Subdomain X of the kinase domain of Lck binds CD45 and facilitates dephosphorylation.

CD45 is a transmembrane, two-domain protein-tyrosine phosphatase expressed exclusively in nucleated hematopoietic cells. The Src family kinase, Lck, is a major CD45 substrate in T cells and CD45 dephosphorylation of Lck is important for both T cell development and activation. However, how the substrate specificity of phosphatases such as CD45 is achieved is not well understood. Analysis of the interaction between the cytoplasmic domain of CD45 and its substrate, Lck, revealed that the active, membrane-proximal phosphatase domain of CD45 (CD45-D1) bound to the phosphorylated Lck kinase domain, the SH2 domain, and the unique N-terminal region of Lck. The second, inactive phosphatase domain (CD45-D2) bound only to the kinase domain of Lck. CD45-D2 was unable to bind phosphotyrosine, and its interaction with the kinase domain of Lck was independent of tyrosine phosphorylation. The binding of CD45-D2 was localized to subdomain X (SD10) of Lck. CD45-D2 bound similarly to Src family kinases but bound Csk to a lesser extent and did not bind significantly to the less related kinase, Erk1. CD45 dephosphorylated Lck and Src at similar rates but dephosphorylated Csk and Erk1 at lower rates. Replacement of Erk1 SD10 with that of Lck resulted in the binding of CD45-D2 and the conversion of Erk1 to a more efficient CD45 substrate. This demonstrates a role for CD45-D2 in binding substrate and identifies the SD10 region in Lck as a novel site involved in substrate recognition.

pressed predominantly in T lymphocytes, are both required for effective signal transduction from the T cell receptor (TCR) when it engages major histocompatability molecules-presenting antigen (reviewed in Refs. 1 and 2). They are necessary for the initiation of the tyrosine phosphorylation cascade, which is the earliest detectable event occurring upon TCR engagement (reviewed in Ref. 1). CD45 is thought to be required for T cell activation by constitutively dephosphorylating the negative regulatory tyrosine of Lck, Tyr 505 . This creates a "primed" Lck molecule that can become activated by autophosphorylation at Tyr 394 . Lck is thought to be activated upon TCR encounter with antigen, which then initiates the signal transduction cascade by phosphorylating downstream substrates such as CD3 and Zap-70 (reviewed in Refs. 1, 3, and 4). There is accumulating evidence that CD45 can also dephosphorylate the activating autophosphorylation site of Lck, thereby down-regulating Lck activity (Refs. 5-7, and reviewed in Ref. 8). Thus, by regulating the phosphorylation state of Lck, CD45 can up-or down-regulate Lck activity.
Like all members of the Src kinase family, Lck has a unique N-terminal region, a Src homology 3 (SH3) domain, an SH2 domain, a catalytic kinase domain and a short C-terminal region. Biochemical and structural evidence indicates that Src family members are regulated by an intramolecular interaction that occurs when the negative regulatory tyrosine in the Cterminal region is phosphorylated and interacts with the SH2 domain, resulting in a closed, inactive conformation (reviewed in Refs. 4 and 9). In contrast to the Src kinase family, very little is understood about how CD45 PTP activity is regulated or how substrate specificity is achieved. In the cytoplasmic domain of CD45, the membrane proximal PTP domain has been shown to be the active PTP domain, whereas most evidence indicates that the second PTP domain possesses no catalytic activity (10 -13). However, the C-terminal region containing the second PTP domain, CD45-D2, is required for optimal activity and stability of the active PTP domain (13,14). CD45 activity, like Lck activity, may also be regulated by intra-and inter-molecular interactions (13)(14)(15)(16). There is good evidence that Lck and, to a lesser extent, Fyn, are physiological substrates of CD45 in T cells (17)(18)(19)(20). In B cells, Lyn appears to be the major CD45 substrate (21)(22)(23)(24) and in macrophages, the phosphorylation state and activity of Hck and Lyn, but not Fgr, are affected by CD45 (25). Therefore, Src family kinases appear to be the preferred substrates for CD45 in leukocytes, although CD3, Zap-70, and Jak kinases have also been implicated as potential CD45 substrates (26 -28). Why Src family kinases are preferred substrates for CD45 or how the substrate specificity of CD45 is achieved at the molecular level is not well understood. To gain insight into these questions, we analyzed the enzymesubstrate interaction between CD45 and Lck using purified recombinant Lck and CD45 cytoplasmic domain proteins. We have determined that the second inactive PTP domain (CD45-D2) interacts with substrate and have identified a distinct region in the kinase domain of Lck, away from the catalytic site interaction, that is a docking site for CD45. The interaction of CD45 with this region was sufficient to promote substrate dephosphorylation, implicating a role for this region in substrate recognition and in contributing toward the determination of Lck as a preferred substrate for CD45.
Recombinant DNA Constructs-The plasmid encoding GST-human Erk1 was from S. Pelech (31), GST-SHC-SH2 was from T. Pawson (32), GST-Grb2 (33) was obtained from M. Gold and pGEX-3X used to express GST alone was from Amersham Biosciences. Murine Fyn T cDNA was obtained from R. Perlmutter (34) and subcloned into the pGEX-4T-2 vector (Amersham Biosciences). The cDNA for murine Src, modified by removal of six unique amino acids found in the neuronal form of Src was provided by T. Hunter (35) and the cDNA for rat Csk (36) was obtained from F. Jirik. These were subcloned into pGEX-2T (Amersham Biosciences).
In Vitro Binding Assays of Immobilized GST Fusion Proteins with Soluble 6His-CD45 Proteins-Either equimolar amounts or equal amounts of protein (as indicated in each individual experiment) of soluble 6His-C817S, 6His-D2, or 6His-D1 were added to 2 g of washed, immobilized GST fusion proteins (bound to glutathione-Sepharose 4B beads). Immobilized GST alone or irrelevant GST fusion proteins were included as negative controls. The assays were incubated shaking for 2 h at 4°C in 40 l of binding buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 0.025% ␤-mercaptoethanol) with protease inhibitors (1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, and 0.2 mM phenylmethylsulfonyl fluoride). In 6His-D1 binding assays, 500 M sodium orthovanadate was included in the buffer to inhibit PTP activity. After incubation, the samples were washed four times by adding 1 ml of radioimmune precipitation assay buffer (20 mM Tris-Cl, pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 1% Nonidet P-40, 0.1% SDS, 0.025% ␤-mercaptoethanol) and vortexing for 30 -60 s. Reducing sample buffer was added, and the samples were separated by SDS-PAGE, transferred to polyvinylidene difluoride (PVDF) membrane (Immobilon P, Millipore Corp.), and the amount of added soluble protein remaining bound to the beads was detected by Western blot analysis with 1/5,000 anti-CD45 antiserum (R02.2) as the primary, and 1/10,000 horseradish peroxidase (HRP)-labeled protein A (Bio-Rad) as the secondary. Membranes were subsequently stained with Coomassie Blue to visualize the immobilized proteins and subjected to densitometry scan analysis using Alphaimager TM software (Alpha Innotech Corp.). In experiments where tyrosine phosphorylation levels of the immobilized proteins were determined, the samples were divided into two, and one membrane was probed with R02.2 and the other with anti-phosphotyrosine antibody, 4G10, diluted 1/5,000 as the primary antibody and 1/10,000 dilution of HRP-labeled goat anti-mouse antibody as the secondary antibody (Southern Biotechnology Associates, Inc.).
Binding Assays Using Immobilized Phosphotyrosine and Soluble 6His-CD45 Proteins-O-Phospho-L-tyrosine (10 mol) was covalently coupled to 1 ml of packed CNBr-activated Sepharose CL-4B beads according to manufacturer's instructions. Approximately 65% coupling efficiency was achieved. 1 g of 6His-C817S or 6His-D2 (300 nM and 555 nM, respectively) were added to ϳ1 mM phosphotyrosine bound to Sepharose CL-4B beads (2.5 l beads) in 40 l of binding buffer plus protease inhibitors. These were incubated and shaken for 2 h at 4°C, then washed as described above. Reducing sample buffer was added, and the samples were separated by SDS-PAGE, transferred to PVDF membrane, and immunoblotted with anti-CD45 antibody (R02.2). Sepharose CL-4B beads were included as a negative control.
Dephosphorylation Assays-These were performed at 30°C in 10 l of PTP buffer (50 mM Tris, pH 7.2, 1 mM EDTA, 0.1% ␤-mercaptoethanol, with or without 0.01% Triton X-100) essentially as described (29). 60 nM of purified recombinant 6His-CD45 was incubated with autophosphorylated recombinant GST fusion proteins: either (a) 245 nM GST-Lck or GST-Src or (b) 25 nM GST-Lck, 130 nM GST-Csk, and 380 nM GST-Erk1 or GST-Erk(L-SD10) for increasing amounts of time. The concentrations of the latter were determined by densitometric analysis using Alphaimager TM software to contain approximately equal levels of phosphotyrosine. The reactions were stopped by immersion in a dry ice/ethanol bath, and samples were electrophoresed on a 10% SDSpolyacrylamide gel and transferred to PVDF membrane, and the amount of phosphorylated substrate remaining was determined by Western blot analysis. The tyrosine-phosphorylated bands were analyzed by densitometric analysis using Alphaimager TM software and subsequently graphed. The amount of tyrosine phosphorylation at time 0 was taken as 100% for all of the substrates.

CD45-D1 Binds the Unique N-terminal, SH2, and Kinase Domains of Lck, Whereas CD45-D2 Binds Only the Kinase
Domain of Lck-To analyze the interaction between CD45 and its substrate, Lck, we first determined which regions of the cytoplasmic domain of CD45 were interacting with the different regions of Lck. Purified, soluble, recombinant CD45 proteins, 6His-D1, 6His-D2, and 6His-C817S, were added to various immobilized GST-Lck domain proteins (Fig. 1), and a binding assay was performed (see "Experimental Procedures" for details). The full-length, inactive CD45 cytoplasmic domain protein (6His-C817S) interacted with both the catalytic kinase domain of Lck, GST-kinase, and the non-catalytic portion of the protein, GST-N32 ( Fig. 2A). 6His-C817S bound at low levels to GST-N and GST-SH2, but it did not bind to GST-SH3, as reported previously for inactive CD45 cytoplasmic domain protein (30). The highest level of binding was observed between 6His-C817S and GST-kinase ( Fig. 2A). 6His-D1 bound to the same regions of Lck as the full-length CD45 cytoplasmic domain protein, 6His-C817S, although less binding was observed to GST-kinase (Fig. 2B). This may be partly due to the inclusion of sodium orthovanadate in the 6His-D1 binding assays, added to inhibit 6His-D1 PTP activity. Attempts to make a catalytically inactive 6His-D1 recombinant protein did not result in sufficient protein for analysis. In contrast, the catalytically inactive 6His-D2 did not bind to any of the non-catalytic regions of Lck but bound significantly to GST-kinase (Fig. 2C). Even upon further exposure, no binding was observed between 6His-D2 and any of the non-catalytic domains above the background level. This indicates that the interaction between the enzyme, CD45, and the substrate, Lck, involves more than just the catalytic site of CD45 and the phosphorylated tyrosine of Lck. It implicates the non-catalytic regions of both CD45 and Lck in this interaction. In particular, it suggests that the second, non-catalytic PTP domain of CD45 is involved in binding substrate.
A comparison of equimolar amounts of 6His-D2 binding to Lck with the inactive CD45 cytoplasmic domain (6His-C817S) revealed that ϳ5-fold less 6His-D2 bound compared with 6His-C817S (data not shown). This indicates that 6His-D2 can interact with full-length Lck. It also suggests that the binding of 6His-C817S to Lck might be a product of multiple interactions involving both the catalytic and non-catalytic regions of CD45.
Binding of CD45-D2 to the Lck Kinase Domain Is Phosphorylation-independent and Does Not Require the Acidic Region or C-terminal Region of CD45-D2-Because 6His-D2 bound to the kinase domain of Lck and has significant sequence identity to active PTP domains, it was possible that the CD45-D2 PTP domain interacted with the phosphorylated tyrosines on Lck. To test this possibility, binding assays were performed between 6His-D2 and specific phosphorylated and non-phosphorylated forms of the Lck kinase domain expressed as GST fusion proteins. In E. coli, the active Lck kinase domain is normally phosphorylated at the autophosphorylation site, Tyr 394 , as is a phosphorylation mutant where the negative regulatory tyrosine, Tyr 505 , has been mutated (Y505F). Mutation of Tyr 394 (Y394F) prevents its autophosphorylation and co-expression of this construct with Csk (C-terminal Src kinase) promotes the phosphorylation at the negative regulatory site, Tyr 505 . Mutation of the catalytic lysine, Lys 273 , generates an inactive Lck mutant (K273R) that is not phosphorylated. Fig. 3A indicates that 6His-D2 bound to all Lck kinase domain proteins, irrespective of their tyrosine phosphorylation status (Fig. 3B). Fig.  3C confirms that equivalent amounts of GST-kinase domain proteins were used in each assay. The results show no significant difference in the binding of 6His-D2 to the various phosphorylated and non-phosphorylated forms of the Lck kinase domain.
To investigate whether the acidic region unique to CD45-D2 (residues 958 -978) was involved in binding to the Lck kinase domain, a 6His-D2 protein lacking the acidic region, 6His-D2-⌬Acidic, was made and purified. Equimolar amounts of 6His-D2 and 6His-D2-⌬Acidic were compared in their ability to bind to the GST-kinase phosphorylation mutants. As can be seen in the right panel of Fig. 3, the binding of CD45-D2 to the Lck kinase domain did not require the acidic region. In fact, more 6His-D2-⌬Acidic bound to the kinase domain than 6His-D2, suggesting that the acidic residues have a negative effect on binding. The binding of 6His-D2-⌬Acidic is also independent of the tyrosine phosphorylation state of the Lck kinase domain.
To determine if the C-terminal 78 amino acids of CD45-D2 A, schematic representation and SDS-PAGE of GST-Lck proteins purified from E. coli. The numbers at the side of the constructs correspond to the numbers above the gel. The numbers above the constructs indicate the first and last residues of the various recombinant proteins, with the individual domains named above the full-length construct, and the positive (Tyr 394 ) and negative (Tyr 505 ) regulatory tyrosines indicated below. The proteins were separated on a 10% SDS-polyacrylamide gel, transferred to PVDF membrane, and stained with Coomassie Blue. Prestained molecular mass standards are indicated on the left in kDa. B, schematic representation and Coomassie Blue-stained 10% SDSpolyacrylamide gel of the His 6 -tagged murine CD45 cytoplasmic domain proteins. MP is the membrane proximal region, PTP-D1 is the first, N-terminal PTP domain, SP is the spacer region, PTP-D2 is the second, C-terminal PTP domain, and CT is the carboxyl tail. The starred residue above construct 7 indicates an inactivating point mutation of the catalytic Cys 817 to Ser. The acidic region unique to CD45-D2 is indicated under construct 9. 6His-D1 typically purifies as a doublet (13). Molecular mass standards are indicated on the left in kDa.

FIG. 2.
Binding of His 6 -tagged CD45 cytoplasmic domain proteins to individual GST fusion Lck domain proteins. Equimolar amounts (300 nM) of 6His-C817S, 6His-D1, and 6His-D2 were added to 2 g of GST-Lck proteins bound to glutathione-Sepharose 4B beads in 40 l of 20 mM Tris, pH 7.5, 150 mM NaCl, 0.025% ␤-mercaptoethanol and incubated for 2 h at 4°C. The beads were washed with radioimmune precipitation assay buffer, reducing sample buffer added, and the samples were separated by 10% SDS-PAGE and transferred to PVDF membrane. The CD45 protein remaining bound to the GST-Lck proteins was detected by Western blot with anti-CD45 antiserum (R02.2; see "Experimental Procedures"). Immobilized Lck proteins are indicated above A. Negative controls are GST alone, GST, or an irrelevant GST fusion protein, GSTϩ: GST-Grb2 for 6His-C817S and 6His-D2, and the GST-SHC SH2 domain for 6His-D1. Coomassie Blue staining of the membranes after Western blotting confirmed that approximately equal amounts of Lck protein was used in each condition (data not shown). A, 6His-C817S added to immobilized Lck domains. B, 6His-D1 added to Lck domains. 500 M sodium orthovanadate was included in the buffer to inhibit PTP activity. C, 6His-D2 added to Lck domains.
were involved in binding to Lck, a GST-D2 protein lacking these amino acids, GST-D2⌬CT, was made. Both GST-D2 and GST-D2⌬CT were cleaved with factor Xa from GST and used in a binding assay with immobilized GST-Lck. The C-terminal deletion of 78 amino acids did not have any significant effect on the binding of CD45-D2 to Lck (data not shown). Therefore, the C-terminal region and the unique acidic region are not required for CD45-D2 to bind to Lck, indicating that the interaction with Lck is mediated by amino acid residues close to or within the D2 PTP domain.
CD45-D2 Does Not Bind Phosphotyrosine-The phosphorylation-independent binding of 6His-D2 to the differentially phosphorylated GST-kinase domain proteins suggested that CD45-D2 may not be binding to Lck via its proposed catalytic pocket, or if it is, the highly substituted CD45-D2 catalytic pocket may not bind to phosphotyrosine. To investigate whether CD45-D2 can bind tyrosine phosphate, phosphotyrosine was immobilized to CNBr-activated Sepharose CL-4B beads and soluble 6His-D2 or 6His-C817S was added to the beads in an in vitro binding assay. The full-length inactive cytoplasmic domain, 6His-C817S, bound to the phosphotyrosine beads, but 6His-D2 did not (Fig. 4). In a total of eight experiments, with varying concentrations of proteins, phosphotyrosine beads, and buffer components, 6His-C817S consistently bound to phosphotyrosine, whereas 6His-D2 did not. Upon longer exposure of the film, 6His-D2 was sometimes detectable in the phosphotyrosine bead sample relative to the Sepharose bead control, but this was insignificant relative to the binding of 6His-C817S. Therefore, unlike the catalytic PTP domain of CD45, CD45-D2 does not bind significantly to phosphotyrosine.  (Fig. 5A). Quantitative densitometric analyses from seven experiments, correcting for the amount of immobilized fusion protein and standardizing the binding to GST-Lck to 1, indicated that the relative binding of 6His-D2 to GST-kinase was 1.1 Ϯ 0.3. The similar levels of binding indicate that the non-catalytic domains of Lck do not interfere with this interaction. In this case, full-length Lck is active, phosphorylated primarily at the autophosphorylation site, Tyr 394 , and predicted to be in an open, unconstrained conformation (reviewed in Ref. 9). When Src family kinases are phosphorylated at their negative regulatory site (Tyr 505 for Lck), the phosphorylated tyrosine binds to the SH2 domain causing the kinase domain to adopt a constrained position with the autophosphorylation loop blocking the active site (44,45). Thus it was possible that CD45-D2 may bind to the unconstrained kinase domain but not to the constrained form. To evaluate this possibility, 6His-D2 binding to GST-Lck Y505F (not phosphorylated at Tyr 505 and therefore predicted to be in the open conformation) and Csk-phosphorylated GST-Lck Y394F (phosphorylated at Tyr 505 and predicted to be in the closed or constrained conformation) was assessed (Fig. 5B). 6His-D2 bound equally well to both GST-Lck proteins, implying that CD45-D2 can bind to the Lck kinase domain independent of the phosphorylation or conformational state of Lck.
6His-D2 Does Not Discriminate Between Binding to Lck, Fyn or Src-Previous work indicates that the absence of CD45-D2 does not affect substrate specificity at the peptide level (13). However, it is possible that an effect of CD45-D2 on substrate specificity may only be observed at the protein level. To determine if the binding ability of CD45 or CD45-D2 reflected the substrate preference observed in the T cell where Lck is preferred over Fyn, and Src is not a substrate when retrovirally expressed in T cells (46,47), in vitro binding assays were performed. Soluble 6His-D2 or 6His-C817S was incubated with immobilized GST fusion proteins of Lck, Fyn, and Src. Both CD45 proteins bound to all three Src family kinases (Fig. 6). Although there appears to be a slight preference for the binding of 6His-D2 to Lck in this experiment, when immobilized protein amounts are taken into account and densitometric analysis performed and averaged from at least three experiments, overlapping standard deviations indicated that the differences in binding were not significant. This demonstrates that both CD45 and CD45-D2 do not discriminate between different members of Src family kinases in an in vitro binding assay, suggesting that the substrate preference observed in T cells for particular Src family members is dictated by cellular factors.  4. Binding of 6His-D2 and 6His-C817S to immobilized phosphotyrosine. 1 g of 6His-C817S or 6His-D2 (300 and 555 nM, respectively) were added to ϳ1 mM phosphotyrosine bound to Sepharose CL-4B beads (P-Y) in a 40-l volume of 20 mM Tris, pH 7.5, 150 mM NaCl, 0.025% ␤-mercaptoethanol. These were incubated for 2 h at 4°C then washed vigorously with radioimmune precipitation assay buffer, and reducing sample buffer was added. The samples were separated by SDS-PAGE and immunoblotted with anti-CD45 antibody (R02.2; see "Experimental Procedures"). Sepharose CL-4B beads were used as a negative control (B). 6His-C817S and 6His-D2 are indicated on the right. 20 ng of each protein was included for reference. Prestained molecular mass standards are indicated on the left in kDa.

6His-D2 Preferentially Binds to GST-Lck Compared with GST-Csk and GST-Erk1
Kinases-To determine if CD45-D2 showed preferential binding to Lck over non-Src family kinases, in vitro binding assays were performed between 6His-D2 and GST-Lck and compared with 6His-D2 binding to a related tyrosine kinase, GST-Csk, and to a less-related serine/threonine kinase, GST-Erk1. 6His-D2 did bind to GST-Csk, but to a much lesser extent than to GST-Lck (Fig. 7). Densitometric analysis from four separate experiments, correcting for the amount of immobilized protein, indicated that the level of binding of 6His-D2 to GST-Csk, relative to GST-Lck, is 24 Ϯ 15%. 6His-D2 did not bind significantly to the more distantly related kinase, GST-Erk1 (Fig. 7). Therefore, although CD45-D2 did not show preferential binding between the Src family kinases, it did discriminate between Src family kinases and less related kinases, and may therefore contribute toward the preference of CD45 for Src family kinases over other less related kinases in the cell.

CD45 Preferentially Dephosphorylates Src Family Kinases over the More Distantly Related Kinases, Csk and Erk1-To
correlate the binding preference of CD45-D2 to substrate specificity, the in vitro dephosphorylation of autophosphorylated Lck by active recombinant CD45 cytoplasmic domain protein (6His-CD45) was compared with the dephosphorylation rate of autophosphorylated Src. Consistent with the 6His-D2 binding assay, no significant difference was observed (Fig. 8A). As previously reported, both GST-Csk and GST-Erk1 were autophosphorylated to low stoichiometry on tyrosine residues when produced from E. coli (31,48,49). Tyrosine phosphorylation of Csk has also been reported to occur in vitro (50) and in HeLa cells (51). The tyrosine phosphorylation level of each kinase was assessed by Western blotting. Dephosphorylation of approximately equivalent levels of tyrosine phosphorylated Lck, Csk, and Erk1 by CD45 demonstrated that Lck was preferentially dephosphorylated compared with Csk and Erk1 (Fig. 8B). The trend observed in Fig. 8B was consistent with results from eight other experiments with different conditions, including varying CD45 concentration, using equal amounts of protein instead of equal amounts of phosphorylated substrate, or including all substrates in the same assay tubes (data not shown). These data support the hypothesis that the binding of  6. Binding assay of 6His-D2 and 6His-C817S to Src family GST fusion proteins. 1 g of soluble 6His-D2 or 6His-C817S was incubated with 2 g of immobilized GST fusion proteins in 50 l of 20 mM Tris, pH 7.5, 150 mM NaCl, and 0.025% ␤-mercaptoethanol for 2 h at 4°C, and washed with radioimmune precipitation assay buffer (see "Experimental Procedures"). The samples were separated by 10% SDS-PAGE and transferred to PVDF membrane. A, 6His-C817S added to immobilized GST-Src family kinases. B, 6His-D2 added to immobilized GST-Src family kinases. The top panels in both A and B are anti-CD45 Western blots (R02.2) to detect relative amounts of 6His-D2 and 6His-C817S remaining bound. The lanes are as indicated above the gels. GST alone was included as a negative control. Prestained molecular mass standards are indicated on the left in kDa. The bottom panels are the membranes after Coomassie Blue staining to show the relative amounts of immobilized GST fusion proteins present. GST-Src was always heavily degraded, however, GST-Lck degradation varied between different purification preparations. FIG. 7. Relative binding of 6His-D2 to GST-Lck, GST-Csk, and GST-Erk1. 0.54 g of 6His-D2 was added to 2 g of immobilized GST fusion protein in a binding assay as described in Fig. 2. The immobilized proteins are indicated above. GST-Grb2 and GST alone were included as negative controls. After the beads were washed, reducing sample buffer was added to each assay, and the samples were electrophoresed on an SDS-polyacrylamide gel and transferred to PVDF membrane for immunoblot analysis. A, anti-CD45 Western blot (R02.2) showing the amount of 6His-D2 remaining bound to the GST fusion proteins. B, Coomassie Blue stain. The membrane from A was stained with Coomassie Blue to show the relative amounts of immobilized proteins. Prestained molecular mass standards are indicated on the left in kDa.
CD45-D2 to substrate facilitates substrate dephosphorylation and contributes towards the observed substrate specificity for Src family kinases.
The SD10 Region of Lck Binds 6His-D2 and Is a Contributing Factor in Determining Substrate Specificity-To determine the region in Lck involved in mediating binding to 6His-D2, we used SWISS-MODEL (52) to model CD45 and Lck (based on the in vitro data and the three-dimensional structures of the twodomain PTP, LAR (53) and the Src family kinases (54 -56)) to look for possible exposed sites of interaction. Because the catalytic domains of tyrosine and serine/threonine kinases are highly conserved in three-dimensional structure and in the existence of 12 subdomains (57,58), the overall structures of Lck, Csk, and Erk1 are quite similar. Given the 6His-D2 binding data, the interacting region was predicted to be conserved within the Src family kinases, Lck, Fyn, and Src, to be less conserved in Csk and not conserved in Erk1. In addition, many kinases are regulated by phosphorylation (59), thereby implicating interactions with other regulatory protein kinases and phosphatases. It is possible that a specific region within the kinase domain may have evolved as a substrate recognition domain, to mediate the regulatory interaction with other kinases and phosphatases. Such a region would be predicted to be divergent between distantly related kinases. This rationale focused our attention on SD10 (57,58). This subdomain is located and exposed at the base of the large kinase lobe. It is the most poorly conserved subdomain in the protein kinase superfamily, and its function is not known (57,58). Comparison of this region in Lck with Src, Fyn, Csk, and Erk1 revealed 60% identity with Src and Fyn, 40% with Csk, and essentially no identity with Erk1 (Fig. 9). This was consistent with the observed binding of 6His-D2 to these kinases.
To determine if the SD10 of Lck played a role in binding 6His-D2, a chimeric protein was made in which the Erk1 SD10 region was replaced by that of Lck (see "Experimental Procedures" for details). This chimeric protein was expressed in E. coli as a soluble GST fusion protein (GST-Erk(L-SD10)). Like GST-Erk1 (31,49), GST-Erk(L-SD10) was capable of in vitro tyrosine autophosphorylation, indicating that the chimeric kinase was active. Fig. 10A shows that significantly more 6His-D2 bound to the chimeric GST-Erk(L-SD10) protein than to the GST-Erk protein, and Fig. 10B shows that equal amounts of GST-Erk1 and GST-Erk(L-SD10) were present. Calculation of the amount of 6His-D2 binding to equivalent amounts of kinase (Fig. 10D) indicated that the increased binding of 6His-D2 to GST-Erk(L-SD10) over GST-Erk approached the level of binding of 6His-D2 to GST-Lck. Specifically, the binding of 6His-D2 to GST-Erk (L-SD10) was approximately two-thirds the level of 6His-D2 binding to GST-Lck. This demonstrates that Lck SD10 is sufficient to mediate binding to 6His-D2 as transfer of this region to a distantly related kinase confers binding to a kinase that does not normally bind to 6His-D2.
Given the binding of 6His-D2 to the Erk(L-SD10) chimera but not to Erk1, it was next investigated whether the presence of Lck SD10 in the Erk(L-SD10) chimera also affected its dephosphorylation by CD45. GST-Erk1 and GST-Erk(L-SD10) were phosphorylated to approximately equal levels but were not autophosphorylated as efficiently as GST-Lck. To compare the rate of dephosphorylation of Erk and the Erk chimera with equivalent amounts of phosphorylated Lck (Fig. 11), one tenth the amount of Lck was used. Unlike GST-Erk1, the GST-Erk(L-SD10) chimera was dephosphorylated at a faster rate, similar to the dephosphorylation rate of GST-Lck (Fig. 11), indicating that the exchange of Erk SD10 with Lck SD10 is sufficient to convert Erk1 into an efficient substrate for CD45. This indicates that the SD10 region of Lck plays a significant role in binding CD45 and in facilitating substrate dephosphorylation. Identification of this region as a substrate recognition domain for CD45 helps explain why Src family kinases, but not more distantly related kinases, are preferred substrates for the tyrosine phosphatase, CD45.  9. Comparison of SD10 sequences. The 23-amino acid sequence of Lck containing the SD10 region is compared with the SD10 regions of Src, Fyn, Csk, Jak2, and Erk-1. The SD10 region of Erk-1 is 53 amino acids and is therefore represented on three lines. Amino acids in common with Lck are highlighted in gray. The helix G region of SD10 is indicated above by the line.
manner that was independent of Lck phosphorylation and conformation.
The reason for the presence of two tandem repeated PTP domains in several transmembrane phosphatases has been enigmatic, particularly as the second domain often has a less well conserved catalytic cleft and many have no detectable PTP activity (reviewed in Ref. 60). One possibility is that, like STYX domains (61), an inactive PTP domain may have evolved to bind phosphotyrosine-containing proteins. However, CD45-D2 bound to Lck independently of its tyrosine phosphorylation status and did not bind detectable levels of phosphotyrosine, making this mode of binding unlikely for the second domain of CD45. Two proteins, a CD45-associated protein, CD45-AP, and the ␤ chain of the interleukin 2 receptor, have been shown to interact with the Lck kinase domain (62,63). In both these cases an acidic region was shown to mediate their interaction with Lck. Because the acidic region of CD45-D2 was not required for its interaction with Lck, the mode of binding to CD45 must be distinct from the interaction between Lck and these two molecules.
CD45 is a very potent tyrosine phosphatase, and over time it will dephosphorylate all tyrosine-phosphorylated proteins in a pervanadate-stimulated T cell lysate. 2 In this study, differences in the rate of dephosphorylation of certain protein substrates by CD45 indicated that CD45 does exhibit substrate specificity in vitro. Evidence was presented for a role of the SD10 region of Lck and the non-catalytic PTP domain of CD45 in facilitating the enzyme-substrate interaction and in promoting substrate dephosphorylation. Accumulating evidence indicates that PTPs are not as promiscuous as originally thought, and cytosolic PTPs, PTP1B and T cell-PTP, exhibit very selective substrate specificity in vivo. This is thought to be achieved by a combination of intrinsic catalytic domain specificity and by phosphatase targeting of the non-catalytic C-terminal region of PTP1B and the alternatively spliced forms of T cell-PTP to specific regions in the cell (reviewed in Ref. 64). Here we provide evidence of a non-catalytic interaction between Lck and CD45 that helps to explain the preferred dephosphorylation of Src family kinases by CD45 observed in leukocytes.
The data reported here demonstrate that one function of the second phosphatase domain of CD45 (CD45-D2) is to bind substrate. Consistent with this, CD45-D2 has also been implicated in binding of CD3, a potential CD45 substrate in T cells, because exchange of CD45-D2 for LAR-D2 resulted in the loss of binding of CD3 to a CD45 substrate-trapping mutant (26). Recently, another kinase family, the Jak kinases, was implicated as CD45 substrates. In this example, CD45-D2 was shown to bind Jak2 kinase (28). Comparison of the SD10 region of Jak2 with Lck shows 35% sequence identity, making it similar to that of Csk. It will be of interest to compare the relative binding of CD45-D2 to Lck and Jak2. CD45-D2 has previously been implicated in influencing substrate specificity, because serine phosphorylation of D2 can enhance CD45 activity for certain artificial substrates (65,66) and CD45-D2 deletions can differentially influence enzymatic activity, depending upon the substrate used (11,12).
We determined that a 23-amino acid sequence from the SD10 region of Lck is sufficient to mediate the interaction with CD45-D2. This region is located at the bottom of the large lobe of the kinase domain and contains Helix G and the connecting loop to Helix H and is exposed in both the active and inactive conformations (54 -57). This region is well conserved between 2 J. Felberg and P. Johnson, unpublished observations. FIG. 10. Binding of 6His D2 to GST-Lck, GST-Erk1, and GST-Erk(L-SD10). A, 0.54 g of 6His-D2 was added to 2 g of immobilized GST fusion protein in 40 l of 20 mM Tris, pH 7.5, 150 mM NaCl, 0.025% ␤-mercaptoethanol, and 0.05% Triton X-100 and incubated for 2 h at 4°C in a binding assay as described in Fig. 2. After the beads were washed, reducing sample buffer was added to each assay, and the samples were electrophoresed on an SDS-polyacrylamide gel and transferred to PVDF membrane for immunoblot analysis. A, anti-CD45 Western blot (R02.2) showing the amount of 6His-D2 remaining bound to the GST fusion proteins. GST fusion proteins used in the binding assay are shown in an Erk Western blot (B), and in a Coomassie Blue protein stain (C). Prestained molecular mass standards are indicated on the left in kDa. D, graph showing the relative binding of 6His-D2 to equivalent amounts of GST-kinase (indicated at the top of panel A). Relative binding was determined by dividing the level of 6His-D2 binding by the amount of GST-kinase (derived from the Coomassie Blue stain (C)). The data are the average from four experiments, and the error bars represent S.E.
FIG. 11. Dephosphorylation of tyrosine-phosphorylated GST fusion proteins. 60 nM 6His-CD45 was incubated with tyrosine-phosphorylated 25 nM GST-Lck, 380 nM GST-Erk1, or 380 nM Erk(L-SD10) in 10 l of PTP buffer for the time points indicated, and samples were treated as described in Fig. 8. Beads were then washed in PTP buffer and used in dephosphorylation assays. A, graphical analysis of data averaged from three separate dephosphorylation assays. The amount of phosphorylation at time 0 for each substrate was taken as 100%. B, anti-phosphotyrosine Western blot analysis of GST fusion protein substrates used in dephosphorylation assays.
Src family kinases (ϳ60% sequence identity), is less conserved for Csk (40% sequence identity), and is not conserved for more distantly related kinases such as Erk1 (no significant sequence identity; see Fig. 9). The crystal structure of an Erk family kinase, Erk2, indicates that the larger SD10 region of Erk kinases contains two additional small helices in addition to Helix G (67), making this region quite distinct from Lck SD10. This report identifies the SD10 region of Lck as an important docking site for CD45. The binding of CD45 to this site facilitates substrate dephosphorylation implying that this region plays a significant role in optimizing the enzyme-substrate interaction and thereby facilitating substrate dephosphorylation. The more divergent the sequence of SD10 was from that of Src family kinases, the less CD45-D2 bound and the slower the kinase was dephosphorylated by CD45. Because many protein kinases are regulated by phosphorylation and must therefore interact with protein phosphatases, it will be interesting to determine if the SD10 region of other kinases influences the binding and substrate selection of other protein phosphatases. Interestingly, the crystal structure of the serine/threonine kinase, Cdk2, complexed with its regulatory phosphatase, KAP, revealed an extensive interaction away from the catalytic site between the C-terminal region of the kinase (encompassing the SD10 region) and the C-terminal helix of the phosphatase (68). The structural and biochemical data suggest that this interaction provides the dominant specificity site for recognition of the kinase by KAP, which also provides the correct alignment for the dephosphorylation of the activation segment of the kinase by KAP (68). This, together with the data reported here, raises the intriguing possibility that the SD10 region of protein kinases may have evolved to be a general recognition/specificity domain that binds regulatory phosphatases.