Prediction of hot spot residues at protein-protein interfaces by combining machine learning and energy-based methods

Background  

Alanine scanning mutagenesis is a powerful experimental methodology for investigating the structural and energetic characteristics of protein complexes. Individual amino-acids are systematically mutated to alanine and changes in free energy of binding (ΔΔG) measured. Several experiments have shown that protein-protein interactions are critically dependent on just a few residues ("hot spots") at the interface. Hot spots make a dominant contribution to the free energy of binding and if mutated they can disrupt the interaction. As mutagenesis studies require significant experimental efforts, there is a need for accurate and reliable computational methods. Such methods would also add to our understanding of the determinants of affinity and specificity in protein-protein recognition.     

Evidence of Conformational Selection Driving the Formation of Ligand Binding Sites in Protein-Protein Interfaces

Many protein-protein interactions (PPIs) are compelling targets for drug discovery, and in a number of cases can be disrupted by small molecules. The main goal of this study is to examine the mechanism of binding site formation in the interface region of proteins that are PPI targets by comparing ligand-free and ligand-bound structures. To avoid any potential bias, we focus on ensembles of ligand-free protein conformations obtained by nuclear magnetic resonance (NMR) techniques and deposited in the Protein Data Bank, rather than on ensembles specifically generated for this study. The measures used for structure comparison are based on detecting binding hot spots, i.e., protein regions that are major contributors to the binding free energy. The main tool of the analysis is computational solvent mapping, which explores the surface of proteins by docking a large number of small “probe” molecules. Although we consider conformational ensembles obtained by NMR techniques, the analysis is independent of the method used for generating the structures. Finding the energetically most important regions, mapping can identify binding site residues using ligand-free models based on NMR data. In addition, the method selects conformations that are similar to some peptide-bound or ligand-bound structure in terms of the properties of the binding site. This agrees with the conformational selection model of molecular recognition, which assumes such pre-existing conformations. The analysis also shows the maximum level of similarity between unbound and bound states that is achieved without any influence from a ligand. Further shift toward the bound structure assumes protein-peptide or protein-ligand interactions, either selecting higher energy conformations that are not part of the NMR ensemble, or leading to induced fit. Thus, forming the sites in protein-protein interfaces that bind peptides and can be targeted by small ligands always includes conformational selection, although other recognition mechanisms may also be involved.     

Hot regions in protein--protein interactions: the organization and contribution of structurally conserved hot spot residues

Structurally conserved residues at protein-protein interfaces correlate with the experimental alanine-scanning hot spots. Here, we investigate the organization of these conserved, computational hot spots and their contribution to the stability of protein associations. We find that computational hot spots are not homogeneously distributed along the protein interfaces; rather they are clustered within locally tightly packed regions. Within the dense clusters, they form a network of interactions and consequently their contributions to the stability of the complex are cooperative; however the contributions of independent clusters are additive. This suggests that the binding free energy is not a simple summation of the single hot spot residue contributions. As expected, around the hot spot residues we observe moderately conserved residues, further highlighting the crucial role of the conserved interactions in the local densely packed environment. The conserved occurrence of these organizations suggests that they are advantageous for protein-protein associations. Interestingly, the total number of hydrogen bonds and salt bridges contributed by hot spots is as expected. Thus, H-bond forming residues may use a "hot spot for water exclusion" mechanism. Since conserved residues are located within highly packed regions, water molecules are easily removed upon binding, strengthening electrostatic contributions of charge-charge interactions. Hence, the picture that emerges is that protein-protein associations are optimized locally, with the clustered, networked, highly packed structurally conserved residues contributing dominantly and cooperatively to the stability of the complex. When addressing the crucial question of "what are the preferred ways of proteins to associate", these findings point toward a critical involvement of hot regions in protein-protein interactions.     

Molecular principles of the interactions of disordered proteins

Thorough knowledge of the molecular principles of protein-protein recognition is essential to our understanding of protein function at the cellular level. Whereas interactions of ordered proteins have been analyzed in great detail, complexes of intrinsically unstructured/disordered proteins (IUPs) have hardly been addressed so far. Here, we have collected a database of 39 complexes of experimentally verified IUPs, and compared their interfaces with those of 72 complexes of ordered, globular proteins. The characteristic differences found between the two types of complexes suggest that IUPs represent a distinct molecular implementation of the principles of protein-protein recognition. The interfaces do not differ in size, but those of IUPs cover a much larger part of the surface of the protein than for their ordered counterparts. Moreover, IUP interfaces are significantly more hydrophobic relative to their overall amino acid composition, but also in absolute terms. They rely more on hydrophobic-hydrophobic than on polar-polar interactions. Their amino acids in the interface realize more intermolecular contacts, which suggests a better fit with the partner due to induced folding upon binding that results in a better adaptation to the partner. The two modes of interaction also differ in that IUPs usually use only a single continuous segment for partner binding, whereas the binding sites of ordered proteins are more segmented. Probably, all these features contribute to the increased evolutionary conservation of IUP interface residues. These noted molecular differences are also manifested in the interaction energies of IUPs. Our approximation of these by low-resolution force-fields shows that IUPs gain much more stabilization energy from intermolecular contacts, than from folding, i.e. they use their binding energy for folding. Overall, our findings provide a structural rationale to the prior suggestions that many IUPs are specialized for functions realized by protein-protein interactions.     

Small-world network approach to identify key residues in protein-protein interaction

We show that protein complexes can be represented as small-world networks, exhibiting a relatively small number of highly central amino-acid residues occurring frequently at protein-protein interfaces. We further base our analysis on a set of different biological examples of protein-protein interactions with experimentally validated hot spots, and show that 83% of these predicted highly central residues, which are conserved in sequence alignments and nonexposed to the solvent in the protein complex, correspond to or are in direct contact with an experimentally annotated hot spot. The remaining 17% show a general tendency to be close to an annotated hot spot. On the other hand, although there is no available experimental information on their contribution to the binding free energy, detailed analysis of their properties shows that they are good candidates for being hot spots. Thus, highly central residues have a clear tendency to be located in regions that include hot spots. We also show that some of the central residues in the protein complex interfaces are central in the monomeric structures before dimerization and that possible information relating to hot spots of binding free energy could be obtained from the unbound structures.     

o-Nitrotyrosine and p-iodophenylalanine as spectroscopic probes for structural characterization of SH3 complexes

High-throughput screening of protein-protein and protein-peptide interactions is of high interest both for biotechnological and pharmacological applications. Here, we propose the use of the noncoded amino acids o-nitrotyrosine and p-iodophenylalanine as spectroscopic probes in combination with circular dichroism and fluorescence quenching techniques (i.e., collisional quenching and resonance energy transfer) as a means to determine the peptide orientation in complexes with SH3 domains. Proline-rich peptides bind SH3 modules in two alternative orientations, according to their sequence motifs, classified as class I and class II. The method was tested on an SH3 domain from a yeast myosin that is known to recognize specifically class I peptides. We exploited the fluorescence quenching effects induced by o-nitrotyrosine and p-iodophenylalanine on the fluorescence signal of a highly conserved Trp residue, which is the signature of SH3 domains and sits directly in the binding pocket. In particular, we studied how the introduction of the two probes at different positions of the peptide sequence (i.e., N-terminally or C-terminally) influences the spectroscopic properties of the complex. This approach provides clear-cut evidence of the orientation of the binding peptide in the SH3 pocket. The chemical strategy outlined here can be easily extended to other protein modules, known to bind linear sequence motifs in a highly directional manner.     

Identification of hot spot residues at protein-protein interface

It is known that binding free energy of protein-protein interaction is mainly contributed by hot spot (high energy) interface residues. Here, we investigate the characteristics of hot spots by examining inter-atomic sidechain-sidechain interactions using a dataset of 296 alanine-mutated interface residues. Results show that hot spots participate in strong and energetically favorable sidechain-sidechain interactions. Subsequently, we describe a novel, yet simple 'hot spot' prediction model with an accuracy that is similar to many available approaches. The model is also shown to efficiently distinguish specific protein-protein interactions from non-specific interactions.     

Role of β-lactamase residues in a common interface for binding the structurally unrelated inhibitory proteins BLIP and BLIP-II

The β-lactamase inhibitory proteins (BLIPs) are a model system for examining molecular recognition in protein-protein interactions. BLIP and BLIP-II are structurally unrelated proteins that bind and inhibit TEM-1 β-lactamase. Both BLIPs share a common binding interface on TEM-1 and make contacts with many of the same TEM-1 surface residues. BLIP-II, however, binds TEM-1 over 150-fold tighter than BLIP despite the fact that it has fewer contact residues and a smaller binding interface. The role of eleven TEM-1 amino acid residues that contact both BLIP and BLIP-II was examined by alanine mutagenesis and determination of the association (k_on) and dissociation (k_off) rate constants for binding each partner. The substitutions had little impact on association rates and resulted in a wide range of dissociation rates as previously observed for substitutions on the BLIP side of the interface. The substitutions also had less effect on binding affinity for BLIP than BLIP-II. This is consistent with the high affinity and small binding interface of the TEM-1-BLIP-II complex, which predicts per residue contributions should be higher for TEM-1 binding to BLIP-II versus BLIP. Two TEM-1 residues (E104 and M129) were found to be hotspots for binding BLIP while five (L102, Y105, P107, K111, and M129) are hotspots for binding BLIP-II with only M129 as a common hotspot for both. Thus, although the same TEM-1 surface binds to both BLIP and BLIP-II, the distribution of binding energy on the surface is different for the two target proteins, that is, different binding strategies are employed.     

Quantifying the importance of protein conformation on ligand migration in myoglobin

Myoglobin (Mb) is a model system for ligand binding and migration. The energy barriers (ΔG) for ligand migration in Mb have been studied in the past by experiment and theory and significant differences between different approaches were found. From experiment, it is known that Mb can assume a large number of conformational substates. In this work, these substates are investigated as a possible source of the differences in migration barriers. We show that the initial structure significantly affects the calculated ΔG for a particular transition and that fluctuations in barrier heights δΔG are of similar magnitude as the free energy barriers themselves. The sensitivity of ΔG to the initial structure is compared to other sources of errors. Different protein structures can affect the calculated ΔG by up to 4 kcal/mol, whereas differences between simple point charge models and more elaborate multipolar charge models for the CO-ligand are smaller by a factor of two. Analysis of the structural changes underlying the large effect of the conformational substate reveals the importance of coupling between protein and ligand motion for migration.     

Effective prediction model and determination of binding residues influential for inhibitors targeting HIV-1 integrase-LEDGF/p75 interface by employing solvent accessible surface area energy as key determinant

Abstract

Development of a highly accurate prediction model for protein–ligand inhibition has been a major challenge in drug discovery. Herein, we describe a novel predictive model for the inhibition of HIV-1 integrase (IN)-LEDGF/p75 protein-protein interaction. The model was constructed using energy parameters approximated from molecular dynamics (MD) simulations and molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) calculations. Chemometric analysis using partial least squares (PLS) regression revealed that solvent accessible surface area energy (ΔG_SASA) is the major determinant parameter contributing greatly to the prediction accuracy. PLS prediction model on the ΔG_SASA values collected from 41 complexes yielded a strong correlation between the predicted and the actual inhibitory activities (R² = 0.9666, RMSEC of pIC₅₀ values = 0.0890). Additionally, for the test set of 14 complexes, the model performed satisfactorily with very low pIC₅₀ errors (Q² = 0.5168, RMSEP = 0.3325). A strong correlation between the buried surface areas on the IN protein, when bound with IN-LEDGF/p75 inhibitors, and the respective ΔG_SASA values was also obtained. Furthermore, the current method could identify ‘hot spots’of amino acid residues highly influential to the inhibitory activity prediction. This could present fruitful implications in binding site determination and future inhibitor developments targeting protein-protein interactions.

Communicated by Ramaswamy H. Sarma     

Detergents as intrinsic P-glycoprotein substrates and inhibitors

We assessed the interaction of three electrically neutral detergents (Triton X-100, C₁₂EO₈, and Tween 80) with P-glycoprotein (ABCB1, MDR1) and identified the molecular elements responsible for this interaction. To this purpose we titrated P-glycoprotein in inside-out plasma membrane vesicles of MDR1-transfected mouse embryo fibroblasts (NIH-MDR1-G185) with the detergents below their critical micelle concentration, CMC. The P-glycoprotein ATPase measured as a function of the detergent concentration yielded bell-shaped activity curves which were evaluated with a two-site binding model. The lipid-water partition coefficient and the transporter-water binding constant of the detergents were measured independently. Knowledge of these two parameters allowed assessment of the free energy of detergent binding to P-glycoprotein in the lipid membrane, ΔG_tl⁰, that reflects the direct detergent-transporter affinity. It increased as the number of ethoxyl groups increased, suggesting that these hydrogen bond acceptor groups are the key elements for the detergent-transporter interaction in the lipid membrane. The free energy of binding to P-glycoprotein per ethoxyl group (EO) was determined as approximately ΔG_EO⁰ = − 1.6 kJ/mol. The present findings moreover document that, depending on the concentration applied, detergents are intrinsic substrates for, or inhibitors of P-glycoprotein.     

Free energy contributions to direct readout of a DNA sequence

The energetic contributions of individual DNA-contacting side chains to specific DNA recognition in the human papillomavirus 16 E2C-DNA complex is small (less than 1.0 kcal mol(-1)), independent of the physical and chemical nature of the interaction, and is strictly additive. The sum of the individual contributions differs 1.0 kcal mol(-1) from the binding energy of the wild-type protein. This difference corresponds to the contribution from the deformability of the DNA, known as "indirect readout." Thus, we can dissect the energetic contribution to DNA binding into 90% direct and 10% indirect readout components. The lack of high energy interactions indicates the absence of "hot spots," such as those found in protein-protein interfaces. These results are compatible with a highly dynamic and "wet" protein-DNA interface, yet highly specific and tight, where individual interactions are constantly being formed and broken.     

Hydrophile scanning as a complement to alanine scanning for exploring and manipulating protein-protein recognition: application to the Bim BH3 domain

Alanine scanning has been widely employed as a method of identifying side chains that play important roles in protein-protein and protein-peptide interactions. Here we show how an analogous and complementary technique, hydrophile scanning, can provide additional insight on such interactions. Mutation of a wild-type residue to alanine removes most of the side-chain atoms, and the effect of this removal is typically interpreted to indicate contribution of the deleted side chain to the stability of the complex. Hydrophile scanning involves systematic mutation of wild-type residues to a cationic or anionic residue (lysine or glutamic acid, in this case). We find that the results of these mutations provide insights on interactions between polypeptide surfaces that are complementary to the information obtained via alanine scanning. We have applied this technique to a peptide that corresponds to the BH3 domain of the pro-apoptotic protein Bim. The wild-type Bim BH3 domain binds strongly to the anti-apoptotic proteins Bcl-x(L) and Mcl-1. Combining information from the alanine, lysine, and glutamic acid scans has enabled us to identify Bim BH3 domain mutants containing only two or three sequence changes that bind very selectively either to Bcl-x(L) or Mcl-1. Our findings suggest that hydrophile scanning may prove to be a broadly useful tool for revealing sources of protein-protein recognition and for engineering selectivity into natural sequences.     

Protein hot spots: the islands of stability

Understanding the structural basis of protein-protein interactions (PPIs) may shed light on the organization and functioning of signal transduction and metabolic networks and may assist in structure-based design of ligands (drugs) targeting protein-protein interfaces. The residues at the bimolecular interface, designated as the hot spots, contribute most of the free binding energy of PPI. To date, there is no conclusive atomistic explanation for the unique functional properties of the hot spots. We hypothesized that backbone compliance may play a role in protein-protein recognition and in the mechanism of binding of small-molecule compounds to protein surfaces. We used a steered molecular dynamics simulation to explore the compliance properties of the backbone of surface-exposed residues in several model proteins: interleukin-2, mouse double minute protein 2 and proliferating cell nuclear antigen. We demonstrated that protein surfaces exhibit distinct patterns in which highly immobile residues form defined clusters ("stability patches") alternating with areas of moderate to high mobility. These "stability patches" tend to localize in functionally important regions involved in protein-protein recognition. We propose a mechanism by which the distinct structural organization of the hot spots may contribute to their role in mediating PPI and facilitating binding of structurally diverse small-molecule compounds to protein surfaces.     

Protein-protein recognition and interaction hot spots in an antigen-antibody complex: free energy decomposition identifies "efficient amino acids"

The molecular mechanics Poisson-Boltzmann surface area (MM/PBSA) method was applied to the study of the protein-protein complex between a camelid single chain variable domain (cAb-Lys3) and hen egg white lysozyme (HEL), and between cAb-Lys3 and turkey egg white lysozyme (TEL). The electrostatic energy was estimated by solving the linear Poisson-Boltzmann equation. A free energy decomposition scheme was developed to determine binding energy hot spots of each complex. The calculations identified amino acids of the antibody that make important contributions to the interaction with lysozyme. They further showed the influence of small structural variations on the energetics of binding and they showed that the antibody amino acids that make up the hot spots are organized in such a way as to mimic the lysozyme substrate. Through further analysis of the results, we define the concept of "efficient amino acids," which can provide an assessment of the binding potential of a particular hot spot interaction. This information, in turn, can be useful in the rational design of small molecules that mimic the antibody. The implications of using free energy decomposition to identify regions of a protein-protein complex that could be targeted by small molecules inhibitors are discussed.     

Analysis of molecular recognition features (MoRFs)

Several proteomic studies in the last decade revealed that many proteins are either completely disordered or possess long structurally flexible regions. Many such regions were shown to be of functional importance, often allowing a protein to interact with a large number of diverse partners. Parallel to these findings, during the last five years structural bioinformatics has produced an explosion of results regarding protein-protein interactions and their importance for cell signaling. We studied the occurrence of relatively short (10-70 residues), loosely structured protein regions within longer, largely disordered sequences that were characterized as bound to larger proteins. We call these regions molecular recognition features (MoRFs, also known as molecular recognition elements, MoREs). Interestingly, upon binding to their partner(s), MoRFs undergo disorder-to-order transitions. Thus, in our interpretation, MoRFs represent a class of disordered region that exhibits molecular recognition and binding functions. This work extends previous research showing the importance of flexibility and disorder for molecular recognition. We describe the development of a database of MoRFs derived from the RCSB Protein Data Bank and present preliminary results of bioinformatics analyses of these sequences. Based on the structure adopted upon binding, at least three basic types of MoRFs are found: α-MoRFs, β-MoRFs, and ι-MoRFs, which form α-helices, β-strands, and irregular secondary structure when bound, respectively. Our data suggest that functionally significant residual structure can exist in MoRF regions prior to the actual binding event. The contribution of intrinsic protein disorder to the nature and function of MoRFs has also been addressed. The results of this study will advance the understanding of protein-protein interactions and help towards the future development of useful protein-protein binding site predictors.     

Revisiting the Association of Cationic Groove-Binding Drugs to DNA Using a Poisson-Boltzmann Approach

Proper modeling of nonspecific salt-mediated electrostatic interactions is essential to understanding the binding of charged ligands to nucleic acids. Because the linear Poisson-Boltzmann equation (PBE) and the more approximate generalized Born approach are applied routinely to nucleic acids and their interactions with charged ligands, the reliability of these methods is examined vis-à-vis an efficient nonlinear PBE method. For moderate salt concentrations, the negative derivative, SK_pred, of the electrostatic binding free energy, ΔG_el, with respect to the logarithm of the 1:1 salt concentration, [M⁺], for 33 cationic minor groove drugs binding to AT-rich DNA sequences is shown to be consistently negative and virtually constant over the salt range considered (0.1-0.4 M NaCl). The magnitude of SK_pred is approximately equal to the charge on the drug, as predicted by counterion condensation theory (CCT) and observed in thermodynamic binding studies. The linear PBE is shown to overestimate the magnitude of SK_pred, whereas the nonlinear PBE closely matches the experimental results. The PBE predictions of SK_pred were not correlated with ΔG_el in the presence of a dielectric discontinuity, as would be expected from the CCT. Because this correlation does not hold, parameterizing the PBE predictions of ΔG_el against the reported experimental data is not possible. Moreover, the common practice of extracting the electrostatic and nonelectrostatic contributions to the binding of charged ligands to biopolyelectrolytes based on the simple relation between experimental SK values and the electrostatic binding free energy that is based on CCT is called into question by the results presented here. Although the rigid-docking nonlinear PB calculations provide reliable predictions of SK_pred, at least for the charged ligand-nucleic acid complexes studied here, accurate estimates of ΔG_el will require further development in theoretical and experimental approaches.     

Protein-protein interactions; coupling of structurally conserved residues and of hot spots across interfaces. Implications for docking

Hot spot residues contribute dominantly to protein-protein interactions. Statistically, conserved residues correlate with hot spots, and their occurrence can distinguish between binding sites and the remainder of the protein surface. The hot spot and conservation analyses have been carried out on one side of the interface. Here, we show that both experimental hot spots and conserved residues tend to couple across two-chain interfaces. Intriguingly, the local packing density around both hot spots and conserved residues is higher than expected. We further observe a correlation between local packing density and experimental deltadeltaG. Favorable conserved pairs include Gly coupled with aromatics, charged and polar residues, as well as aromatic residue coupling. Remarkably, charged residue couples are underrepresented. Overall, protein-protein interactions appear to consist of regions of high and low packing density, with the hot spots organized in the former. The high local packing density in binding interfaces is reminiscent of protein cores.     

Thermodynamic Profiling of HIV RREIIB RNA-Zinc Finger Interactions

The interactions between the HIV Rev-responsive element (RRE) RNA and the HIV regulatory protein Rev, are crucial for the HIV life-cycle. Earlier, we showed that single C₂H₂ zinc fingers (znfs) have the same binding site as the Rev peptide and exhibit nanomolar affinity. In this study, the specific role of amino acid side chains and molecular processes involved with complex formation were investigated by perturbation of the binding energetics via changes in temperature, pH, buffers, and salt concentrations, as well as znf and RNA mutations, by isothermal titration calorimetry. Interestingly, despite the large cationic charge on the znfs, the number of interactions with the RNA phosphate backbone was lower than intuitively expected. The presence of binding induced protonation was established by ITC and localized by NMR to a histidine on the znf β-sheet. The ΔC_p of znf-RNA binding was observed to be substantially negative and could not be accounted for by conventional solvent-accessible surface area models. An alternative model, based on the extent of hydrogen bond changes as a result of differences in ligand-induced water displacement at the binding site, provided reasonable explanation of the trends in ΔC_p, as well as ΔH and ΔS. Our studies show that incorporation of favorable interactions at the solvent-excluded binding interface can be used to alleviate the unfavorable enthalpic penalties of displacing water molecules from the hydrated RNA surface.