Predicting important residues and interaction pathways in proteins using Gaussian Network Model: binding and stability of HLA proteins

Design of Protein Segments and Peptides for Binding to Protein Targets

Recent years have witnessed a rise in methods for accurate prediction of structure and design of novel functional proteins. Design of functional protein fragments and peptides occupy a small, albeit unique, space within the general field of protein design. While the smaller size of these peptides allows for more exhaustive computational methods, flexibility in their structure and sparsity of data compared to proteins, as well as presence of noncanonical building blocks, add additional challenges to their design. This review summarizes the current advances in the design of protein fragments and peptides for binding to targets and discusses the challenges in the field, with an eye toward future directions.

Sequence-structure-function relationships in class I MHC: A local frustration perspective

Class I Major Histocompatibility Complex (MHC) binds short antigenic peptides with the help of Peptide Loading Complex (PLC), and presents them to T-cell Receptors (TCRs) of cytotoxic T-cells and Killer-cell Immunglobulin-like Receptors (KIRs) of Natural Killer (NK) cells. With more than 10000 alleles, human MHC (Human Leukocyte Antigen, HLA) is the most polymorphic protein in humans. This allelic diversity provides a wide coverage of peptide sequence space, yet does not affect the three-dimensional structure of the complex. Moreover, TCRs mostly interact with HLA in a common diagonal binding mode, and KIR-HLA interaction is allele-dependent. With the aim of establishing a framework for understanding the relationships between polymorphism (sequence), structure (conserved fold) and function (protein interactions) of the human MHC, we performed here a local frustration analysis on pMHC homology models covering 1436 HLA I alleles. An analysis of local frustration profiles indicated that (1) variations in MHC fold are unlikely due to minimally-frustrated and relatively conserved residues within the HLA peptide-binding groove, (2) high frustration patches on HLA helices are either involved in or near interaction sites of MHC with the TCR, KIR, or tapasin of the PLC, and (3) peptide ligands mainly stabilize the F-pocket of HLA binding groove.

Gaussian network model can be enhanced by combining solvent accessibility in proteins

Gaussian network model (GNM), regarded as the simplest and most representative coarse-grained model, has been widely adopted to analyze and reveal protein dynamics and functions. Designing a variation of the classical GNM, by defining a new Kirchhoff matrix, is the way to improve the residue flexibility modeling. We combined information arising from local relative solvent accessibility (RSA) between two residues into the Kirchhoff matrix of the parameter-free GNM. The undetermined parameters in the new Kirchhoff matrix were estimated by using particle swarm optimization. The usage of RSA was motivated by the fact that our previous work using RSA based linear regression model resulted out higher prediction quality of the residue flexibility when compared with the classical GNM and the parameter free GNM. Computational experiments, conducted based on one training dataset, two independent datasets and one additional small set derived by molecular dynamics simulations, demonstrated that the average correlation coefficients of the proposed RSA based parameter-free GNM, called RpfGNM, were significantly increased when compared with the parameter-free GNM. Our empirical results indicated that a variation of the classical GNMs by combining other protein structural properties is an attractive way to improve the quality of flexibility modeling.

Identification of Hot Spots in Protein Structures Using Gaussian Network Model and Gaussian Naive Bayes

Residue fluctuations in protein structures have been shown to be highly associated with various protein functions. Gaussian network model (GNM), a simple representative coarse-grained model, was widely adopted to reveal function-related protein dynamics. We directly utilized the high frequency modes generated by GNM and further performed Gaussian Naive Bayes (GNB) to identify hot spot residues. Two coding schemes about the feature vectors were implemented with varying distance cutoffs for GNM and sliding window sizes for GNB based on tenfold cross validations: one by using only a single high mode and the other by combining multiple modes with the highest frequency. Our proposed methods outperformed the previous work that did not directly utilize the high frequency modes generated by GNM, with regard to overall performance evaluated using F1 measure. Moreover, we found that inclusion of more high frequency modes for a GNB classifier can significantly improve the sensitivity. The present study provided additional valuable insights into the relation between the hot spots and the residue fluctuations.

Recent findings and future directions for interpolar mitotic kinesin inhibitors in cancer therapy

The kinesin class of microtubule-associated motor proteins present attractive anti-cancer targets owing to their roles in key functions in dividing cells. Two interpolar mitotic kinesins Eg5 and HSET have opposing motor functions in mitotic spindle assembly with respect to microtubule movement, but both offer opportunities to develop cancer selective therapeutic agents. Here, we summarize the progress to date in developing inhibitors of Eg5 and HSET, with an emphasis on structural biology insights into the binding modes of allosteric inhibitors, compound selectivity and mechanisms of action of different chemical scaffolds. We discuss translation of preclinical studies to clinical experience with Eg5 inhibitors, recent findings on potential resistance mechanisms, and explore the implications for future anticancer drug development against these targets.

'MHC-I-opathy'-unified concept for spondyloarthritis and Behçet disease

The spondyloarthropathies comprise ankylosing spondylitis (AS), reactive arthritis, psoriatic arthritis (PsA) and arthritis associated with inflammatory bowel disease. In this Perspectives article, we describe how Behçet disease and several clinically distinct spondyloarthropathies-all associated with MHC class I (MHC-I) alleles such as HLA-B(*)51, HLA-C(*)0602 and HLA-B(*)27 and epistatic ERAP-1 interactions-have a shared immunopathogenetic basis. As a unifying concept, we propose that barrier dysfunction in environmentally exposed organs such as the skin, and aberrant innate immune reactions at sites of mechanical stress, can often trigger secondary adaptive immune CD8(+) T-cell responses with prominent neutrophilic inflammation that culminate in exacerbation and recurrence of these diseases. Of note, these 'MHC-I-opathies' show a differential immunopathology, probably reflecting antigenic differences within target tissues: HLA-B(*)51 is linked to ocular and mucocutaneous disease but not gut involvement, and HLA-C(*)0602 is linked to type I psoriasis but not scalp or nail disease.

Conformational Motions and Functionally Key Residues for Vitamin B12 Transporter BtuCD-BtuF Revealed by Elastic Network Model with a Function-Related Internal Coordinate

BtuCD-BtuF from Escherichia coli is a binding protein-dependent adenosine triphosphate (ATP)-binding cassette (ABC) transporter system that uses the energy of ATP hydrolysis to transmit vitamin B12 across cellular membranes. Experimental studies have showed that during the transport cycle, the transporter undergoes conformational transitions between the "inward-facing" and "outward-facing" states, which results in the open-closed motions of the cytoplasmic gate of the transport channel. The opening-closing of the channel gate play critical roles for the function of the transporter, which enables the substrate vitamin B12 to be translocated into the cell. In the present work, the extent of opening of the cytoplasmic gate was chosen as a function-related internal coordinate. Then the mean-square fluctuation of the internal coordinate, as well as the cross-correlation between the displacement of the internal coordinate and the movement of each residue in the protein, were calculated based on the normal mode analysis of the elastic network model to analyze the function-related motions encoded in the structure of the system. In addition, the key residues important for the functional motions of the transporter were predicted by using a perturbation method. In order to facilitate the calculations, the internal coordinate was introduced as one of the axes of the coordinate space and the conventional Cartesian coordinate space was transformed into the internal/Cartesian space with linear approximation. All the calculations were carried out in this internal/Cartesian space. Our method can successfully identify the functional motions and key residues for the transporter BtuCD-BtuF, which are well consistent with the experimental observations.

The role of binding site on the mechanical unfolding mechanism of ubiquitin

We apply novel atomistic simulations based on potential energy surface exploration to investigate the constant force-induced unfolding of ubiquitin. At the experimentally-studied force clamping level of 100 pN, we find a new unfolding mechanism starting with the detachment between β₅ and β₃ involving the binding site of ubiquitin, the Ile44 residue. This new unfolding pathway leads to the discovery of new intermediate configurations, which correspond to the end-to-end extensions previously seen experimentally. More importantly, it demonstrates the novel finding that the binding site of ubiquitin can be responsible not only for its biological functions, but also its unfolding dynamics. We also report in contrast to previous single molecule constant force experiments that when the clamping force becomes smaller than about 300 pN, the number of intermediate configurations increases dramatically, where almost all unfolding events at 100 pN involve an intermediate configuration. By directly calculating the life times of the intermediate configurations from the height of the barriers that were crossed on the potential energy surface, we demonstrate that these intermediate states were likely not observed experimentally due to their lifetimes typically being about two orders of magnitude smaller than the experimental temporal resolution.

Regulation of ryanodine receptor RyR2 by protein-protein interactions: prediction of a PKA binding site on the N-terminal domain of RyR2 and its relation to disease causing mutations

Protein-protein interactions are the key processes responsible for signaling and function in complex networks. Determining the correct binding partners and predicting the ligand binding sites in the absence of experimental data require predictive models. Hybrid models that combine quantitative atomistic calculations with statistical thermodynamics formulations are valuable tools for bioinformatics predictions. We present a hybrid prediction and analysis model for determining putative binding partners and interpreting the resulting correlations in the yet functionally uncharacterized interactions of the ryanodine RyR2 N-terminal domain. Using extensive docking calculations and libraries of hexameric peptides generated from regulator proteins of the RyR2 channel, we show that the residues 318-323 of protein kinase A, PKA, have a very high affinity for the N-terminal of RyR2. Using a coarse grained Elastic Net Model, we show that the binding site lies at the end of a pathway of evolutionarily conserved residues in RyR2. The two disease causing mutations are also on this path. The program for the prediction of the energetically responsive residues by the Elastic Net Model is freely available on request from the corresponding author.

Study, by use of coarse-grained models, of the functionally crucial residues and allosteric pathway of anesthetic regulation of the Gloeobacter violaceus ligand-gated ion channel

Although pentameric ligand-gated ion channels (pLGICs) have been found to be the targets of general anesthetics, the mechanism of the effects of anesthetics on pLGICs remains elusive. pLGICs from Gloeobacter violaceus (GLIC) can be inhibited by the anesthetic ketamine. X-ray crystallography has shown that the ketamine binding site is distant from the channel gate of the GLIC. It is still not clear how ketamine controls the function of the GLIC by long-range allosteric regulation. In this work, the functionally crucial residues and allosteric pathway of anesthetic regulation of the GLIC were identified by use of a coarse-grained thermodynamic method developed by our group. In our method, the functionally crucial sites were identified as the residues thermodynamically coupled with binding of ketamine. The results from calculation were highly consistent with experimental data. Our study aids understanding of the mechanism of the anesthetic action of ketamine on the GLIC by long-range allosteric modulation.

Prediction of allosteric sites on protein surfaces with an elastic-network-model-based thermodynamic method

Allostery is a rapid and efficient way in many biological processes to regulate protein functions, where binding of an effector at the allosteric site alters the activity and function at a distant active site. Allosteric regulation of protein biological functions provides a promising strategy for novel drug design. However, how to effectively identify the allosteric sites remains one of the major challenges for allosteric drug design. In the present work, a thermodynamic method based on the elastic network model was proposed to predict the allosteric sites on the protein surface. In our method, the thermodynamic coupling between the allosteric and active sites was considered, and then the allosteric sites were identified as those where the binding of an effector molecule induces a large change in the binding free energy of the protein with its ligand. Using the proposed method, two proteins, i.e., the 70 kD heat shock protein (Hsp70) and GluA2 alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor, were studied and the allosteric sites on the protein surface were successfully identified. The predicted results are consistent with the available experimental data, which indicates that our method is a simple yet effective approach for the identification of allosteric sites on proteins.

Specialized Dynamical Properties of Promiscuous Residues Revealed by Simulated Conformational Ensembles

The ability to interact with different partners is one of the most important features in proteins. Proteins that bind a large number of partners (hubs) have been often associated with intrinsic disorder. However, many examples exist of hubs with an ordered structure, and evidence of a general mechanism promoting promiscuity in ordered proteins is still elusive. An intriguing hypothesis is that promiscuous binding sites have specific dynamical properties, distinct from the rest of the interface and pre-existing in the protein isolated state. Here, we present the first comprehensive study of the intrinsic dynamics of promiscuous residues in a large protein data set. Different computational methods, from coarse-grained elastic models to geometry-based sampling methods and to full-atom Molecular Dynamics simulations, were used to generate conformational ensembles for the isolated proteins. The flexibility and dynamic correlations of interface residues with a different degree of binding promiscuity were calculated and compared considering side chain and backbone motions, the latter both on a local and on a global scale. The study revealed that (a) promiscuous residues tend to be more flexible than nonpromiscuous ones, (b) this additional flexibility has a higher degree of organization, and (c) evolutionary conservation and binding promiscuity have opposite effects on intrinsic dynamics. Findings on simulated ensembles were also validated on ensembles of experimental structures extracted from the Protein Data Bank (PDB). Additionally, the low occurrence of single nucleotide polymorphisms observed for promiscuous residues indicated a tendency to preserve binding diversity at these positions. A case study on two ubiquitin-like proteins exemplifies how binding promiscuity in evolutionary related proteins can be modulated by the fine-tuning of the interface dynamics. The interplay between promiscuity and flexibility highlighted here can inspire new directions in protein–protein interaction prediction and design methods.

Hot spots in a network of functional sites

It is of significant interest to understand how proteins interact, which holds the key phenomenon in biological functions. Using dynamic fluctuations in high frequency modes, we show that the Gaussian Network Model (GNM) predicts hot spot residues with success rates ranging between S 8–58%, C 84–95%, P 5–19% and A 81–92% on unbound structures and S 8–51%, C 97–99%, P 14–50%, A 94–97% on complex structures for sensitivity, specificity, precision and accuracy, respectively. High specificity and accuracy rates with a single property on unbound protein structures suggest that hot spots are predefined in the dynamics of unbound structures and forming the binding core of interfaces, whereas the prediction of other functional residues with similar dynamic behavior explains the lower precision values. The latter is demonstrated with the case studies; ubiquitin, hen egg-white lysozyme and M2 proton channel. The dynamic fluctuations suggest a pseudo network of residues with high frequency fluctuations, which could be plausible for the mechanism of biological interactions and allosteric regulation.

Identification of functionally key residues in AMPA receptor with a thermodynamic method

AMPA receptor mediates the fast excitatory synaptic transmission in the central nervous system, and it is activated by the binding of glutamate that results in the opening of the transmembrane ion channel. In the present work, the thermodynamic method developed by our group was improved and then applied to identify the functionally key residues that regulate the glutamate-binding affinity of AMPA receptor. In our method, the key residues are identified as those whose perturbation largely changes the ligand binding free energy of the protein. It is found that besides the ligand binding sites, other residues distant from the binding cleft can also influence the glutamate binding affinity through a long-range allosteric regulation. These allosteric sites include the hinge region of the ligand binding cleft, the dimer interface of the ligand binding domain, the linkers between the ligand binding domain and the transmembrane domain, and the interface between the N-terminal domain and the ligand binding domain. Our calculation results are consistent with the available experimental data. The results are helpful for our understanding of the mechanism of long-range allosteric communication in the AMPA receptor and the mechanism of channel opening triggered by glutamate binding.

Dynamics of endoglucanase catalytic domains: implications towards thermostability

Thermostable endoglucanases play a crucial role in the production of biofuels to breakdown plant cellulose. Analyzing their structure-dynamics relationship can inform about the origins of their thermostability. Although tertiary structures of many endoglucanase proteins are available, the relationship between thermostability, structure, and dynamics is not explored fully. We have generated elastic network models for thermostable and mesostable endoglucanases with the (αβ)₈ fold in substrate bound and unbound states. The comparative analyses shed light on the relation between protein dynamics, thermostability, and substrate binding. We observed specific differences in the dynamic behavior of catalytic residues in slow modes: while both the nucleophile and the acid/base donor residues show positively correlated motions in the thermophile, their dynamics is uncoupled in the mesophile. Our proof-of-concept comparison study suggests that global dynamics can be harnessed to further our understanding of thermostability.

The elastic network model reveals a consistent picture on intrinsic functional dynamics of type II restriction endonucleases

The vibrational dynamics of various type II restriction endonucleases, in complex with cognate/non-cognate DNA and in the apo form, are investigated with the elastic network model in order to reveal common functional mechanisms in this enzyme family. Scissor-like and tong-like motions observed in the slowest modes of all enzymes and their complexes point to common DNA recognition and cleavage mechanisms. Normal mode analysis further points out that the scissor-like motion has an important role in differentiating between cognate and non-cognate sequences at the recognition site, thus implying its catalytic relevance. Flexible regions observed around the DNA-binding site of the enzyme usually concentrate on the highly conserved β-strands, especially after DNA binding. These β-strands may have a structurally stabilizing role in functional dynamics for target site recognition and cleavage. In addition, hot spot residues based on high-frequency modes reveal possible communication pathways between the two distant cleavage sites in the enzyme family. Some of these hot spots also exist on the shortest path between the catalytic sites and are highly conserved.

Computational design of peptide ligands

Peptides possess several attractive features when compared to small molecule and protein therapeutics, such as high structural compatibility with target proteins, the ability to disrupt protein-protein interfaces, and small size. Efficient design of high-affinity peptide ligands via rational methods has been a major obstacle to the development of this potential drug class. However, structural insights into the architecture of protein-peptide interfaces have recently culminated in several computational approaches for the rational design of peptides that target proteins. These methods provide a valuable alternative to experimental high-resolution structures of target protein-peptide complexes, bringing closer the dream of in silico designed peptides for therapeutic applications.

Identification of ligand binding sites of proteins using the Gaussian Network Model

The nonlocal nature of the protein-ligand binding problem is investigated via the Gaussian Network Model with which the residues lying along interaction pathways in a protein and the residues at the binding site are predicted. The predictions of the binding site residues are verified by using several benchmark systems where the topology of the unbound protein and the bound protein-ligand complex are known. Predictions are made on the unbound protein. Agreement of results with the bound complexes indicates that the information for binding resides in the unbound protein. Cliques that consist of three or more residues that are far apart along the primary structure but are in contact in the folded structure are shown to be important determinants of the binding problem. Comparison with known structures shows that the predictive capability of the method is significant.

VitAL: Viterbi algorithm for de novo peptide design

Background

Drug design against proteins to cure various diseases has been studied for several years. Numerous design techniques were discovered for small organic molecules for specific protein targets. The specificity, toxicity and selectivity of small molecules are hard problems to solve. The use of peptide drugs enables a partial solution to the toxicity problem. There has been a wide interest in peptide design, but the design techniques of a specific and selective peptide inhibitor against a protein target have not yet been established.

Methodology/Principal Findings

A novel de novo peptide design approach is developed to block activities of disease related protein targets. No prior training, based on known peptides, is necessary. The method sequentially generates the peptide by docking its residues pair by pair along a chosen path on a protein. The binding site on the protein is determined via the coarse grained Gaussian Network Model. A binding path is determined. The best fitting peptide is constructed by generating all possible peptide pairs at each point along the path and determining the binding energies between these pairs and the specific location on the protein using AutoDock. The Markov based partition function for all possible choices of the peptides along the path is generated by a matrix multiplication scheme. The best fitting peptide for the given surface is obtained by a Hidden Markov model using Viterbi decoding. The suitability of the conformations of the peptides that result upon binding on the surface are included in the algorithm by considering the intrinsic Ramachandran potentials.

Conclusions/Significance

The model is tested on known protein-peptide inhibitor complexes. The present algorithm predicts peptides that have better binding energies than those of the existing ones. Finally, a heptapeptide is designed for a protein that has excellent binding affinity according to AutoDock results.