Characterization of PDZ domain-peptide interaction interface based on energetic patterns

PDZ domain is one of the abundant modular domains that recognize short peptide sequences to mediate protein-protein interactions. To decipher the binding specificity of PDZ domain, we analyzed the interactions between 11 mouse PDZ domains and 2387 peptides using a method called MIEC-SVM, which energetically characterizes the domain-peptide interaction using molecular interaction energy components (MIECs) and predicts binding specificity using support vector machine (SVM). Cross-validation and leave-one-domain-out test showed that the MIEC-SVM using all 44 PDZ-peptide residue pairs at the interaction interface outperformed the sequence-based methods in the literature. A further feature (residue pair) selection procedure illustrated that 16 residue pairs were uninformative to the binding specificity, even though they contributed significantly (~50%) to the binding energy. If only using the 28 informative residue pairs, the performance of the MIEC-SVM on predicting the PDZ binding specificity was significantly improved. This analysis suggests that the informative and uninformative residue interactions between the PDZ domain and the peptide may represent those contributing to binding specificity and affinity, respectively. We performed additional structural and energetic analyses to shed light on understanding how the PDZ-peptide recognition is established. The success of the MIEC-SVM method on PDZ domains in this study and SH3 domains in our previous studies illustrates its generality on characterizing protein-peptide interactions and understanding protein recognition from a structural and energetic viewpoint.     

Prediction of binding affinities between the human amphiphysin-1 SH3 domain and its peptide ligands using homology modeling, molecular dynamics and molecular field analysis

The SH3 domain of the human protein amphiphysin-1, which plays important roles in clathrin-mediated endocytosis, actin function and signaling transduction, can recognize peptide motif PXRPXR (X is any amino acid) with high affinity and specificity. We have constructed a complex structure of the amphiphysin-1 SH3 domain and a high-affinity peptide ligand PLPRRPPRA using homology modeling and molecular docking, which was optimized by molecular dynamics (MD). Three-dimensional quantitative structure-affinity relationship (3D-QSAR) analyses on the 200 peptides with known binding affinities to the amphiphysin-1 SH3 domain was then performed using comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). The best CoMSIA model showed promising predictive power, giving good predictions for about 95% of the peptides in the test set (absolute prediction errors less than 1.0). It was used to validate peptide-SH3 binding structure and provide insight into the structural requirements for binding of peptides to SH3 domains. Finally, MD simulations were performed to analyze the interaction between the SH3 domain and another peptide GFPRRPPPRG that contains with the PXRPXsR (s represents residues with small side chains) motif. MD simulations demonstrated that the binding conformation of GFPRRPPPRG is quite different from that of PLPRRPPRAA especially the four residues at the C terminal, which may explain why the CoMSIA model cannot give good predictions on the peptides of the PXRPXsR motif. Because of its efficiency and predictive power, the 3D-QSAR model can be used as a scoring filter for predicting peptide sequences bound to SH3 domains.     

Modeling and prediction of binding affinities between the human amphiphysin SH3 domain and its peptide ligands using genetic algorithm-Gaussian processes

In this article, we discuss the application of the Gaussian process (GP) and other statistical methods (PLS, ANN, and SVM) for the modeling and prediction of binding affinities between the human amphiphysin SH3 domain and its peptide ligands. Divided physicochemical property scores of amino acids, involving significant hydrogen bond, electronic, hydrophobic, and steric properties, was used to characterize the peptide structures, and quantitative structure-affinity relationship models were then constructed by PLS, ANN, SVM, and GP coupled with genetic algorithm-variable selection. The results show that: (i) since the significant flexibility and high complexity possessed in polypeptide structures, linear PLS method was incapable of fulfilling a satisfying behavior on SH3 domain binding peptide dataset; (ii) the overfitting involved in training process has decreased the predictive power of ANN model to some extent; (iii) both SVM and GP have a good performance for SH3 domain binding peptide dataset. Moreover, by combining linear and nonlinear terms in the covariance function, the GP is capable of handling linear and nonlinear-hybrid relationship, and which thus obtained a more stable and predictable model than SVM. Analyses of GP models showed that diversified properties contribute remarkable effect to the interactions between the SH3 domain and the peptides. Particularly, steric property and hydrophobicity of P(2), electronic property of P(0), and electronic property and hydrogen bond property of P(-3) in decapeptide (P(4)P(3)P(2)P(1)P(0)P(-1)P(-2)P(-3)P(-4)P(-5)) significantly contribute to the binding affinities of SH3 domain-peptide interactions.     

Characterization of domain-peptide interaction interface: prediction of SH3 domain-mediated protein-protein interaction network in yeast by generic structure-based models

Determination of the binding specificity of SH3 domain, a peptide recognition module (PRM), is important to understand their biological functions and reconstruct the SH3-mediated protein-protein interaction network. In the present study, the SH3-peptide interactions for both class I and II SH3 domains were characterized by the intermolecular residue-residue interaction network. We developed generic MIEC-SVM models to infer SH3 domain-peptide recognition specificity that achieved satisfactory prediction accuracy. By investigating the domain-peptide recognition mechanisms at the residue level, we found that the class-I and class-II binding peptides have different binding modes even though they occupy the same binding site of SH3. Furthermore, we predicted the potential binding partners of SH3 domains in the yeast proteome and constructed the SH3-mediated protein-protein interaction network. Comparison with the experimentally determined interactions confirmed the effectiveness of our approach. This study showed that our sophisticated computational approach not only provides a powerful platform to decipher protein recognition code at the molecular level but also allows identification of peptide-mediated protein interactions at a proteomic scale. We believe that such an approach is general to be applicable to other domain-peptide interactions.     

Characterization of PDZ domain‐peptide interaction interface based on energetic patterns

PDZ domain is one of the abundant modular domains that recognize short peptide sequences to mediate protein–protein interactions. To decipher the binding specificity of PDZ domain, we analyzed the interactions between 11 mouse PDZ domains and 217 peptides using a method called MIECSVM, which energetically characterizes the domain‐peptide interaction using molecular interaction energy components (MIECs) and predicts binding specificity using support vector machine (SVM). Cross‐validation and leave‐one‐domain‐out test showed that the MIEC‐SVM using all 44 PDZ‐peptide residue pairs at the interaction interface outperformed the sequence‐based methods in the literature. A further feature (residue pair) selection procedure illustrated that 16 residue pairs were uninformative to the binding specificity, even though they contributed significantly (～50%) to the binding energy. If only using the 28 informative residue pairs, the performance of the MIEC‐SVM on predicting the PDZ binding specificity was significantly improved. This analysis suggests that the informative and uninformative residue interactions between the PDZ domain and the peptide may represent those contributing to binding specificity and affinity, respectively. We performed additional structural and energetic analyses to shed light on understanding how the PDZ‐peptide recognition is established. The success of the MIEC‐SVM method on PDZ domains in this study and SH3 domains in our previous studies illustrates its generality on characterizing protein‐ peptide interactions and understanding protein recognition from a structural and energetic viewpoint.     

Distinct peptide binding specificities of Src homology 3 (SH3) protein domains can be determined by modulation of local energetics across the binding interface

The yeast Nbp2p SH3 and Bem1p SH3b domains bind certain target peptides with similar high affinities, yet display vastly different affinities for other targets. To investigate this unusual behavior, we have solved the structure of the Nbp2p SH3-Ste20 peptide complex and compared it with the previously determined structure of the Bem1p SH3b bound to the same peptide. Although the Ste20 peptide interacts with both domains in a structurally similar manner, extensive in vitro studies with domain and peptide mutants revealed large variations in interaction strength across the binding interface of the two complexes. Whereas the Nbp2p SH3 made stronger contacts with the peptide core RXXPXXP motif, the Bem1p SH3b domain made stronger contacts with residues flanking the core motif. Remarkably, this modulation of local binding energetics can explain the distinct and highly nuanced binding specificities of these two domains.     

A fluid salt-bridging cluster and the stabilization of p53

p53 is a homotetrameric tumor suppressor protein that is found to be mutated in most human cancers. Some of these mutations, particularly mutations to R337, fall in the tetramerization domain and cause defects in tetramer formation leading to loss of function. Mutation to His at this site has been found to destabilize the tetramer in a pH-dependent fashion. In structures of the tetramerization domain determined by crystallography, R337 from one monomer makes a salt bridge with D352 from another monomer, apparently helping to stabilize the tetramer. Here we present molecular dynamics simulations of wild-type p53 and the R337His mutant at several different pH and salt conditions. We find that the 337-352 salt bridge is joined by two other charged side chains, R333 and E349. These four residues do not settle into a fixed pattern of salt bridging, but continue to exchange salt-bridging partners on the nanosecond time scale throughout the simulation. This unusual system of fluid salt bridging may explain the previous finding from alanine scanning experiments that R333 contributes significantly to protein stability, even though in the crystal structure it is extended outward into solvent. This extended conformation of R333 appears to be the result of a specific crystal contact and, this contact being absent in the simulation, R333 turns inward to join its interaction partners. When R337 is mutated to His but remains positively charged, it maintains the original interaction with D352, but the newly observed interaction with E349 is weakened, accounting for the reduced stability of R337H even under mildly acidic conditions. When this His is deprotonated, the interaction with D352 is also lost, accounting for the further destabilization observed under mildly alkaline conditions. Simulations were carried out using both explicit and implicit solvent models, and both displayed similar behavior of the fluid salt-bridging cluster, suggesting that implicit solvent models can capture at least the qualitative features of this phenomenon as well as explicit solvent. Simulations under strongly acidic conditions in implicit solvent displayed the beginnings of the unfolding process, a destabilization of the hydrophobic dimer-dimer interface. Computational alanine scanning using the molecular mechanics Poisson-Boltzmann surface area method showed significant correlation to experimental unfolding data for charged and polar residues, but much weaker correlation for hydrophobic residues.     

Novel Src homology 3 domain-binding motifs identified from proteomic screen of a Pro-rich region

The Src homology (SH) 3 domain has been shown recently to bind peptide sequences that lack the canonical PXXP motif. The diverse specificity in ligand recognition for a group of 15 SH3 domains has now been investigated using arrays of peptides derived from the proline-rich region of the SH2 domain-containing leukocyte protein of 76 kDa (SLP-76). A screen of the peptide arrays using individual or mixed SH3 domains has allowed the identification of a number of candidate SH3-binding peptides. Although some peptides contain the conventional PXXP motif, most are devoid of such a motif and are instead enriched in basic residues. Fluorescent polarization measurements using soluble peptides and purified SH3 domains demonstrated that several SH3 domains, including those from growth factor receptor-bound protein 2 (Grb2), NCK, and phospholipase C (PLC)-gamma1, bound with moderate affinities (10-100 microm) to a group of non-conventional peptides. Of particular interest, the PLC-gamma1 SH3 domain was found to associate with SLP-76 through at least three distinct sites, two of which bore a novel KKPP motif and the other contained the classic PXXP sequence. Intriguingly mutation of critical residues for the three sites not only affected binding of SLP-76 to the PLC-gamma1 SH3 domain but also to the Grb2 C-terminal SH3 domain, indicating that the binding sites in SLP-76 for the two SH3 domains are overlapped. Our studies suggest that the SH3 domain is an inherently promiscuous interaction module capable of binding to peptides that may or may not contain a PXXP motif. Furthermore the identification of numerous non-conventional SH3-binding peptides in SLP-76 implies that the global ligand pool for SH3 domains in a mammalian proteome may be significantly greater than previously acknowledged.     

Versatile retargeting of SH3 domain binding by modification of non-conserved loop residues

Src-homology (SH3) domain belongs to a class of ubiquitous modular protein domains found in nature. SH3 domains have a conserved surface that recognises proline-rich peptides in ligand proteins, but additional contacts also contribute to binding. Using the SH3 domain of hematopoietic cell kinase as a test case, we show that SH3 binding properties can be profoundly altered by modifications within a hexapeptide sequence in the RT-loop region that is not involved in recognition of currently known consensus SH3 target peptides. These results highlight the role of non-conserved regions in SH3 target selection, and introduce a strategy that may be generally feasible for generating artificial SH3 domains with desired ligand binding properties.     

Hierarchy of simulation models in predicting molecular recognition mechanisms from the binding energy landscapes: structural analysis of the peptide complexes with SH2 domains.

Computer simulations using the simplified energy function and simulated tempering dynamics have accurately determined the native structure of the pYVPML, SVLpYTAVQPNE, and SPGEpYVNIEF peptides in the complexes with SH2 domains. Structural and equilibrium aspects of the peptide binding with SH2 domains have been studied by generating temperature-dependent binding free energy landscapes. Once some native peptide-SH2 domain contacts are constrained, the underlying binding free energy profile has the funnel-like shape that leads to a rapid and consistent acquisition of the native structure. The dominant native topology of the peptide-SH2 domain complexes represents an extended peptide conformation with strong specific interactions in the phosphotyrosine pocket and hydrophobic interactions of the peptide residues C-terminal to the pTyr group. The topological features of the peptide-protein interface are primarily determined by the thermodynamically stable phosphotyrosyl group. A diversity of structurally different binding orientations has been observed for the amino-terminal residues to the phosphotyrosine. The dominant native topology for the peptide residues carboxy-terminal to the phosphotyrosine is tolerant to flexibility in this region of the peptide-SH2 domain interface observed in equilibrium simulations. The energy landscape analysis has revealed a broad, entropically favorable topology of the native binding mode for the bound peptides, which is robust to structural perturbations. This could provide an additional positive mechanism underlying tolerance of the SH2 domains to hydrophobic conservative substitutions in the peptide specificity region.     

SH3 Domain-Peptide Binding Energy Calculations Based on Structural Ensemble and Multiple Peptide Templates

SH3 domains mediate signal transduction by recognizing short peptides. Understanding of the driving forces in peptide recognitions will help us to predict the binding specificity of the domain-peptide recognition and to understand the molecular interaction networks of cells. However, accurate calculation of the binding energy is a tough challenge. In this study, we propose three ideas for improving our ability to predict the binding energy between SH3 domains and peptides: (1) utilizing the structural ensembles sampled from a molecular dynamics simulation trajectory, (2) utilizing multiple peptide templates, and (3) optimizing the sequence-structure mapping. We tested these three ideas on ten previously studied SH3 domains for which SPOT analysis data were available. The results indicate that calculating binding energy using the structural ensemble was most effective, clearly increasing the prediction accuracy, while the second and third ideas tended to give better binding energy predictions. We applied our method to the five SH3 targets in DREAM4 Challenge and selected the best performing method.     

All-Atom Monte Carlo Approach to Protein-Peptide Binding

We develop a procedure for exploring the free energy landscape of protein-peptide binding at atomic detail and apply it to PDZ domain-peptide interactions. The procedure involves soft constraints on receptor proteins providing limited chain flexibility, including backbone motions. Peptide chains are left fully flexible and kept in spatial proximity of the protein through periodic boundary conditions. By extensive Monte Carlo simulations, full representative conformational ensembles at temperatures where bound and unbound states coexist are obtained. To make this approach computationally feasible, we develop an effective all-atom energy function centering on hydrophobicity, hydrogen bonding, and electrostatic interactions. Our initial focus is a set of 11 PDZ domain-peptide pairs with experimentally determined complex structures. Minimum-energy conformations are found to be highly similar to the respective native structures in eight of the cases (all-atom peptide RMSDs < 6 Å). Having achieved that, we turn to a more complete characterization of the bound peptide state through a clustering scheme applied on the full ensembles of peptide structures. We find a significant diversity among bound peptide conformations for several PDZ domains, in particular involving the N terminal side of the peptide chains. Our computational model is then tested further on a set of nine PDZ domain-peptide pairs where the peptides are not originally present in the experimentally determined structures. We find a similar success rate in terms of the nativeness of minimum-energy conformations. Finally, we investigate the ability of our approach to capture variations in binding affinities for different peptide sequences. This is done in particular for a set of related sequences binding to the third PDZ domain of PSD-95 with encouraging results.     

Structural basis of PxxDY motif recognition in SH3 binding

We have determined the solution structure of epidermal growth factor receptor pathway substrate 8 (Eps8) L1 Src homology 3 (SH3) domain in complex with the PPVPNPDYEPIR peptide from the CD3ε cytoplasmic tail. Our structure reveals the distinct structural features that account for the unusual specificity of the Eps8 family SH3 domains for ligands containing a PxxDY motif instead of canonical PxxP ligands. The CD3ε peptide binds Eps8L1 SH3 in a class II orientation, but neither adopts a polyproline II helical conformation nor engages the first proline-binding pocket of the SH3 ligand binding interface. Ile531 of Eps8L1 SH3, instead of Tyr or Phe residues typically found in this position in SH3 domains, renders this hydrophobic pocket smaller and nonoptimal for binding to conventional PxxP peptides. A positively charged arginine at position 512 in the n-Src loop of Eps8L1 SH3 plays a key role in PxxDY motif recognition by forming a salt bridge to D7 of the CD3ε peptide. In addition, our structural model suggests a hydrogen bond between the hydroxyl group of the aromatic ring of Y8 and the carboxyl group of E496, thus explaining the critical role of the PxxDY motif tyrosine residue in binding to Eps8 family SH3. These finding have direct implications also for understanding the atypical binding specificity of the amino-terminal SH3 of the Nck family proteins.     

The recognition specificity of the CHD1 chromodomain with modified histone H3 peptides

Histone tail peptides comprise the flexible portion of chromatin, the substance which serves as the packaging for the eukaryotic genome. According to the histone code hypothesis, reader protein domains (chromodomains) can recognize modifications of amino acid residues within these peptides, regulating the expression of genes. We have performed simulations on models of chromodomain helicase DNA-binding protein 1 complexed with a variety of histone H3 modifications. Binding free energies for both the overall complexes and the individual residues within the protein and peptides were computed with molecular mechanics-generalized Born surface area. The simulation results agree well with experimental data and identify several chromodomain helicase DNA-binding protein 1 residues that play key roles in the interaction with each of the H3 modifications. We identified one class of protein residues that bind to H3 in all of the complexes (generally interacting hydrophobically), and a second class of residues that bind only to particular H3 modifications (generally interacting electrostatically). Additionally, we found that modifications of H3R2 and H3T3 have a dominant effect on the binding affinity; methylation of H3K4 has little effect on the interaction strength when H3R2 or H3T3 is modified. Our findings with regard to the specificity shown by the latter class of protein residues in their binding affinity to certain modifications of H3 support the histone code hypothesis.     

Solution structure of the first SH3 domain of human vinexin and its interaction with vinculin peptides

Solution structure of the first Src homology (SH) 3 domain of human vinexin (V_SH3_1) was determined using nuclear magnetic resonance (NMR) method and revealed that it was a canonical SH3 domain, which has a typical beta-beta-beta-beta-alpha-beta fold. Using chemical shift perturbation and surface plasmon resonance experiments, we studied the binding properties of the SH3 domain with two different peptides from vinculin hinge regions: P856 and P868. The observations illustrated slightly different affinities of the two peptides binding to V_SH3_1. The interaction between P868 and V_SH3_1 belonged to intermediate exchange with a modest binding affinity, while the interaction between P856 and V_SH3_1 had a low binding affinity. The structure and ligand-binding interface of V_SH3_1 provide a structural basis for the further functional study of this important molecule.     

Molecular dynamics-solvated interaction energy studies of protein-protein interactions: the MP1-p14 scaffolding complex

Using the MP1-p14 scaffolding complex from the mitogen-activated protein kinase signaling pathway as model system, we explored a structure-based computational protocol to probe and characterize binding affinity hot spots at protein-protein interfaces. Hot spots are located by virtual alanine-scanning consensus predictions over three different energy functions and two different single-structure representations of the complex. Refined binding affinity predictions for select hot-spot mutations are carried out by applying first-principle methods such as the molecular mechanics generalized Born surface area (MM-GBSA) and solvated interaction energy (SIE) to the molecular dynamics (MD) trajectories for mutated and wild-type complexes. Here, predicted hot-spot residues were actually mutated to alanine, and crystal structures of the mutated complexes were determined. Two mutated MP1-p14 complexes were investigated, the p14(Y56A)-mutated complex and the MP1(L63A,L65A)-mutated complex. Alternative ways to generate MD ensembles for mutant complexes, not relying on crystal structures for mutated complexes, were also investigated. The SIE function, fitted on protein-ligand binding affinities, gave absolute binding affinity predictions in excellent agreement with experiment and outperformed standard MM-GBSA predictions when tested on the MD ensembles of Ras-Raf and Ras-RalGDS protein-protein complexes. For wild-type and mutant MP1-p14 complexes, SIE predictions of relative binding affinities were supported by a yeast two-hybrid assay that provided semiquantitative relative interaction strengths. Results on the MP1-mutated complex suggested that SIE predictions deteriorate if mutant MD ensembles are approximated by just mutating the wild-type MD trajectory. The SIE data on the p14-mutated complex indicated feasibility for generating mutant MD ensembles from mutated wild-type crystal structure, despite local structural differences observed upon mutation. For energetic considerations, this would circumvent costly needs to produce and crystallize mutated complexes. The sensitized protein-protein interface afforded by the p14(Y56A) mutation identified here has practical applications in screening-based discovery of first-generation small-molecule hits for further development into specific modulators of the mitogen-activated protein kinase signaling pathway.     

Isoform-Selective Interaction of the Adaptor Protein Tks5/FISH with Sos1 and Dynamins

The adaptor protein Tks5/FISH (tyrosine kinase substrate 5/five SH3 domains, hereafter termed Tks5) is a crucial component of a protein network that controls the invasiveness of cancer cells and progression of Alzheimer's disease. Tks5 consists of an amino-terminal PX domain that is followed by five SH3 domains (SH3A-E), and two different splice variants are expressed. We identified son of sevenless-1 (Sos1) as a novel binding partner of Tks5 and found colocalization of Tks5 with Sos1 in human epithelial lung carcinoma (A549) cells and in podosomes of Src-transformed NIH 3T3 cells. We observe synergistic binding of SH3A and SH3B to Sos1 when peptide arrays are used, indicating that the tandem SH3A and SH3B domains of Tks5 can potentially bind in a superSH3 binding mode, as was described for the homologous protein p47phox. These results are further corroborated by pull-down assays and isothermal titration calorimetry showing that both intact SH3 domains are required for efficient binding to the entire proline-rich domain of Sos1. The presence of a basic insertion between the SH3A and SH3B domains in the long splice variant of Tks5 decreases the affinity to Sos1 isoforms about 10-fold as determined by analytical ultracentrifugation. Furthermore, it leads to an alteration in the recognition of binding motifs for the interaction with Sos1: While the insertion abrogates the interaction with the majority of peptides derived from the proline-rich domains of Sos1 and dynamin that are recognized by the short splice isoform, it enables binding to a different set of peptides including a sequence comprising the splice insertion in the long isoform of Sos1 (Sos1_2). In the absence of the basic insertion, Tks5 was found to bind a range of Sos1 and dynamin peptides including conventional proline-rich motifs and atypical recognition sequences. Hereby, the tandem SH3 domains in Tks5 employ two distinct types of binding modes: One class of peptides is recognized by single SH3 domains, whereas a second class of peptides requires the presence of both domains to bind synergistically. We conclude that the tandem SH3A and SH3B domains of Tks5 constitute a versatile module for the implementation of isoform-specific protein-protein interactions.     

A novel structure-based encoding for machine-learning applied to the inference of SH3 domain specificity

MOTIVATION: Unravelling the rules underlying protein-protein and protein-ligand interactions is a crucial step in understanding cell machinery. Peptide recognition modules (PRMs) are globular protein domains which focus their binding targets on short protein sequences and play a key role in the frame of protein-protein interactions. High-throughput techniques permit the whole proteome scanning of each domain, but they are characterized by a high incidence of false positives. In this context, there is a pressing need for the development of in silico experiments to validate experimental results and of computational tools for the inference of domain-peptide interactions. RESULTS: We focused on the SH3 domain family and developed a machine-learning approach for inferring interaction specificity. SH3 domains are well-studied PRMs which typically bind proline-rich short sequences characterized by the PxxP consensus. The binding information is known to be held in the conformation of the domain surface and in the short sequence of the peptide. Our method relies on interaction data from high-throughput techniques and benefits from the integration of sequence and structure data of the interacting partners. Here, we propose a novel encoding technique aimed at representing binding information on the basis of the domain-peptide contact residues in complexes of known structure. Remarkably, the new encoding requires few variables to represent an interaction, thus avoiding the 'curse of dimension'. Our results display an accuracy >90% in detecting new binders of known SH3 domains, thus outperforming neural models on standard binary encodings, profile methods and recent statistical predictors. The method, moreover, shows a generalization capability, inferring specificity of unknown SH3 domains displaying some degree of similarity with the known data.