Structural Basis for Substrate Specificity of Protein-tyrosine Phosphatase SHP-1*

The substrate specificity of the catalytic domain of SHP-1, an important regulator in the proliferation and development of hematopoietic cells, is critical for understanding the physiological functions of SHP-1. Here we report the crystal structures of the catalytic domain of SHP-1 complexed with two peptide substrates derived from SIRPα, a member of the signal-regulatory proteins. We show that the variable β5-loop-β6 motif confers SHP-1 substrate specificity at the P-4 and further N-terminal subpockets. We also observe a novel residue shift at P-2, the highly conserved subpocket in protein- tyrosine phosphatases. Our observations provide new insight into the substrate specificity of SHP-1.

Protein-tyrosine phosphatases (PTPs) 1 consist of a diverse family of enzymes that play crucial roles in cell growth, differentiation, and transformation (1)(2)(3). They can be broadly divided into membrane-bound, receptor-like PTPs, and cytosolic PTPs. The cytosolic PTPs contain only one catalytic domain, whereas the membrane-bound receptor-like PTPs usually contain two tandem catalytic domains. The catalytic domains of PTPs are highly conserved in their three-dimensional structures (4 -7). However, they have remarkably different substrate specificity (3, 8 -10), which is still not well understood. Previous studies using various synthetic phosphotyrosyl peptides failed to identify a shared by PTP substrate because the peptides studied were not derived from physiological substrates of PTPs. In the present study, we have addressed the structural basis for the substrate specificity of PTPs using SHP-1 and its physiological substrate SIRP␣/SHPS-1 as a model. SIRP␣ is a transmembrane protein of the signal-regulatory protein family. Its extracellular domain contains three immunoglobulin domains, and its cytoplasmic domain contains four phosphotyrosine sites (Tyr(P) 427 , Tyr(P) 452 , Tyr(P) 469 , and Tyr(P) 495 ).
SHP-1 is expressed primarily in hematopoietic cells, and contains two Src homology 2 (SH2) domains, a neighboring catalytic domain, and a C-terminal tail. Its phosphatase activity is inhibited by both the SH2 domains and the C-terminal tail (11,12). SHP-1 is activated upon the binding of its tandem SH2 domains to immunoreceptor tyrosine-based inhibitory motifs. Domain-swapping studies on SHP-1 and its analogue, SHP-2, have shown that the catalytic domains of SHP-1 and SHP-2 have distinct substrate specificity (9,10), and therefore illustrate that the dissection of the structural basis for the substrate specificity of SHP-1 is fundamental to the understanding of its physiological functions. The identification of the substrates of SHP-1 (i.e. SIRP␣, CD22, and CD72; Refs. [13][14][15] has made it possible for us to probe this structural basis. The results of this probe are presented below.

EXPERIMENTAL PROCEDURES
Crystallization and Data Collection-The C455S mutant of the SHP-1 catalytic domain (245-532) was cloned, expressed, and purified as described elsewhere (16). The phosphotyrosyl decapeptides were synthesized and purified to 90% purity by SynPep Inc. (Dublin, CA). Crystals of the native enzyme were obtained by vapor diffusion method with 5 mg/ml protein drops equilibrating over reservoir solution containing 1.9 M (NH 4 ) 2 SO 4 , 0.1 M Tris-HCl, pH 8.5. The Tyr(P) 469 complex crystals were obtained by co-crystallization method with the protein and the peptide at a 1:1 molar ratio. The Tyr(P) 495 complex crystals were grown by soaking native crystals in cryosolvent (30% sucrose, 1.9 M (NH 4 ) 2 SO 4 , 0.1 M Tris-HCl, pH 8.5) containing 0.5 mM peptide Tyr(P) 495 for 2-3 days. A data set of the Tyr(P) 469 complex diffracted up to 2.5 Å was collected at Ϫ188°C using an ADSC cryosystem and MAR imaging plate (Area Detector System Corp., Poway, CA); the x-ray source was generated from a Rigaku (Japan) RU-300 operating at 50 kV and 100 mA. A data set diffracted to 2.3 Å resolution for Tyr(P) 495 complex was collected at Ϫ180°C on F1 beamline at CHESS (Cornell University). Both data sets were processed with DENZO and SCALE-PACK (17). The crystal statistics are summarized in Table I.
Structure Determination-Crystal structures of both complexes were solved by molecular replacement method with program AMoRe (18), using coordinates of the native enzyme (Protein Data Bank code 1GWZ) as the search model. The molecular replacement solution for each complex structure was refined by rigid body and positional protocols in X-PLOR (19). The peptide positions were identified from the initial electron density difference map (F o Ϫ F c ). However, the electron densities for the peptides were not quite continuous. After several cycles of model rebuilding and refinements, the electron densities for the peptides became clearer, and the peptides were added to the models. The complex models were then subjected to X-PLOR refinement, using the slow cool and positional refinement protocols with the 2 cutoff data between 6 Å and the highest resolution. The R-free value was also calculated from the beginning to the penultimate cycle by randomly selecting 10% data as the test set. The model building interspersed with X-PLOR refinement was done by the program TURBO-FRODO (20). In the penultimate cycle, the R-free values were 30.3% and 27.2% for the Tyr(P) 469 and Tyr(P) 495 complexes, respectively. The final crystallographic R-values for the Tyr(P) 469 and Tyr(P) 495 complexes were 19.9 and 20.5%, respectively, with all reflections used in the refinement. Fifty-three water molecules were added in the final Tyr(P) 495 complex model; however, no water molecules were added in the final Tyr(P) 469 complex model. The final refinement results and geometry analyses for each complex are also summarized in Table I.

RESULTS
Identification of in Vitro Substrates for SHP-1-SIRP␣ has been identified as a potential physiological substrate of SHP-1 in macrophages (13). However, the dephosphorylation sites of SHP-1 on SIRP␣ are still unknown. To identify these sites, we synthesized four phosphotyrosyl decapeptides corresponding to the four phosphotyrosine sites in the cytoplasmic domain of SIRP␣ and measured the kinetic parameters of the catalytic domain of SHP-1 toward each of them (Table II). The results demonstrated that peptides Tyr(P) 469 and Tyr(P) 495 were effective in vitro substrates for SHP-1. With respect to amino acid sequence, the four synthetic peptides fall into two groups: Tyr(P) 427 and Tyr(P) 469 , and Tyr(P) 452 and Tyr(P) 495 . The two peptides in each group have very similar sequences for the residues on the C-terminal side of residue Tyr(P). However, only one peptide in each group (Tyr(P) 469 and Tyr(P) 495 , respectively) is the in vitro substrate of SHP-1. Therefore, substrate specificity of the catalytic domain of SHP-1 is determined primarily by residues N-terminal to residue Tyr(P) in the peptide substrates.
Structures of the Two Complexes-Crystal structures of the catalytic domain of SHP-1 (with C455S mutation) complexed with peptides Tyr(P) 469 and Tyr(P) 495 were determined at 2.5 and 2.3 Å resolution, respectively (Table I). The final electron density maps around the Tyr(P) peptides in both complex structures are shown in Fig. 1 (A and B). The average B factor for peptide Tyr(P) 495 was higher than the average B factors for peptide Tyr(P) 469 and the SHP-1 catalytic domain, indicating that peptide Tyr(P) 495 was very flexible.
Both peptides bound to the catalytic domain of SHP-1 in extended conformations (Fig. 1, C and D). Backbones of the peptides were approximately 45 degrees to the central twist ␤-sheet that formed the core of the catalytic domain of SHP-1. Binding peptides Tyr(P) 469 and Tyr(P) 495 to the catalytic domain of SHP-1 buried 800 and 660 Å 2 solvent-accessible surface, respectively. The binding of peptides Tyr(P) 469 and Tyr(P) 495 to the catalytic domain of SHP-1 did not bring the WPD loop (Trp 419 -Pro 428 ) into the closed conformation. In contrast, the WPD loop moved from the half-open/half-closed conformation observed in the native enzyme structure (7) to the open conformation in the complex structures. In addition to the Tyr(P) subpocket, six substrate-binding subpockets were well defined (Fig. 2). Unlike the Tyr(P) subpocket that protruded into the center of the catalytic domain, these subpockets were relatively shallow and open to the solvent. For the convenience of discussion, we name each subpocket according to the residue numbers within the structure of the Tyr(P) 469 complex. In both peptides, residues at the C-terminal side of residue Tyr(P) bound to the same binding subpockets of SHP-1 with similar orientations and conformations, except for a slight difference at the Pϩ4 subpocket. The C-terminal residues of the peptides interacted mainly with loop ␣1/␤1 and the ␣5-loop-␣6 motif of SHP-1 (defined in Fig. 2).
The P-2 Subpocket-P-2 was the most distinct subpocket at the N-terminal side of the Tyr(P) subpocket. It was formed by hydrocarbons from the side chains of Lys 362 and Arg 277 and by side chain of Tyr 278 . In the Tyr(P) 469 complex, the P-2 subpocket was occupied by Leu P-2 ( Fig. 1, C and E). Surprisingly, it was occupied by Phe P-3 in the Tyr(P) 495 complex structure, indicating a likely residue shift at the N-terminal side of residue Tyr(P) in the Tyr(P) 495 complex relative to the Tyr(P) 469 complex. Because of this novel residue shift, the conformations of residue Tyr(P) were also different between the two complex structures. By shifting into the P-2 subpocket, Phe P-3 pushed Ser P-2 of peptide Tyr(P) 495 into the P-1 subpocket, which was occupied by Thr P-1 in the Tyr(P) 469 complex structure. At the same time, it pushed Gln P-1 of peptide Tyr(P) 495 toward the solvent (Fig. 1D).
The Subpockets N-terminal to the P-2 Subpocket-The kinetic studies showed that the substrate specificity of SHP-1 catalytic domain was determined mainly by residues N-terminal to residue Tyr(P) in the peptides. Between the Tyr(P) 469 and Tyr(P) 495 complexes, significant conformational differences were observed for peptide residues N-terminal to the P-2 subpocket (Fig. 1E). These N-terminal residues interacted mainly with the ␤5-loop-␤6 motif of SHP-1. Therefore, the ␤5-loop-␤6 motif determines, at least in part, the substrate specificity of SHP-1. In both complex structures, the ␤5-loop-␤6 motif moved approximately 6.5 Å toward the N termini of the peptides and formed different interactions with the substrates. In the Tyr(P) 469 complex, Asp P-4 formed a salt bridge with residue Arg 360 of the ␤5-loop-␤6 motif. This salt bridge was further stabilized by two hydrogen bonds between Arg 360 and Asn 361 (Fig. 2A). These suggest that either an aspartate or a glutamate will be preferred for the P-4 position of the peptide substrates of SHP-1. In the Tyr(P) 495 complex structure, Pro P-5 was the corresponding residue to Asp P-4 . Because of the repulsion between the guanidine group of residue Arg 360 and the hydrocarbons of residue Pro P-5 , Pro P-5 swept away from the SHP-1 molecule into the solvent. Residue Arg 360 also underwent a conformational change and formed weak van der Waals' interactions with residue Pro P-5 . Other N-terminal residues, such as Thr P-3 and Glu P-5 of peptide Tyr(P) 469 and Ser P-4 in peptide Tyr(P) 495 , were exposed to the solvent.
The Tyr(P)-binding Subpocket-Like other PTPs, the signature motif of SHP-1, i.e. HCXAGXGR(S/T), formed the base for the Tyr(P) subpocket and interacted with the phosphate group of residue Tyr(P) by extensive hydrogen bonds in both complex structures (Fig. 2). Because of different binding at the P-2  interaction between its phenyl ring and the side chain of residue Tyr(P) in the Tyr(P) 469 complex structure. However, this Ϫ interaction was much weaker in the Tyr(P) 495 complex structure because of the different conformations of Tyr 278 and Tyr(P). The results indicated that peptide Tyr(P) 469 had stronger binding affinity and was a more favorable in vitro substrate for SHP-1, which is consistent with the kinetic study results (Table II). In addition, Tyr 278 of the enzyme was involved in defining the depth of the Tyr(P) subpocket, making the sub- pocket too deep to bind either phosphoserine or phosphothreonine residue.

DISCUSSION
Substrate Specificity of SHP-1-We have determined that the ␤5-loop-␤6 motif and loop ␣1/␤1 determined the substrate specificity of SHP-1. We also observed the novel residue shift at the P-2 subpocket. The P-2 subpocket can bind either the P-2 or the P-3 residue of phosphotyrosyl substrates, provided that they are hydrophobic residues. To our knowledge, this is the first time that a residue shift has been observed in proteinprotein recognition. This novel residue shift suggests peptides containing the sequence (L/I/V)X n pYXX(L/I/V) (n ϭ 1 or 2) be potential substrates for SHP-1. The formation of the salt bridge between residue Arg 360 of SHP-1 and residue Asp P-4 of peptide Tyr(P) 469 in the Tyr(P) 469 complex structure suggests that peptides having the consensus sequence (D/E)X(L/I/V)X n pYXX(L/ I/V) (n ϭ 1 or 2) are favored as substrates of SHP-1. This consensus sequence explains why peptides Tyr(P) 427 and Tyr(P) 452 are not in vitro substrates for SHP-1. Recently, signal transduction co-receptors CD22 and CD72 were identified as both the activators and substrates of SHP-1 (14,15). Using the consensus sequence, we predict that sites Tyr 777 , Tyr 837 , and Tyr 857 of CD22 and sites Tyr 7 and Tyr 39 of CD72 are the dephosphorylation sites of SHP-1.
The catalytic domain swapping studies on SHP-1 and SHP-2 have demonstrated the different substrate specificity of the catalytic domains of SHP-1 and SHP-2 (9, 10). Among the six identified substrate-binding subpockets other than the Tyr(P) subpocket, only the P-4 subpocket differs between SHP-1 and SHP-2 (see Fig. 4). Residue 360, which is located at the ␤5loop-␤6 motif, is an arginine in SHP-1 and a lysine in SHP-2. This residue difference is one of the main reasons for the different substrate specificity of SHP-1 and SHP-2. In addition to residue 360, several other residues in the ␤5-loop-␤6 motif also differ between SHP-1 and SHP-2. These residues confer specific recognition for P-4 and further N-terminal residues of the substrates.
Comparison with PTP1B-Hexapeptide Complex-Comparison of the Tyr(P) 469 and Tyr(P) 495 complex structures with the PTP1B-hexapeptide structure (21) (Fig. 3) indicates significant conformational differences for residues on the N-terminal side of residue Tyr(P) between the Tyr(P) 469 and Tyr(P) 495 complex structures and the PTP1B-hexapeptide structure. The N terminus of the hexapeptide was positioned away from the PTP1B enzyme molecule and extended into the solvent, whereas the N termini of peptides Tyr(P) 469 and Tyr(P) 495 bound much closer to the catalytic domain of SHP-1. Based on the PTP1Bhexapeptide structure, Arg 47 of PTP1B (equivalent to Arg 279 of SHP-1) was proposed to recognize and stabilize the hexapeptide through two hydrogen bonds (34). However, we propose that the different substrate specificity of PTPs is not determined by Arg 47 , because it is highly conserved among PTPs (Fig. 4). In contrast to PTP1B, which shows indiscriminate low K m and high k cat /K m toward phosphotysosyl peptides (22,23), SHP-1 exhibits much higher k cat /K m toward peptides Tyr(P) 469 and Tyr(P) 495 than other phosphotyrosyl peptides (24). The high K m values of SHP-1 toward peptide Tyr(P) 469 and Tyr(P) 495 are due to the fact that the substrate specificity of Residues labeled with asterisks are from symmetry-related molecules. In addition to the hydrogen bonds, peptides Tyr(P) 469 and Tyr(P) 495 also interacted with the catalytic domain of SHP-1 by van der Waals' interactions. Besides the Tyr(P) subpocket, we also identified six well defined substrate-binding subpockets (P-4, P-2, Pϩ1, Pϩ2, Pϩ3, and Pϩ4). The P-4 subpocket was formed mainly by residue Arg 360 from the ␤5-loop-␤6 motif. The P-2 subpocket was formed by residue Lys 362 from the ␤5-loop-␤6 motif and residues Arg 277 and Tyr 278 from loop ␣1/␤1. The Tyr(P) subpocket was formed mainly by the PTP signature motif (HCXAGXGR(S/T)) and residue Tyr 278 from loop ␣1/␤1. The Pϩ1 subpocket was formed by residues Tyr 278 , Asn 280 , and Ile 281 from loop ␣1/␤1. The Pϩ2 subpocket was formed by residue Gln 502 from the ␣5-loop-␣6 motif. The Pϩ3 subpocket was formed by residue Ser 498 from the ␣5-loop-␣6 motif, residue Ile 281 from loop ␣1/␤1, and residue Lys 259 . The Pϩ4 subpocket was formed mainly by residues Gln 491 and Arg 494 from the ␣5-loop-␣6 motif.
SHP-1 is conferred by both the catalytic domain and the two tandem SH2 domains. Subcellular relocation of SHP-1 by its SH2 domains would significantly increase the local substrate concentrations. Therefore, the SHP-1-peptide complex structures are better representations of the in vivo PTP-substrate interactions than the PTP1B-hexapeptide structure.
Comparison with Other PTPs-Residues forming the six substrate-binding subpockets other than the Tyr(P) subpocket were distributed around loop ␣1/␤1, the ␤5-loop-␤6 motif, and the ␣5-loop-␣6 motif. Because the catalytic domains of PTPs share a highly conserved three-dimensional structure, we aligned the amino acid sequences of 21 different PTPs around these three substrate-binding regions (Fig. 4). In addition to the hyper-variable ␤5-loop-␤6 motif, the Pϩ3 and Pϩ4 subpocket-forming residues in the ␣5-loop-␣6 motif were also variable in PTPs. This finding strongly suggests that Pϩ3 and Pϩ4 residues in the substrates also can be the specific recognition sites for PTPs and that the recognition is conferred by the ␣5-loop-␣6 motif of PTPs.
Residues forming the P-2 subpocket are highly conserved among PTPs, except for Yersinia PTP, suggesting that the observed residue shift at the P-2 subpocket is also present in other PTPs. Therefore, the P-2 subpocket contributes to the binding affinity of PTPs toward their substrates, without necessarily determining their substrate specificity. Residue 360, the P-4 subpocket-forming residue located in the hyper-variable ␤5-loop-␤6 motif, is either an arginine or a lysine in most PTPs. Therefore, those PTPs will prefer either an aspartate or a glutamate residue at the P-4 position of the peptide substrates. Other residues within the ␤5-loop-␤6 motif are highly variable among PTPs. These residues likely are involved in the formation of the P-4 subpocket and its interaction with further N-terminal residues of the peptide substrates. The above analyses indicate that PTP substrate specificity comes from two regions: the P-4 and further N-terminal sites and the Pϩ3 and Pϩ4 sites of the substrates. The recognition of these two regions was conferred by the ␤5-loop-␤6 and ␣5-loop-␣6 motifs, respectively.  469 and Tyr(P) 495 complex structures with the PTP1B-hexapeptide complex structure. The peptides are shown as stick models, and the catalytic domains are shown as ribbons. The Tyr(P) 469 , Tyr(P) 495 , and PTP1B complex structures are shown in yellow, blue, and gray, respectively. The only similarity between the SHP-1-peptide complex structures and PTP1B-hexapeptide structure was that residue Tyr(P) fit into the Tyr(P)-binding subpocket, and the phosphate group of residue Tyr(P) made extensive hydrogen bonds with the PTP signature motif. In the PTP1B-hexapeptide structure, the side chain of residue Leu Pϩ1 pointed in the direction of the main chain of peptides Tyr(P) 469 and Tyr(P) 495 in the SHP-1 complex structures. This directional change may have been caused by the end effects (i.e. Leu Pϩ1 is the C-terminal residue). However, at the N terminus the peptide was positioned away from the PTP1B molecule and extended into the solvent.