Research ArticleBiochemistry

Superbinder SH2 Domains Act as Antagonists of Cell Signaling

See allHide authors and affiliations

Science Signaling  25 Sep 2012:
Vol. 5, Issue 243, pp. ra68
DOI: 10.1126/scisignal.2003021

Abstract

Protein-ligand interactions mediated by modular domains, which often play important roles in regulating cellular functions, are generally of moderate affinities. We examined the Src homology 2 (SH2) domain, a modular domain that recognizes phosphorylated tyrosine (pTyr) residues, to investigate how the binding affinity of a modular domain for its ligand influences the structure and cellular function of the protein. We used the phage display method to perform directed evolution of the pTyr-binding residues in the SH2 domain of the tyrosine kinase Fyn and identified three amino acid substitutions that critically affected binding. We generated three SH2 domain triple-point mutants that were “superbinders” with much higher affinities for pTyr-containing peptides than the natural domain. Crystallographic analysis of one of these superbinders revealed that the superbinder SH2 domain recognized the pTyr moiety in a bipartite binding mode: A hydrophobic surface encompassed the phenyl ring, and a positively charged site engaged the phosphate. When expressed in mammalian cells, the superbinder SH2 domains blocked epidermal growth factor receptor signaling and inhibited anchorage-independent cell proliferation, suggesting that pTyr superbinders might be explored for therapeutic applications and useful as biological research tools. Although the SH2 domain fold can support much higher affinity for its ligand than is observed in nature, our results suggest that natural SH2 domains are not optimized for ligand binding but for specificity and flexibility, which are likely properties important for their function in signaling and regulatory processes.

Introduction

Cellular signal transduction relies on regulated, coordinated, and dynamic protein-protein interactions (1, 2). Once a signal is terminated, the interacting signaling proteins dissociate in a timely manner to allow cells to return to a basal state poised to respond to the next stimulus. These dynamic protein-protein interactions are frequently mediated by modular domains, many of which are capable of binding to defined sequence motifs containing posttranslationally modified amino acids (3). One example of interaction domains of this type is the Src homology 2 (SH2) domain, a protein module of about 100 amino acids, first identified in the oncoprotein v-Src (4) and subsequently found in numerous other proteins, including 111 human proteins (5). All known SH2 domain structures conform to a conserved domain fold comprising a central β sheet flanked by two α helices (6, 7) (see Fig. 1).

Fig. 1

Locations of the 15 pTyr-binding pocket residues on the SH2 domain. (A) Amino acid sequences and secondary structure elements of the Fyn and Src SH2 domains. The 15 residues that form the pTyr-binding pocket are renumbered from 1 to 15 for convenience. The three Cys residues in the Fyn SH2 domain were replaced with Ser (colored red) without affecting the structure or ligand-binding properties (26, 56). The human Src SH2 domain is shown for comparison. Asterisks indicate identical residues. (B) The Fyn SH2 domain in complex with the MidT-pTyr324 peptide [PDB: 1AOT (26)]. The SH2 domain is shown as ribbons with the 15 pocket residues as magenta spheres. Square labels indicate secondary structure elements. The peptide is depicted as a stick model. For clarity, only the four-residue stretch pTyr-Glu-Glu-Ile of the bound peptide is drawn. The Fyn SH2 domain has positional selectivity for the pTyr +3 residue (Ile in the peptide), conferred by a hydrophobic pocket located between the EF and the BG loops.

In general, SH2 domains bind only to targets containing phosphotyrosine (pTyr) (810). The conserved pTyr-binding pocket of the SH2 domain provides the fundamental specificity for ligands with a pTyr and basal affinity for binding. Most SH2 domains display preference for a recognition sequence C terminal to pTyr in the ligand (1113), conferred by a second pocket, or subsite, which enhances specificity (6, 14, 15). Despite this seemingly optimal two-pronged binding mode of ligand recognition, a typical SH2-pTyr ligand interaction is moderate in strength. Affinity can be quantified by determining the equilibrium dissociation constant (Kd) of a binding reaction. A low Kd indicates a high affinity. A typical SH2-pTyr ligand interaction has a Kd in the 0.1 to 10 μM range (1618). This value is orders of magnitude higher than for a typical antibody-antigen interaction, indicating lower affinity. Although it is conceivable that moderate affinity may be required for transient and dynamic protein-protein interactions occurring in a cell, how this is achieved at the molecular level is not fully understood. From another perspective, it is not clear why the SH2 domain and other modular interaction domains have moderate affinities for their cognate ligands (19, 20). Is the moderate affinity required by cellular functions or dictated by the domain structure? Using the SH2 domain as an example, we addressed whether the SH2 fold can support stronger binding to its ligand than observed in nature.

Because the pTyr residue contributes about half of the binding free energy (ΔG°) of a ligand to an SH2 domain (7), we wanted to test whether the pTyr-binding pocket on an SH2 domain may be optimized to allow high-affinity binding. Because the pTyr-binding pocket in a typical SH2 domain comprises a dozen or more residues, making it impractical to conduct systematic mutagenesis on a per-residue basis, we adopted a phage display approach to generate a mutant SH2 domain with improved affinity through directed evolution. Specifically, we created a library of mutant SH2 domains from the tyrosine kinase Fyn in which the pTyr-binding pocket residues were randomized, displayed the library on the M13 bacteriophage, and screened the library for binding to a series of pTyr peptides derived from physiological phosphorylation sites. Of the numerous clones recovered, one variant SH2 domain that contained substitutions of three residues displayed greatly enhanced affinities for the pTyr peptides. This pTyr “superbinder” exhibited submicromolar Kd values for the GGpYGG peptide, which gives an indication of its affinity for the pTyr moiety alone because no side-chain contribution to binding is expected from a Gly residue (7). To the best of our knowledge, no natural SH2 domain has been shown to bind to the pTyr moiety with a micromolar affinity to date. Equivalent residue substitutions, when introduced into the SH2 domain framework of a tyrosine kinase, Src, or an adaptor protein, Grb2, yielded variants with similarly enhanced affinities. These pTyr superbinders, when expressed in mammalian cells, inhibited epidermal growth factor receptor (EGFR) signaling and blocked anchorage-independent cell growth. The SH2 fold is thus capable of supporting much stronger binding to its ligand than observed in nature, and the superbinder SH2 domains obtained by directed in vitro evolution may be used as potential protein therapeutic agents for cancer.

Results

The pTyr-binding pocket is evolvable

The two-pronged binding mode for most SH2 domains makes it possible to modulate affinity and specificity separately using directed evolution. We tested the evolvability (the ability to evolve under selective pressure) of the pTyr-binding pocket residues by creating a phage-displayed SH2 domain library (>1010 unique phage clones). Fifteen residues that form the pTyr-binding pocket of the Fyn SH2 domain were selected on the basis of distances between their side chains and the pTyr side chain of the ligand in the SH2-peptide complex structures (table S1). These residues are renumbered from 1 to 15 for convenience (Fig. 1). Throughout the text, residue numbers based on the absolute position of the residue in a protein sequence will be superscripted, but residue numbers indicating the identity of the 15 residues that underwent directed evolution will not be superscripted. Although the tyrosine residue between positions 14 and 15 also contributes to formation of the pTyr-binding pocket, we did not apply randomization to this position because it also forms part of a specificity pocket for the recognition of a C-terminal residue to the pTyr (14, 15, 21). By sparing residues that are involved in specificity determination, we hoped to obtain a clear picture of how residues that are dedicated to pTyr recognition might be evolved to allow tighter binding to a ligand. To this end, the pocket residues were systematically altered by the soft randomization mutagenesis technique (22), which allows for 50% mutation rate at each position (Fig. 1). The phage-displayed SH2 domain library was then subjected to independent binding selections against 33 pTyr-containing peptides that represent all major specificity groups (peptides that have a specificity-determining residue at the second, third, or fourth residue C terminal to the pTyr) (6, 13), including those corresponding to the physiological pTyr sites within the receptor tyrosine kinases EGFR and ErbB4 (23, 24) (table S2). The selections yielded 63 unique variants, but no wild-type Fyn SH2 domain sequence was recovered (Fig. 2A and fig. S1). This suggested that the pTyr-binding pocket is not optimized for high affinity, and moreover, the pocket is highly evolvable.

Fig. 2

Directed evolution of the pTyr-binding pocket of an SH2 domain by phage display. (A) Screening of pTyr-binding pocket phage display library with immobilized peptides yielded 63 unique SH2 variants. The wild-type residues are shaded magenta at the top. Residues different from the wild type in each variant are identified in green. (B) The pTyr-binding pocket of the Fyn SH2 domain. The upper panel is the domain surface representation, whereas the lower panel shows the SH2 domain backbone with the 15 pocket residues as spheres. The pTyr-binding pocket residues are colored as a magenta-to-green gradient according to substitution frequency. Orientation of the structure is the same as in Fig. 1B. (C) Substitution frequency at each position for the 63 variants. (D) A sequence logo (74) generated from the 63 SH2 variants. The corresponding positions of the 15 pocket-forming residues are shown below the logo. (E) A sequence logo generated for the 26 variants that contain an Ile or Val at position 8 to show its coevolution with position 15, which is frequently replaced by a Leu.

Different combinations of substitutions were obtained for the pocket-forming residues, but not all positions were equally evolvable. Some positions, such as Ala3, Arg4, Leu11, and Ser12, were resistant to substitution (Fig. 2, B to D). Arg4, corresponding to the βB5 residue, is absolutely conserved in natural human SH2 domains (fig. S2) and is indispensable for pTyr binding (8, 25). Leu11 is involved in the formation of the hydrophobic core of the SH2 domain (26). In contrast, other pocket residues, including Thr7, Thr8, Ser10, and Lys15, are highly evolvable (Fig. 2, B to D). Most variants isolated contain two to five substitutions (fig. S1B).

Substitutions of three pocket residues led to a pTyr superbinder

Because the phage selections were based on affinity, we predicted that the isolated SH2 variants would exhibit enhanced affinities for pTyr peptides. Variants that contain mutations at positions Thr8, Ser10, and Lys15 were particularly interesting because these three positions showed the highest substitution frequencies (Fig. 2C). Moreover, non-native residues were more frequently observed than the native ones for these positions (Fig. 2D). These hydrophilic residues were substituted with hydrophobic ones in most of the variants identified. For example, 42 of 63 variants contained a Leu or an Ile instead of a Lys at position 15, suggesting that the wild-type residue (Lys) is not optimized for binding. We found that substitutions at positions 8, 10, and 15 were highly correlated. Variants that contained an Ile or a Val residue at position 8 favored a Leu at position 15 and a Ser, Ala, or Val at position 10, suggesting that these positions coevolved to favor strong binding to ligands (Fig. 2E). A single Cys10Ala substitution within the v-Src SH2 domain results in a sevenfold increase in affinity for a pTyr-containing peptide (8). Our phage display screening results suggested that further affinity improvement may be achieved by combining multiple substitutions in the pTyr-binding pocket of the Fyn SH2 domain.

To investigate how the most frequently observed mutations (at positions 8, 10, and 15) affect pTyr binding, we prepared Fyn SH2 domain mutants in which the three positions were mutated individually or in different combinations. We measured binding of the resulting mutants to a panel of eight pTyr-containing peptides by fluorescence polarization. The affinities of the mutants for pTyr peptides (table S3) were progressively augmented by the single, double, and triple substitutions (Table 1 and table S4), suggesting that the substitutions enhance binding in an additive or synergistic manner. The most marked improvement in affinity was observed when all three substitutions were combined in the triple-mutant Thr8Val/Ser10Ala/Lys15Leu. This mutant exhibited superbinding characteristics for pTyr peptides with affinities ten to hundred times greater than the wild-type domain. For example, although the wild-type domain binds with moderate affinity to the EGFR-pTyr978 peptide (Kd = 3.7 μM), the triple mutant binds 380-fold tighter (Kd = 0.0097 μM) (Fig. 3A). Notably, the triple mutant also exhibited high affinity for the pTyr moiety itself. Whereas the wild-type domain bound only weakly to the GGpYGG peptide, the triple mutant binds it with a submicromolar Kd of 0.71 μM (Fig. 3B). The observed high affinity for the mutant was phosphorylation-dependent, as the unphosphorylated peptide failed to bind (Fig. 3B).

Table 1

Binding affinity of the wild-type or mutant Fyn SH2 domains to a panel of pTyr peptides labeled with fluorescein as measured by fluorescence polarization assay. Kd values are shown in the micromolar unit. See table S4 for curve fitting statistics. The Kd values with >20-fold increase in affinity compared to the wild type are highlighted with bold fonts. The variant numbers are found in Fig. 2A.

View this table:
Fig. 3

Triple-mutant SH2 domains exhibit markedly enhanced affinities for phosphotyrosyl ligands. Affinities were measured by fluorescence polarization (ΔFP). The corresponding Kd values are shown below the binding curves. (A) Binding of the wild-type or mutant Fyn SH2 domain to an EGFR-derived pTyr978 peptide. (B) Binding of the wild-type or triple-mutant Fyn SH2 domain to the GGpYGG peptide or the nonphosphorylated GGYGG peptide. (C) Binding of the wild-type or triple-mutant Src SH2 domain to the ShcA pTyr239 peptide.

Because the pTyr-binding pocket residue composition is highly conserved within the SH2 domain family (6) (fig. S2), we tested whether the same triple substitutions would enhance affinity in other SH2 domains. Introduction of the equivalent substitutions, either individually or in combination, into the Src SH2 domain produced mutants that display a similar trend of affinity enhancement, with the triple mutant exhibiting an average affinity increase of 240-fold for a panel of seven peptides (Table 2 and table S5). The tightest binding (Kd = 0.0038 μM) was observed between the triple-mutant Src SH2 domain and the peptides derived from the pTyr239 site of the adaptor protein ShcA (ShcA-pTyr239) (27) and the pTyr324 site of the hamster polyomavirus middle T antigen (MidT-pTyr324) (28) (Table 2 and Fig. 3C). In sum, these results demonstrate that the affinity-enhancing effect of the substitutions identified in the context of the Fyn SH2 domain can be conferred on the Src SH2 domain, yielding similar Kd values for the Fyn and the Src SH2 domain mutants (Fig. 4A).

Table 2

Binding affinity of the wild-type or mutant Src SH2 domains to a panel of pTyr peptides. Kd values are shown in the micromolar unit. See table S5 for curve fitting statistics. See also notes on Table 1.

View this table:
Fig. 4

Synergistic contributions of the substituted amino acids in the triple mutant to pTyr binding. (A) A plot of Kd values derived from the triple-mutant Fyn and Src SH2 domains, binding to the same set of eight peptides. See also Tables 1 and 2 for Kd data. (B) Binding free energy contribution of substitutions at positions 8, 10, and 15 to pTyr peptide binding. See Materials and Methods for calculation of the ΔΔG° values. The GGpYGG peptide data were excluded because the binding was too weak to derive Kd values for some mutants. Average ΔΔG° values calculated from the data are shown in the gray box. (C) Additive and synergistic effects for the three substitutions toward ligand binding. Boxes show average ΔΔG° values for the Fyn SH2 domain mutants, and arrows show relationships between the substitution combinations. Thin arrows indicate apparent additive contributions, and thick arrows indicate apparent synergistic effects.

Although our study focused on the in vitro evolution of affinity that led to the isolation of a pTyr superbinder, we also wanted to ask whether enhanced affinity for an SH2 domain would compromise its specificity. To address this question, we used the wild-type and triple-mutant Fyn SH2 domains to pull down tyrosine phosphorylated proteins from lysates of mammalian cells (fig. S3A). It is apparent that the mutant pulled down more phosphoproteins than the wild-type SH2 domain likely due to increased affinity, rather than specificity, of the mutant domain. This assertion is supported by the general correlation (R2 = 0.7) in ΔG°, derived from Kd values, of the wild-type and triple-mutant Fyn SH2 domain for a group of pTyr peptides (fig. S3B). Thus, the wild-type and triple-mutant Fyn SH2 domains have similar relative specificity toward pTyr peptides despite large differences in affinity.

The three substitutions in the pTyr-binding pocket act synergistically to increase affinity

To understand the physicochemical basis underlying the increased affinities of the mutant Fyn and Src SH2 domains, we calculated the difference in ΔG° between the wild type and a mutant for binding to each peptide (ΔΔG°) (see Materials and Methods) (Fig. 4B). This analysis suggests that, although the three substitutions play different roles in pTyr binding, they function in a synergistic manner to achieve high-affinity ligand binding. Individually, the Lys15Leu and Ser10Ala substitutions, in particular the former, of the Fyn SH2 domain contributed significantly to binding with average ΔΔG° values of −1.4 and −1.0 kcal/mol, respectively. In contrast, the single Thr8Val substitution alone showed little effect on binding with the average ΔΔG° = −0.1 kcal/mol. When paired with the Lys15Leu substitution, the resulting Thr8Val/Lys15Leu double mutant binds to pTyr peptides with an average reduction of 2.2 kcal/mol in the binding free energy compared to the wild-type SH2 domain (Fig. 4, B and C). This indicates that the two substitutions act synergistically to reduce the binding free energy of the mutant-pTyr interaction. In line with these data, the phage display library screens showed that the evolution of positions 8 and 15 are highly correlated (Fig. 2E). Of the 120 human SH2 domains (13, 29) (fig. S2), most (89 in total, including the Src, Fyn, and Grb2 SH2 domains) have a five-residue BC loop, but none contains the optimal triple hydrophobic combinations at positions 8, 10, and 15 of the pTyr-binding pocket, nor the double combinations obtained from our in vitro evolution (fig. S4).

The triple substitutions form a novel hydrophobic surface to engage the phenyl ring of the pTyr residue

To understand the structural basis underpinning the marked affinity enhancement exhibited by the triple-mutant SH2 domains, we determined the crystal structures of the Src SH2 triple mutant in the apo, phosphate ion-bound, and pTyr-bound forms (Fig. 5A, table S6, and figs. S5A and S6). The structure of the triple mutant–pTyr complex revealed that the three substituted residues were located on one side of the pTyr-binding pocket and formed an extended hydrophobic surface to engage the phenyl ring of the pTyr moiety (Fig. 5, A and B). The side chain of Leu15 is aligned with the pTyr aromatic ring, and the two residues form extensive hydrophobic contacts. Cys10 of the wild-type Src SH2 domain has been suggested to be slightly detrimental to ligand binding due to its bulkiness, charge, or both (8, 30). Its replacement by the smaller, apolar alanine may alleviate this effect and enhance pTyr binding. The double substitutions at positions 10 and 15 showed an approximately additive free energy contribution of the corresponding single substitutions for the Fyn SH2 domain (Fig. 4, B and C).

Fig. 5

Structural basis for the superbinding properties exhibited by the triple-mutant Src SH2 domain. (A) The pTyr-binding pocket, shown as a surface representation, from the crystal structure of the triple-mutant Src SH2 domain bound to pTyr, determined at a resolution of 1.57 Å. Carbon atoms of the three substituted residues are colored green, oxygen in red, and nitrogen in blue. The pTyr is shown as a stick representation (carbon atoms in yellow). The FoFc omit map for the pTyr is shown as a brown mesh, contoured at 3.5 σ. (B) The two faces of pTyr recognition by the triple-mutant Src SH2 domain. On one face, the phenyl ring of pTyr makes hydrophobic contacts with residues Val8, Ala10, and Leu15 (side-chain carbon atoms are drawn as green space-fill spheres). On the other face, the phosphate moiety forms electrostatic interactions or hydrogen bonds with Arg1 (αA2), Arg4 (βB5), Ser5 (βB7), Glu6 (BC1), and Thr7 (BC2). (C and D) A closeup view of the β-turn hydrogen bond in the BC loop, formed between the carbonyl oxygen of BC3 and the amide nitrogen of βC1 (nonhydrogen atomic distances shown in the panels). Residues at positions 8, 10, and 15 are colored green. (C) The triple-mutant Src SH2 domain in complex with pTyr (yellow). (D) The wild-type human Src kinase bound to its phosphorylated tail [PDB: 1FMK (55)]. (E) Superimposition of the two structures shown in (C) and (D). Residues 8, 10, and 15 of the pTyr-binding pocket are depicted as sticks and colored green or pink in the mutant or wild-type structure, respectively.

The triple substitutions provided a stably molded hydrophobic curvature to optimally capture one side of the pTyr while leaving essential polar contacts intact on its other side (Fig. 5B). The pTyr was drawn closer to the hydrophobic surface formed by the three substituted residues in the triple-mutant than in the wild-type SH2 domain (Fig. 5, C to E). The triple-mutant SH2 domain bound to the pTyr moiety in a double-faced binding mode: One face was hydrophobic, and the other was polar (Fig. 5B). A hydrophobic patch formed by Val8, Ala10, and Leu15 engaged the aromatic ring with hydrophobic interactions, whereas the polar residues on the opposite side tethered the phosphate group in place with electrostatic interactions and hydrogen bonds (Fig. 5B and fig. S6C).

Given the synergistic effect of the Thr8Val and Lys15Leu substitutions (Fig. 4C), the main role of Val8 might be to stabilize and optimize the positioning of the Leu15 side chain through a hydrophobic interaction (Fig. 5). This appeared to pull the BC loop residue Val8 closer to Leu15 (βD6). Compared to the wild-type structure, Val8 (BC3) in the mutant was shifted inward to allow the formation of a more compact β-turn conformation in the BC loop (Fig. 5, C to E). The crystallographic B-factor distribution for the BC loop region of the triple-mutant structures implied that the BC loop in the mutant was less flexible than in the wild-type SH2 domain (fig. S5B). Together, the triple substitutions in the superbinder SH2 mutant use optimized hydrophobic contacts and reduced entropy for the free SH2 domain to achieve enhanced affinities for the pTyr residue.

The pTyr superbinders effectively inhibited cell signaling and growth

Dysregulation in the expression or activity of EGFR family members is associated with many epithelial cancers (3133). The markedly enhanced affinities of the triple-mutant Fyn and Src SH2 domains for the EGFR-derived peptides implied that they may function as inhibitors of EGFR signaling and EGF-dependent cell growth. To test this hypothesis, we expressed the wild-type or triple-mutant Fyn or Src SH2 domain in human embryonic kidney (HEK) 293 cells, a transformed cell line that expresses EGFR (34). Although neither wild-type SH2 domain bound to the EGFR, both triple mutants did (Fig. 6A, upper panel). Because Grb2 is a critical adaptor protein in EGFR signaling (35), we also created the triple-mutant Grb2 SH2 domain (see Materials and Methods), which bound substantially tighter to EGFR than the wild-type domain (Fig. 6A, upper panel). Moreover, all three mutant SH2 domains reduced substantially the phosphorylation of extracellular signal–regulated kinase (ERK), a downstream kinase of the EGFR signaling pathway, whereas the wild-type domains had no effect (Fig. 6A, lower panel). This indicated that the triple-mutant SH2 domains inhibited the mitogen-activated protein kinase (MAPK) pathway, likely by competing for binding to the pTyr sites on either the EGFR receptor or an adaptor or both. The SH2 mutants showed strong binding (nanomolar affinity) to pTyr peptides of the adaptor protein ShcA (Fig. 4A), which is immediately downstream of EGFR and upstream of ERK (27, 36). In agreement with these biochemical data, cells expressing the triple-mutant SH2 domains are less viable than those expressing a wild-type SH2 domain (Fig. 6B). All three SH2 mutants, but not the wild-type counterparts, effectively blocked colony formation of HEK293 cells in soft agar (Fig. 6, C and D). Because anchorage-independent cell growth is a hallmark of tumorigenicity (37), our results suggest that pTyr superbinders such as the triple-mutant SH2 domains may be potential anticancer agents.

Fig. 6

Inhibition of EGFR signaling and cell growth by the superbinder SH2 domains. HEK293 cells expressing a wild-type (Wt) or a triple-mutant (TrM) Fyn, Grb2, or Src SH2 domain (as GFP fusions) were subjected to Western blotting and cell growth assays. (A) The upper four panels show reciprocal immunoprecipitation of the SH2 domains and EGFR. IP, immunoprecipitation; IB, immunoblot. The lower two panels show the presence or absence of phosphorylated and total ERK in whole-cell lysates with the indicated SH2 domains. WCL, whole-cell lysate. Data shown are representative of three independent experiments. (B) The effects of triple-mutant SH2 domains on cell viability. The percentages of viable cells expressing a wild-type (open column) or mutant (filled column) SH2 domain relative to viable cells transfected with the pEGFP empty vector as a control (set at 100%) are shown here. The P value is less than 0.005 for each wild type–mutant pair (Student’s t test, two-tailed, n = 5). Error bars indicate the SD. (C) The effects of triple-mutant SH2 domains on anchorage-independent cell growth. The percentages of colonies formed in soft agar by cells expressing a wild-type (open column) or a mutant (filled column) SH2 domain relative to the empty vector control (set at 100%) are shown. Error bars indicate the SD (n = 5). (D) Representative images of soft agar plates showing colonies formed by cells expressing a wild-type or a mutant SH2 domain.

Discussion

Our study demonstrated that the SH2 domain fold can support much higher affinities for pTyr targets than seen in natural SH2 domains. By introducing three residue substitutions into the pTyr-binding pocket, we created pTyr superbinders using three different SH2 domains as backbones. These superbinders have markedly enhanced affinities for the pTyr moiety and physiological pTyr-containing peptides when compared to wild type. Our observation that these superbinders, but not the corresponding wild-type SH2 domains, effectively blocked EGFR signaling and inhibited cell growth suggested that unnaturally tight binding between an SH2 domain and its physiological targets was detrimental to the cell. That is, natural SH2 domains are likely optimized for function, not affinity. Why does nature favor moderate over high affinity for modular domain-ligand interactions? One plausible explanation is that nature may favor specificity and flexibility over tight binding to mediate the dynamic protein-protein interactions required for physiological regulatory processes. None of natural human SH2 domain sequences contained triple- or double-residue combinations that contribute to affinity increase (fig. S4). This implied that human cellular signaling systems may avoid such combinations to maintain moderate affinities that preserve the dynamic nature of the physiological interactions between SH2 domains and pTyr ligands (1, 38). It is likely that other modular domain scaffolds, such as the SH3 and PDZ domains (39, 40), may also be able to support much greater affinities for their cognate ligands than displayed by the natural domains.

Our observation that the pTyr superbinders, but not the wild-type SH2 domains, effectively inhibited EGFR signaling and cell growth suggests that a delicate balance between specificity and affinity needs to be struck for normal cellular function. Given the complexity of the pTyr signaling network, finding such a balancing point would be a difficult task—but one that is solved effortlessly by the cell. A human cell may contain 120 SH2 domains (5, 29), 90 tyrosine kinases (41), and more than 10,000 tyrosine phosphorylation sites (42, 43), which together form a complex regulatory network. For an SH2 domain in this signaling network, specificity is regarded as its ability to distinguish between physiological and nonphysiological binding partners, even if both contain a pTyr residue. The optimal affinity-specificity balance may, in part, be determined by the degree of network complexity for which an increase in the number of nodes (for example, SH2 domains or pTyr sites) leads to greater complexity (44). The pTyr signaling network of metazoans generally shows a trend of increasing complexity during the course of evolution (4446).

It has also been proposed that the specificity of domain-ligand recognition is optimized to minimize cross talk in a network (47). The tyrosine content of the proteome in a species is negatively correlated with the number of tyrosine kinases in that species (48), suggesting an evolutionary strategy to systematically attenuate the complexity of pTyr signaling network. Although other factors in the cell such as the spatiotemporal patterns of protein expression, subcellular localization, and scaffolding may contribute to maintain specificity in pTyr signaling (1), tipping the balance of specificity and affinity of a component in the network may have dire consequences for the cell, as we have shown here using the pTyr superbinders. The triple-mutant Fyn and Src SH2 domains gained the capacity to bind EGFR (Fig. 6A), which is not considered to be a physiological target of the corresponding wild-type SH2 domains. However, when SH2 domains gain the capacity to bind more proteins, this could lead to a traffic jam in the pTyr signaling network (fig. S3A).

At least two additional factors besides competition for physiological interactions need to be considered to interpret the data on the inhibitory effect of the pTyr superbinders on cell signaling and growth: loss of dynamics in physiological SH2-ligand interactions because of enhanced affinity and loss of specificity (the ability to discriminate between physiological and nonphysiological partners). Similar to other posttranslational modifications (3), tyrosine phosphorylation is a spatiotemporally controlled signaling event (1, 38). Enhanced SH2-ligand interaction that prevents dissociation could stall signal transmission. Increased affinity could result in unproductive or unintended binding to nonphysiological partners, which may derail the normal course of cellular signaling and may even cause detrimental consequences to cells (Fig. 6, B to D). A recent report on the interaction between phosphorylated bacterial effector proteins and host SH2 domains revealed an alternative strategy, found in nature, for disrupting the pTyr signaling network (49). Bacterial pTyr sites share a unique motif, not found in mammalian proteomes, which is capable of recruiting many human (host) SH2 domains. These bacterial pTyr sites may have evolved to be promiscuous, such that they intercept multiple mammalian host signaling pathways upon infection of the host cell.

Because the pTyr-binding mode is conserved in the SH2 domain family, one may be able to apply the principles obtained from the directed in vitro evolution of the Fyn SH2 domain to other family members, as shown herein for the Src and Grb2 SH2 domains, to create a variety of pTyr superbinders. As we have demonstrated here for the SH2 domain family, combinatorial evolution of the ligand-binding residues in a modular domain may be used not only to understand the molecular basis of domain-ligand interaction but also to derive superbinders with potential therapeutic applications. Given that the pTyr side chain accounts for about a half of the total binding free energy between a natural SH2 domain and a pTyr-containing peptide ligand (7), we predict that further improvement in affinity is attainable by combining optimization of the pTyr-binding pocket, for example, through directed evolution of the BC loop sequence, with functional selection of SH2 domain variants containing an optimized specificity pocket for binding to the pTyr +2, +3, or +4 residue of the peptide ligand. This approach may also yield SH2 domain variants with desired affinity and specificity to be used in dissecting cellular signaling events and in synthetic biology in combination with other modular domains and their binding motifs (50, 51). Because aberrant tyrosine phosphorylation is a hallmark of many cancers (44, 52, 53), SH2 superbinders bear great promise as antagonists of tyrosine kinase signaling and thereby potential diagnostic and therapeutic agents.

Materials and Methods

Peptide synthesis

Peptides were synthesized with the MultiPep synthesizer (Intavis) on the TentaGel amide resin (Intavis) by 9-fluorenyl methoxycarbonyl chemistry. The sequences of peptides synthesized are listed in tables S2 and S3. Unless otherwise noted, all peptides were derived from human proteins. A biotin moiety was coupled to the N terminus of a peptide through a linker consisting of a pair of 6-aminohexanoic acids. N-hydroxysuccinimide–fluorescein was used for N-terminal fluorescein labeling.

Phage display and library screening

To choose pTyr-binding pocket residues to randomize, we calculated non–hydrogen atom minimal distances between side chains of an SH2 domain residue and the pTyr with the program Crystallography and NMR system (CNS) (54). Three complex structures of the Src family SH2 domains from Fyn [PDB: 1AOT (26)], Src [PDB: 1FMK (55)], and Lck [PDB: 1LCJ (14)] were used for the calculation, from which 15 residues were picked (table S1). The βD5 residue was not included in the phage display randomization because it constitutes a part of the wall for the specificity pocket (21). For the phage display experiments, a highly diverse library (diversity greater than 1010) of the Fyn SH2 domain with the 15 pTyr-binding pocket residues randomized was screened for binding to immobilized pTyr peptides (table S2). The template human Fyn SH2 domain (Val138-Gly249) was subcloned into the pFN-OM6 vector. The three cysteines in the domain were replaced with serine (Fig. 1A) without compromising the domain structure and peptide binding properties (26, 56). A soft randomization technique that allows for 50% substitution rate at each residue was applied (22, 57). The phage pool was incubated with a peptide immobilized through biotin on a well of a 96-well plate, and the unbound phages were washed off. The bound phages were eluted and amplified by infection of Escherichia coli XL1-Blue cells, and the amplified pools were used for further rounds of selection (58). The process was repeated for five rounds to enrich the bound phages. Individual phage clones were separated, and the phage clone–peptide interactions were further verified by phage enzyme-linked immunosorbent assay (ELISA). The positive clones were subjected to DNA sequencing analysis to determine the sequences of the displayed SH2 domain variants.

Protein expression in E. coli, purification, and binding assays

The Kunkel mutagenesis method (57) was used to prepare mutant Fyn SH2 domain constructs used for the binding assay. Wild-type or a mutant Fyn SH2 domain was expressed as hexahistidine-tagged proteins in E. coli BL21 (DE3) cells and purified on Ni-NTA beads (Qiagen) according to the manufacturer’s instructions. The eluted protein samples were further purified on a Superdex 75 size exclusion column (GE Healthcare) with buffer composed of 20 mM tris-HCl (pH 7.0) and 150 mM NaCl.

The gene encoding the human Src SH2 domain (Asp144-Lys252) was subcloned into the pETM11 vector (59). The QuikChange II mutagenesis kit (Stratagene) was used to construct the mutant Src SH2 domains. The wild-type and mutant Src SH2 domains were expressed as a hexahistidine-tagged fusion protein. The protein was purified with Ni-NTA beads and dialyzed overnight against 20 mM tris-HCl (pH 7.0), 50 mM NaCl, 0.5 mM EDTA, and 1 mM dithiothreitol (DTT) at 4°C. The hexahistidine tag was then cleaved by the tobacco etch virus protease (60), overnight at room temperature. The sample was further purified on a Superdex 75 size exclusion column (GE Healthcare) with buffer composed of 20 mM tris-HCl (pH 7.0), 150 mM NaCl, and 1 mM DTT.

Fluorescence polarization assays were performed as previously described (6). Titration data points obtained from two independent measurements were used for curve fitting with a single-site binding model. The SE of the mean values was less than 30% of the Kd values for all in-solution binding data presented. The individual Kd and error values are shown in tables S4 and S5. The ΔΔG° values were derived as ΔΔG° = ΔG°Mutant − ΔG°Wt = RTln (Kd[Mutant]/Kd[Wt]), where Kd[Mutant] and Kd[Wt] are the Kd values of the mutant and wild-type SH2 domains, respectively.

Crystallographic analysis of the Src SH2 triple mutant

Crystals of the triple-mutant Src SH2 domain without a bound ligand were initially obtained from the PACT Suite crystallization screening kit (Qiagen) by the sitting drop vapor diffusion method. The optimized crystals were obtained in a buffer containing 0.1 M tris-HCl (pH 7.6), 15 to 21% polyethylene glycol (PEG) 6000 or PEG 8000, and 0.2 M lithium chloride after incubation for 10 days at 5° or 22°C. The apo form crystal was soaked for 15 min in the cryoprotectant composed of 0.1 M tris-HCl (pH 7.6), 21% PEG 8000, 0.2 M lithium chloride, and 14% PEG 400. The phosphate ion complex structure was obtained by soaking a crystal for 40 min in the cryoprotectant that contained 0.1 M tris-HCl (pH 7.6), 20% PEG 6000, 15% PEG 400, and 10 mM sodium phosphate. The pTyr complex structure was obtained by soaking a crystal for 40 hours in the cryoprotectant that contained 0.1 M tris-HCl (pH 7.6), 20% PEG 6000, 15% PEG 400, and 6.7 mM pTyr (Bachem Inc.). The crystals were flash-frozen under nitrogen gas at 114 K for data collection on a RUH3R x-ray generator (Rigaku) and the mar345 detector (MarResearch). The data sets were processed with iMosflm (61) and Scala (62). The initial phases for the apo structure were determined by molecular replacement with MOLREP (63), with PDB 1O43 (64) as the search model. The crystal contains one SH2 domain per asymmetric unit. The structures were modeled and refined with CNS (54), Refmac5 (65), LAFIRE (66), Coot (67), and the utility programs of the CCP4 suite (68). The coordinates were validated with PROCHECK (69) and NQ-Flipper (70). See table S6 for crystallographic statistics. There were no outlier residues in the Ramachandran plot. The structure figures were generated with MacPyMOL (DeLano Scientific).

Antibodies and SH2 domain constructs for expression in mammalian cells

A rabbit polyclonal antibody to green fluorescent protein (GFP) was purchased from Sigma-Aldrich. The rabbit polyclonal antibody to EGFR and the antibody to pTyr, 4G10, were purchased from Millipore. Mouse monoclonal antibody to p44/42MAPK (ERK1/2) and mouse monoclonal antibody to phospho-p44/42MAPK (Thr202/Tyr204) were purchased from Cell Signaling Technology. The wild-type and triple-mutant SH2 domains were subcloned into a eukaryotic expression vector pEGFP-C3 (Clontech), with Xho I and Bam HI sites. For the human Grb2 SH2 domain, the gene encoding the region between Met55 and Pro158 was subcloned. The triple-mutant Grb2 SH2 domain contains three substitutions, namely, A91V (position 8), S96A (position 10), and K109L (position 15). The positions were identified on the basis of a sequence alignment (figs. S2 and S4). The program PROMALS3D (71) was used to generate the alignment.

Cell culture and transfection

HEK293 and HeLa cells were obtained from American Type Culture Collection. The cells were grown in Dulbecco’s modified Eagle’s medium (DMEM; Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS; SAFC Biosciences), penicillin (50 U/ml), and streptomycin (50 μg/ml; Gibco Invitrogen Corp.) in a humidified atmosphere of 5% CO2 in air at 37°C. For stimulation of HEK293 cells, EGF (Invitrogen) at a final concentration of 100 ng/ml was added to the medium at an indicated time point. Transient transfections were carried out with jetPEI (PolyPlus-transfection) according to the manufacturer’s instructions.

Glutathione S-transferase pulldown and Western blots

For the glutathione S-transferase (GST) pulldown assay, the wild-type and triple-mutant Fyn SH2 domains (Ala139-Gly249) were, respectively, subcloned into the pETM30 vector (59). The Fyn SH2 domain constructs contain a FLAG tag sequence (dykddddkc) at the C terminus. To create the GST control vector, we inserted a stop codon after the GST tag sequence of the original pETM30 vector. The GST and GST-SH2 proteins were expressed in E. coli BL21 (DE3). HeLa cells were treated with 50 mM pervanadate for 10 min at 37°C. HeLa cells were lysed on ice in lysis buffer containing 0.5% NP-40, 50 mM Hepes (pH 7.4), 1 mM magnesium chloride, 150 mM KCl, and the Complete protease inhibitor cocktail. For the immunoprecipitation and immunoblotting experiments, HEK293 cells were lysed on ice in lysis buffer containing 1% NP-40, 50 mM tris-HCl (pH 7.4), 150 mM NaCl, 2 mM EDTA, 50 mM NaF, 10% glycerol, and the Complete protease inhibitor cocktail (Roche). After centrifugation at 13,000g for 15 min, the supernatants were collected. The GST pulldown assay, immunoprecipitation, and immunoblotting experiments were performed as previously reported (72). Before conducting the immunoprecipitation assay, the lysate was cleared with appropriate preimmune serum and protein G (Roche). Glutathione Sepharose beads (GE Healthcare) were used for the GST pulldown experiments.

Cell signaling and viability assays

HEK293 cells were transfected with indicated constructs and incubated in serum-containing full medium for 24 hours followed by serum starvation for 16 hours, EGF (100 ng/ml) treatment for 10 min, and then pervanadate treatment for 10 min. The whole-cell lysate was prepared and subjected to immunoblotting to detect phosphorylated ERK and total ERK. To examine the effect of the SH2 domain on cell growth and viability, we transfected HEK293 cells with the indicated plasmids and then incubated in full medium with EGF (100 ng/ml) for 36 hours. Cells were trypsinized and stained with 0.4% trypan blue (Sigma-Aldrich), and viable cell numbers were counted in a hemocytometer by microscopy according to the manufacturer’s instruction.

Colony formation assays in soft agar

Soft agar assays were performed as described (73). Briefly, HEK293 cells were transfected with the indicated SH2 constructs by PEI (polyethylenimine) and grown overnight. On the following day, cells were trypsinized and plated at a density of 1 × 104 cells in 0.25% agarose in DMEM (10% FBS) on top of 0.5% agarose in DMEM (10% FBS) in 60-mm dishes. Cells were maintained at 37°C in 5% CO2 for 21 days and stained overnight with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/5/243/ra68/DC1

Fig. S1. SH2 domain variants obtained by screening a phage-displayed library.

Fig. S2. An alignment of the human SH2 domains showing the region for pTyr binding.

Fig. S3. The specificity and affinity of the Fyn SH2 triple mutant in comparison to those of the wild-type domain.

Fig. S4. Amino acid combinations of the pTyr-binding pocket in natural SH2 domains.

Fig. S5. The dynamics of the BC loop and its stabilization in the Src SH2 triple-mutant domain.

Fig. S6. The structure of the pTyr-binding pocket of the Src SH2 domain triple mutant.

Table S1. Minimal distances between pocket-forming residues in an SH2 domain and the pTyr residue of the ligand.

Table S2. A list of biotinylated peptides used for screening the phage-displayed Fyn SH2 domain library.

Table S3. A list of fluorescein-labeled peptides used for the in-solution binding assay.

Table S4. Fitting error statistics for the Kd values reported in Table 1.

Table S5. Fitting error statistics for the Kd values reported in Table 2.

Table S6. Data collection and refinement statistics for x-ray crystallography.

References and Notes

Acknowledgments: We thank H. Liu, B. Zhao, I. Sochirca, and B. H. Shilton for technical support. Funding: This work was supported by grants from the Canadian Cancer Society (to S.S.C.L.), the Ontario Research Fund—Global Leadership (to S.S.C.L.), and the Canadian Institute of Health Research (grant MOP-93684 to S.S.S.). S.S.C.L. holds a Canada Research Chair in Functional Genomics and Cellular Proteomics. Author Contributions: T.K., H.H., S.S.S., and S.S.C.L. designed the project. T.K., H.H., C.L., and C.V. prepared DNA constructs. H.H. performed phage display experiments. T.K., H.H., S.S.S., and S.S.C.L. analyzed the data. T.K. and X.L. synthesized peptides and performed in-solution binding assays. T.K. conducted crystallographic analysis. X.C. performed cell signaling and growth assay. T.K., X.C., S.S.S., and S.S.C.L. wrote the paper. Competing interests: S.S.C.L., S.S.S., T.K., H.H., and X.C. filed a provisional U.S. patent application that is related to this work. Data and materials availability: Complementary DNA constructs for the reported mutant SH2 domains require a signed material transfer agreement from the University of Western Ontario and are limited to use for research purposes only. Atomic coordinates and structure factors have been deposited to the Protein Data Bank, with accession codes 4F59 (apo form), 4F5A (phosphate ion complex), and 4F5B (pTyr complex).
View Abstract

Stay Connected to Science Signaling

Navigate This Article