Intra- and Intermolecular Interactions between Intracellular Domains of Receptor Protein-tyrosine Phosphatases*

The presence of two protein-tyrosine phosphatase (PTP) domains is a striking feature in most transmembrane receptor PTPs (RPTPs). The generally inactive membrane-distal PTP domains (RPTP-D2s) bind and are proposed to regulate the membrane-proximal PTP domains (RPTP-D1s). We set out to characterize the interactions between RPTP-D1s and RPTP-D2s in vivo by co-immunoprecipitation of hemagglutinin-tagged fusion proteins encoding the transmembrane domain and RPTP-D1 and myc-tagged RPTP-D2. Seven RPTPs from four different subfamilies were used: RPTP , RPTP , LAR, RPTP , RPTP , CD45, and RPTP . We found that RPTP-D2s bound to RPTPs with different affinities. The presence of intrinsic RPTP-D2 altered the binding specificity toward other RPTP-D2s positively or negatively, depending on the identity of the RPTPs. Furthermore, the C terminus of RPTP-D2s and the “wedge” in RPTPD1s played a central role in binding specificity. Finally, full-length RPTP and LAR heterodimerized in an oxidative stress-dependent manner. Like RPTP -D2, the LAR-D2 conformation was affected by oxidative stress, suggesting a common regulatory mechanism for RPTP complex formation. Taken together, interactions between RPTP-D1s and RPTP-D2s are a common but specific mechanism that is likely to be regulated. The RPTP-D2s and the wedge structures are crucial determinants of binding specificity, thus regulating crosstalk between RPTPs.

Protein-tyrosine phosphorylation is of major importance for cell migration, proliferation, differentiation, and transformation within higher eukaryotic organisms. A common way to transmit extracellular signals into the cytoplasm is through receptor protein-tyrosine kinase (RPTK) 1 activation that consequently activates cytosolic proteins by protein-tyrosine phosphorylation. Tyrosine phosphorylation levels are negatively regulated by the protein-tyrosine phosphatases (PTPs) (1). The family of PTPs is divided into two large groups, the cytosolic PTPs and the transmembrane, receptor-like PTPs (RPTPs) (2,3). Most of the RPTPs include two PTP domains, of which the membrane proximal domain (RPTP-D1) contains all or most of the PTP activity. An understanding of the conservation of the membrane distal PTP domain (RPTP-D2) has remained elusive for a long time. Recent reports suggest a regulatory instead of a catalytic function. Indeed, most of the RPTP-D2s are inactive or very weakly active (4). However, the structure of these RPTP-D2s is similar to that of RPTP-D1s (5), and mutation of only two residues, which are otherwise highly conserved in active PTPs, restored catalytic activity in several RPTP-D2s (5)(6)(7)(8), suggesting that there is evolutionary pressure to keep RPTP-D2s inactive (8).
Biochemical and structural studies show that RPTP-D2s bind to RPTP-D1s in an intra-and intermolecular fashion. RPTP␦-D2 was found to directly inhibit RPTP-D1 activity through binding to the juxtamembrane region of RPTP-D1 (9). In addition, RPTP␣ binds to various RPTP-D2s, suggesting that cross-talk between RPTPs may be a shared mechanism of regulation (10). Inter-and intramolecular interactions between purified CD45-D1 and CD45-D2 suggest they are direct. CD45-D2 binding to CD45-D1 may disrupt CD45-D1/CD45-D1 homodimerization, perhaps leading to the CD45-D1 activation detected (11). Recently, the juxtamembrane region of RPTP was shown to bind in an intramolecular fashion with both PTP domains of RPTP, thus regulating RPTP-D1 activity (12,13). Finally, the crystal structure of the complete cytoplasmic domain of LAR, containing LAR-D1 and LAR-D2, provided some structural evidence for intramolecular D1/D2 binding (5). LAR-D1 interacts extensively with LAR-D2 through the spacer region. Using intramolecular FRET, the conformation of RPTP␣-D2 was found to change from a "closed" to an "open" conformation in an oxidative stress-dependent manner. The change in conformation has functional consequences, because only the open RPTP␣-D2 binds intermolecularly with RPTP domains (14). However, whether RPTP-D1/RPTP-D2 binding is a general mechanism and, if so, with what specificity, is not known.
Like RPTKs, RPTPs may form heterodimers in vivo. Studies on the Erb family of RPTKs showed that heterodimerization induced specific downstream events different than the ones induced by homodimerization (20,21). To get basic insight into the relationship between RPTPs, we set out to identify binding between various RPTPs, including RPTP␣, RPTP⑀, LAR, RPTP␦, RPTP, RPTP, and CD45, and their RPTP-D2s in vivo. We found that specific RPTP-D1/RPTP-D2 interactions were favored. The presence of intrinsic RPTP-D2 affected the binding specificity. Furthermore, we found that the C-terminal sequence, at least in RPTP␦-D2, and the wedge in RPTP␣ were critical to direct RPTP␦-D2 binding specificity suggesting multiple sites of interaction. Finally, we also show heterodimerization between full-length RPTP␣ and LAR in an oxidative stressdependent manner. Taken together, we show that binding of RPTPs to RPTP-D2s are common but specific and that RPTP-D2s contain all the features necessary to drive the specificity of RPTP dimerization. Furthermore, our results suggest a specific mechanism of cross-talk between RPTPs.
Cell Cultures and Transfections-293 cells were routinely grown in DF medium (a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F12 medium) supplemented with 7.5% fetal calf serum. Cells were transfected using the standard calcium-phosphate precipitation method. Briefly, 10-cm dishes were transfected with a total of 20 g of DNA. The next day, the medium was refreshed and left for another 16 h before harvesting (22).
FRET Analysis-CFP, YFP2.1 (24), and LAR-SpD2 (residues 1609 -1897) were amplified by PCR and cloned in-frame using restriction sites. The resulting construct (CFP-LAR-SpD2-YFP) was checked by sequencing. 293 HEK cells were cultured on glass coverslips and transfected with CFP-LAR-SpD2-YFP. FRET was measured according to a previous study (14). In short, single cells were excited at 430 nm (a suboptimal excitation of CFP at 430 nm was chosen to minimize direct excitation of YFP), and the emission spectrum was recorded.

RPTP-D1s and RPTP-D2s
-We previously showed that RPTP␣ was able to bind to multiple RPTP-D2s, including RPTP␣-D2, LAR-D2, RPTP-D2, and RPTP␦-D2 with various affinities (10). Multiple sites are involved in RPTP interactions (see the introduction). To investigate the possible role of RPTP-D2s in RPTP (hetero)dimerization, we asked how different isolated RPTP-D2s bound to different RPTP-D1s. For this purpose we used RPTP␣, RPTP⑀, LAR, RPTP, RPTP␦, RPTP, and CD45. All RPTP-D2s were myc-tagged and were equivalent, in that they all started at residues equivalent to RPTP␣ residue 537 and did not contain the spacer region (region between D1 and D2) (Fig. 1B). The absence of the spacer region was crucial, because it released the RPTP-D2s from their inhibitory closed conformation and allowed interaction with RPTPs independently of stimulation (14). The differences in size between RPTP-D2s were due to small differences in size of the PTP domains as well as to small differences in the cloning procedures. The transmembrane domain and the first PTP domain (RPTP-D1) of RPTP␣, RPTP⑀, LAR, RPTP, RPTP␦, RPTP, and CD45 were cloned in-frame with the signal sequence of RPTP␣ and a HA tag. The signal sequence ensured membrane targeting but did not interfere with the protein, because it is cleaved off during processing of the fusion protein at the membrane, exposing the HA tag (Fig. 1A). Small extracellular domains were included (for RPTP␣ and RPTP⑀), whereas large extracellular domains from the other RPTPs were not included to avoid possible disturbance. For LAR, RPTP, and RPTP␦ whose ectodomain is cleaved, the sequence from the cleavage site onward was used (25). Similarly, constructs encoding the transmembrane and full intracellular domain (RPTP-D1D2) of RPTP␣, RPTP⑀, and LAR were cloned (Fig. 1B).
Interaction between RPTP-D1s and RPTP-D2s-Routinely, 293 cells were transiently co-transfected with HA-tagged RPTP-D1 constructs together with the panel of myc-tagged RPTP-D2s to detect differences in affinities. The HA tag was used for immunoprecipitation, and co-immunoprecipitated myc-tagged RPTP-D2s were detected. Typical results of a coimmunoprecipitation assay are depicted in Fig. 2. Note that an intrinsic difficulty of such an in vivo approach is to reach equivalent protein expression. Therefore, the expression levels were monitored to allow for corrections. RPTP␣-D1 bound all RPTP-D2s tested, although with different affinities (Fig. 2A). RPTP⑀-D2 and RPTP␦-D2 bound with higher affinity to RPTP␣-D1, relative to other RPTP-D2s. The differences in binding affinity were more extreme for LAR-D1D2 (Fig. 2B).
RPTP␦-D2 bound strongly to LAR-D1D2, as did RPTP⑀-D2, whereas CD45-D2 and LAR-D2 bound only weakly. Furthermore, RPTP-D1 and CD45-D1 bound efficiently to RPTP⑀-D2 and RPTP-D2, although RPTP-D1 was not always equally expressed (Fig. 2C). Binding to other RPTP-D2s was observed as well (e.g. RPTP␣-D2, Fig. 2C), but the relatively low expression of the HA-tagged constructs made it difficult to detect. It is noteworthy that RPTP-D1 bound to RPTP-D2 (and RPTP⑀-D2, Table I) but not to RPTP␦-D2, contrary to results from Wallace et al. (9). Several reasons may explain these apparent differences: we used (i) membrane-localized RPTP-D1s that (ii) did not contain the spacer region (region between RPTP-D1 and RPTP-D2). Because the spacer region is involved in the stabilization and regulation of the intramolecular structure, this region may be of particular importance in the differences observed. In conclusion, we developed a system that can be used routinely for the detection of the in vivo interaction of RPTP-D1s with RPTP-D2s and show that RPTP-D1 binding to RPTP-D2s may be a common (but specific) mechanism within the RPTP family.
Effect of Endogenous D2 on Binding to Other RPTP-D2s-We further investigated the binding of RPTP-D2s to the full intracellular domain of RPTP␣, RPTP⑀, and LAR. For most RPTP-D2s the binding pattern to the full intracellular domain (RPTP-D1D2) was similar to that of binding to RPTP-D1. Significant differences were found for LAR and RPTP⑀ binding to RPTP␦-D2 and RPTP-D2. Although full-length RPTP⑀ (RPTP⑀-D1D2) bound RPTP-D2 relatively well and RPTP␦-D2 more weakly, RPTP⑀-D1 bound more efficiently to RPTP␦-D2 than RPTP-D2 ( Fig. 3A and Table I). The opposite effect was found with LAR. LAR-D1D2 bound preferentially to RPTP␦-D2, whereas LAR-D1 bound preferentially to RPTP-D2 ( Fig. 3B and Table I). As a control, RPTP⑀-D2 always bound strongly to LAR-D1D2 and LAR-D1 (Fig. 3B). The same was true for RPTP⑀ binding to RPTP⑀-D2 (Table I). This suggests that the presence of intrinsic D2 alters binding to some but not all RPTP-D2s. Whether the effect is positive or negative depends on the identity of the RPTPs.
A Matrix of Interactions between RPTP-D1s and RPTP-D2s-We repeated the co-immunoprecipitation experiments several times for each RPTP, leading to a good understanding of the interactions between RPTP-D1s (or RPTP-D1D2) and RPTP-D2s. The results presented in Table I were derived from the comparison of binding of RPTP-D2s to one HA-tagged RPTP at a time. However, because of the high number of variables, these data should be used qualitatively rather than quantitatively. We did not attempt to compare binding of one RPTP-D2 to all HA-tagged constructs because of the variation in HA-tagged protein expression and their pattern after immunoblotting (some proteins are single sharp bands, whereas some are more or less diffuse). Furthermore, each experiment was done separately, making comparison between blots impossible. Therefore, the data in the rows in Table I may be compared directly, whereas the data inside a column cannot be compared directly. However, patterns in rows may be compared with each other.
All RPTP-D2s tested interacted with multiple RPTP-D1s. RPTP⑀-D2 and RPTP-D2 always bound relatively well to all RPTP-D1s tested. RPTP␦-D2 bound either strongly (to RPTP␣, RPTP⑀, or LAR-D1D2) or not at all (to the other RPTPs). RPTP␣-D2 and RPTP-D2 bound with relatively similar but weak affinity to all RPTP-D1s tested. LAR-D2 and CD45-D2 also have a similar binding pattern and bound weakly to some but not to other RPTPs. In conclusion, our results suggest that RPTP-D2s have different patterns of binding to RPTP-D1s with different affinities. Taken together, our results support the idea that RPTP-D1s and RPTP-D2s of different RPTPs interact with each other suggesting specific cross-talk between RPTPs.
Specificity Determinants-We investigated which residues in RPTP-D2s were likely to be involved in the difference in affinity observed. RPTP␦-D2 bound better to RPTP␣ than RPTP␣-D2 but also than RPTP-D2 and LAR-D2 despite their very high sequence homology (Fig. 2 and Ref. 10). Sequence comparison of RPTP-D2s revealed one particular residue that was conserved between mRPTP␦ and rRPTP␦ (9) and that at the same time was highly divergent between other RPTPs: LAR, RPTP, and RPTP␣ (data not shown). Thus, this residue, RPTP␦-Met-1257, was a good candidate to explain the differences in affinity observed. Unfortunately, point mutation of

TABLE I A matrix of interactions between RPTP-D1s and RPTP-D2s
293 cells were transiently cotransfected with HA-tagged RPTP (HA tag) constructs together with a panel of Myc-tagged RPTP-D2s as indicated (see Fig. 1). After anti-HA co-immunoprecipitation, the co-immunoprecipitated Myc-tagged RPTP-D2s were detected by immunoblotting. The interactions in the same row were scored relative to each other from not detectable (Ϫ), weak (ϩ/Ϫ), and strong to very strong (gradually from ϩ to ϩϩϩϩ). The table depicted is the result of multiple independent experiments (n) as indicated. Because expression of the different constructs was variable, only results within experiments should be compared. Therefore, the data in the rows in this table may be compared directly, whereas the data in the columns should not be compared directly. However, patterns in rows may be compared to each other.

HA tag
Myc-tagged RPTP-D2s RPTP␦-Met-1257 to Ser, the corresponding residue in RPTP␣, had no effect (data not shown). Introduction of a frameshift in RPTP␦ at Met-1257 by insertion of a single base led to replacement of the C-terminal 35 residues of RPTP␦ (␣6 in the LAR crystal structure (5)) by non-related sequence of similar length (RPTP␦-D2fs). After co-immunoprecipitation with HARPTP␣, only RPTP␦-D2 bound very strongly to HARPTP␣, whereas binding of the frameshift mutant RPTP␦-D2fs was significantly reduced to levels in the same order of magnitude as RPTP␣-D2 (Fig. 4A). The frameshift mutation not only reduced but also enhanced binding to some RPTP-D1s. Although no interaction between RPTP-D1 and RPTP␦-D2 was detected, RPTP-D1 bound efficiently to RPTP␦-D2fs (Fig. 4B). Similarly, RPTP␦-D2 did not bind to LAR-D1, whereas RPTP␦-D2fs did (Table II). In conclusion, these results show that the C terminus of RPTP␦-D2 affects binding of RPTP␦-D2 to RPTPs positively and negatively, and we conclude that the C terminus of RPTP-D2s is an important determinant of the specificity for RPTP-D1 and RPTP-D2 binding.
To get better insight into the function of RPTP-D2s binding to RPTP-D1, it is important to define the binding site within RPTP-D1. We previously showed that although the wedge structure of RPTP␣-D1 is a binding site for RPTP-D2s in yeast two-hybrid and in glutathione S-transferase pull-down experiments (10), multiple sites of interactions exist in vivo. Indeed, deletion of the wedge of RPTP␣ (⌬204 -235) does not prevent binding to most RPTP-D2s tested and even seems to favor RPTP-⑀D2 and RPTP-D2 binding ( Fig. 4C and Ref. 10). In contrast, the same deletion abolished binding of RPTP␦-D2 to RPTP␣. It is noteworthy that a point mutation in the wedge of RPTP␣ (D228A) was not sufficient to affect RPTP␦-D2 binding. These results suggest that multiple binding sites exist between RPTP-D1 and RPTP-D2s and that some RPTP-D2 have different preferences for specific types of interactions.
Inducible Heterodimerization of Full-length RPTPs-Because there are multiple sites of interaction between RPTPs, it is necessary to study the RPTP domains separately to understand their binding specificity. However, it remained to be established whether heterodimerization occurs in vivo between (full-length) RPTPs. Although RPTP␣ dimers have been detected in living cells (26), we found that oxidative stress stimulation greatly increased co-immunoprecipitation of full-length RPTP␣ (14). Oxidative stress induced a conformational change in RPTP␣ leading to stabilization of RPTP␣ dimers and to complete inactivation of RPTP␣. Reactive oxygen species are produced in response to many stimuli, and reactive oxygen species-mediated enhanced dimerization and inactivation of RPTPs may be an important regulatory mechanism for RPTPs. Therefore, we asked if heterodimerization between RPTPs might also be dependent on extracellular stimuli. LAR and HA-tagged RPTP␣ were transiently co-transfected in 293 cells, and binding was assessed by co-immunoprecipitation. Although no co-immunoprecipitation was detected in unstimulated cells, oxidative stress led to a large increase in RPTP␣/ LAR heterodimers (Fig. 5). Binding was specific, because an increase in the amount of immunoprecipitated HARPTP␣ also increased the amount of co-immunoprecipitated LAR (Fig. 5,  compare lanes 2 and 3). Furthermore, mutation of the second catalytic site cysteine (the residue required for the oxidative stress-induced conformational change) in RPTP␣ largely reduced the formation of a stable interaction between RPTP␣ and LAR (data not shown). These results show that full-length RPTPs can be induced to form heterodimers.
The requirement of oxidative stress may be explained by the "opening" of RPTP␣ allowing intermolecular interactions with LAR. Whether similar changes in the conformation of LAR also occur is not known. We used fluorescence resonance energy transfer (FRET) to study the effect of oxidative stress on the conformation of LAR-SpD2 ("Sp" represents the presence of the spacer region, the region between RPTP-D1 and RPTP-D2, see Fig. 1). Reminiscent of what was found with RPTP␣-SpD2 (14), FRET was observed when a chimeric protein was engineered, CFP-LAR-SpD2-YFP, where CFP and YFP were fused (N-and C-terminally, respectively) to LAR-SpD2 (Fig. 6). This indicates that CFP and YFP are located very close to each other (Ͻ40 -50 Å) in the protein, consistent with the LAR crystal structure (5). The weaker FRET prior to stimulation observed within CFP-LAR-SpD2-YFP (ratio 520/480 nm is around 1.4, see Fig. 6, inset) compared with CFP-RPTP␣-SpD2-YFP (ratio 520/480 nm is ϳ2.5, see Ref. 14) might reflect a less optimal orientation of CFP toward YFP or less flexibility in the structure of RPTP␣-SpD2. Oxidative stress reduced FRET (as detected by a reduction of the YFP emission and a concomitant increase in CFP emission due to the reduction in energy transfer from CFP to YFP) within CFP-LAR-SpD2-YFP (ratio 520/ 480 nm after H 2 O 2 was ϳ1.2, which is similar to the value observed with CFP-RPTP␣-SpD2-YFP after peroxide (14)), in-  The interactions were scored as in Table I, from not detectable (Ϫ), weak (ϩ/Ϫ), and strong to very strong (gradually from ϩ to ϩϩϩϩ).

HA tag
Myc-tagged RPTP-D2s dicating that a conformational change also occurred within LAR-SpD2 (Fig. 6). Taken together, these results suggest that a conformational change in both RPTP␣ and LAR is involved in oxidative stress-induced heterodimerization of full-length RPTP␣ and LAR.

DISCUSSION
Like RPTKs, RPTPs may be regulated by dimerization (27). Even though multiple regions may be involved in RPTP dimerization, an important component for dimerization is RPTP-D2. RPTP␦-D2 binds to RPTP-D1 (9), different RPTP-D2s bind to RPTP␣ (10), and CD45-D2 and RPTP-D2 bind to their respective RPTP-D1s. Furthermore, the fact, that the so-called wedge structure, which interacts with the catalytic site of RPTP-D1 in the RPTP␣ dimer (15), also binds to RPTP-D2s (9, 10), suggests that RPTP-D2s may be involved in the regulation of dimerization (in trans) (10). Our results form the basis of a matrix of in vivo binding between RPTPs and RPTP-D2s. Such a matrix shows that all RPTP-D2s tested bound to membrane-localized RPTP-D1s but with different affinities. Apparently, four different patterns exist for RPTP/RPTP-D2 binding (Table I): 1) RPTP⑀-D2 and RPTP-D2 always bound relatively well to all RPTP-D1s tested; 2) RPTP␦-D2 bound either strongly (to RPTP␣, RPTP⑀, or LAR-D1D2) or not at all (to the other RPTPs); 3) RPTP␣-D2 and RPTP-D2 bound with relatively similar but weak affinity to all RPTP-D1s tested; and 4) LAR-D2 and CD45-D2 also have a similar binding pattern and bound weakly to some but not to other RPTPs. Very closely related RPTP-D2s have significantly different binding properties. RPTP⑀-D2 bound strongly to all RPTP-D1s, whereas RPTP␣-D2, the closest homologue of RPTP⑀-D2, bound with much less affinity. Furthermore, LAR-D2, RPTP-D2, and RPTP␦-D2 have very different binding affinities, while they all share very high sequence homology. LAR-D2 bound weakly but specifically to RPTPs. RPTP-D2 bound strongly to all RPTPs tested, whereas RPTP␦-D2 bound strongly to some RPTPs, but not at all to others. Taken together, our results show that RPTP-D2s have all the characteristics for being at the base of the specificity of RPTP dimerization. Furthermore, our results show that accurate predictions of RPTP-D1/RPTP-D2 interactions based on sequence comparison is still not possible, putting emphasis on our experimental approach to unravel RPTP-D1/RPTP-D2 interactions and function.
The presence or absence of intrinsic RPTP-D2 affected the binding efficiency between the remaining RPTP-D1 and other RPTP-D2s. For instance, LAR-D1D2 preferentially bound RPTP␦-D2 while LAR-D1 preferentially bound RPTP-D2 (Fig.  3B). These results are consistent with the model that intramolecular D1/D2 binding regulates intermolecular D1/D2 interactions. Indeed, it was shown for CD45 and RPTP that the juxtamembrane region and the spacer region, when fused to RPTP-D1, alter the binding specificity of D1 toward RPTP-D2 (12,30). Furthermore, we previously showed that the spacer region and the C-terminal region of RPTP␣-D2 bind intramolecularly and regulate the interaction between RPTP␣-D1 and RPTP␣-D2 (14). Because the binding specificity of RPTP-D1s is different than RPTP-D1D2s, our results suggest that stimuli may induce a reshuffling of RPTP dimers.
We have previously shown that the C terminus of RPTP␣-D2 was involved in binding to RPTP␣ (10). Using RPTP␦-D2, we further show the importance of the C terminus in RPTP-D2s and of the wedge in RPTP-D1s. Replacement of the C-terminal sequence of RPTP␦-D2 changed the binding specificity and affinity. The corresponding region in LAR forms a helix that is localized at the interface between LAR-D1 and LAR-D2 in the crystal structure (5), providing structural support that this region is somehow involved in intramolecular RPTPD1/ RPTPD2 binding. The fact that mutation of this region positively and negatively altered the binding efficiency of different RPTPs may suggest that the C terminus forms (part of) the binding site for RPTP-D1s or plays an indirect role (by regulating RPTP-D2 opening). Other sites are involved in RPTP-D1/RPTP-D2 binding as well, because replacement of the C terminus increased binding in some cases. These results suggest that multiple binding sites between RPTP-D1 and RPTP-D2s exist and may compete and/or cooperate. The juxtamembrane region (Fig. 4C and Refs. 9 and 10), the C-terminal region ( Fig. 4 and Ref. 10), and the spacer region (5,12,14,30) are all involved in RPTP-D1/RPTP-D2 binding. The differences in binding between constructs may be due to differences in binding efficiency to one or more binding sites. A matrix such as the one described here may help to pinpoint specific binding sites or help to define these sites.
Homodimerization of RPTP␣ and CD45 is now well established. However, we found that oxidative stress stimulation was necessary to allow co-immunoprecipitation of full-length RPTP␣ dimers (14). Oxidative stress induced a change in the conformation of RPTP␣-D2 that released RPTP␣ from a closed to an open conformation allowing stabilization of dimers. Here we show that oxidative stress induced formation of heterodimers between full-length RPTP␣ and LAR as well. Using FRET, we demonstrated that oxidative stress induced a conformational change in LAR-D2 as well, albeit to a lesser extent than in RPTP␣-D2. Although this was the first time that fulllength RPTP heterodimers were detected, cross-talk between RPTPs has been proposed before. Importantly, several elaborate studies on Drosophila axon pathfinding clearly indicate functional cooperation and competition between RPTPs (28,29). Although these studies only show genetic interactions, it is now tempting to speculate that cross-talk between Drosophila RPTPs may be mediated by direct, stimulation-dependent heterodimerization.
A burning question that remains is: what is the effect of RPTP-D2 on RPTP-D1 activity? RPTP␦-D2 binding inhibits RPTP-D1 (9), and RPTP-D2 binding decreases RPTP-D1 catalytic activity (12,13). In contrast, the presence of CD45-D2 fused to CD45-D1 led to an increase in total PTP catalytic activity, presumably due to an increase in monomerization (11). Furthermore, in vitro no changes in RPTP␣-D1 activity were detected after addition of RPTP␣-D2 (data not shown). Taken together, these results suggest that the effect of RPTP-D2s on RPTP-D1s may depend on the RPTP, or on the RPTP-D1/RPTP-D2 combination. Indeed, the strength and site(s) of the interaction may be important. Strong binding of RPTP-D2 to the wedge may lead to inactivation of RPTP-D1, whereas weak and dynamic RPTP-D2 binding to the wedge may lead to monomerization and activation of RPTP-D1. For these reasons, all RPTP-D1/RPTP-D2 combinations will need to be studied for changes in activity. However, these activity assays are technically difficult, because the stoichiometry of binding of RPTP-D1 to RPTP-D2 will influence the effect of RPTP-D2 on RPTP-D1 activity in assays in vitro. A much-preferred configuration would require analysis of dephosphorylation of physiological substrates in living cells, which is not feasible yet for all RPTPs.
In conclusion, our results suggest that specific and regulated heterodimerization between RPTPs occurs in vivo. RPTP-D2s have specific affinity for RPTP-D1s and consequently may be at the base of functional and regulatory cross-talk between RPTPs.