An algorithm for predicting protein-protein interaction sites: Abnormally exposed amino acid residues and secondary structure elements

Multiprotein systems mediate most regulatory processes in living organisms. Although the structures of the individual proteins are often defined, less is known of the structures of multiprotein systems. Computational methods for predicting interfaces, using evolutionary conservation and/or physicochemical data, have been developed. Here we consider the use of solvent accessibility, residue propensity, and hydrophobicity, in conjunction with secondary structure data, as prediction parameters. We analyze the influence of residue type and secondary structure on solvent accessibility and define a measure of "relative exposedness." Clustering abnormally high scoring residues provides a basis for predicting interaction sites. The analysis is extended to investigate abnormally exposed secondary structure elements, particularly beta-sheet strands. We show that surface-exposed beta-strands lacking protective features are more likely to be found at protein-protein interfaces, allowing us to create an algorithm with approximately 68% and approximately 75% accuracy in differentiating between interacting and edge strands in isolated beta-strands and beta-sheet strands, respectively. These methods of identifying abnormally exposed surface regions are combined in an algorithm, which, on a data set of 77 unbound and disjoint (single chain extracted from complex) structures, predicts 79% of the protein-protein interfaces correctly. If enzyme-inhibitor complexes, where the inhibitor mimics a nonprotein substrate, are excluded, the accuracy increases to 85%.     

Assessing predictions of protein-protein interaction: the CAPRI experiment

The Critical Assessment of PRedicted Interactions (CAPRI) experiment was designed in 2000 to test protein docking algorithms in blind predictions of the structure of protein-protein complexes. In four years, 17 complexes offered by crystallographers as targets prior to publication, have been subjected to structure prediction by docking their two components. Models of these complexes were submitted by predictor groups and assessed by comparing their geometry to the X-ray structure and by evaluating the quality of the prediction of the regions of interaction and of the pair wise residue contacts. Prediction was successful on 12 of the 17 targets, most of the failures being due to large conformation changes that the algorithms could not cope with. Progress in the prediction quality observed in four years indicates that the experiment is a powerful incentive to develop new procedures that allow for flexibility during docking and incorporate nonstructural information. We therefore call upon structural biologists who study protein-protein complexes to provide targets for further rounds of CAPRI predictions.     

A new, structurally nonredundant, diverse data set of protein-protein interfaces and its implications

Here, we present a diverse, structurally nonredundant data set of two-chain protein-protein interfaces derived from the PDB. Using a sequence order-independent structural comparison algorithm and hierarchical clustering, 3799 interface clusters are obtained. These yield 103 clusters with at least five nonhomologous members. We divide the clusters into three types. In Type I clusters, the global structures of the chains from which the interfaces are derived are also similar. This cluster type is expected because, in general, related proteins associate in similar ways. In Type II, the interfaces are similar; however, remarkably, the overall structures and functions of the chains are different. The functional spectrum is broad, from enzymes/inhibitors to immunoglobulins and toxins. The fact that structurally different monomers associate in similar ways, suggests "good" binding architectures. This observation extends a paradigm in protein science: It has been well known that proteins with similar structures may have different functions. Here, we show that it extends to interfaces. In Type III clusters, only one side of the interface is similar across the cluster. This structurally nonredundant data set provides rich data for studies of protein-protein interactions and recognition, cellular networks and drug design. In particular, it may be useful in addressing the difficult question of what are the favorable ways for proteins to interact. (The data set is available at http://protein3d.ncifcrf.gov/~keskino/ and http://home.ku.edu.tr/~okeskin/INTERFACE/INTERFACES.html.)     

Localization of protein-binding sites within families of proteins

We address the question of whether or not the positions of protein-binding sites on homologous protein structures are conserved irrespective of the identities of their binding partners. First, for each domain family in the Structural Classification of Proteins (SCOP), protein-binding sites are extracted from our comprehensive database of structurally defined binary domain interactions (PIBASE). Second, the binding sites within each family are superposed using a structural alignment of its members. Finally, the degree of localization of binding sites within each family is quantified by comparing it with localization expected by chance. We found that 72% of the 1847 SCOP domain families in PIBASE have binding sites with localization values greater than expected by chance. Moreover, 554 (30%) of these families have localizations that are statistically significant (i.e., more than four standard deviations away from the mean expected by chance). In contrast, only 144 (8%) families have significantly low localization. The absence of a significant correlation of the binding site localization with the average sequence and structural conservations in a family suggests that localization can be helpful for describing the functional diversity of protein-protein interactions, complementing measures of sequence and structural conservation. Consideration of the binding site localization may also result in spatial restraints for the modeling of protein assembly structures.     

Improved side-chain modeling for protein-protein docking

Success in high-resolution protein-protein docking requires accurate modeling of side-chain conformations at the interface. Most current methods either leave side chains fixed in the conformations observed in the unbound protein structures or allow the side chains to sample a set of discrete rotamer conformations. Here we describe a rapid and efficient method for sampling off-rotamer side-chain conformations by torsion space minimization during protein-protein docking starting from discrete rotamer libraries supplemented with side-chain conformations taken from the unbound structures, and show that the new method improves side-chain modeling and increases the energetic discrimination between good and bad models. Analysis of the distribution of side-chain interaction energies within and between the two protein partners shows that the new method leads to more native-like distributions of interaction energies and that the neglect of side-chain entropy produces a small but measurable increase in the number of residues whose interaction energy cannot compensate for the entropic cost of side-chain freezing at the interface. The power of the method is highlighted by a number of predictions of unprecedented accuracy in the recent CAPRI (Critical Assessment of PRedicted Interactions) blind test of protein-protein docking methods.     

High resolution crystallographic studies of alpha-hemolysin-phospholipid complexes define heptamer-lipid head group interactions: implication for understanding protein-lipid interactions

The alpha-hemolysin is an archetypal pore-forming protein that is secreted from Staphylococcus aureus as a water-soluble monomer. When the monomer binds to the membrane of a susceptible cell, the membrane-bound molecules assemble into the lytic heptamer. Although a bilayer or a bilayer-like environment are essential to toxin assembly, there is no high resolution information on toxin-phospholipid complexes. We have determined the structures of detergent-solubilized alpha-hemolysin heptamer bound to glycerophosphocholine or dipropanoyl glycerophosphocholine at 1.75-1.80 A resolution and 110 K. The phosphocholine head group binds to each subunit in a crevice between the rim and the stem domains. The quaternary ammonium group interacts primarily with aromatic residues, whereas the phosphodiester moiety interacts with a conserved arginine residue. These structures provide a molecular basis for understanding why alpha-hemolysin preferentially assembles on membranes comprised of phosphocholine lipids.     

Association of putative concave protein-binding sites with the fluctuation behavior of residues

Here, we propose a binding site prediction method based on the high frequency end of the spectrum in the native state of the protein structural dynamics. The spectrum is obtained using an elastic network model (GNM). High frequency vibrating (HFV) residues are determined from the fastest modes dynamics. HFV residue clusters and the associated surface patch residues are tested for their likelihood to locate at the binding interfaces using two different data sets, the Benchmark Set of mainly enzymes and antigen/antibodies and the Cluster Set of more diverse structures. The binding interface is identified to be within 7.5 A of the HFV residue clusters in the Benchmark Set and Cluster Set, for 77% and 70% of the structures, respectively. The success rate increases to 88% and 84%, respectively, by using the surface patches. The results suggest that concave binding interfaces, typically those of enzyme-binding sites, are enriched by the HFV residues. Thus, we expect that the association of HFV residues with the interfaces is mostly for enzymes. If, however, a binding region has invaginations and cavities, as in some of the antigen/antibodies and in cases in the Cluster data set, we expect it would be detected there too. This implies that binding sites possess several (inter-related) properties such as cavities, high packing density, conservation, and disposition for hotspots at binding surfaces. It further suggests that the high frequency vibrating residue-based approach is a potential tool for identification of regions likely to serve as protein-binding sites. The software is available at http://www.prc.boun.edu.tr/PRC/software.html.     

Potential of mean force for protein-protein interaction studies.

Calculating protein-protein interaction energies is crucial for understanding protein-protein associations. On the basis of the methodology of mean-field potential, we have developed an empirical approach to estimate binding free energy for protein-protein interactions. This knowledge-based approach has been used to derive distance-dependent free energies of protein complexes from a nonredundant training set in the Protein Data Bank (PDB), with a careful treatment of homology. We calculate atom pair potentials for 16 pair interactions, which can reflect the importance of hydrophobic interactions and specific hydrogen-bonding interactions. The derived potentials for hydrogen-bonding interactions show a valley of favorable interactions at a distance of approximately 3 A, corresponding to that of an established hydrogen bond. For the test set of 28 protein complexes, the calculated energies have a correlation coefficient of 0.75 compared with experimental binding free energies. The performance of the method in ranking the binding energies of different protein-protein complexes shows that the energy estimation can be applied to value binding free energies for protein-protein associations.     

Revisiting the Voronoi description of protein-protein interfaces

We developed a model of macromolecular interfaces based on the Voronoi diagram and the related alpha-complex, and we tested its properties on a set of 96 protein-protein complexes taken from the Protein Data Bank. The Voronoi model provides a natural definition of the interfaces, and it yields values of the number of interface atoms and of the interface area that have excellent correlation coefficients with those of the classical model based on solvent accessibility. Nevertheless, some atoms that do not lose solvent accessibility are part of the interface defined by the Voronoi model. The Voronoi model provides robust definitions of the curvature and of the connectivity of the interfaces, and leads to estimates of these features that generally agree with other approaches. Our implementation of the model allows an analysis of protein-water contacts that highlights the role of structural water molecules at protein-protein interfaces.     

Protein conformational changes studied by diffusion NMR spectroscopy: application to helix-loop-helix calcium binding proteins

Pulsed-field gradient (PFG) diffusion NMR spectroscopy studies were conducted with several helix-loop-helix regulatory Ca(2+)-binding proteins to characterize the conformational changes associated with Ca(2+)-saturation and/or binding targets. The calmodulin (CaM) system was used as a basis for evaluation, with similar hydrodynamic radii (R(h)) obtained for apo- and Ca(2+)-CaM, consistent with previously reported R(h) data. In addition, conformational changes associated with CaM binding to target peptides from myosin light chain kinase (MLCK), phosphodiesterase (PDE), and simian immunodeficiency virus (SIV) were accurately determined compared with small-angle X-ray scattering results. Both sets of data demonstrate the well-established collapse of the extended Ca(2+)-CaM molecule into a globular complex upon peptide binding. The R(h) of CaM complexes with target peptides from CaM-dependent protein kinase I (CaMKI) and an N-terminal portion of the SIV peptide (SIV-N), as well as the anticancer drug cisplatin were also determined. The CaMKI complex demonstrates a collapse analogous to that observed for MLCK, PDE, and SIV, while the SIV-N shows only a partial collapse. Interestingly, the covalent CaM-cisplatin complex shows a near complete collapse, not expected from previous studies. The method was extended to related calcium binding proteins to show that the R(h) of calcium and integrin binding protein (CIB), calbrain, and the calcium-binding region from soybean calcium-dependent protein kinase (CDPK) decrease on Ca(2+)-binding to various extents. Heteronuclear NMR spectroscopy suggests that for CIB and calbrain this is likely because of shifting the equilibrium from unfolded to folded conformations, with calbrain forming a dimer structure. These results demonstrate the utility of PFG-diffusion NMR to rapidly and accurately screen for molecular size changes on protein-ligand and protein-protein interactions for this class of proteins.     

The C-terminal domain of biotin protein ligase from E. coli is required for catalytic activity

Biotin protein ligase of Escherichia coli, the BirA protein, catalyses the covalent attachment of the biotin prosthetic group to a specific lysine of the biotin carboxyl carrier protein (BCCP) subunit of acetyl-CoA carboxylase. BirA also functions to repress the biotin biosynthetic operon and synthesizes its own corepressor, biotinyl-5'-AMP, the catalytic intermediate in the biotinylation reaction. We have previously identified two charge substitution mutants in BCCP, E119K, and E147K that are poorly biotinylated by BirA. Here we used site-directed mutagenesis to investigate residues in BirA that may interact with E119 or E147 in BCCP. None of the complementary charge substitution mutations at selected residues in BirA restored activity to wild-type levels when assayed with our BCCP mutant substrates. However, a BirA variant, in which K277 of the C-terminal domain was substituted with Glu, had significantly higher activity with E119K BCCP than did wild-type BirA. No function has been identified previously for the BirA C-terminal domain, which is distinct from the central domain thought to contain the ATP binding site and is known to contain the biotin binding site. Kinetic analysis of several purified mutant enzymes indicated that a single amino acid substitution within the C-terminal domain (R317E) and located some distance from the presumptive ATP binding site resulted in a 25-fold decrease in the affinity for ATP. Our data indicate that the C-terminal domain of BirA is essential for the catalytic activity of the enzyme and contributes to the interaction with ATP and the protein substrate, the BCCP biotin domain.     

Realistic protein-protein association rates from a simple diffusional model neglecting long-range interactions, free energy barriers, and landscape ruggedness

We develop a simple but rigorous model of protein-protein association kinetics based on diffusional association on free energy landscapes obtained by sampling configurations within and surrounding the native complex binding funnels. Guided by results obtained on exactly solvable model problems, we transform the problem of diffusion in a potential into free diffusion in the presence of an absorbing zone spanning the entrance to the binding funnel. The free diffusion problem is solved using a recently derived analytic expression for the rate of association of asymmetrically oriented molecules. Despite the required high steric specificity and the absence of long-range attractive interactions, the computed rates are typically on the order of 10(4)-10(6) M(-1) sec(-1), several orders of magnitude higher than rates obtained using a purely probabilistic model in which the association rate for free diffusion of uniformly reactive molecules is multiplied by the probability of a correct alignment of the two partners in a random collision. As the association rates of many protein-protein complexes are also in the 10(5)-10(6) M(-1) sec(-1) range, our results suggest that free energy barriers arising from desolvation and/or side-chain freezing during complex formation or increased ruggedness within the binding funnel, which are completely neglected in our simple diffusional model, do not contribute significantly to the dynamics of protein-protein association. The transparent physical interpretation of our approach that computes association rates directly from the size and geometry of protein-protein binding funnels makes it a useful complement to Brownian dynamics simulations.     

Determination of the human type I interferon receptor binding site on human interferon-alpha2 by cross saturation and an NMR-based model of the complex

Type I interferons (IFNs) are a family of homologous helical cytokines that exhibit pleiotropic effects on a wide variety of cell types, including antiviral activity and antibacterial, antiprozoal, immunomodulatory, and cell growth regulatory functions. Consequently, IFNs are the human proteins most widely used in the treatment of several kinds of cancer, hepatitis C, and multiple sclerosis. All type I IFNs bind to a cell surface receptor consisting of two subunits, IFNAR1 and IFNAR2, associating upon binding of interferon. The structure of the extracellular domain of IFNAR2 (R2-EC) was solved recently. Here we study the complex and the binding interface of IFNalpha2 with R2-EC using multidimensional NMR techniques. NMR shows that IFNalpha2 does not undergo significant structural changes upon binding to its receptor, suggesting a lock-and-key mechanism for binding. Cross saturation experiments were used to determine the receptor binding site upon IFNalpha2. The NMR data and previously published mutagenesis data were used to derive a docking model of the complex with an RMSD of 1 Angstrom, and its well-defined orientation between IFNalpha2 and R2-EC and the structural quality greatly improve upon previously suggested models. The relative ligand-receptor orientation is believed to be important for interferon signaling and possibly one of the parameters that distinguish the different IFN I subtypes. This structural information provides important insight into interferon signaling processes and may allow improvement in the development of therapeutically used IFNs and IFN-like molecules.     

"Zero-length" cross-linking in solid state as an approach for analysis of protein-protein interactions

We have developed a new approach for the analysis of interacting interfaces in protein complexes and protein quaternary structure based on cross-linking in the solid state. Protein complexes are freeze-dried under vacuum, and cross-links are introduced in the solid phase by dehydrating the protein in a nonaqueous solvent creating peptide bonds between amino and carboxyl groups of the interacting peptides. Cross-linked proteins are digested into peptides with trypsin in both H2(16)O and H(2)18O and then readily distinguished in mass spectra by characteristic 8 atomic mass unit (amu) shifts reflecting incorporation of two 18O atoms into each C terminus of proteolytic peptides. Computer analysis of mass spectrometry (MS) and MS/MS data is used to identify the cross-linked peptides. We demonstrated specificity and reproducibility of our method by cross-linking homo-oligomeric protein complexes of glutathione-S-transferase (GST) from Schistosoma japonicum alone or in a mixture of many other proteins. Identified cross-links were predominantly of amide origin, but six esters and thioesters were also found. The cross-linked peptides were validated against the GST monomer and dimer X-ray structures and by experimental (MS/MS) analyses. Some of the identified cross-links matched interacting peptides in the native 3D structure of GST, indicating that the structure of GST and its oligomeric complex remained primarily intact after freeze-drying. The pattern of oligomeric GST obtained in solid state was the same as that obtained in solution by Ru (II) Bpy(3)2+ catalyzed, oxidative "zero-length" cross-linking, confirming that it is feasible to use our strategy for analyzing the molecular interfaces of interacting proteins or peptides.     

Crystal structural analysis of protein-protein interactions drastically destabilized by a single mutation

The complex of barnase (bn) and barstar (bs), which has been widely studied as a model for quantitative analysis of protein-protein interactions, is significantly destabilized by a single mutation, namely, bs Asp39 --> Ala, which corresponds to a change of 7.7 kcal x mol(-1) in the free energy of binding. However, there has been no structural information available to explain such a drastic destabilization. In the present study, we determined the structure of the mutant complex at 1.58 A resolution by X-ray crystallography. The complex was similar to the wild-type complex in terms of overall and interface structures; however, the hydrogen bond network mediated by water molecules at the interface was significantly different. Several water molecules filled the cavity created by the mutation and consequently caused rearrangement of the hydrated water molecules at the interface. The water molecules were redistributed into a channel-like structure that penetrated into the complex. Furthermore, molecular dynamics simulations showed that the mutation increased the mobility of water molecules at the interface. Since such a drastic change in hydration was not observed in other mutant complexes of bn and bs, the significant destabilization of the interaction may be due to this channel-like structure of hydrated water molecules.     

Free energies of protein-protein association determined by electrospray ionization mass spectrometry correlate accurately with values obtained by solution methods

The advantages of electrospray ionization mass spectrometry (ESIMS) to measure relative solution-phase affinities of tightly bound protein-protein complexes are demonstrated with selected variants of the Bacillus amyloliquefaciens protein barstar (b*) and the RNAase barnase (bn), which form protein-protein complexes with a range of picomolar to nanomolar dissociation constants. A novel chemical annealing procedure rapidly establishes equilibrium in solutions containing competing b* variants with limiting bn. The relative ion abundances of the complexes and those of the competing unbound monomers are shown to reflect the relative solution-phase concentrations of those respective species. No measurable dissociation of the complexes occurs either during ESI or mass detection, nor is there any evidence for nonspecific binding at protein concentrations < 25 microM. Differences in DeltaDeltaG of dissociation between variants were determined with precisions < 0.1 kcal/mol. The DeltaDeltaG values obtained deviate on average by 0.26 kcal/mol from those measured with a solution-phase enzyme assay. It is demonstrated that information about the protein conformation and covalent modifications can be obtained from differences in mass and charge state distributions. This method serves as a rapid and precise means to interrogate protein-protein-binding surfaces for complexes that have affinities in the picomolar to nanomolar range.     

Structural basis for docking of peroxisomal membrane protein carrier Pex19p onto its receptor Pex3p

Peroxisomes require peroxin (Pex) proteins for their biogenesis. The interaction between Pex3p, which resides on the peroxisomal membrane, and Pex19p, which resides in the cytosol, is crucial for peroxisome formation and the post-translational targeting of peroxisomal membrane proteins (PMPs). It is not known how Pex3p promotes the specific interaction with Pex19p for the purpose of PMP translocation. Here, we present the three-dimensional structure of the complex between a cytosolic domain of Pex3p and the binding-region peptide of Pex19p. The overall shape of Pex3p is a prolate spheroid with a novel fold, the 'twisted six-helix bundle.' The Pex19p-binding site is at an apex of the Pex3p spheroid. A 16-residue region of the Pex19p peptide forms an α-helix and makes a contact with Pex3p; this helix is disordered in the unbound state. The Pex19p peptide contains a characteristic motif, consisting of the leucine triad (Leu18, Leu21, Leu22), and Phe29, which are critical for the Pex3p binding and peroxisome biogenesis.     

The nature and character of the transition state for the ADP-ribosyltransferase reaction

Exotoxin A (ExoA) from Pseudomonas aeruginosa is an important virulence factor that belongs to a class of exotoxins that are secreted by pathogenic bacteria which cause human diseases such as cholera, diphtheria, pneumonia and whooping cough. We present the first crystal structures, to our knowledge, of ExoA in complex with elongation factor 2 (eEF2) and intact NAD(+), which indicate a direct role of two active-site loops in ExoA during the catalytic cycle. One loop moves to form a solvent cover for the active site of the enzyme and reaches towards the target residue (diphthamide) in eEF2 forming an important hydrogen bond. The NAD(+) substrate adopts a conformation remarkably different from that of the NAD(+) analogue, betaTAD, observed in previous structures, and fails to trigger any loop movements. Mutational studies of the two loops in the toxin identify several residues important for catalytic activity, in particular Glu 546 and Arg 551, clearly supporting the new complex structures. On the basis of these data, we propose a transition-state model for the toxin-catalysed reaction.     

Structural and thermodynamic effects of ANS binding to human interleukin-1 receptor antagonist

Although 8-anilinonaphthalene-1-sulfonic acid (ANS) is frequently used in protein folding studies, the structural and thermodynamic effects of its binding to proteins are not well understood. Using high-resolution two-dimensional NMR and human interleukin-1 receptor antagonist (IL-1ra) as a model protein, we obtained detailed information on ANS-protein interactions in the absence and presence of urea. The effects of ambient to elevated temperatures on the affinity and specificity of ANS binding were assessed from experiments performed at 25 degrees C and 37 degrees C. Overall, the affinity of ANS was lower at 37 degrees C compared to 25 degrees C, but no significant change in the site specificity of binding was observed from the chemical shift perturbation data. The same site-specific binding was evident in the presence of 5.2 M urea, well within the unfolding transition region, and resulted in selective stabilization of the folded state. Based on the two-state denaturation mechanism, ANS-dependent changes in the protein stability were estimated from relative intensities of two amide resonances specific to the folded and unfolded states of IL-1ra. No evidence was found for any ANS-induced partially denatured or aggregated forms of IL-1ra throughout the experimental conditions, consistent with a cooperative and reversible denaturation process. The NMR results support earlier observations on the tendency of ANS to interact with solvent-exposed positively charged sites on proteins. Under denaturing conditions, ANS binding appears to be selective to structured states rather than unfolded conformations. Interestingly, the binding occurs within a previously identified aggregation-critical region in IL-1ra, thus providing an insight into ligand-dependent protein aggregation.