Multiple Interactions between Receptor Protein-tyrosine Phosphatase (RPTP) a and Membrane-distal Protein-tyrosine Phosphatase Domains of Various RPTPs*

Receptor protein-tyrosine phosphatase (RPTP) a belongs to the large family of receptor protein-tyrosine phosphatases containing two tandem phosphatase domains. Most of the catalytic activity is retained in the first, membrane-proximal domain (RPTPa-D1), and little is known about the function of the second, membrane-distal domain (RPTPa-D2). We investigated whether proteins bound to RPTPa using the two-hybrid system and found that the second domain of RPTPs interacted with the juxtamembrane domain of RPTPa. We confirmed this interaction by co-immunoprecipitation experiments. Furthermore, RPTPa not only interacted with RPTPs-D2 but also with RPTPa-D2, LAR-D2, RPTPd-D2, and RPTPm-D2, members of various RPTP subfamilies, although with different affinities. In the yeast two-hybrid system and in glutathione S-transferase pull-down assays, we show that the RPTP-D2s interacted directly with the wedge structure of RPTPa-D1 that has been demonstrated to be involved in inactivation of the RPTPa-D1/RPTPa-D1 homodimer. The interaction was specific because the equivalent wedge structure in LAR was unable to interact with RPTPa-D2 or LAR-D2. In vivo, we show that other interaction sites exist as well, including the C terminus of RPTPa-D2. The observation that RPTPa, but not LAR, bound to multiple RPTP-D2s with varying affinities suggests a specific mechanism of cross-talk between RPTPs that may regulate their biological function.

cellular domains, which can be very large, containing domains such as fibronectin type III-like domains and Immunoglobulinlike repeats in the case of for instance LAR, or very short (e.g. RPTP⑀) and heavily glycosylated such as RPTP␣ (2, 3).
RPTP␣ is a widely but dynamically expressed RPTP that dephosphorylates Tyr 527 of c-Src in vitro and that increases c-Src kinase activity when overexpressed in vivo (4,5). Furthermore, cells derived from RPTP␣ knock-out mice have reduced Src and Fyn activity, indicating that RPTP␣ activates Fyn as well as Src in vivo (6,7). RPTP␣ is also involved in m1 muscarinic acetylcholine receptor-dependent regulation of Kv1.2 channel activity. Presumably, RPTP␣ directly dephosphorylates Kv1.2, leading to desuppression of its K ϩ channel activity (8).
RPTP␣ is itself phosphorylated on Tyr 789 , a GRB2-SH2 consensus binding site. Both the SH2 and the C-terminal SH3 domain of GRB2 cooperate for tight binding to RPTP␣. Although it is clear that RPTP␣ interacts with GRB2 in vivo, the function of the binding remains elusive because it is exclusive of Sos, another partner for the SH3 domain of GRB2 (9 -12).
Although it is now established that receptor protein-tyrosine kinases are activated by ligand induced dimerization, little is known about the regulation of RPTPs. One of the few examples of modulation of RPTP activity is activation of RPTP␣ by serine phosphorylation in response to phorbol ester-mediated activation of protein kinase C (13,14). Although ligands have been found for some RPTPs that induced cellular changes, neither changes in activity of the RPTPs nor changes in phosphorylation state of a direct substrate were reported (15)(16)(17). This might suggest that the extracellular domain has another function in addition to regulating RPTP activity. Increasing structural (18) and functional (19 -21) evidence suggests that dimerization negatively regulates RPTP activity, at least for RPTP␣ and CD45. In the RPTP␣ dimer, a helix-loop-helix so-called "wedge" structure interacted directly with and occluded the catalytic site of the opposing monomer. The inhibition is reciprocal in that the wedge of the second monomer interacts with the catalytic site of the first monomer (18). Mutations that destabilized the wedge structure abolished the inactivation (20,21). Furthermore, we have used fluorescence resonance energy transfer between RPTP␣ fusion proteins, fused to two derivatives of green fluorescent protein mutants, cyan and yellow fluorescent protein to show that RPTP␣ formed homodimers in living cells. 2 All these results indicate that RPTPs, as opposed to the receptor protein-tyrosine kinases, would be inactivated by dimerization in which a key component of the inactivation is the interaction between the wedge of one monomer and the catalytic site of the other monomer in RPTP-D1 dimers. How-* 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.
A striking feature of the RPTPs is the high conservation of a second PTP domain, distal to the membrane (D2), that contains no or very low catalytic activity. The second domains of LAR (23) and RPTP␣ 3 have the conserved PTP structure. For both RPTPs, the lack of activity in D2 resides in two amino acids that are highly conserved in active PTPs but not in RPTP-D2s: the Tyr in the KNRY motif and the Asp in the WPD loop. PTP activity was restored in D2 by mutating these key residues into their wild type D1 counterpart (23)(24)(25). The conservation of an inactive PTP domain (D2) in the RPTP family raised questions about the biological function of D2.
We used the yeast two-hybrid system to screen for proteins that interacted with RPTP␣, in search of possible effector proteins of RPTP␣. We found that the D2 of RPTP interacted with RPTP␣. The observation was extended to other D2s, including RPTP␣-D2, LAR-D2, RPTP␦-D2, and RPTP-D2 in vitro and in vivo. Using the two-hybrid system as well as GST pull-downs, we show for the first time that the wedge in RPTP␣-D1, but not the equivalent wedge of LAR-D1, is involved in the interaction with D2s, indicating that these interactions are specific. In vivo binding of RPTP␣ to RPTP-D2s was further confirmed by co-immunoprecipitation. Furthermore, the affinity of binding between RPTP␣ and the different D2s appeared to be in the following order: RPTP␦-D2 Ͼ RPTP-D2/ LAR-D2 Ͼ RPTP␣-D2/RPTP-D2. Our results suggest a role for RPTP-D2s in the regulation of RPTP␣ because the wedge is involved in the interaction between RPTP␣-D1 and RPTP␣-D1 as well as between RPTP␣-D1 and RPTP-D2s. Furthermore, our results suggest cross-talk between different nonrelated RPTPs that might regulate their biological function.
Two-hybrid Screen and Assays-For the library screening, the reporter strain YGH-1 transformed with pGBT8-RPTP␣-JXT was retransformed as described (28) with a reamplified mouse brain cDNA library cloned in pGAD and plated on synthetic medium lacking leucine, tryptophane, and histidine (SC ϪLTH ). The number of transformants was approximately 1 ϫ 10 7 . After 4 days 150 colonies were recovered after selection and replated on new SC ϪLTH plates for analysis of ␤-galactosidase activity, using a filter lift assay. The plasmid DNA from the His and 32 ␤-galactosidase positives colonies was isolated from liquid cultures, transformed into bacteria, and retested in yeast for specificity. For interaction assays, 3 g of relevant plasmids and 40 g of carrier DNA were introduced in yeast using the lithiumacetate protocol and plated on medium lacking leucine and tryptophane (SC ϪLT ) or SC ϪLTH . After 3-4 days, independent colonies from both kind of plate (SC ϪLT and SC ϪLTH ), were replated and tested for ␤-galactosidase activity using a filter lift assay.
Cell Cultures and Transfections-293 and COS cells were routinely grown in DF medium (a 1:1 mixture of Dulbecco's minimum essential medium and Ham's F12 medium) supplemented with 7.5% fetal calf serum. Cells were transfected using the standard calcium-phosphate method (10). Briefly, 10-cm dishes were transfected with a total of 20 g of DNA. The next day, the medium was refreshed and left another 16 h before harvesting.
Immunoprecipitation and Western Blotting-Subconfluent transfected cells were washed twice with ice-cold phosphate-buffered saline, and lysed with cell lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 10% glycerol, 1% Triton X-100, and protease inhibitors, including benzamidine, aprotinin, and leupeptin) for 20 min on ice, harvested, and centrifuged at 14,000 ϫ g for 15 min to remove the insoluble fraction. The supernatant was added to 12CA5 antibodies that had been covalently linked to protein A-Sepharose beads, using dimethylpimelimidate. After 2 h of incubation, the beads were washed four times with HNTG buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, 10% glycerol), mixed with 2ϫ Laemmli buffer and loaded on a 10% SDS-PAGE gel. The proteins were transferred to PVDF membrane using a semidry transfer system. After Coomassie staining, the membrane was blocked for 1 h with 5% milk in TBS-Tween (50 mM Tris, pH 8.0, 150 mM NaCl, 0.05% Tween-20) at room temperature, incubated with the first antibody for 1.5 h, washed four times with TBS-Tween, incubated with horseradish peroxidase-conjugated anti mouse antibody (Transduction Laboratories) for 1 h, washed with TBS-Tween, and developed using enhanced chemiluminescence.
Purification of GST Fusion Proteins-All the constructs in pGEX-KG were transformed into BL21 bacteria. GST fusion proteins were purified essentially as described (29). Briefly, fresh colonies were allowed to grow until they reached an optical density of 0.6 -0.8 and were then diluted 10 times. After 1 h, isopropyl-1-thio-␤-D-galactopyranoside was added up to 100 M to induce synthesis of the fusion proteins overnight at 28 -30°C. The cells were lysed by addition of lysozyme and further by sonication in TBS (50 mM Tris, pH 8.0, 150 mM NaCl) containing Triton X-100 (1%) and protease inhibitors at 4°C. After centrifugation at 14,000 ϫ g, the GST fusion proteins were collected using Sepharose-GSH beads and eluted with 10 mM reduced glutatione in TBS. The supernatant containing pure GST fusion protein was further dialyzed against a solution of TBS containing 10% glycerol.

RPTP␣ Interacts with Different RPTP-D2s in the Yeast Two-
hybrid System-To find proteins interacting with RPTP␣ we used the yeast two-hybrid system. Different fragments of RPTP␣ containing the full intracellular domain or the individual PTP domains fused to the GAL4 DNA-binding domain were used to screen a mouse brain cDNA library as well as an 11.5-day-old mouse embryo cDNA library without success. The presence of RPTP␣-D1 appeared to be detrimental in the twohybrid system. Possibly, the PTP activity of RPTP␣ was toxic to the yeast or RPTP␣-D1 homodimerization inhibited further binding of proteins. The juxtamembrane domain of RPTP␣ (RPTP␣-JXT, aa 175-240, encompassing the region just Cterminal to the transmembrane domain and a small part of RPTP␣-D1) was used to screen a mouse brain cDNA library. Three clones out of 10 7 transformants were found to bind strongly and specifically to RPTP␣ and were sequenced. All three clones were identical and encoded the second domain of RPTP-D2. To investigate whether other RPTP-D2s might interact with the juxtamembrane domain of RPTP␣ as well, we generated a construct of the GAL4 activation domain fused to RPTP␣-D2 and found that RPTP␣-D2, like RPTP-D2, interacted with the juxtamembrane domain of RPTP␣ (Table I). To validate the latter interaction, the GAL4 DNA-binding domain and GAL4 activation domain were switched, and again, a strong interaction was detected suggesting a specific interaction (Table I). To further investigate the interaction of RPTP␣ with other RPTP-D2s, we introduced LAR-D2, RPTP␦-D2, and RPTP-D2 in the two-hybrid vectors and tested for specific interaction with RPTP␣-JXT. As depicted in Table I, all D2s that we tested interacted with the juxtamembrane domain of RPTP␣, with RPTP␦-D2 apparently binding better. The inter-actions were not nonspecific because the RPTP-D2s did not interact with the empty vectors nor with each other. In conclusion, we show that the juxtamembrane domain of RPTP␣ interacted specifically with many RPTP-D2s in the two-hybrid system.
RPTP␣-JXT/RPTP-D2 Interactions Are Specific and Dependent on the Wedge-The juxtamembrane domain of RPTP␣ that was used in the two-hybrid system contains the helix-loophelix, so-called wedge, structure that interacted with the other monomer in the RPTP␣-D1 homodimers of the crystal structure (18). Based on the crystal structure, we speculated that the wedge would also be important for the interaction between RPTP␣-D1 and RPTP-D2s. We assessed which residues in the juxtamembrane domain of RPTP␣ were responsible for the interaction with D2s and substituted residues in the wedge that form contacts in the crystal structure with the other monomer with Ala. Single amino acid substitutions in the wedge, K230A or E234A, were sufficient to abolish binding in the yeast two-hybrid system, demonstrating that the wedge is important for the interaction (Table II). Moreover, the RPTP-D2s bound specifically to RPTP␣-JXT, because none of the RPTP-D2s tested bound to the equivalent juxtamembrane domain of LAR, indicating an intrinsic difference between these two RPTPs (Table III). Taken together, these results demonstrate that the juxtamembrane domain of RPTP␣, but not of LAR, specifically bound to RPTP-D2s and that the interactions were dependent on the wedge.

RPTP␣-JXT
In Vivo Binding between RPTP␣-D1 and RPTP␣-D2 Is Partially Dependent on the Wedge-To investigate the role of the wedge in the D1-D2 interaction in vivo, we co-expressed Myctagged RPTP␣-D2 with HA-RPTP␣ wedge mutants. Relatively similar amounts of MtRPTP␣-D2 coimmunoprecipitated with HA-RPTP␣ or with the single mutants E227A, K230A, E231A, or E234A (data not shown). Similarly, single mutations in the wedge of HA-RPTP␣-⌬D2 had no effect on MtRPTP-␣D2 binding (data not shown). However, single mutations might be too subtle to induce changes in in vivo binding. Therefore, we deleted the wedge completely from residues 207 to 238. This deletion was based on the crystal structure to minimize any possible conformational perturbation of the PTP domain. In the HA-RPTP␣-⌬D2 context, complete deletion of the wedge (HA-RPTP␣-⌬D2⌬207/238) led to a reproducible reduction in binding but not a complete loss (Fig. 5, second and third lanes), indicating that the in vivo binding was partially dependent on the wedge structure. Surprisingly, in the full-length context, the same deletion led to a reproducible increase in binding (Fig.  5, fourth and fifth lanes), suggesting that deletion of the wedge opened up a binding site for RPTP␣-D2 in the full-length background. Interestingly, mutation of Glu 228 to Ala or Arg seemed to affect the binding of RPTP␣ to MtRPTP␣-D2 in a similar way, albeit the effects were not as pronounced (data not shown). Similar effects were found with other RPTP-D2s, including RPTP-LARD2, RPTP-D2, and RPTP-D2 (data not shown). These results indicate that the interaction between RPTP␣-D1 and RPTP-D2s is partially dependent on the wedge.
Region of RPTP␣-D2 Involved in Binding-To get better insight into the region in RPTP␣-D2 that was important for binding to RPTP␣, we made single point mutations and deletions in D2 and investigated the effect on co-immunoprecipitation. Mutations in the catalytic site of D2 (V555Y, C723S, and E690D) (24) had no effect on the binding (data not shown). Because RPTP␣ has autodephosphorylating activity on tyrosine 789 (Tyr 789 ), a Src phosphorylation site (9), we investigated whether the RPTP␣-D1/D2 interaction was merely an enzyme-substrate interaction. Mutation of Tyr 789 to Phe in the Myc-tagged RPTP␣-D2 did not change the binding efficiency nor did the mutation of C433S in HA-RPTP␣ or another substrate trapping mutant, RPTP␣D401A (data not shown), indicating that binding to RPTP␣-D2 was not an enzyme-substrate interaction. Interestingly, deletion of the C-terminal part of RPTP␣-D2 (aa 774 -792) reduced the binding to RPTP␣ and to RPTP␣-⌬D2, indicating that the C terminus is involved in the interaction between RPTP␣D1 and RPTP␣D2 (Fig. 6). Furthermore, when HA-RPTP␣-⌬D2⌬207/238 was coexpressed with RPTP␣-D2⌬Ct, most, if not all of the interaction was abolished, indicating independent involvement of the wedge in D1 on the one hand and the C terminus of D2 on the other (Fig.  6). These results suggest that there are at least two independent sites of binding between RPTP␣ and RPTP␣-D2. One involves the wedge in RPTP␣-D1 and the other involves the C-terminal part of RPTP␣-D2. DISCUSSION The conservation of an inactive D2 in most RPTPs remains elusive. Here we show that RPTP␣ interacts directly with multiple RPTP-D2s in vitro and in vivo with varying affinities. The interactions were reproduced in three different assays: (i) in the two-hybrid system, (ii) in GST pull-downs, and (iii) by co-immunoprecipitations. Using the two first assays we showed that the integrity of the wedge in the juxtamembrane domain of RPTP␣ is important for efficient binding. Furthermore, we show that the binding is specific, because the equivalent wedge of LAR did not interact with any of the RPTP-D2s tested. These results are consistent with those of Wallace et al. (30), who showed that the juxtamembrane domain of LAR did not bind to RPTP␦-D2. Specificity was further demonstrated by variable affinities between RPTP␣ and different D2s. In vivo, multiple interaction sites were mapped between RPTP␣-D1 and RPTP␣-D2, in that not only the wedge in RPTP␣-D1 was involved but also the C-terminal tail of RPTP␣-D2.
RPTPs belong to the large PTP family, and little is known about regulation of RPTPs. Cytosolic PTPs contain specific localization sequences or protein modules that have been shown to be involved in their regulation (31). For instance, SHP-1 and SHP-2 contain a unique C-terminal PTP domain and two SH2 domains able to bind to phosphotyrosine. The crystal structure demonstrated how SHP-2 is inactivated by binding of its SH2 domain to the catalytic site of the PTP. Phosphotyrosine containing proteins that bind to the SH2 domain release the interaction leading to the opening of the structure and activation of the PTP (32), thereby confirming previous reports that Tyr(P)-containing peptides activate SHP-2 (33).
In addition to an active, membrane proximal PTP domain (D1), most RPTPs contain a second PTP domain (D2). D2 generally has no or very low activity and, at least for LAR and RPTP␣, has a highly conserved three-dimensional structure in which only two residues that are absolutely conserved in active RPTP-D1s are responsible for the low D2 activity (23)(24)(25). The structural conservation of an inactive D2 in RPTPs raised the question of their biological function. Here we demonstrate that not only RPTP␣-D2, but also LAR-D2, RPTP-D2, RPTP␦-D2, and RPTP-D2 interacted with RPTP␣, suggesting that RPTP-D2 binding to RPTPs may play an important role in RPTP function. We were unsuccessful in detecting interactions between different full-length RPTPs in coimmunoprecipitation experiments, which was probably due to low expression levels and poor detection of some of these RPTPs.
We show here that there are multiple interaction sites between RPTP␣ and RPTP-D2s. For instance, the wedge to the N-terminal side of RPTP␣-D1 was involved but not sufficient for the interaction. In addition, the C terminus was involved in the interaction. Deletion of the C-terminal tail of RPTP␣-D2 reduced the binding to RPTP␣ and, coupled to the deletion of the wedge, abolished binding completely, indicating there are two distinct sites of interaction. The presence of one of the sites apparently is sufficient to detect binding in vivo. It remains to be determined to which regions the wedge and the C terminus of D2 bind in RPTP␣.
The role of the wedge in RPTP-D2 binding is complex (Fig. 5). Deletion of the wedge in HA-RPTP␣-⌬D2 reduced RPTP␣-D2 binding, suggesting the wedge directly binds RPTP␣-D2, consistent with the results from our two-hybrid experiments. However, deletion of the wedge in full-length HA-RPTP␣ led to an increase in RPTP␣-D2 binding, which suggests that a binding site for RPTP␣-D2 becomes (more) available in full-length RPTP␣ upon deletion of the wedge. Apparently, this binding site does not become (more) available in RPTP␣-⌬D2. It is noteworthy that the wedge is also involved in RPTP␣-D1/ RPTP␣-D1 homodimerization. Deletion of the wedge will destabilize the D1/D1 interaction, which may perhaps render the second binding site more accessible to D2s. Because the wedge is involved in homodimerization of RPTP␣-D1, as well as in binding to RPTP␣-D2, we speculate that D2 binding to RPTP␣ affects RPTP␣-D1 homodimerization.
Recently, the crystal structure of the tandem phosphatase domain of LAR was solved (23). The arrangement of the domains in the LAR structure indicates clearly that intramolecular interaction between the juxtamembrane domain of D1 and D2 is highly improbable, suggesting that the interaction between RPTP␣-D1 and RPTP␣-D2 is a reflection of an interaction in trans. Nevertheless, the LAR crystal structure showed an interface between LAR-D1 and LAR-D2, which mainly involves the interdomain loop, stabilized with multiple hydrogen bonds between the two PTP domains (23). Whether this interaction is strong enough to be detected in co-immunoprecipitation is not known. The residues in the interdomain loop that are involved in the LAR-D1/LAR-D2 interaction are conserved in RPTPs, including RPTP␣, and they were present in all the HA-tagged constructs that we used. It is possible that, in addition to the two regions that we identified, this LAR-like interdomain interaction between RPTP␣-D1 and RPTP␣-D2 occurred as well. In fact, this third interface might explain the residual binding that we detected in some cases (Fig. 6). However, the presence of endogenous RPTP␣ (through dimerization) or a third protein might explain residual binding as well.
Our results suggest that there is competition between RPTP␣-D1 and RPTP-D2s, because both RPTP␣-D1 and RPTP-D2s interacted with the wedge. Such competition between D1 and D2 was already suspected for CD45 in vitro because homodimerization of CD45-D1 was not detected in a protein containing D1 and D2 (34). Furthermore, in vivo, it was shown that RPTP␦-D2 interacts with RPTP-D1 (30). Involvement of the wedge in the RPTP␦-D2/RPTP-D1 interaction was suggested on the basis of a large deletion of the juxtamembrane domain that abolished the interaction. We demonstrate, using mutagenesis of single residues, that the wedge is directly involved in binding to RPTP-D2s. The functional significance of the interaction between D2 and D1 is controversial. Wallace et al. (30) demonstrated that the interaction between RPTP␦-D2 and RPTP-D1 led to the inactivation of RPTP-D1. However, CD45-D1 showed an increase in catalytic activity in vitro when fused to CD45-D2, which was presumably due to an increase in the monomeric form (34). We were unable to detect any differences in RPTP␣-D1 activity in the presence of RPTP␣-D2 in vitro using p-nitrophenylphosphate as a substrate (data not shown). However, whether RPTP␣-D1/D2 interactions persist during the activity assays is not known. Furthermore, interpretation of these in vitro PTP assays should be done with most care, because subtle differences in the experimental conditions will gravely affect the outcome of the experiments.
Because the interaction between RPTP␣-D1 and RPTP␣-D2 directly involved the wedge that is also involved in RPTP␣-D1/ RPTP␣-D1 dimerization and because deletion of the wedge had striking effects on D1/D2 binding, we propose that similarly to CD45, RPTP␣-D2 might regulate the dimerization state of RPTP␣-D1 and thus the activity of RPTP␣ in vivo. Indeed, RPTP␣ dimerization, which is dependent on the wedge, negatively regulates its activity (20). Furthermore, the ability of distant RPTP-D2s (LAR, RPTP, RPTP␦, and RPTP) to bind to RPTP␣ raises the possibility that RPTP␣ is regulated by other RPTPs or regulates other RPTPs. The differences in binding efficiency that we detected are surprising, given the high homology between RPTP-D2s and suggest a precise interplay between RPTPs. It is noteworthy that the binding efficiencies appear to be low, in that only a small proportion of the RPTP-D2s bound to RPTP␣ in the coimmunoprecipitation experiments (Figs. 2-6). Perhaps the low binding affinities of the RPTP-D2s reflect that binding is dynamic, allowing rapid changes in the equilibrium between bound and not bound because of subtle changes in conditions.
For the interactions we describe here to occur, the different RPTPs have to be expressed in the same cell. It is noteworthy that RPTP␣, LAR, and RPTP have different but overlapping expression patterns during mouse embryogenesis. For instance, RPTP␣ and LAR are both expressed in mouse dorsal root ganglia (35-37). Furthermore, both LAR and RPTP␣ have been shown, albeit in different studies, to localize to the growth cone of growing neurons (36,38,39). From Drosophila work, it is clear that different RPTPs cooperate in a fashion that is still unclear (40). Here, we show that directional interactions between different subfamilies of RPTPs exist with a central role for RPTP␣. We further show that the interactions involve the wedge structure of RPTP␣, suggesting a direct role for these interactions in regulation of catalytic activity and thus in crosstalk between RPTPs.