the mutation C1095S in FLAG-RPTPC1 did not affect association with HA-cl.2 (Fig.   3A). We also find that the mutation E896R in HA-cl.2 does not affect the association with the C1 domain (Fig. 3B). Taken together, catalytic activity and glutamate 896 are not essential for the association between the juxtamembrane and the C1 domain of RPTP.
The Juxtamembrane Domain Is Essential for Catalytic Activity of the C1 Domain-To examine how the distinct domains of RPTP contribute to catalytic activity, we determined tyrosine phosphatase activity in immune complexes using a nonradioactive tyrosine phosphatase assay (see "Experimental Procedures"). We measured the activity of both full-length RPTP and different epitope-tagged constructs of RPTP (Fig.  4, A and B) expressed in COS cells. We found that the isolated C1 and C2 domains as well as C1C2 are inactive (Fig. 4B). In marked contrast, however, N-terminal extension of the C1 domain leads to phosphatase activity as inferred from the JC1 polypeptide being active in the assay. In other words, the juxtamembrane domain is required for activity of the C1 domain. This is consistent with reports on LAR and RPTP␣, which show that the isolated C1 domains require at least part The numbers correspond to the residues of full-length RPTP. B, staining for ␤-galactosidase activity. The yeast colonies tested express RPTPC1, clone 1 with or without RPTPC1, or clone 2 with or without RPTPC1. of the juxtamembrane domain for activity in vitro (12,25). The present data also indicate that the C2 domain is not required for activity of the C1 domain, although we cannot exclude the possibility that the C2 domain may somehow contribute to the activity of C1. Finally, we found that the E896R mutation in the juxtamembrane domain of HA-RPTPJC1 does not affect  5 and 6). The blot was probed with anti-RPTP antibody 3D7 directed against the extracellular domain of RPTP. The upper and middle panels show expression controls. D, overview of co-immunoprecipitation analysis with HA-tagged clone 2 or HA-tagged RPTPJC1 and different parts of RPTP. Experiments were carried out as described above. HA-cl.2 or HA-RPTPJC1 (as shown on the left) was co-expressed with different parts of RPTP (shown on the right). Plus (ϩ) signs indicate that co-precipitation was detected; minus (Ϫ) signs indicate no detectable co-precipitation under the same conditions. IP, immunoprecipitate. phosphatase activity when compared with the wild-type JC1 polypeptide (Fig. 4B), indicating that glutamate 896 is not essential for activity of the C1 domain. DISCUSSION In the present study, we have shown that the juxtamembrane domain of RPTP can bind to both the first and the second phosphatase domain (C1 and C2) and that this interaction is likely to be intramolecular rather than intermolecular. Furthermore, we have presented evidence that the juxtamembrane domain is required for the C1 domain to become fully active.
Through yeast two-hybrid analysis, we found that the RPTPC1 domain binds to an RPTP segment, termed JM, consisting of a large part of the juxtamembrane domain and a small part of the C1 domain. Our COS cell experiments revealed that the JM segment interacts not only with C1 but also with the C2 domain of RPTP. These results would be consistent with RPTP forming dimers, in which the JM region of one molecule interacts with the juxtaposed C1 and/or C2 domains in the partner RPTP molecule, analogous to what has been proposed for the C1 domain of RPTP␣ (13). Inconsistent with a homodimerization model, however, is our finding that the JM segment fails to interact with extended, JM-containing versions of RPTPC1. JM also fails to interact with full-length RPTP. We also did not detect any interdomain interactions between versions of RPTP that comprise both JM and C1. These results are most readily explained by a model in which JM-C1/C2 binding represents an intramolecular interaction within one single RPTP molecule. Mutational analysis indicates that the interaction is independent of RPTP catalytic activity and of glutamate 896 in the helix-turn-helix segment, which is analogous to that in the corresponding motif of CD45, where it has been implicated in dimerization-dependent inhibition of CD45 activity caused by dimerization (17).
In a recent crystallographic study on the RPTPC1 domain (residues 874 -1168), it was concluded that the protein behaves as a monomer in solution and that C1-C1 dimerization is most likely a consequence of crystallization (18). The C1 crystal structure revealed that the catalytic site is unhindered and adopts an open conformation. Caution is needed, however, to extrapolate findings obtained with RPTPC1 to the full-length molecule, particularly because the juxtamembrane domain was excluded from crystallographic analysis and, hence, any JM-C1 interaction would go undetected. The N terminus of the C1 domain used for crystallization starts at the helix-turn-helix segment (at the membrane-distal end) very close to the boundary of the C1 domain. This would imply that a more membraneproximal part of the juxtamembrane domain (immediately Nterminal to the helix-turn-helix structure) is involved in the observed JM-C1/C2 interaction. Further crystallization studies using N-terminally extended versions of the C1 domain are required to clarify this point.
It seems likely that the interdomain interactions found in RPTP also occur in other members of the RPTP family. It is of note that the juxtamembrane domain of RPTP, in common with the other MAM domain containing RPTPs, is about 70 residues longer than that in all other RPTPs (10); the significance of this extension remains unknown. It will be interesting to see whether JM-C1/C2 interactions are a specific feature of the MAM domain-containing subfamily of RPTPs. Our results support the view that dimerization is not involved in the regulation of RPTP activity, in contrast to what has been proposed for the regulation of RPTP␣ and CD45 (14,20). In fact, there is no direct evidence that RPTP␣ dimers are catalytically inactive. Parts of RPTP␣, containing the inhibitory wedge and the C1 domain, are catalytically active and probably act as active monomers in solution (25,26). The RPTP dimerization concept has become even more complex since the C1 domain of RPTP was reported to interact with the C2 domain of RPTP␦ but not the RPTPC2 with RPTP␦C1 (27). This apparent C1-C2 heterodimerization requires the wedge region of RPTP, which was thought to bind the "pseudo-active" site in the juxtaposed RPTP␦C2 domain (27). Although the precise cellular role of the C2 domain remains unknown, the latter result does suggest that the C2 domain is involved in a variety of protein-protein interactions. Very recently, structural studies on the tandem phosphatase domains of RPTP LAR revealed a monomeric con- figuration without any indication of dimer formation either in the crystal structure or in solution (28). The LAR crystal structure further reveals that the N-terminal helix wedge is not involved in any intermolecular interaction and that the catalytic sites of both C1 and C2 are accessible, a configuration that is in direct contrast to the previous model of dimeric-blocked orientation based on the crystal structure of the RPTP␣C1 domain alone (28).
We also have shown that the juxtamembrane region of RPTP is required for the C1 domain to be catalytically active, consistent with an intramolecular JM-C1 interaction regulating catalytic activity. Previous reports have shown that LAR and RPTP␣ similarly need the juxtamembrane domain for full catalytic activity (12,25), suggesting a general regulatory mechanism of the juxtamembrane domain among RPTPs.
Based on our findings, we propose a model explaining how the observed interdomain interactions may contribute to the regulation of catalytic activity (Fig. 5). In this model, RPTP can adopt two different conformations. In one conformation, the juxtamembrane domain interacts with the regulatory, catalytically inactive C2 domain. In this way, the C1 domain lacks interaction with the juxtamembrane domain and thereby remains inactive. When a proper tyrosine-phosphorylated substrate is presented, RPTP adopts a new conformation, in which the JM-C2 domain interaction dissociates to promote the formation of a JM-C1 intramolecular complex thereby stimulating catalytic activity and allowing substrate dephosphorylation. The precise function of the juxtamembrane domain remains to be elucidated. It could be important for proper folding of the C1 domain, but it might also be involved in substrate recognition and/or binding. Similarly, the role of the C2 domain remains poorly understood. Interactions between the C2 domain and other signaling molecules might be involved but still remain to be established. Recent mutational analysis has raised the intriguing possibility that C2 might in fact be an active PTPase domain in the correct cellular context (28). The role of the juxtamembrane domain and the C2 domain are key issues that need to be addressed for better understanding of the regulation of RPTP activity. In conclusion, our findings reveal the occurrence of interdomain interactions in RPTP, and given the lack of any indication for intermolecular interactions, they support the view that the dimerization model might not be applicable for the regulation of RPTP activity in general and that of RPTP in particular.