A single point mutation switches the specificity of group III Src homology (SH) 2 domains to that of group I SH2 domains.

Src homology 2 (SH2) domains recognize phosphotyrosine-containing sequences, and thereby mediate the association of specific signaling proteins in response to tyrosine phosphorylation (Pawson, T., and Schlessinger, J.(1993) Curr. Biol. 3, 434-442). We have shown that specific binding of SH2 domains to tyrosine-phosphorylated sites is determined by sequences adjacent to the phosphotyrosine. Based on the phosphopeptide specificity and crystal structures, SH2 domains were classified into four different groups (Songyang, Z., Shoelson, S. E., Chaudhuri, M., Gish, G., Pawson, T., Haser, W. G., King, F., Roberts, T., Ratnofsky, S., Lechleider, R. J., Neel, B. G., R. B. B., Fajardo, J. E., Chou, M. M., Hanafusa, H., Schaffhausen, B., and Cantley, L. C.(1993) Cell 72, 767-778). The [Medline] βD5 residues of SH2 domains were predicted to be critical in distinguishing these groups (Songyang, Z., Shoelson, S. E., Chaudhuri, M., Gish, G., Pawson, T., Haser, W. G., King, F., Roberts, T., Ratnofsky, S., Lechleider, R. J., Neel, B. G., R. B. B., Fajardo, J. E., Chou, M. M., Hanafusa, H., Schaffhausen, B., and Cantley, L. C.(1993) Cell 72, 767-778; Eck, M. J., Shoelson, S. E., and Harrison, S. C.(1993) Nature 362, 87-91). We report here that replacing the aliphatic residues at the βD5 positions of two Group III SH2 domains (phosphoinositide 3-kinase N-terminal SH2 domain and phospholipase C- C-terminal SH2 domain) with Tyr (as found in Group I SH2 domains) results in a switch in phosphopeptide selectivity, consistent with the specificities of Group I SH2 domains. These results establish the importance of the βD5 residue in determining specificities of SH2 domains.

Stimulation of cellular responses by growth factors and cy-tokines is often accomplished by the activation of protein-tyrosine kinases (1,6). One of the major consequences of tyrosine phosphorylation is to induce a specific set of protein-protein interactions, and thereby initiate a series of intracellular signaling cascades. The SH2 1 domains of cytosolic signaling proteins mediate the assembly of such complexes by binding to phosphotyrosine moieties within specific sequence contexts (1,6,7). Therefore, in order to understand signal transduction by protein-tyrosine kinases, it is important to decode the mechanisms by which SH2 domains achieve specificity in their recognition of phosphotyrosine sites.
The specificity of SH2 domains was first systematically studied using a degenerate phosphopeptide library (2,3). Subsequently, crystal structures of SH2 domains complexed with their high affinity ligands were obtained, providing a structural basis for phosphopeptide recognition by SH2 domains (4,5,8,9). From these studies and related experiments analyzing the in vivo binding sites of various SH2 domains, it has become evident that 3-6 residues C-terminal of phosphotyrosine dictate the specificity of SH2 domain binding. Importantly, SH2 domains can be divided into four subgroups on the basis of their specificity and primary sequences (2). For example, Group I SH2 domains (SH2 domains of Src family, Abl, Crk, GRB2, Nck SH2 domains, etc.) select the general motif Tyr(P)-hydrophilichydrophilic-hydrophobic and have an aromatic amino acid at the ␤D5 position. However, Group III SH2 domains (e.g. SH2 domains of phosphoinositide 3-kinase p85, phospholipase C-␥, and Syp/SHPTP2) select the general motif Tyr(P)-hydrophobic-X-hydrophobic and have Ile or Cys at the ␤D5 position (Fig.  1A). In the three-dimensional structures of Src and Lck SH2 domains, the ␤D5 Tyr is at the surface and makes contacts with the side chains of both the pYϩ1 and pYϩ3 residues of the bound phosphopeptide (4,5). However, in the three-dimensional structures of Syp and PLC-␥ SH2 domains (Group III), the aliphatic residue at the ␤D5 position is buried deeper in the protein, opening a hydrophobic cavity for the pYϩ1 through pYϩ6 residues (8,9) (Fig. 1B). To address the importance of the ␤D5 position, we have investigated the effect of substituting this residue on the specificity of two Group III SH2 domains.
Mapping the Specificity of SH2 Domains-The wild-type and mutant SH2 domains were expressed and purified as glutathione S-transferase (GST) fusion proteins from E. coli. The specificities of GST SH2 domains were deduced using a phosphopeptide library as described * This work was supported in part by National Institutes of Health Grant GM36624 (to L. C. C.) and grants from the Protein Engineering Network Center of Excellence and the Medical Research Council of Canada (to T. P.). 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.
Peptide Synthesis and Surface Plasmon Resonance Affinity Measurements-Peptides were synthesized using Fmoc (9-fluorenyl methoxycarbonyl) solid phase chemistry with direct incorporation of phosphotyrosine as the N ␣ -fluorenylmethoxycarbonyl-O-dimethylphosphono-Ltyrosine derivative. Cleavage of the peptide from the resin and deprotection was achieved through an 8-h incubation at 4°C in trifluoroacetic acid containing 2 M bromotrimethyl sìlane and a scavenger mixture composed of thioanisole, m-cresol, and 1,2-ethanedithiol (1.0: 0.5:0.1% by volume). The product was precipitated with cold t-butyl ethylether and collected by centrifugation. Following desalting of the crude material, pure phosphopeptide was isolated using reverse phase high performance liquid chromatography. The authenticity of the phosphopeptide was confirmed by amino acid analysis and mass spectroscopy.
Surface plasmon resonance analysis was carried out using a Biacore apparatus (Pharmacia Biosensor) as described previously (11). The phosphotyrosine-containing peptides were immobilized to a biosensor chip through injection of a 0.5 mM solution of the phosphopeptide, in 50 mM HEPES, pH 7.5, and 2 M NaCl, across the chip surface previously activated following procedures outlined by the manufacturer. Injection of anti-phosphotyrosine antibody was used to confirm that successful immobilization of the peptide was achieved. To obtain K d values for GST-SH2 domain binding to various phosphopeptides, solutions (100 l) containing varying concentrations of GST-SH2 domain fusion protein in 50 mM sodium phosphate, pH 7.5, 150 mM NaCl, 0.1 mM EDTA, and 2 mM dithiothreitol, were injected across a surface containing the immobilized phosphopeptide.
The amount of bound GST-SH2 domain was estimated from the steady-state surface plasmon resonance signal (RU). The data were analyzed using the equation where RU is the steady-state signal, [GST-SH2] is the concentration of SH2 domain, and C is a constant. Prior to each run the phosphopeptide surface of the Biosensor chip was regenerated using 2 M guanidinium HCl. Peptide inhibition experiments were performed using solutions (100 l) containing 1 M GST-SH2 domain fusion protein and the indicated concentrations of soluble phosphopeptide that were injected across a Biosensor surface to which the phosphopeptide DNDpY-EEFLPDPK was immobilized. The amount of bound GST-SH2 domain was estimated from the surface plasmon resonance signal at a fixed time following the end of the injection and the percentage bound, relative to injection of GST-SH2 domain alone, calculated.

Phosphopeptide Selectivity of Mutant Phosphoinositide 3-Kinase p85 SH2 and Phospholipase C-␥ (PLC-␥) SH2 Domains-
The p85 N-terminal SH2 (NSH2) domain and PLC-␥1 C-terminal SH2 (CSH2) domain were selected to study the importance of the ␤D5 residue for SH2 domain specificity because their three-dimensional structures, with associated high affinity phosphopeptides have been solved (8,13). 2 The wild-type p85 NSH2 domain contains a Ile (Ile-383) residue at the ␤D5 position and preferentially binds the sequence Tyr(P)-Met-X-Met (2) (Fig. 1A). The wild-type PLC-␥ CSH2 domain has a Cys (Cys-715) residue at the ␤D5 position and preferentially selects a Tyr(P)-Val/Ile-Ile-Pro motif (2) (Fig. 1B). Both these SH2 domains prefer a general motif Tyr(P)-hydrophobic-X-hydrophobic and belong to Group III SH2 domains. To test the idea that the ␤D5 residue is critical in the selectivity of SH2 domains, we replaced the ␤D5 residues of p85 NSH2 domain and PLC-␥ CSH2 domain with tyrosine, which is found in ␤D5 positions of Group I SH2 domains such as Src (Fig. 1A). The mutant SH2 domains were expressed as GST fusion proteins, and the specificities of these domains were tested using a degenerate phosphopeptide library in which all amino acids except Trp and Cys are available at the pYϩ1, pYϩ2, and pYϩ3 positions (2). As shown in Fig. 2, the conversion of Cys-715 to Tyr in the PLC-␥ CSH2 domain resulted in a distinct selectivity from that of the wild-type SH2 domain. At the three degenerate positions tested, the wild-type PLC-␥ SH2 domain selects all hydrophobic amino acids, with the optimal sequence being Tyr(P)-Val/Ile-Ile-Pro. However, the mutant PLC-␥ CSH2 domain (C715Y) binds preferentially to peptides with acidic residues, instead of hydrophobic amino acids, at the pYϩ1 and ϩ2 positions. At the pYϩ3 position, a different set of hydrophobic amino acids were selected by the mutant SH2 domain, with Phe being the preferred amino acid. Hence, the C715Y PLC-␥ CSH2 domain selects peptides with the optimal motif Tyr(P)-Glu-Glu-Phe, which resembles the general motif for Group I SH2 domains (2,3).
Conversion of the ␤D5 Ile of the p85 NSH2 domain to Tyr resulted in a domain with similar selectivity to the mutant PLC-␥ CSH2 domain (C715Y). While the wild-type p85 NSH2 domain selected for Tyr(P)-Met-X-Met, the I383Y mutant selected for Tyr(P)-Glu-Gln-Phe (Table I).
Evaluation of Phosphopeptide Binding to Wild-type and Mutant PLC-␥ CSH2 Domains-While the peptide library technique predicts optimal amino acids at specific positions Cterminal of phosphotyrosine, it does not determine the affinity for specific peptide sequences or address the complexity of cooperative effects of adjacent amino acids. To investigate the affinities of specific peptides for wild-type and mutant PLC-␥ 2 S. Harrison personal communication. CSH2 domains, phosphopeptides containing the optimal binding sequences of wild-type and mutant PLC-␥ SH2 domains were synthesized and studied using surface plasmon resonance (Table II). These peptides were constructed based on sequences surrounding Y1021 of the human platelet-derived growth factor receptor (DNDpYIIPLPDPK; named pYIIP peptide), an in vivo binding site of PLC-␥ SH2 domains. In close agreement with previous studies, the wild-type PLC-␥ CSH2 domain bound tightly to the pYIIP peptide (K d ϳ 80 nM). However, no apparent binding of this peptide to mutant PLC-␥ CSH2 domain or Src SH2 domain (a Group I SH2 domain) was detected at concentrations up to 50 M. In contrast, the mutant PLC-␥ CSH2 domain and Src SH2 domain were able to bind the phosphopeptide named pYEEF (DNDpYEEFLPDPK) with dissociation constants of approximately 398 and 1300 nM, respectively. The pYEEF peptide also bound to the wild-type PLC-␥ SH2 domain but with lower affinity (K d ϳ180 nM) than the optimal pYIIP peptide. Thus, consistent with the prediction in Fig. 2, the C715Y mutation changed the specificity of PLC-␥ CSH2 domain. These data confirm the suggestion that the ␤D5 Cys to Tyr mutant of the PLC-␥ SH2 domain behaves more like a Group I SH2 domain.  To further test the specificity of the mutant PLC-␥ CSH2 domain and the prediction of the peptide library, two additional peptides were constructed. The pYIIF (DNDpYIIFLPDPK) peptide has a Phe rather than a Pro at the pYϩ3 position, which is predicted to increase the affinity for the mutant PLC-␥ (compared to pYIIP) but lower the affinity for the wild-type protein (Fig. 2). The results in Table II (part B) are in agreement with this prediction. Similarly, substituting a Glu at the pYϩ1 position (pYEIP: DNDpYEIPLPDPK) increases the affinity for the mutant protein (compared to pYIIP) and lowers the affinity for the wild-type protein, as predicted from the results in Fig. 2.
One result in Table II (part B) is not predicted by the peptide library result. The pYEEF peptide had a significantly higher affinity than the pYEIP peptide for the wild-type PLC-␥ CSH2 domain. One possible explanation for this result is that the Glu at pYϩ1 forces the peptide out of the normal binding groove such that favorable interactions for the Ile and Pro at the pYϩ2 and pYϩ3 positions are precluded. If the peptide is on the surface of the SH2 domain then the more hydrophilic Glu at pYϩ2 would be selected (as is the case for Group I SH2 domains such as Src and Lck).
In summary, we have demonstrated here that the ␤D5 residue is crucial in determining the binding specificity of SH2 domains. We were able to switch the specificity of two group III SH2 domains (p85 NSH2 and PLC-␥ CSH2 domains) to that of Group I SH2 domains by substituting their wild-type ␤D5 residues (Ile or Cys) with tyrosine found at the ␤D5 positions of Group I SH2 domains. A comparison of three-dimensional structures of Group I (Src and Lck) and Group III SH2 domains (Syp and PLC-␥) with bound ligands indicates a major difference in ligand binding (4,5,8,9). In the structures of Src and Lck SH2 domains complexed with pYEEI peptide, only the phosphotyrosine and three residues (Glu-Glu-Ile) immediately C-terminal to it make specific contacts with the SH2 domain backbone. Among these three residues, pYϩ3 Ile plugs into a hydrophobic binding pocket while pYϩ1 and ϩ2 Glu residues form salt bridges and hydrogen-bonds with amino acids on the surfaces of the SH2 domain. The Syp and PLC-␥ SH2 domain structures, however, show a rather "open" groove configuration. The aliphatic residues at the ␤D5 position are buried in the protein such that a cavity is opened between the phosphotyrosine pocket and the pYϩ3 binding site. This allows the accommodation of hydrophobic amino acids at the pYϩ1 position (Fig. 1B). In addition, the distance between the BG and EF loops is wider than that of Src and Lck SH2 domains providing additional binding pockets for pYϩ4 and pYϩ5 residues. Therefore, up to five residues of the bound peptides are embedded in a hydrophobic channel.
The prediction of our model is that substituting a more bulky aromatic residue (Tyr) at the ␤D5 position, as found in Group I SH2 domains, will disrupt the hydrophobic cavity and force the pYϩ1 and pYϩ2 residues to lie on the surface, as found for Group I SH2 domains. The selection of the ␤D5 Tyr mutants of PLC-␥ CSH2 and p85 NSH2 for Glu residues (rather than hydrophobic residues as in the wild-type proteins) at pYϩ1 and pYϩ2 are consistent with this model. The results in Table II argue that substituting Tyr at ␤D5 did not create new contacts for the phosphopeptide but rather eliminated the selection for hydrophobic amino acids at the pYϩ1 and pYϩ2 positions. This conclusion is supported by the observation that the optimal phosphopeptide for the PLC-␥ CSH2 mutant (pYEEF) had a slightly higher affinity for wild-type PLC-␥ CSH2 domain than for the mutant. As discussed above, we suspect that this is due to the ability of the pYEEF peptide to bind in a conformation in which the glutamate residues at pYϩ1 and pYϩ2 are on the surface of the SH2 domains. However, this peptide does not have as high affinity as the pYIIP peptide for wild-type PLC-␥ CSH2.
The data presented here, together with our earlier work involving mutations in the Src SH2 domain (10) and p85 NSH2 domain (12), indicate the potential to modulate the specificity of SH2 domains by making substitutions at sites predicted to bind side chains of associated phosphopeptides. This approach deepens our understanding of SH2-binding specificity. Furthermore, it should be feasible in the nearest future to design novel SH2 domains with pharmacological applications.