Exploiting the Basis of Proline Recognition by SH3 and WW Domains: Design of N-Substituted Inhibitors

See allHide authors and affiliations

Science  11 Dec 1998:
Vol. 282, Issue 5396, pp. 2088-2092
DOI: 10.1126/science.282.5396.2088


Src homology 3 (SH3) and WW protein interaction domains bind specific proline-rich sequences. However, instead of recognizing critical prolines on the basis of side chain shape or rigidity, these domains broadly accepted amide N-substituted residues. Proline is apparently specifically selected in vivo, despite low complementarity, because it is the only endogenous N-substituted amino acid. This discriminatory mechanism explains how these domains achieve specific but low-affinity recognition, a property that is necessary for transient signaling interactions. The mechanism can be exploited: screening a series of ligands in which key prolines were replaced by nonnatural N-substituted residues yielded a ligand that selectively bound the Grb2 SH3 domain with 100 times greater affinity.

Protein-protein interaction domains, such as Src homology 3 (SH3) and WW domains, participate in diverse signaling pathways and are important targets in drug design (1, 2). These domains specifically recognize unique proline-rich peptide motifs but bind them with low affinities (K d = 1 to 200 μM) compared with other peptide recognition proteins such as antibodies and receptors (K d = nanomolar to picomolar concentrations). SH3 domains recognize sequences bearing the core element, PXXP (P = proline, X = any amino acid), flanked by other domain-specific residues (3). Identification of compounds that potently interrupt these interactions has proven difficult: extensive screening of natural and nonnatural combinatorial libraries has not yielded compounds that bind as well as or better than PXXP peptides (4,5). Here we show that the essential ligand feature recognized by both SH3 and WW domains is an irregular backbone substitution pattern: N-substituted residues placed at key positions along an otherwise normal Cα-substituted peptide scaffold. Prolines are required at these sites, not on the basis of side chain shape but simply because they are the only naturally available N-substituted residue. This unusual recognition code has been used to guide design of SH3 inhibitors with improved affinity and selectivity.

We used a chemical minimization scheme to identify essential ligand recognition elements for two domains, the COOH-terminal SH3 domain from the Caenorhabditis elegans adapter protein Sem5, and the WW domain from the human signaling protein Yap. The Sem5 SH3 domain recognizes the core PXXP sequence, flanked by a specific arginine residue (1). WW domains recognize the consensus motif PPXY (Y = tyrosine) (2). We scanned through the proline-rich core of each ligand and made the following substitutions:

Both substitutions destroy the proline ring, but each leaves a single methyl group bonded to a different main chain atom.

Proline recognition by the Sem5 SH3 domain and the Yap WW domain was almost exclusively based on amide N-substitution (Table 1)—other unique properties of proline, such as its unusual side chain shape and conformational rigidity, were dispensable (6). Critical prolines of the SH3 core PXXP motif (sites P2 and P−1) are sites where alanine or other amino acid replacements are not tolerated (1,4). However, these sites tolerated sarcosine replacement. Thus nearly complete deletion of the proline side chain was acceptable as long as N-substitution was maintained. An identical tolerance pattern was seen at site P−2 of the WW domain ligand. The scanning results also revealed a second backbone requirement: a Cα-substituted residue must precede the required N-substituted residue. At Site P0 in the SH3 ligand and site P−3 in the WW ligand, alanine and other Cα-substituted residues were acceptable, but sarcosine was not (7).

Table 1

Reduction in binding affinity (21) caused by alanine (A) or sarcosine (A*) substitutions within proline-rich ligands of Sem5 SH3 domain and Yap WW domain (22). “Required” prolines are underlined.

View this table:

Examination of the crystal structure of the Sem5 SH3 domain complex (8) and the NMR (nuclear magnetic resonance) structure of the Yap WW domain complex (9) reveals the basis for these requirements. In both complexes, the ligand binds in a polyproline II (PPII) helical conformation, a left-handed helix with three residues per turn. Turns on one face of the helix pack into a series of grooves on the domain surface (Fig. 1A). Each groove accommodates two peptide residues. The minimal and sufficient requirement at each binding groove is a pair of sequential Cα- and N-substituted residues (10). The Cα/N-substituted pair may be required because, in this arrangement, substituent groups are separated by only a single backbone carbon atom, forming a relatively continuous ridge that can pack efficiently into the domain grooves (Fig. 1B) (11).

Figure 1

Backbone substitution requirements for SH3 and WW domain recognition. (A) Structural mapping of alanine and sarcosine scanning results (Table 1). Peptide/domain complex interfaces (8, 9) shown schematically. Ligands adopt a PPII conformation, depicted schematically as a triangular prism. Residue positions (spheres) are color-coded by class: white—does not require either Cα- or N-substitution (alanine and sarcosine tolerant); green—requires Cα-substitution (alanine tolerant, sarcosine intolerant); orange—requires N-substitution (sarcosine tolerant, alanine intolerant). (B) Minimally sufficient recognition unit for SH3 and WW domain binding grooves. Schematic view of a single binding groove cross-section, looking down the PPII helical axis (viewed from left side of Fig. 1A). Minimally required atoms defined in this study, a sequential pair of Cα- and N-substituted residues, are solid black. The van der Waals binding surface that these atoms present is shaded. (C) Distinct mechanisms of proline recognition. Proline can be recognized by a lock and key mechanism, utilizing the full chemical potential of the side chain. In contrast, SH3 and WW domains recognized key prolines based on N-substitution. This mechanism utilizes relatively little of the binding potential of ligand or protein (hatched surface) but is still highly discriminatory for proline among natural amino acids.

SH3 and WW domains appear to read their signature sequences by a mechanism fundamentally different from the “lock and key” mechanism (12) of canonical sequence-specific recognition proteins (Fig. 1C). Such interactions utilize an array of surface pockets optimized to fit the shape and size of anchor side chains displayed along the ligand peptide backbone (13). In such cases, even small perturbations in side chain properties can be deleterious. Proline is recognized in this fashion at many protein interfaces. In contrast, the proline-requiring pockets of SH3 and WW domains actually recognize a unique backbone property: N-substitution. Specificity is achieved, not by favoring binding to proline but by disfavoring binding to any other natural amino acid, all of which lack N-substitution.

This backbone discrimination mechanism reveals a strategy for inhibitor design: maintain the required hybrid Cα- and N-substituted scaffold, but vary side chain identity along this scaffold to optimize complementarity. We tested this strategy by synthesizing a series of SH3 ligands in which each of the two “required” PXXP prolines was replaced by a diverse set of 11 nonnatural N-substituted glycine, or “peptoid,” residues (Fig. 2A). Such groups could exploit the untapped chemical potential of this interface. We tested binding of these ligands to four SH3 domains—Sem5, Crk, Grb2, and Src—all of which share a preference for ligands with the consensus sequence PXXPXR (14). More than half the 22 N-substituted ligands bound as well as or better than natural proline-containing peptides. In contrast, no other natural amino acids are tolerated at these sites (4).

Figure 2

Replacing required ligand prolines with peptoids can increase affinity and selectivity for SH3 domains. (A) Effects of peptoid substitutions at proline-requiring sites of SH3 ligand (23). Wild-type background is YEVPPPVPPRRR (24). Required proline sites (P2and P−1) are shown shaded in the chemical structure. Binding was measured to the Sem5 COOH-terminal SH3 domain, the mouse Crk NH2-terminal SH3 domain, the human Grb2 NH2-terminal SH3 domain, and the mouse Src SH3 domain (21). Changes in free energy of binding upon mutation (ΔΔG) relative to proline are color coded (orange—favorable; blue—unfavorable). Dissociation constants for the wild-type 12-mer peptide are as follows: Sem5, K d = 48 μM; Crk, K d = 6 μM; Grb2,K d = 5 μM; Src, K d = 25 μM. (B) Peptide 45 selectively binds Grb2 SH3 with 102-fold improved affinity. Binding curve of peptide 45 [N-(S)-phenylethyl peptoid at P-1] to Grb2 NH2-terminal SH3 domain (filled circles), Src SH3 domain (white circles), and Crk SH3 domain (triangles), as measured by fluorescence perturbation. Data were fit (solid lines) as described (21). Data for the Sem5 SH3 domain are not shown, as this domain binds with 50-fold lower affinity than the other domains tested. For reference, isotherm of wild-type peptide binding to Grb2 is shown by a dashed line (overlaps with Crk binding curve). (C) Inhibition of Grb2 SH3 binding by peptide 45. Binding of biotinylated Grb2 SH3 domain to Sos peptide/GST fusion protein in the presence of peptide 45 or wild-type peptide (25). The K i of inhibitor is estimated to be 1/10th of the observed IC50.

Peptide 45 (Fig. 2B) bound the Grb2 SH3 domain with an affinity (K d = 40 nM) > 100 times that of the wild-type peptide (15). This substitution of anN-(S)-phenylethyl group at site P−1 results in a favorable increase in binding energy of ΔΔG = −2.8 kcal/mol, a 40% increase in total interaction energy. Peptide 41, which has an N-(4-hydroxy)phenyl substitution, bound the Sem5 SH3 domain with 25-fold improved affinity. Four other specific domain-ligand pairs showed 5- to 10-fold improved affinity.

Recognition of these peptoid side chains is stereospecific, as is typical for interactions with high complementarity. For example, the Grb2 SH3 domain bound peptide 45 with 103-fold greater affinity than the related R-stereoisomer (peptide 44). In addition, peptide 45 acted as a potent competitive inhibitor (Fig. 2C), blocking binding of the Grb2 SH3 domain to a Sos peptide fusion protein with an IC50 about 1/50th that of the wild-type proline peptide.

The peptoid ligands have improved domain selectivity, overcoming a second major problem posed by SH3 domains as drug targets—members of the family are highly cross-reactive (16). This enlarged range of N-substituted residues can be used to exploit subtle differences between individual SH3 domains. For example, peptide 45 binds potently to Grb2 but shows only modest to negligible improvement in binding to the other domains, resulting in about 102-fold selectivity for Grb2 (Fig. 2B).

We crystallized and solved the structures of three SH3-peptoid ligand complexes (17). These structures (Fig. 3) confirmed that the peptoid side chains bound at the proline-requiring sites and suggest how peptoids increase affinity and domain discrimination. The peptoid side chains insert into these sites more deeply than proline, packing slightly differently. Thus specific side chains can make better fitting and more extensive contacts with the domain, including contacts with regions on the SH3 surface that show higher sequence or structural variation. The chemistry of peptoid synthesis allows for exploration of greater side chain diversity than examined here and could yield optimized ligands for other SH3 domains.

Figure 3

Structural basis of peptoid recognition. (A) Structure of wild-type Sos peptide (PPPVPPRRR) bound to Crk SH3 domain (20). Proline-rich core binding grooves are indicated by dashed boxes. Highly conserved surface residues among the four SH3 domains studied here (one or two conservative amino acid types) are green. Variable surface residues (3+ amino acid types) are brown. The ligand PXXP core binds at the most conserved surface on the protein. (B) Structure of peptide 34 bound to Crk SH3 domain. N-(S)-1-Phenylethyl peptoid side chain (orange) bound at site P2. Close-up view from the same perspective as above. (C) Structure of peptide 39 bound to the Sem5 SH3 domain. N-Cyclopropylmethyl peptoid side chain (orange) bound at site P−1. Close-up view from the same perspective as above.

The recognition strategy of SH3 and WW domains allows for high specificity binding that need not be of high affinity. In vivo, binding to proline peptides is highly selective, despite suboptimal shape complementarity, because there are no other natural sequences that can satisfy the minimal ligand backbone requirements. The resultant weak but specific interactions are ideal for intracellular signaling domains. These modules must recognize ligands with high enough selectivity to maintain proper information flow but with low enough affinity to allow for sensitive and dynamic modulation in response to changing signals. In contrast the high-affinity and high-specificity interactions that result from typical lock and key recognition are ill-suited for such a function. The ability to recognize proline in this way may explain why proline-rich motifs are so commonly used in regulatory interactions.

  • * To whom correspondence should be addressed E-mail: wlim{at}


View Abstract

Navigate This Article