Ranking of binding and nonbinding peptides to MHC class I molecules using inverse folding approach: Implications for vaccine design

T cell recognition of the peptide–MHC complex initiates a cascade of immunological events necessary for immune responses. Accurate T-cell epitope prediction is an important part of the vaccine designing. Development of predictive algorithms based on sequence profile requires a very large number of experimental binding peptide data to major histocompatibility complex (MHC) molecules. Here we used inverse folding approach to study the peptide specificity of MHC Class-I molecule with the aim of obtaining a better differentiation between binding and nonbinding sequence. Overlapping peptides, spanning the entire protein sequence, are threaded through the backbone coordinates of a known peptide fold in the MHC groove, and their interaction energies are evaluated using statistical pairwise contact potentials. We used the Miyazawa & Jernigan and Betancourt & Thirumalai tables for pairwise contact potentials, and two distance criteria (Nearest atom ≫ 4.0 Å & C-beta ≫ 7.0 Å) for ranking the peptides in an ascending order according to their energy values, and in most cases, known antigenic peptides are highly ranked. The predictions from threading improved when used multiple templates and average scoring scheme. In general, when structural information about a protein-peptide complex is available, the current application of the threading approach can be used to screen a large library of peptides for selection of the best binders to the target protein. The proposed scheme may significantly reduce the number of peptides to be tested in wet laboratory for epitope based vaccine design.     

Structural prediction of peptides binding to MHC class I molecules

Peptide binding to class I major histocompatibility complex (MHCI) molecules is a key step in the immune response and the structural details of this interaction are of importance in the design of peptide vaccines. Algorithms based on primary sequence have had success in predicting potential antigenic peptides for MHCI, but such algorithms have limited accuracy and provide no structural information. Here, we present an algorithm, PePSSI (peptide-MHC prediction of structure through solvated interfaces), for the prediction of peptide structure when bound to the MHCI molecule, HLA-A2. The algorithm combines sampling of peptide backbone conformations and flexible movement of MHC side chains and is unique among other prediction algorithms in its incorporation of explicit water molecules at the peptide-MHC interface. In an initial test of the algorithm, PePSSI was used to predict the conformation of eight peptides bound to HLA-A2, for which X-ray data are available. Comparison of the predicted and X-ray conformations of these peptides gave RMSD values between 1.301 and 2.475 A. Binding conformations of 266 peptides with known binding affinities for HLA-A2 were then predicted using PePSSI. Structural analyses of these peptide-HLA-A2 conformations showed that peptide binding affinity is positively correlated with the number of peptide-MHC contacts and negatively correlated with the number of interfacial water molecules. These results are consistent with the relatively hydrophobic binding nature of the HLA-A2 peptide binding interface. In summary, PePSSI is capable of rapid and accurate prediction of peptide-MHC binding conformations, which may in turn allow estimation of MHCI-peptide binding affinity.     

A structure-based approach for prediction of MHC-binding peptides

Identification of immunodominant peptides is the first step in the rational design of peptide vaccines aimed at T-cell immunity. The advances in sequencing techniques and the accumulation of many protein sequences without the purified protein challenge the development of computer algorithms to identify dominant T-cell epitopes based on sequence data alone. Here, we focus on antigenic peptides recognized by cytotoxic T cells. The selection of T-cell epitopes along a protein sequence is influenced by the specificity of each of the processing stages that precede antigen presentation. The most selective of these processing stages is the binding of the peptides to the major histocompatibility complex molecules, and therefore many of the predictive algorithms focus on this stage. Most of these algorithms are based on known binding peptides whose sequences have been used for the characterization of binding motifs or profiles. Here, we describe a structure-based algorithm that does not rely on previous binding data. It is based on observations from crystal structures that many of the bound peptides adopt similar conformations and placements within the MHC groove. The algorithm uses a structural template of the peptide in the MHC groove upon which peptide candidates are threaded and their fit to the MHC groove is evaluated by statistical pairwise potentials. It can rank all possible peptides along a protein sequence or within a suspected group of peptides, directing the experimental efforts towards the most promising peptides. This approach is especially useful when no previous peptide binding data are available.     

MHC Class I/peptide interactions: Binding specificity and kinetics

Recent developments in the preparation of soluble analogues of the major histocompatibility complex (MHC) class l molecules as well as in the applications of real time biosensor technology have permitted the direct analysis of the binding of MHC class l molecules to antigenic peptides. Using synthetic peptide analogues with cysteine substitutions at appropriate positions, peptides can be immobilized on a dextran-modified gold biosensor surface with a specific spatial orientation. A full set of such substituted peptides (known as ‘pepsicles’, as they are peptides on a stick) representing antigenic or self peptides can be used in the functional mapping of the MHC class l peptide binding site. Scans of sets of peptide analogues reveal that some amino acid side chains of the peptide are critical to stable binding to the MHC molecule, while others are not. This is consistent with functional experiments using substituted peptides and three-dimensional molecular models of MHC/peptide complexes. Details analysis of the kinetic dissociation rates (k_d) of the MHC molecules from the specifically coupled solid phase peptides revels that the stability of the complex is a function of the particular peptide, its coupling position, and the MHC molecule. Measured k_d values for antigenic peptide/class I interactions at 25°C are in the range of ca 10^?4–10^?6/s. Biosensor methodology for the analysis of the binding of MHC class I molecules to solid-phase peptides using real time surface plasmon resonance offers a rational approach to the general analysis of protein/peptide interactions.     

Improving MHC binding peptide prediction by incorporating binding data of auxiliary MHC molecules

MOTIVATION: Various computational methods have been proposed to tackle the problem of predicting the peptide binding ability for a specific MHC molecule. These methods are based on known binding peptide sequences. However, current available peptide databases do not have very abundant amounts of examples and are highly redundant. Existing studies show that MHC molecules can be classified into supertypes in terms of peptide-binding specificities. Therefore, we first give a method for reducing the redundancy in a given dataset based on information entropy, then present a novel approach for prediction by learning a predictive model from a dataset of binders for not only the molecule of interest but also for other MHC molecules. RESULTS: We experimented on the HLA-A family with the binding nonamers of A1 supertype (HLA-A*0101, A*2601, A*2902, A*3002), A2 supertype (A*0201, A*0202, A*0203, A*0206, A*6802), A3 supertype (A*0301, A*1101, A*3101, A*3301, A*6801) and A24 supertype (A*2301 and A*2402), whose data were collected from six publicly available peptide databases and two private sources. The results show that our approach significantly improves the prediction accuracy of peptides that bind a specific HLA molecule when we combine binding data of HLA molecules in the same supertype. Our approach can thus be used to help find new binders for MHC molecules.     

Prediction of MHC class II binding affinity using SMM-align,a novel stabilization matrix alignment method

Background  

Antigen presenting cells (APCs) sample the extra cellular space and present peptides from here to T helper cells, which can be activated if the peptides are of foreign origin. The peptides are presented on the surface of the cells in complex with major histocompatibility class II (MHC II) molecules. Identification of peptides that bind MHC II molecules is thus a key step in rational vaccine design and developing methods for accurate prediction of the peptide:MHC interactions play a central role in epitope discovery. The MHC class II binding groove is open at both ends making the correct alignment of a peptide in the binding groove a crucial part of identifying the core of an MHC class II binding motif. Here, we present a novel stabilization matrix alignment method, SMM-align, that allows for direct prediction of peptide:MHC binding affinities. The predictive performance of the method is validated on a large MHC class II benchmark data set covering 14 HLA-DR (human MHC) and three mouse H2-IA alleles.     

A combined bioinformatic approach oriented to the analysis and design of peptides with high affinity to MHC class I molecules

We report on a new method to compute the antigenic degree of peptides from available experimental data on peptide binding affinity to class I MHC molecules. The methodology is a combination of two strategies at different levels of information. The first, at the primary structure level, consists in expressing the peptides binding activity as a profile of amino acid contributions, amino acid similarity being accounted for by their characteristic physicochemical properties and their position within the sequence. The higher level of the strategy is based on a meticulous analysis of the contact interface of the peptides with the cleft constituting the receptor region of a particular class I MHC molecule. Interaction interfaces are inferred by docking the peptide onto the receptor groove of the MHC molecule; evaluation of the affinity of the peptide to the receptor is then performed by analysis of the electrostatic and hydrophobic energies on points of the interaction interface.The result is a robust system for analysis of peptide affinity to class I MHC molecules since while the first analysis dictates the composition of active sequences at the amino acid level, the second translates this information to the atomic level, where the molecular interaction can be analyzed in terms of the intrinsic interatomic forces and energies. Evaluation results for the methodology are encouraging since high affinity peptides are reflected by high scores at both levels of information, and are proportionally lower for peptides of medium and lower affinity for which interaction surfaces show relatively lower electrostatic complementarity and hydrophobic correlation than for the former.     

Binding of longer peptides to the H-2Kb heterodimer is restricted to peptides extended at their C terminus: refinement of the inherent MHC class I peptide binding criteria.

MHC class I molecules usually bind short peptides of 8-10 amino acids, and binding is dependent on allele-specific anchor residues. However, in a number of cellular systems, class I molecules have been found containing peptides longer than the canonical size. To understand the structural requirements for MHC binding of longer peptides, we used an in vitro class I MHC folding assay to examine peptide variants of the antigenic VSV 8 mer core peptide containing length extensions at either their N or C terminus. This approach allowed us to determine the ability of each peptide to productively form Kb/beta2-microglobulin/peptide complexes. We found that H-2Kb molecules can accommodate extended peptides, but only if the extension occurs at the C-terminal peptide end, and that hydrophobic flanking regions are preferred. Peptides extended at their N terminus did not promote productive formation of the trimolecular complex. A structural basis for such findings comes from molecular modeling of a H-2Kb/12 mer complex and comparative analysis of MHC class I structures. These analyses revealed that structural constraints in the A pocket of the class I peptide binding groove hinder the binding of N-terminal-extended peptides, whereas structural features at the C-terminal peptide residue pocket allow C-terminal peptide extensions to reach out of the cleft. These findings broaden our understanding of the inherent peptide binding and epitope selection criteria of the MHC class I molecule. Core peptides extended at their N terminus cannot bind, but peptide extensions at the C terminus are tolerated.     

Methods for prediction of peptide binding to MHC molecules: a comparative study

BACKGROUND: A variety of methods for prediction of peptide binding to major histocompatibility complex (MHC) have been proposed. These methods are based on binding motifs, binding matrices, hidden Markov models (HMM), or artificial neural networks (ANN). There has been little prior work on the comparative analysis of these methods. MATERIALS AND METHODS: We performed a comparison of the performance of six methods applied to the prediction of two human MHC class I molecules, including binding matrices and motifs, ANNs, and HMMs. RESULTS: The selection of the optimal prediction method depends on the amount of available data (the number of peptides of known binding affinity to the MHC molecule of interest), the biases in the data set and the intended purpose of the prediction (screening of a single protein versus mass screening). When little or no peptide data are available, binding motifs are the most useful alternative to random guessing or use of a complete overlapping set of peptides for selection of candidate binders. As the number of known peptide binders increases, binding matrices and HMM become more useful predictors. ANN and HMM are the predictive methods of choice for MHC alleles with more than 100 known binding peptides. CONCLUSION: The ability of bioinformatic methods to reliably predict MHC binding peptides, and thereby potential T-cell epitopes, has major implications for clinical immunology, particularly in the area of vaccine design.     

Cooperativity of hydrophobic anchor interactions: evidence for epitope selection by MHC class II as a folding process

Peptide binding to MHC class II (MHCII) molecules is stabilized by hydrophobic anchoring and hydrogen bond formation. We view peptide binding as a process in which the peptide folds into the binding groove and to some extent the groove folds around the peptide. Our previous observation of cooperativity when analyzing binding properties of peptides modified at side chains with medium to high solvent accessibility is compatible with such a view. However, a large component of peptide binding is mediated by residues with strong hydrophobic interactions that bind to their respective pockets. If these reflect initial nucleation events they may be upstream of the folding process and not show cooperativity. To test whether the folding hypothesis extends to these anchor interactions, we measured dissociation and affinity to HLA-DR1 of an influenza hemagglutinin-derived peptide with multiple substitutions at major anchor residues. Our results show both negative and positive cooperative effects between hydrophobic pocket interactions. Cooperativity was also observed between hydrophobic pockets and positions with intermediate solvent accessibility, indicating that hydrophobic interactions participate in the overall folding process. These findings point out that predicting the binding potential of epitopes cannot assume additive and independent contributions of the interactions between major MHCII pockets and corresponding peptide side chains.     

C-terminal anchoring of a peptide to class II MHC via the P10 residue is compatible with a peptide bulge.

The binding of antigenic peptide to class II MHC is mediated by hydrogen bonds between the MHC and the peptide, by salt bridges, and by hydrophobic interactions. The latter are confined to a number of deeper pockets within the peptide binding groove, and peptide side chains that interact with these pockets are referred to as anchor residues. T cell recognition involves solvent-accessible peptide residues along with minor changes in MHC helical pitch induced by the anchor residues. In class I MHC there is an added level of epitope complexity that results from binding of longer peptides that bulge out into the solvent-accessible, T cell contact area. Unlike class I MHC, class II MHC does not bind peptides of discrete length, and the possibility of peptide bulging has not been clearly addressed. A peptide derived from position 24-37 of integrin beta(3) can either bind or not bind to the class II MHC molecule HLA DRB3*0101 based on a polymorphism at the P9 anchor. We show that the loss of binding can be compensated by changes at the P10 position. We propose that this could be an example of a class II peptide bulge. Although not as efficient as P9 anchoring, the use of P10 as an anchor adds another possible mechanism by which T cell epitopes can be generated in the class II presentation system.     

Automated generation and evaluation of specific MHC binding predictive tools: ARB matrix applications

Prediction of which peptides can bind major histocompatibility complex (MHC) molecules is commonly used to assist in the identification of T cell epitopes. However, because of the large numbers of different MHC molecules of interest, each associated with different predictive tools, tool generation and evaluation can be a very resource intensive task. A methodology commonly used to predict MHC binding affinity is the matrix or linear coefficients method. Herein, we described Average Relative Binding (ARB) matrix methods that directly predict IC₅₀ values allowing combination of searches involving different peptide sizes and alleles into a single global prediction. A computer program was developed to automate the generation and evaluation of ARB predictive tools. Using an in-house MHC binding database, we generated a total of 85 and 13 MHC class I and class II matrices, respectively. Results from the automated evaluation of tool efficiency are presented. We anticipate that this automation framework will be generally applicable to the generation and evaluation of large numbers of MHC predictive methods and tools, and will be of value to centralize and rationalize the process of evaluation of MHC predictions. MHC binding predictions based on ARB matrices were made available at web server.     

MHC class II binding of peptides derived from HLA-DR 1.

The aim of these studies was to determine whether auto- and alloreactivity can arise from T cell recognition of MHC-peptides in context of syngeneic MHC. Four synthetic peptides derived from the first domain of the HLA-DR beta 1 * 0101 chain were used in limiting dilution analysis to prime T cells from HLA-DR1- and HLA-DR1+ responders. The frequency of T cells responding to these four peptides was similar in individuals with or without HLA-DR1. In both cases, the peptide corresponding to the nonpolymorphic sequence 43-62, was less immunogenic than peptides corresponding to the three hypervariable regions 1-20, 21-42, and 66-90, eliciting a lower number of reactive T cells. Experiments using a T cell line with specific reactivity to peptide 21-42 showed, however, that this response can be efficiently blocked by adding to the culture a nonpolymorphic sequence peptide. This suggests that alloreactivity can be blocked by use of monomorphic (self) peptides. The binding of both "monomorphic" and "polymorphic" synthetic DR1 peptides to affinity purified HLA-DR 1 and DR 11 molecules was measured using radiolabeled peptides and high performance size exclusion chromatography. The data showed that the polymorphic as well as monomorphic synthetic DR1 peptides bound to both DR1 and DR11 molecules. Competitive inhibition studies indicated that the monomorphic 43-62 peptide can block the binding of the polymorphic peptides, consistent with the results obtained in T cell cultures. Taken together these data suggest that anti-MHC autoreactive T cells are present in the periphery and that both auto and alloreactivity can be elicited by MHC peptides binding to MHC class II molecules.     

Predicting peptides that bind to MHC molecules using supervised learning of hidden markov models

The binding of a major histocompatibility complex (MHC) molecule to a peptide originating in an antigen is essential to recognizing antigens in immune systems, and it has proved to be important to use computers to predict the peptides that will bind to an MHC molecule. The purpose of this paper is twofold: First, we propose to apply supervised learning of hidden Markov models (HMMs) to this problem, which can surpass existing methods for the problem of predicting MHC-binding peptides. Second, we generate peptides that have high probabilities to bind to a certain MHC molecule, based on our proposed method using peptides binding to MHC molecules as a set of training data. From our experiments, in a type of cross-validation test, the discrimination accuracy of our supervised learning method is usually approximately 2–15% better than those of other methods, including backpropagation neural networks, which have been regarded as the most effective approach to this problem. Furthermore, using an HMM trained for HLA-A2, we present new peptide sequences that are provided with high binding probabilities by the HMM and that are thus expected to bind to HLA-A2 proteins. Peptide sequences not shown in this paper but with rather high binding probabilities can be obtained from the author (E-mail: mami@ccm.cl.nec.co.jp). Proteins 33:460–474, 1998. © 1998 Wiley-Liss, Inc.     

A computational pipeline to generate MHC binding motifs

BACKGROUND: Major histocompatibility complex (MHC) class I molecules play key roles in host immunity against pathogens by presenting peptide antigens to CD8+ T-cells. Many variants of MHC molecules exist, and each has a unique preference for certain peptide ligands. Both experimental approaches and computational algorithms have been utilized to analyze these peptide MHC binding characteristics. Traditionally, MHC binding specificities have been described in terms of binding motifs. Such motifs classify certain peptide positions as primary and secondary anchors according to their impact on binding, and they list the preferred and deleterious residues at these positions. This provides a concise and easily communicatable summary of MHC binding specificities. However, so far there has been no algorithm to generate such binding motifs in an automated and uniform fashion. In this paper, we present a computational pipeline that takes peptide MHC binding data as input and produces a concise MHC binding motif. We tested our pipeline on a set of 18 MHC class I molecules and showed that the derived motifs are consistent with historic expert assignments. We have implemented a pipeline that formally codifies rules to generate MHC binding motifs. The pipeline has been incorporated into the immune epitope database and analysis resource (IEDB) and motifs can be visualized while browsing MHC alleles in the IEDB.     

A rapid method for quantitative prediction of high affinity CTL epitopes: QSAR studies on peptides having affinity with the class I MHC molecular HLA-A*0201

Summary It would be useful to develop a method to rapidly identify peptide epitopes for vaccine development. In this paper, empirical three-dimensional quantitative structure-affinity relationship (3D-QSAR) methods were used to study the relationship between the three dimensional structural parameters (the isotropic surface area, ISA, and the electronic charge index, ECI) of the HLA-A^*0201 binding peptide and the HLA-A^*0201/peptide binding affinities. A set of 102 peptides having affinity with the class I MHC HLA-A^*0201 molecule was used as training set. A test set of 40 peptides was used to determine the predictive value of the models. The 3D-QSAR models gave aq ²=0.5724 and highr _pred ² =0.6955. According to the standard regression coefficients, it is known that the hydrophobic interactions (in these studies, the ISA is highly correlative with the hydrophobic property) play a dominant role in peptide-MHC molecule binding, and also which amino acid residue with what property is needed at specific position of the peptide. The approach we have taken is highly complementary to the many excellent methods described in references and appears highly predictive. It is a rapid and convenient method for detecting high affinity peptide epitopes.     

Limitations of Ab Initio Predictions of Peptide Binding to MHC Class II Molecules

Successful predictions of peptide MHC binding typically require a large set of binding data for the specific MHC molecule that is examined. Structure based prediction methods promise to circumvent this requirement by evaluating the physical contacts a peptide can make with an MHC molecule based on the highly conserved 3D structure of peptide:MHC complexes. While several such methods have been described before, most are not publicly available and have not been independently tested for their performance. We here implemented and evaluated three prediction methods for MHC class II molecules: statistical potentials derived from the analysis of known protein structures; energetic evaluation of different peptide snapshots in a molecular dynamics simulation; and direct analysis of contacts made in known 3D structures of peptide:MHC complexes. These methods are ab initio in that they require structural data of the MHC molecule examined, but no specific peptide:MHC binding data. Moreover, these methods retain the ability to make predictions in a sufficiently short time scale to be useful in a real world application, such as screening a whole proteome for candidate binding peptides. A rigorous evaluation of each methods prediction performance showed that these are significantly better than random, but still substantially lower than the best performing sequence based class II prediction methods available. While the approaches presented here were developed independently, we have chosen to present our results together in order to support the notion that generating structure based predictions of peptide:MHC binding without using binding data is unlikely to give satisfactory results.     

Prediction of MHC Class I Binding Peptides by a Query Learning Algorithm Based on Hidden Markov Models

A query learning algorithm based on hidden Markov models (HMMs) isdeveloped to design experiments for string analysis and prediction of MHCclass I binding peptides. Query learning is introduced to aim at reducingthe number of peptide binding data for training of HMMs. A multiple numberof HMMs, which will collectively serve as a committee, are trained withbinding data and used for prediction in real-number values. The universeof peptides is randomly sampled and subjected to judgement by the HMMs.Peptides whose prediction is least consistent among committee HMMs aretested by experiment. By iterating the feedback cycle of computationalanalysis and experiment the most wanted information is effectivelyextracted. After 7 rounds of active learning with 181 peptides in all,predictive performance of the algorithm surpassed the so far bestperforming matrix based prediction. Moreover, by combining the bothmethods binder peptides (log Kd < -6) could be predicted with84% accuracy. Parameter distribution of the HMMs that can be inspectedvisually after training further offers a glimpse of dynamic specificity ofthe MHC molecules.     

Unique Pathogen Peptidomes Facilitate Pathogen-Specific Selection and Specialization of MHC Alleles

A key component of pathogen-specific adaptive immunity in vertebrates is the presentation of pathogen-derived antigenic peptides by major histocompatibility complex (MHC) molecules. The excessive polymorphism observed at MHC genes is widely presumed to result from the need to recognize diverse pathogens, a process called pathogen-driven balancing selection. This process assumes that pathogens differ in their peptidomes—the pool of short peptides derived from the pathogen’s proteome—so that different pathogens select for different MHC variants with distinct peptide-binding properties. Here, we tested this assumption in a comprehensive data set of 51.9 Mio peptides, derived from the peptidomes of 36 representative human pathogens. Strikingly, we found that 39.7% of the 630 pairwise comparisons among pathogens yielded not a single shared peptide and only 1.8% of pathogen pairs shared more than 1% of their peptides. Indeed, 98.8% of all peptides were unique to a single pathogen species. Using computational binding prediction to characterize the binding specificities of 321 common human MHC class-I variants, we investigated quantitative differences among MHC variants with regard to binding peptides from distinct pathogens. Our analysis showed signatures of specialization toward specific pathogens especially by MHC variants with narrow peptide-binding repertoires. This supports the hypothesis that such fastidious MHC variants might be maintained in the population because they provide an advantage against particular pathogens. Overall, our results establish a key selection factor for the excessive allelic diversity at MHC genes observed in natural populations and illuminate the evolution of variable peptide-binding repertoires among MHC variants.