Identifiying human MHC supertypes using bioinformatic methods

MULTIPRED2: a computational system for large-scale identification of peptides predicted to bind to HLA supertypes and alleles

MULTIPRED2 is a computational system for facile prediction of peptide binding to multiple alleles belonging to human leukocyte antigen (HLA) class I and class II DR molecules. It enables prediction of peptide binding to products of individual HLA alleles, combination of alleles, or HLA supertypes. NetMHCpan and NetMHCIIpan are used as prediction engines. The 13 HLA Class I supertypes are A1, A2, A3, A24, B7, B8, B27, B44, B58, B62, C1, and C4. The 13 HLA Class II DR supertypes are DR1, DR3, DR4, DR6, DR7, DR8, DR9, DR11, DR12, DR13, DR14, DR15, and DR16. In total, MULTIPRED2 enables prediction of peptide binding to 1077 variants representing 26 HLA supertypes. MULTIPRED2 has visualization modules for mapping promiscuous T-cell epitopes as well as those regions of high target concentration – referred to as T-cell epitope hotspots. Novel graphic representations are employed to display the predicted binding peptides and immunological hotspots in an intuitive manner and also to provide a global view of results as heat maps. Another function of MULTIPRED2, which has direct relevance to vaccine design, is the calculation of population coverage. Currently it calculates population coverage in five major groups in North America. MULTIPRED2 is an important tool to complement wet-lab experimental methods for identification of T-cell epitopes. It is available at http://cvc.dfci.harvard.edu/multipred2/.

State of the art and challenges in sequence based T-cell epitope prediction

Sequence based T-cell epitope predictions have improved immensely in the last decade. From predictions of peptide binding to major histocompatibility complex molecules with moderate accuracy, limited allele coverage, and no good estimates of the other events in the antigen-processing pathway, the field has evolved significantly. Methods have now been developed that produce highly accurate binding predictions for many alleles and integrate both proteasomal cleavage and transport events. Moreover have so-called pan-specific methods been developed, which allow for prediction of peptide binding to MHC alleles characterized by limited or no peptide binding data. Most of the developed methods are publicly available, and have proven to be very useful as a shortcut in epitope discovery. Here, we will go through some of the history of sequence-based predictions of helper as well as cytotoxic T cell epitopes. We will focus on some of the most accurate methods and their basic background.

Reverse vaccinology: developing vaccines in the era of genomics

The sequence of microbial genomes made all potential antigens of each pathogen available for vaccine development. This increased by orders of magnitude potential vaccine targets in bacteria, parasites, and large viruses and revealed virtually all their CD4(+) and CD8(+) T cell epitopes. The genomic information was first used for the development of a vaccine against serogroup B meningococcus, and it is now being used for several other bacterial vaccines. In this review, we will first summarize the impact that genome sequencing has had on vaccine development, and then we will analyze how the genomic information can help further our understanding of immunity to infection or vaccination and lead to the design of better vaccines by diving into the world of T cell immunity.

Major histocompatibility complex class I binding predictions as a tool in epitope discovery

SUMMARY

Over the last decade, in silico models of the major histocompatibility complex (MHC) class I pathway have developed significantly. Before, peptide binding could only be reliably modelled for a few major human or mouse histocompatibility molecules; now, high-accuracy predictions are available for any human leucocyte antigen (HLA) -A or -B molecule with known protein sequence. Furthermore, peptide binding to MHC molecules from several non-human primates, mouse strains and other mammals can now be predicted. In this review, a number of different prediction methods are briefly explained, highlighting the most useful and historically important. Selected case stories, where these 'reverse immunology' systems have been used in actual epitope discovery, are briefly reviewed. We conclude that this new generation of epitope discovery systems has become a highly efficient tool for epitope discovery, and recommend that the less accurate prediction systems of the past be abandoned, as these are obsolete.

MHC class II epitope predictive algorithms

SUMMARY

Major histocompatibility complex class II (MHC-II) molecules sample peptides from the extracellular space, allowing the immune system to detect the presence of foreign microbes from this compartment. To be able to predict the immune response to given pathogens, a number of methods have been developed to predict peptide-MHC binding. However, few methods other than the pioneering TEPITOPE/ProPred method have been developed for MHC-II. Despite recent progress in method development, the predictive performance for MHC-II remains significantly lower than what can be obtained for MHC-I. One reason for this is that the MHC-II molecule is open at both ends allowing binding of peptides extending out of the groove. The binding core of MHC-II-bound peptides is therefore not known a priori and the binding motif is hence not readily discernible. Recent progress has been obtained by including the flanking residues in the predictions. All attempts to make ab initio predictions based on protein structure have failed to reach predictive performances similar to those that can be obtained by data-driven methods. Thousands of different MHC-II alleles exist in humans. Recently developed pan-specific methods have been able to make reasonably accurate predictions for alleles that were not included in the training data. These methods can be used to define supertypes (clusters) of MHC-II alleles where alleles within each supertype have similar binding specificities. Furthermore, the pan-specific methods have been used to make a graphical atlas such as the MHCMotifviewer, which allows for visual comparison of specificities of different alleles.

Predicting MHC class I epitopes in large datasets

Background

Experimental screening of large sets of peptides with respect to their MHC binding capabilities is still very demanding due to the large number of possible peptide sequences and the extensive polymorphism of the MHC proteins. Therefore, there is significant interest in the development of computational methods for predicting the binding capability of peptides to MHC molecules, as a first step towards selecting peptides for actual screening.

Results

We have examined the performance of four diverse MHC Class I prediction methods on comparatively large HLA-A and HLA-B allele peptide binding datasets extracted from the Immune Epitope Database and Analysis resource (IEDB). The chosen methods span a representative cross-section of available methodology for MHC binding predictions. Until the development of IEDB, such an analysis was not possible, as the available peptide sequence datasets were small and spread out over many separate efforts. We tested three datasets which differ in the IC₅₀ cutoff criteria used to select the binders and non-binders. The best performance was achieved when predictions were performed on the dataset consisting only of strong binders (IC₅₀ less than 10 nM) and clear non-binders (IC₅₀ greater than 10,000 nM). In addition, robustness of the predictions was only achieved for alleles that were represented with a sufficiently large (greater than 200), balanced set of binders and non-binders.

Conclusions

All four methods show good to excellent performance on the comprehensive datasets, with the artificial neural networks based method outperforming the other methods. However, all methods show pronounced difficulties in correctly categorizing intermediate binders.

Human leukocyte antigen class I supertypes and HIV-1 control in African Americans

The role of human leukocyte antigen (HLA) class I supertypes in controlling human immunodeficiency virus type 1 (HIV-1) infection in African Americans has not been established. We examined the effects of the HLA-A and HLA-B alleles and supertypes on the outcomes of HIV-1 clade B infection among 338 African American women and adolescents. HLA-B58 and -B62 supertypes (B58s and B62s) were associated with favorable HIV-1 disease control (proportional odds ratio [POR] of 0.33 and 95% confidence interval [95% CI] of 0.21 to 0.52 for the former and POR of 0.26 and 95% CI of 0.09 to 0.73 for the latter); B7s and B44s were associated with unfavorable disease control (POR of 2.39 and 95% CI of 1.54 to 3.73 for the former and POR of 1.63 and 95% CI of 1.08 to 2.47 for the latter). In general, individual alleles within specific B supertypes exerted relatively homogeneous effects. A notable exception was B27s, whose protective influence (POR, 0.58; 95% CI, 0.35 to 0.94) was masked by the opposing effect of its member allele B*1510. The associations of most B supertypes (e.g., B58s and B7s) were largely explained either by well-known effects of constituent B alleles or by effects of previously unimplicated B alleles aggregated into a particular supertype (e.g., B44s and B62s). A higher frequency of HLA-B genotypic supertypes correlated with a higher mean viral load (VL) and lower mean CD4 count (Pearson's r = 0.63 and 0.62, respectively; P = 0.03). Among the genotypic supertypes, B58s and its member allele B*57 contributed disproportionately to the explainable VL variation. The study demonstrated the dominant role of HLA-B supertypes in HIV-1 clade B-infected African Americans and further dissected the contributions of individual class I alleles and their population frequencies to the supertype effects.

HLA-C cell surface expression and control of HIV/AIDS correlate with a variant upstream of HLA-C

A variant 35 kb upstream of the HLA-C gene (-35C/T) was previously shown to associate with HLA-C mRNA expression level and steady-state plasma HIV RNA levels. We genotyped this variant in 1,698 patients of European ancestry with HIV. Individuals with known seroconversion dates were used for disease progression analysis and those with longitudinal viral load data were used for viral load analysis. We further tested cell surface expression of HLA-C in normal donors using an HLA-C-specific antibody. We show that the -35C allele is a proxy for high HLA-C cell surface expression, and that individuals with high-expressing HLA-C alleles progress more slowly to AIDS and control viremia significantly better than individuals with low HLA-C expressing alleles. These data strongly implicate high HLA-C expression levels in more effective control of HIV-1, potentially through better antigen presentation to cytotoxic T lymphocytes or recognition and killing of infected cells by natural killer cells.

Integrating in silico and in vitro analysis of peptide binding affinity to HLA-Cw*0102: a bioinformatic approach to the prediction of new epitopes

Background

Predictive models of peptide-Major Histocompatibility Complex (MHC) binding affinity are important components of modern computational immunovaccinology. Here, we describe the development and deployment of a reliable peptide-binding prediction method for a previously poorly-characterized human MHC class I allele, HLA-Cw*0102.

Methodology/Findings

Using an in-house, flow cytometry-based MHC stabilization assay we generated novel peptide binding data, from which we derived a precise two-dimensional quantitative structure-activity relationship (2D-QSAR) binding model. This allowed us to explore the peptide specificity of HLA-Cw*0102 molecule in detail. We used this model to design peptides optimized for HLA-Cw*0102-binding. Experimental analysis showed these peptides to have high binding affinities for the HLA-Cw*0102 molecule. As a functional validation of our approach, we also predicted HLA-Cw*0102-binding peptides within the HIV-1 genome, identifying a set of potent binding peptides. The most affine of these binding peptides was subsequently determined to be an epitope recognized in a subset of HLA-Cw*0102-positive individuals chronically infected with HIV-1.

Conclusions/Significance

A functionally-validated in silico-in vitro approach to the reliable and efficient prediction of peptide binding to a previously uncharacterized human MHC allele HLA-Cw*0102 was developed. This technique is generally applicable to all T cell epitope identification problems in immunology and vaccinology.

Peptide binding to the HLA-DRB1 supertype: a proteochemometrics analysis

A proteochemometrics approach was applied to a set of 2666 peptides binding to 12 HLA-DRB1 proteins. Sequences of both peptide and protein were described using three z-descriptors. Cross terms accounting for adjacent positions and for every second position in the peptides were included in the models, as well as cross terms for peptide/protein interactions. Models were derived based on combinations of different blocks of variables. These models had moderate goodness of fit, as expressed by r2, which ranged from 0.685 to 0.732; and good cross-validated predictive ability, as expressed by q2, which varied from 0.678 to 0.719. The external predictive ability was tested using a set of 356 HLA-DRB1 binders, which showed an r2(pred) in the range 0.364-0.530. Peptide and protein positions involved in the interactions were analyzed in terms of hydrophobicity, steric bulk and polarity.

HLA class I supertypes in type 1 diabetic children in an urban children's hospital

Type 1 diabetes is an autoimmune disorder characterized by progressive destruction of insulin-secreting beta cells of the pancreas, in which CD8(+) T cells play a critical role. The diversity in the HLA alleles expressed among various racial and ethnic groups leads to great variability in antigen presentation and recognition by CD8(+) T cells in the context of MHC class I molecules. To date, studies aimed at identifying disease-relevant antigenic epitopes have focused on using mice transgenic for HLA-A*0201, a common allele, particularly among Caucasians. We present HLA class I typing data from 88 children with type 1 diabetes at the Children's Hospital at Montefiore, where the patient population is ethnically diverse, but largely minority. When categorized into the HLA supertypes A2, A3, B7, and C1, 77% of those studied have alleles belonging to at least one supertype, and of these patients, 65% do not belong to the A2 supertype, which is the supertype represented by the HLA-A*0201 allele. These results support the need for studies using HLA transgenic mice expressing MHC molecules representative of a variety of HLA supertypes, particularly when searching for antigenic epitopes applicable for study among largely urban, minority pediatric populations.

Principal component analysis of native ensembles of biomolecular structures (PCA_NEST): insights into functional dynamics

MOTIVATION

To efficiently analyze the 'native ensemble of conformations' accessible to proteins near their folded state and to extract essential information from observed distributions of conformations, reliable mathematical methods and computational tools are needed.

RESULT

Examination of 24 pairs of structures determined by both NMR and X-ray reveals that the differences in the dynamics of the same protein resolved by the two techniques can be tracked to the most robust low frequency modes elucidated by principal component analysis (PCA) of NMR models. The active sites of enzymes are found to be highly constrained in these PCA modes. Furthermore, the residues predicted to be highly immobile are shown to be evolutionarily conserved, lending support to a PCA-based identification of potential functional sites. An online tool, PCA_NEST, is designed to derive the principal modes of conformational changes from structural ensembles resolved by experiments or generated by computations.

AVAILABILITY

http://ignm.ccbb.pitt.edu/oPCA_Online.htm

Essential differences in ligand presentation and T cell epitope recognition among HLA molecules of the HLA-B44 supertype

Human leukocyte antigens (HLA) have long been grouped into supertypes to facilitate peptide-based immunotherapy. Analysis of several hundreds of peptides presented by all nine antigens of the HLA-B44 supertype (HLA-B*18, B*37, B*40, B*41, B*44, B*45, B*47, B*49 and B*50) revealed unique peptide motifs for each of them. Taking all supertype members into consideration only 25 out of 670 natural ligands were found on more than one HLA molecule. Further direct comparisons by two mass spectrometric methods--isotope labeling as well as a label-free approach--consistently demonstrated only minute overlaps of below 3% between the ligandomes of different HLA antigens. In addition, T cell reactions of healthy donors against immunodominant HLA-B*44 and HLA-B*40 epitopes from EBV lacked promiscuous T-cell recognition within the HLA-B44 supertype. Taken together, these results challenge the common paradigm of broadly presented epitopes within this supertype.

HLAMatchmaker-defined triplet matching is not associated with better survival rates of patients with class I HLA allele mismatched hematopoietic cell transplants from unrelated donors

This report addresses the concept that permissible HLA mismatching, that is, mismatches that do not generate an allogeneic response, in hematopoietic stem cell transplantation (HCT) can be determined with structural similarity of polymorphic regions. We have applied the triplet version of a structural algorithm called HLAMatchmaker, which considers short sequences involving polymorphic amino acid residues on the molecular surface as key elements of immunogenic epitopes. The triplet matching effect was analyzed in a National Marrow Donor Program dataset consisting of 744 unrelated hematopoietic cell transplantation cases with 1 HLA-A, -B, or -C mismatch and 1690 fully HLA-A, -B, -C, -DR, or -DQ allele matched cases. In multivariate models adjusting for other significant clinical risk factors, the degree of triplet mismatching did not significantly correlate with patient survival, engraftment, or acute graft-versus-host disease (aGVHD). Other structurally based strategies should be pursued to identify permissible HLA mismatches in HCT.

Structural basis for HLA-A2 supertypes

The human leukocyte antigen (HLA) alleles are extremely polymorphic among ethnic population, and the peptide-binding specificity varies for different alleles in a combinatorial manner. However, it has been suggested that majority of alleles can be covered within few HLA supertypes, where different members of a supertype bind similar peptides, yet exhibiting distinct repertoires. Nonetheless, the structural basis for HLA supertype-like function is not clearly known. Here, we use structural data to explain the molecular basis for HLA-A2 supertypes.

The classification of HLA supertypes by GRID/CPCA and hierarchical clustering methods

Biological experiments often produce enormous amount of data, which are usually analyzed by data clustering. Cluster analysis refers to statistical methods that are used to assign data with similar properties into several smaller, more meaningful groups. Two commonly used clustering techniques are introduced in the following section: principal component analysis (PCA) and hierarchical clustering. PCA calculates the variance between variables and groups them into a few uncorrelated groups or principal components (PCs) that are orthogonal to each other. Hierarchical clustering is carried out by separating data into many clusters and merging similar clusters together. Here, we use an example of human leukocyte antigen (HLA) supertype classification to demonstrate the usage of the two methods. Two programs, Generating Optimal Linear Partial Least Square Estimations (GOLPE) and Sybyl, are used for PCA and hierarchical clustering, respectively. However, the reader should bear in mind that the methods have been incorporated into other software as well, such as SIMCA, statistiXL, and R.

Static energy analysis of MHC class I and class II peptide-binding affinity

Antigenic peptide is presented to a T-cell receptor (TCR) through the formation of a stable complex with a major histocompatibility complex (MHC) molecule. Various predictive algorithms have been developed to estimate a peptide's capacity to form a stable complex with a given MHC class II allele, a technique integral to the strategy of vaccine design. These have previously incorporated such computational techniques as quantitative matrices and neural networks. A novel predictive technique is described, which uses molecular modeling of predetermined crystal structures to estimate the stability of an MHC class II-peptide complex. The structures are remodeled, energy minimized, and annealed before the energetic interaction is calculated.

Quantitative peptide binding motifs for 19 human and mouse MHC class I molecules derived using positional scanning combinatorial peptide libraries

BACKGROUND

It has been previously shown that combinatorial peptide libraries are a useful tool to characterize the binding specificity of class I MHC molecules. Compared to other methodologies, such as pool sequencing or measuring the affinities of individual peptides, utilizing positional scanning combinatorial libraries provides a baseline characterization of MHC molecular specificity that is cost effective, quantitative and unbiased.

RESULTS

Here, we present a large-scale application of this technology to 19 different human and mouse class I alleles. These include very well characterized alleles (e.g. HLA A*0201), alleles with little previous data available (e.g. HLA A*3201), and alleles with conflicting previous reports on specificity (e.g. HLA A*3001). For all alleles, the positional scanning combinatorial libraries were able to elucidate distinct binding patterns defined with a uniform approach, which we make available here. We introduce a heuristic method to translate this data into classical definitions of main and secondary anchor positions and their preferred residues. Finally, we validate that these matrices can be used to identify candidate MHC binding peptides and T cell epitopes in the vaccinia virus and influenza virus systems, respectively.

CONCLUSION

These data confirm, on a large scale, including 15 human and 4 mouse class I alleles, the efficacy of the positional scanning combinatorial library approach for describing MHC class I binding specificity and identifying high affinity binding peptides. These libraries were shown to be useful for identifying specific primary and secondary anchor positions, and thereby simpler motifs, analogous to those described by other approaches. The present study also provides matrices useful for predicting high affinity binders for several alleles for which detailed quantitative descriptions of binding specificity were previously unavailable, including A*3001, A*3201, B*0801, B*1501 and B*1503.

HLA class I supertypes: a revised and updated classification

BACKGROUND

Class I major histocompatibility complex (MHC) molecules bind, and present to T cells, short peptides derived from intracellular processing of proteins. The peptide repertoire of a specific molecule is to a large extent determined by the molecular structure accommodating so-called main anchor positions of the presented peptide. These receptors are extremely polymorphic, and much of the polymorphism influences the peptide-binding repertoire. However, despite this polymorphism, class I molecules can be clustered into sets of molecules that bind largely overlapping peptide repertoires. Almost a decade ago we introduced this concept of clustering human leukocyte antigen (HLA) alleles and defined nine different groups, denominated as supertypes, on the basis of their main anchor specificity. The utility of this original supertype classification, as well several other subsequent arrangements derived by others, has been demonstrated in a large number of epitope identification studies.

RESULTS

Following our original approach, in the present report we provide an updated classification of HLA-A and -B class I alleles into supertypes. The present analysis incorporates the large amount of class I MHC binding data and sequence information that has become available in the last decade. As a result, over 80% of the 945 different HLA-A and -B alleles examined to date can be assigned to one of the original nine supertypes. A few alleles are expected to be associated with repertoires that overlap multiple supertypes. Interestingly, the current analysis did not identify any additional supertype specificities.

CONCLUSION

As a result of this updated analysis, HLA supertype associations have been defined for over 750 different HLA-A and -B alleles. This information is expected to facilitate epitope identification and vaccine design studies, as well as investigations into disease association and correlates of immunity. In addition, the approach utilized has been made more transparent, allowing others to utilize the classification approach going forward.

Efficient peptide-MHC-I binding prediction for alleles with few known binders

MOTIVATION

In silico methods for the prediction of antigenic peptides binding to MHC class I molecules play an increasingly important role in the identification of T-cell epitopes. Statistical and machine learning methods in particular are widely used to score candidate binders based on their similarity with known binders and non-binders. The genes coding for the MHC molecules, however, are highly polymorphic, and statistical methods have difficulties building models for alleles with few known binders. In this context, recent work has demonstrated the utility of leveraging information across alleles to improve the performance of the prediction.

RESULTS

We design a support vector machine algorithm that is able to learn peptide-MHC-I binding models for many alleles simultaneously, by sharing binding information across alleles. The sharing of information is controlled by a user-defined measure of similarity between alleles. We show that this similarity can be defined in terms of supertypes, or more directly by comparing key residues known to play a role in the peptide-MHC binding. We illustrate the potential of this approach on various benchmark experiments where it outperforms other state-of-the-art methods.

AVAILABILITY

The method is implemented on a web server: http://cbio.ensmp.fr/kiss. All data and codes are freely and publicly available from the authors.

An evolutionary approach to major histocompatibility diversity based on allele supertypes

Human leukocyte antigens are traditionally classified by serologic or molecular techniques into a bewildering variety of alleles. It is generally believed that this allelic diversity is maintained by selection pressures for inbreeding avoidance and/or maximal immune system diversity. While the usual antigen-based classification of individual alleles may be most appropriate in the artificial situation of tissue transplantation, we hypothesize that a functional classification based on allele supertypes may represent a more biologically relevant way to view MHC diversity in the contexts of mate choice and disease pathogenesis. Furthermore, immune system diversity could be quantitatively estimated by calculating a Supertype Diversity Index (SDI) which is the number of different MHC supertypes possessed by an individual. This hypothesis generates a number of testable predictions. First, it predicts that a reduced inherited diversity of MHC allele supertypes may predispose to the development of malignancies because of a decreased native ability to present different tumor-associated antigens. Furthermore, specific autoimmune diseases may be associated with the presence or absence of a particular MHC supertype rather than a particular MHC haplotype. In transplant medicine, it is possible that unmatched alleles may trigger a weaker foreign antigen response if they are matched by allele supertype. Finally, there have been several studies documenting dissortative mating in humans for dissimilar MHC alleles. We predict that natural selection should favor maximization of the heterozygosity of allele supertypes instead of the heterozygosity of individual alleles and that the previously observed dissortative mating may actually be an adaptive strategy to maximize allele supertype diversity.

Amino acid biophysical properties in the statistical prediction of peptide-MHC class I binding

Background

A key step in the development of an adaptive immune response to pathogens or vaccines is the binding of short peptides to molecules of the Major Histocompatibility Complex (MHC) for presentation to T lymphocytes, which are thereby activated and differentiate into effector and memory cells. The rational design of vaccines consists in part in the identification of appropriate peptides to effect this process. There are several algorithms currently in use for making such predictions, but these are limited to a small number of MHC molecules and have good but imperfect prediction power.

Results

We have undertaken an exploration of the power gained by taking advantage of a natural representation of the amino acids in terms of their biophysical properties. We used several well-known statistical classifiers using either a naive encoding of amino acids by name or an encoding by biophysical properties. In all cases, the encoding by biophysical properties leads to substantially lower misclassification error.

Conclusion

Representation of amino acids using a few important bio-physio-chemical property provide a natural basis for representing peptides and greatly improves peptide-MHC class I binding prediction.

NetMHCpan, a method for quantitative predictions of peptide binding to any HLA-A and -B locus protein of known sequence

BACKGROUND

Binding of peptides to Major Histocompatibility Complex (MHC) molecules is the single most selective step in the recognition of pathogens by the cellular immune system. The human MHC class I system (HLA-I) is extremely polymorphic. The number of registered HLA-I molecules has now surpassed 1500. Characterizing the specificity of each separately would be a major undertaking.

PRINCIPAL FINDINGS

Here, we have drawn on a large database of known peptide-HLA-I interactions to develop a bioinformatics method, which takes both peptide and HLA sequence information into account, and generates quantitative predictions of the affinity of any peptide-HLA-I interaction. Prospective experimental validation of peptides predicted to bind to previously untested HLA-I molecules, cross-validation, and retrospective prediction of known HIV immune epitopes and endogenous presented peptides, all successfully validate this method. We further demonstrate that the method can be applied to perform a clustering analysis of MHC specificities and suggest using this clustering to select particularly informative novel MHC molecules for future biochemical and functional analysis.

CONCLUSIONS

Encompassing all HLA molecules, this high-throughput computational method lends itself to epitope searches that are not only genome- and pathogen-wide, but also HLA-wide. Thus, it offers a truly global analysis of immune responses supporting rational development of vaccines and immunotherapy. It also promises to provide new basic insights into HLA structure-function relationships. The method is available at http://www.cbs.dtu.dk/services/NetMHCpan.

Prediction of supertype-specific HLA class I binding peptides using support vector machines

Experimental approaches for identifying T-cell epitopes are time-consuming, costly and not applicable to the large scale screening. Computer modeling methods can help to minimize the number of experiments required, enable a systematic scanning for candidate major histocompatibility complex (MHC) binding peptides and thus speed up vaccine development. We developed a prediction system based on a novel data representation of peptide/MHC interaction and support vector machines (SVM) for prediction of peptides that promiscuously bind to multiple Human Leukocyte Antigen (HLA, human MHC) alleles belonging to a HLA supertype. Ten-fold cross-validation results showed that the overall performance of SVM models is improved in comparison to our previously published methods based on hidden Markov models (HMM) and artificial neural networks (ANN), also confirmed by blind testing. At specificity 0.90, sensitivity values of SVM models were 0.90 and 0.92 for HLA-A2 and -A3 dataset respectively. Average area under the receiver operating curve (A(ROC)) of SVM models in blind testing are 0.89 and 0.92 for HLA-A2 and -A3 datasets. A(ROC) of HLA-A2 and -A3 SVM models were 0.94 and 0.95, validated using a full overlapping study of 9-mer peptides from human papillomavirus type 16 E6 and E7 proteins. In addition, a large-scale experimental dataset has been used to validate HLA-A2 and -A3 SVM models. The SVM prediction models were integrated into a web-based computational system MULTIPRED1, accessible at antigen.i2r.a-star.edu.sg/multipred1/.

Identifying HLA supertypes by learning distance functions

MOTIVATION

The development of epitope-based vaccines crucially relies on the ability to classify Human Leukocyte Antigen (HLA) molecules into sets that have similar peptide binding specificities, termed supertypes. In their seminal work, Sette and Sidney defined nine HLA class I supertypes and claimed that these provide an almost perfect coverage of the entire repertoire of HLA class I molecules. HLA alleles are highly polymorphic and polygenic and therefore experimentally classifying each of these molecules to supertypes is at present an impossible task. Recently, a number of computational methods have been proposed for this task. These methods are based on defining protein similarity measures, derived from analysis of binding peptides or from analysis of the proteins themselves.

RESULTS

In this paper we define both peptide derived and protein derived similarity measures, which are based on learning distance functions. The peptide derived measure is defined using a peptide-peptide distance function, which is learned using information about known binding and non-binding peptides. The protein derived similarity measure is defined using a protein-protein distance function, which is learned using information about alleles previously classified to supertypes by Sette and Sidney (1999). We compare the classification obtained by these two complimentary methods to previously suggested classification methods. In general, our results are in excellent agreement with the classifications suggested by Sette and Sidney (1999) and with those reported by Buus et al. (2004). The main important advantage of our proposed distance-based approach is that it makes use of two different and important immunological sources of information-HLA alleles and peptides that are known to bind or not bind to these alleles. Since each of our distance measures is trained using a different source of information, their combination can provide a more confident classification of alleles to supertypes.

Definition of MHC supertypes through clustering of MHC peptide-binding repertoires

Identification of peptides that can bind to major histocompatibility complex (MHC) molecules is important for anticipation of T-cell epitopes and for the design of epitope-based vaccines. Population coverage of epitope vaccines is, however, compromised by the extreme polymorphism of MHC molecules, which is in fact the basis for their differential peptide binding. Therefore, grouping of MHC molecules into supertypes according to peptide-binding specificity is relevant for optimizing the composition of epitope-based vaccines. Despite the fact that the peptide-binding specificity of MHC molecules is linked to their specific amino acid sequences, it is unclear how amino sequence differences correlate with peptide-binding specificities. In this chapter, we detail a method for defining MHC supertypes based on the analysis and subsequent clustering of their peptide-binding repertoires.

Grouping of class I HLA alleles using electrostatic distribution maps of the peptide binding grooves

Human leukocyte antigen (HLA) molecules involved in immune function by binding to short peptides (8-20 residues) have different sequences in different individuals belonging to distinct ethnic population. Hence, the peptide-binding function of HLA alleles is specific. Class I HLA alleles (alternative forms of a gene) are associated with CD8+ T cells, and their allele-specific sequence information is available at the IMGT/HLA database. The available sequences are one-dimensional (ID), and the peptide-binding functional inference often requires 3-dimensional (3D) structural models of respective alleles. Hence, 3D structures were constructed for 1,000 class I HLA alleles (310 A, 570 B, and 120 C) using MODELLER (a comparative protein modeling program for modeling protein structures). The electrostatic distribution maps were generated for each modeled structure using Deep View (Swiss PDB Viewer Version 3.7). The 1,000 models were then grouped into different categories by visual inspection of their electrostatic distribution maps in the peptide binding grooves. The distribution of the models based on electrostatic distribution was 30% negative (300), 1% positive (12), 8% neutral (84), and 60% (604) mixed (random mixture of negative, positive, and neutral). This grouping provides insight toward the inference for functional overlap among HLA alleles.

In silico grouping of peptide/HLA class I complexes using structural interaction characteristics

MOTIVATION

Classification of human leukocyte antigen (HLA) proteins into supertypes underpins the development of epitope-based vaccines with wide population coverage. Current methods for HLA supertype definition, based on common structural features of HLA proteins and/or their functional binding specificities, leave structural interaction characteristics among different HLA supertypes with antigenic peptides unexplored.

METHODS

We describe the use of structural interaction descriptors for the analysis of 68 peptide/HLA class I crystallographic structures. Interaction parameters computed include the number of intermolecular hydrogen bonds between each HLA protein and its corresponding bound peptide, solvent accessibility, gap volume and gap index.

RESULTS

The structural interactions patterns of peptide/HLA class I complexes investigated herein vary among individual alleles and may be grouped in a supertype dependent manner. Using the proposed methodology, eight HLA class I supertypes were defined based on existing experimental crystallographic structures which largely overlaps (77% consensus) with the definitions by binding motifs. This mode of classification, which considers conformational information of both peptide and HLA proteins, provides an alternative to the characterization of supertypes using either peptide or HLA protein information alone.

Differential selection pressure exerted on HIV by CTL targeting identical epitopes but restricted by distinct HLA alleles from the same HLA supertype

HLA diversity is seen as a major challenge to CTL vaccines against HIV. One current approach focuses on "promiscuous" epitopes, presented by multiple HLA alleles from within the same HLA supertype. However, the effectiveness of such supertype vaccines depends upon the functional equivalence of CTL targeting a particular epitope, irrespective of the restricting HLA. In this study, we describe the promiscuous HIV-specific CTL epitopes presented by alleles within the B7 supertype. Substantial differences were observed in the ability of CTL to select for escape mutation when targeting the same epitope but restricted by different HLA. This observation was common to all six promiscuous B7 epitopes identified. Moreover, with one exception, there were no significant differences in the frequency, magnitude, or immunodominance of the CTL responses restricted by different HLA alleles to explain these discrepancies. This suggests that the unique peptide/MHC complexes generated by even closely related HLA induce CTL responses that are qualitatively different. This hypothesis is supported by additional differences observed between CTL targeting identical epitopes but restricted by different HLA: first, the occurrence of distinct, HLA-specific escape mutation; second, the recruitment of distinct TCR repertoires by particular peptide/MHC complexes; and, third, significant differences in the functional avidity of CTL. Taken together, these data indicate that significant functional differences exist between CTL targeting identical epitopes but restricted by different, albeit closely related HLA. These findings are of relevance to vaccine approaches that seek to exploit HLA supertypes to overcome the problem of HLA diversity.

Human leukocyte antigen B58 supertype and human immunodeficiency virus type 1 infection in native Africans

Human leukocyte antigen (HLA) class I alleles can be grouped into supertypes according to their shared peptide binding properties. We examined alleles of the HLA-B58 supertype (B58s) in treatment-naïve human immunodeficiency virus type 1 (HIV-1)-seropositive Africans (423 Zambians and 202 Rwandans). HLA-B and HLA-C alleles were resolved to four digits by a combination of molecular methods, and their respective associations with outcomes of HIV-1 infection were analyzed by statistical procedures appropriate for continuous or categorical data. The effects of the individual alleles on natural HIV-1 infection were heterogeneous. In HIV-1 subtype C-infected Zambians, the mean viral load (VL) was lower among B*5703 (P = 0.01) or B*5703-Cw*18 (P < 0.001) haplotype carriers and higher among B*5802 (P = 0.02) or B*5802-Cw*0602 (P = 0.03) carriers. The B*5801-Cw*03 haplotype showed an association with low VL (P = 0.05), whereas B*5801 as a whole did not. Rwandans with HIV-1 subtype A infection showed associations of B*5703 and B*5802 with slow (P = 0.06) and rapid (P = 0.003) disease progression, respectively. In neither population were B*1516-B*1517 alleles associated with more favorable responses. Overall, B58s alleles, individually or as part of an HLA-B-HLA-C haplotype, appeared to have a distinctive impact on HIV-1 infection among native Africans. As presently defined, B58s alleles cannot be considered uniformly protective against HIV/AIDS in every population.

The role of human leukocyte antigen class I polymorphism in HIV/AIDS

PURPOSE OF REVIEW

Polymorphism at human leukocyte antigen class I loci accounts for a portion of the individual variability in susceptibility to, and evolution of, HIV-1 infection. Evaluation of the impact of class I polymorphism has improved, with refined techniques for genotyping, quantifying peptide binding characteristics of class I molecules, and measuring cytotoxic T lymphocyte responses.

RECENT FINDINGS

Most class I alleles that are particularly advantageous or disadvantageous in controlling HIV/AIDS are human leukocyte antigen-B locus products. Some act more uniformly than others, across racial and viral subtype boundaries, and different alleles may exert their effects at different times during the course of infection. Disease progression may be retarded by alleles of less common supertypes. The virus mutates to escape host pressure exerted by class 1 programmed cytotoxic T lymphocytes, and may or may not revert to its earlier form, depending on the toll taken on its replication capacity.

SUMMARY

Large virologically and immunologically well defined cohorts will be critical for increasingly comprehensive evaluation of the multiple small effects conferred by these highly polymorphic loci. Elucidating these effects at the population, clinical, cellular, and molecular levels poses a formidable challenge, but the reward could be vaccine constructs customized to the genetic profiles of individual populations, along with more effective individualized therapeutic intervention.

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.

Statistical deconvolution of enthalpic energetic contributions to MHC-peptide binding affinity

Background

MHC Class I molecules present antigenic peptides to cytotoxic T cells, which forms an integral part of the adaptive immune response. Peptides are bound within a groove formed by the MHC heavy chain. Previous approaches to MHC Class I-peptide binding prediction have largely concentrated on the peptide anchor residues located at the P2 and C-terminus positions.

Results

A large dataset comprising MHC-peptide structural complexes was created by re-modelling pre-determined x-ray crystallographic structures. Static energetic analysis, following energy minimisation, was performed on the dataset in order to characterise interactions between bound peptides and the MHC Class I molecule, partitioning the interactions within the groove into van der Waals, electrostatic and total non-bonded energy contributions.

Conclusion

The QSAR techniques of Genetic Function Approximation (GFA) and Genetic Partial Least Squares (G/PLS) algorithms were used to identify key interactions between the two molecules by comparing the calculated energy values with experimentally-determined BL₅₀ data. Although the peptide termini binding interactions help ensure the stability of the MHC Class I-peptide complex, the central region of the peptide is also important in defining the specificity of the interaction. As thermodynamic studies indicate that peptide association and dissociation may be driven entropically, it may be necessary to incorporate entropic contributions into future calculations.

Modeling the bound conformation of Pemphigus vulgaris-associated peptides to MHC Class II DR and DQ alleles

Background

Pemphigus vulgaris (PV) is a severe autoimmune blistering disorder characterized by the presence of pathogenic autoantibodies directed against desmoglein-3 (Dsg3), involving specific DR4 and DR6 alleles in Caucasians and DQ5 allele in Asians. The development of sequence-based predictive algorithms to identify potential Dsg3 epitopes has encountered limited success due to the paucity of PV-associated allele-specific peptides as training data.

Results

In this work we constructed atomic models of ten PV associated, non-associated and protective alleles. Nine previously identified stimulatory Dsg3 peptides, Dsg3 96–112, Dsg3 191–205, Dsg3 206–220, Dsg3 252–266, Dsg3 342–356, Dsg3 380–394, Dsg3 763–777, Dsg3 810–824 and Dsg3 963–977, were docked into the binding groove of each model to analyze the structural aspects of allele-specific binding.

Conclusion

Our docking simulations are entirely consistent with functional data obtained from in vitro competitive binding assays and T cell proliferation studies in DR4 and DR6 PV patients. Our findings ascertain that DRB1*0402 plays a crucial role in the selection of specific self-peptides in DR4 PV. DRB1*0402 and DQB1*0503 do not necessarily share the same core residues, indicating that both alleles may have different binding specificities. In addition, our results lend credence to the hypothesis that the alleles DQB1*0201 and *0202 play a protective role by binding Dsg3 peptides with greater affinity than the susceptible alleles, allowing for efficient deletion of autoreactive T cells.

HLA-B polymorphism in Japanese HIV-1-infected long-term surviving hemophiliacs

Approximately 30% of patients with hemophilia in Japan were infected with human immunodeficiency virus (HIV) in early 1980s through contaminated blood products. In 1995, a cohort of HIV-infected, asymptomatic patients with hemophilia was set up for follow-up study. Although the patients met the criteria for long-term non-progressor (LTNP) at the entry to the cohort, some of them later developed lymphopenia during five more years of observation. We collected blood samples from 80 long-term survivors; 42 of them did not require antiviral therapy, but the rest were under treatment. Analysis of HLA-B genotype revealed that carriers of known HIV-resistant alleles such as HLA-B*5701, B*5801, and alleles of B27 antigenic group were not increased in frequency, but that HLA-B*1507 was increased in the cohort (6.25% vs. 1.03%, OR = 6.40, p = 0.039). We also observed the decrease in carriers of HLA-B*5401 (3.75% vs. 14.95%, OR = 0.22, p = 0.016). HLAB* 5401 is a relatively common allele in East Asian populations and belongs to the same B22 antigenic group as B55 and B56 which were reported to associate with rapid progression. Our data indicated that HLA class I is one of the host factors involved in the retardation of HIV disease progression as also reported in the previous studies; however, the alleles associated with this resistance were not the same because of divergent host genetic background.

In silico identification of supertypes for class II MHCs

The development of epitope-based vaccines, which have wide population coverage, is greatly complicated by MHC polymorphism. The grouping of alleles into supertypes, on the basis of common structural and functional features, addresses this problem directly. In the present study we applied a combined bioinformatics approach, based on analysis of both protein sequence and structure, to identify similarities in the peptide binding sites of 2225 human class II MHC molecules, and thus define supertypes and supertype fingerprints. Two chemometric techniques were used: hierarchical clustering using three-dimensional Comparative Similarity Indices Analysis fields and nonhierarchical k-means clustering using sequence-based z-descriptors. An average consensus of 84% was achieved, i.e., 1872 of 2225 class II molecules were classified in the same supertype by both techniques. Twelve class II supertypes were defined: five DRs, three DQs, and four DPs. The HLA class II supertypes and their fingerprints given in parenthesis are DR1 (Trp(9beta)), DR3 (Glu(9beta), Gln(70beta), and Gln/Arg(74beta)), DR4 (Glu(9beta), Gln/Arg(70beta), and Glu/Ala(74beta)), DR5 (Glu(9beta), Asp(70beta)), and DR9 (Lys/Gln(9beta)); DQ1 (Ala/Gly(86beta)), DQ2 (Glu(86beta), Lys(71beta)), and DQ3 (Glu(86beta), Thr/Asp(71beta)); DPw1 (Asp(84beta) and Lys(69beta)), DPw2 (Gly/Val(84beta) and Glu(69beta)), DPw4 (Gly/Val(84beta) and Lys(69beta)), and DPw6 (Asp(84beta) and Glu(69beta)). Apart from the good agreement between known binding motifs and our classification, several new supertypes, and corresponding thematic binding motifs, were also defined.

Classification of A1- and A24-supertype molecules by analysis of their MHC-peptide binding repertoires

At the functional level, the majority of human leukocyte antigen (HLA) class I MHC variants can be classified into about ten different major groups, or supertypes, characterized by overlapping peptide binding motifs and repertoires. Previous studies have detailed the peptide binding specificity of the HLA A2, A3, B7, and B44 supertypes, and predicted, on the basis of MHC pocket structures, known motifs, or the sequence of T cell epitopes, the existence of the HLA A1 and A24 supertypes. Direct experimental validation of the A1 and A24 supertypes, however, has been lacking. In the current study, the peptide-binding repertoires and main anchor specificities of several common HLA A molecules (A*0101, A*2301, A*2402, A*2601, A*2902, and A*3002) predicted to be members of the A1 or A24 supertypes were analyzed and defined using single amino acid substituted peptides and a large peptide library. Based on the present findings, the A1 supertype includes A*0101, A*2601, A*2902, and A*3002, whereas the A24 supertype includes A*2301 and A*2402. Interestingly, A*2902 is associated with a motif and peptide binding repertoire that overlaps significantly with those of all of the A1- and A24-supertype molecules studied, representing-to our knowledge-the first report of significant cross-reactivity among molecules belonging to different supertypes.