Clinical Immunology (2014) 153, 31–39
available at www.sciencedirect.com
Clinical Immunology www.elsevier.com/locate/yclim
Meta-analysis of IgE-binding allergen epitopes Virginie Lollier ⁎, Sandra Denery-Papini, Chantal Brossard, Dominique Tessier UR1268 BIA (Biopolymers Interactions Assemblies), INRA, 44300 Nantes, France
Received 10 March 2014; accepted with revision 19 March 2014 KEYWORDS IgE epitopes; Allergens; Bioinformatics; Database
Abstract IgE-binding epitopes are related to allergic symptoms by eliciting degranulation of special cells and release of molecules that trigger the hypersensitivity reaction. Little is known about what characterises allergen IgE-binding epitopes, although advances in analytical methods have led to the identification of a large number of them. To assess if a binary classification of allergen regions into epitopes or non-epitopes may accurately reflect biological reality, we computed the fraction of allergen amino acids that are involved in epitopes. A relationship between this fraction and the increasing number of literature references was modelled. Due to the wide variety of methods that are used in the literature, a peak in the number of matches between an allergen sequence and its epitopes confirms their validity. Accordingly, our graphical representation of positive assays along sequences provides an overview of epitope localisation, which should help to highlight major positions for IgE binding to allergens. © 2014 Elsevier Inc. All rights reserved.
1. Introduction Specific recognition of epitopes on an allergen by IgE antibodies that are bound to effector cells such as basophils is essential for the development of the allergic response. Consequently, B-cell epitope identifications play an important role in vaccine design, disease diagnosis and allergy research. Epidemiological studies at a molecular level that may highlight the parts of the protein most often involved in specific recognition by IgE antibodies are rare. Indeed, an
Abbreviations: FAE, fraction of amino acids in a protein sequence that are involved in epitopes. ⁎ Corresponding author at: UR1268 BIA (Biopolymers Interactions Assemblies), INRA, 44300 Nantes, France. E-mail address: [email protected] (V. Lollier).
http://dx.doi.org/10.1016/j.clim.2014.03.010 1521-6616/© 2014 Elsevier Inc. All rights reserved.
exhaustive study of the epitope repertoire of an allergen in multiple environments and in a large cohort is currently too expensive and may be technically impossible due to low quantities of the available sera. However, many experimental techniques have been widely used worldwide to identify and map epitopes on allergens. The resulting accumulated knowledge has been partially distributed among several databases [1–3]. However, those databases do not provide a condensed view of the epitope distribution on each allergen. In a previous study [4], we have shown that the cumulative epitope identifications mapped nearly everywhere along the protein sequence of the milk and latex major allergens (i.e., Bos d 5 and Hev b 6). By defining the B-cell epitope that is involved in a normal immunological response, Van Regenmortel [5] indicated that the number of identified B-cell epitopes of an antigen
32 corresponds to the size of the analysed immunological repertoire of a host that is immune to that protein. IgE antibodies tend to be more often cross-reactive than IgG antibodies [6]; therefore, they might be more numerous. In addition, Greenbaum et al. [7] suggested that the antigen surface might be considered a continuous landscape of epitopic regions. In the case of allergens, this surface undergoes unknown modifications during the crossing of the epithelial barrier. The aim of this article is to explore whether B-cell epitopes of allergens can be localised everywhere: not only on the overall protein surface but also all along its primary sequence. In this meta-analysis, we further examined whether the frequencies of amino-acid positions in the primary sequence, which are involved in epitopes could be used to overcome the lack of epidemiological data at the molecular level.
2. Methods 2.1. Selection of epitope data The data related to B-cell epitopes were extracted from the Immune Epitope DataBase (IEDB, August 2013). Only epitopes corresponding to positive assays for the IgE isotype were selected. The epitopes were also filtered based on their unique GenBank identifier, and only epitopes that were identified in allergens were maintained. The epitope data and data related to allergen names and clusters were stored in a relational database with the help of the ETL (Extract Transform and Load) software called Talend Open Studio. In the database, epitopes were described by their continuous or discontinuous sequence, their position within the protein, their bibliographic references and their assay types based on the IEDB ontology. One epitope longer than 100 amino acids was evaluated as too imprecise and removed from the dataset. Data stored in the relational database was used in the web application LocAllEpi.
2.2. Protein sequence selection The sequence identifiers of multiple public databases that corresponded to known allergen sources were collected from the IUIS [8] (allergen.org) and Allergome [9] web sites during August 2013. Based on these identifiers, the allergen sequences were extracted from the requests to the National Center for Biotechnology Information (NCBI) using Entrez Programming Utilities [10] and its cross-referencing system. Protein sequences were selected if they were referred to in IEDB, and their clustering was achieved with the help of the Cd-hit software (version 4.5.4) [11]. The clusters were labelled by the allergen name of the longest sequence. The epitope localisation within clusters was achieved by correcting the amino acid numbering from multiple sequence alignments of the variants using the clustalw software (version 2.1) [12]. Only clusters having at least one sequence member related to an identified IgE-binding epitope were used for further analysis.
V. Lollier et al.
2.3. Computation of the fraction of amino-acids involved in epitopes (FAE) For each allergen and cluster sequence, we computed the number of amino acids that were involved in at least one positive epitope assay and divided this number by the sequence length. We called the resulting value FAE (or fraction of amino-acids involved in epitopes). The FAE represents the fraction of overlap between a protein sequence and its set of epitope sequences. A complete overlap between a protein and its epitope repertoire is expressed by an FAE value close or equal to one. This implies that all amino acids of the allergen cluster or sequence are involved in epitopes. This fraction was graphically represented as a function of how many studies had been conducted in the literature or how many different techniques were performed. On the resulting scatter plots, each point represents a sequence or a cluster of sequences. To compare allergens and clusters regardless of their length, the number of references and of different assay types were also divided by the number of amino acids that constituted the sequence. The assay types were named with the Ontology of Immune Epitopes (ONTIE) used by IEDB to represent experiments that identify and characterize immune epitopes. A fitted curve was computed using the nls function within R for non-linear regression of data.
2.4. Display of the frequencies of overlaps between epitopes along sequences With the help of the start and end positions of the collected epitopes in the allergen sequence or in the cluster alignment, we computed the number of times an amino acid position was found in a positive assay among the references. Numbers obtained for each amino acid were graphically mapped as a function of their position within the different protein sequences. These frequencies of overlaps were associated with additional data in 3 graphics: with the number of positive sera in Appendix A, being labelled with their PubMed identifiers in Fig. 1 and with the assay type that was related to the studied epitope in Appendix B Computations and graphs were produced with the help of the R framework.
3. Results and discussion 3.1. Relationship between the abundance of literature and the FAE Our dataset was composed of 2971 IgE B-cell epitopes of allergens. Although the Immune Epitope DataBase [1] is not dedicated to allergy, we chose this database as a central hub for overall epitope information. It includes nearly 40% of the International Union of Immunological Societies (IUIS) allergens [13]. Compared to the existing databases on B-cell epitopes [2,3], the IEDB ontology provides a standardised description of the experimental conditions under which the epitopes were identified. This database is also frequently updated with continuous and discontinuous B-cell epitopes from many antigens. Because the experimental work on
Meta-analysis of IgE-binding allergen epitopes
33
Figure 1 Fraction of amino-acids involved in epitopes according to the increasing number of references. Proteins and clusters are coloured in blue, green, red and orange, respectively, for food, respiratory, skin and sting allergens.
protein sequences often implies the production of isoforms or variants, the collected sequences were clustered according to their similarity to obtain epitope knowledge along a single representative sequence of the allergen. Each cluster was labelled with the allergen name of the longest sequence. The allergens where further represented by 112 protein and cluster sequences. Assuming that the whole protein sequences have been scanned for epitope localisations, we analysed the FAE (see Methods section for the computation of Fraction of Amino-acids involved in Epitopes) in relation to how often the allergen molecules were studied. Fig. 1 shows that all residues of an allergen sequence were progressively involved in epitopes when the number of references increased. This result concerns food as well as respiratory and skin allergens, with the exception of the Pru a 1 cluster. Indeed, this allergen has been mostly studied with Mal d 1 using site-directed mutagenesis at a few limited, shared locations to explore cross-reactivity of Bet v 1 homologous allergens [14,15]. In this case, because a large part of the protein has not been scanned, the Pru a 1 cluster was considered as an outlier and was removed from the fitted curve computation. The resulting curve shows that the data follows an exponential model with significant parameters, where the number of amino acids constituting epitopes is a function of the number of references. In this model, the maximum FAE is close to 90%. Few references were sufficient to identify nearly 50% of the amino acids that took part in epitopes. By extrapolation from the fitted curve, this value could be obtained from two articles for a protein consisting of 225 amino acids, which was the median length of sequences in our dataset. Above this threshold, additional epitopic elements from further studies were less numerous, suggesting an overlapping of identifications between references.
3.2. Implications of allergen clustering We evaluated the impact of clustering on our results. As defined by Chapman et al. [16], variants could display greater than 70% sequence identity and share identical biochemical features. We chose this threshold because this similarity should result in potential cross-reactions in the context of allergy. Fig. 2 illustrates how the epitope data were spread over multiple isoforms of four major allergens (the complete list of allergen clusters and unclustered sequences are in the Appendix C). Most of the clusters contained isoforms of the same protein, although some clusters could be extended to allergens from the same conserved protein family. For example, the Der f 2 cluster also contained the Der p 2 sequences. However, the extension of sequences beyond isoforms of the same allergen had little influence on our measured FAE. In the example of Der f 2, the gathered sequence clustering data was related to an allergen protein from European and American house dust mites. As shown on Fig. 2, Der f 2 and Der p 2 allergens have similar sequences, with the exception of three residues (in bold face). Epitopes identified on the Der p 2 allergen were distributed nearly entirely along the representative sequence of the cluster (mauve, pink and light blue points, Fig. 3). Epitopes from Der f 2 overlapped with some Der p 2 epitopes, even at the less conserved positions 36 and 105. More generally, when we inspected the distribution of epitope references within the extended clusters shown in the table of Appendix C, few references were provided based on the extent of the clusters to additional allergens from the same protein family. Two clusters, named Bet v 2 and Cor a 9, were exceptions. The allergen sources within these last clusters, belonging, respectively to plant profilins and 11S globulins (legumins), are known to cross-react.
34
V. Lollier et al.
Figure 2 Sequence members of clusters representing four major allergens and multiple sequence alignment of Der f 2–Der p 2 proteins (where “*” means identity of sequences, “:” conserved residues and “.” semi-conserved residues). The identity percentage of a sequence is related to the representative sequence, which is indicated by an “x” within each cluster.
3.3. Influence of techniques Because the epitopes from the literature have been rapidly identified along whole protein sequences, we wondered
if we faced a complementarity of identification techniques or/and some imprecision of the most common of them. Most of the continuous and discontinuous epitopes were published after 1995. Since that time, analytical techniques
Figure 3 Epitope identifications for each sequence belonging to the Der f 2 cluster. The curves labelled 416898, 37958157 and 55859468 correspond to the GenBank identifiers of the Der f 2 variants. The labels 256095984, 157829757 and 1352237 are the GenBank identifiers of the Der p 2 variants. Positions where the amino acids differed in the multiple sequence alignment are indicated by black arrows, and letters correspond to amino acids of the 37958157 sequence id. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
Meta-analysis of IgE-binding allergen epitopes have rapidly evolved from hydrolysis fragmentation to X-ray 3D structure-solving, and the epitope sequence size has varied. In the present dataset, the epitopes were composed of peptides between 1 and 95 amino acids, and 70% of them were smaller than 20 amino acids. Epitopes of the continuous form appeared to predominate in datasets because the protein mutation and antibody/antigen structural resolution imply more complex steps than measuring IgE binding to protein-derived peptides. Nevertheless, the classification of epitopes into continuous and discontinuous forms has been recently revised, suggesting no clear-cut difference between these two types of epitopes [5,6]. For example, the weak binding of the smallest linear peptides could lead to the assumption that such peptides may be parts of larger, discontinuous epitopes [17]. Concerning the ideal epitope size, some results suggest that allergen epitopes are short stretches of the protein sequence [18,19], while other studies observed that different sized peptides display different profiles with reactive sera [17,20]. To assess if one particular technique predominates and explains the gradual overlap of protein sequence with epitopes, we examined the potential relationship between the assay types and the median FAE per sequence for each analysis reference (Fig. 4). The names of the assay types correspond to the IEDB ontology, which addresses either the technique that was used to identify B-cell epitopes (e.g., ELISA, X-ray crystallography) or an experimental scope (e.g., “Reduction of disease after treatment” or “Induction of tolerance”). The most frequent identification techniques were “Western/Immunoblot”, “ELISA” and “Antigen Competition of Antibody Binding”, which are similar because they focus on epitope mapping along the entire protein sequence. However, the FAE brought by these techniques did not exceed 30%. No assay type displayed an extremely high FAE, with the exception of “Phage Display/Immunopanning”.
Figure 4
35 However, this technique was related to only three references, which constitutes a weak sampling. In addition, the curve displayed in Fig. 5 further oriented to a complementarity of assay types. As an exception, the Phl p 2 allergen was plotted out of the regressive curve because a single reference was related to this allergen, whereas multiple analyses were conducted on a focused area (PUBMED: 19201867 [21]). In other cases, epitope matching with all amino acids occurred progressively according to the increasing number of different applied techniques. However, each of these techniques should have been applied on different populations, which should contribute to epitope diversity. The comparison of Figs. 1 and 6 gives some indications about this contribution. When points representing allergens move to the left from Figs. 1 to 6, the FAE is more explained by the number of references than the number of techniques. There is, therefore, an influence of the number of tested populations. This is the case, for example, for Bos d 9. Accordingly, when points move to the right, the FAE should then be linked to the number of different techniques that are involved in epitope identifications (e.g., Bet v 4). If one technique introduces imprecision, we should observe a fast and systematic overlap between epitopes and the protein sequence when this technique is used. This is not the case, as shown in Fig. 4. Moreover, Fig. 5 highlights that this overlap is at least partially due to the combination of several techniques. These results suggest more complementarity of the experimental conditions than an imprecision of some techniques.
3.4. Prediction concerns During our study, we noticed that some proteins were not entirely scanned for epitope localisation. However, two observations tend to confirm that B-cell epitopes of allergens
Frequencies of applied techniques and their corresponding median FAE.
36
V. Lollier et al.
Figure 5 FAE against the number of different assay types per allergen sequence. Proteins and clusters are coloured in blue, green, red and orange, respectively, for food, respiratory, skin and sting allergens. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
can be localised everywhere along primary sequences. The first is provided by the overall tendency that is shown in Fig. 1. The second observation relies on the fact that the FAE remains high for these proteins (e.g., more than 60% for Bet v 1). This localisation of epitopes along allergens raises further questions regarding the application of existing epitope prediction tools in the context of allergy. Their presence all along the sequence of allergens may be related to their ability to cross-react with antibodies. Epitopes of lesser affinity may then be involved in the allergic reaction without having induced the production of specific antibodies by B-cells. Considering the present level of information on epitopes, it is difficult to classify them according to their affinity. Moreover, only the BciPep database stores the immunogenicity of collected epitopes. Unfortunately, this database, which is not dedicated to allergy, gathers only 2% of IgE epitopes (December 2013). In addition, building datasets to classify epitopic amino acids against the non-epitopic amino acids of a sequence will be challenged when identifications all along the allergen are lacking. In the allergy context, the available data should be restricted to major allergens, and major-studied regions should be linked to symptom severity and, eventually, sequence conservation. A more extended meta-analysis of IEDB content has suggested a potential bias in dataset composition because a relationship between the number of epitopes and the morbidity of antigens was observed [22]. In the general context of antigenicity, prediction tools often consider epitopes as intrinsic features of the protein because they are based on their structure [7]. This is the case when the relationship between the antigenicity of peptides and their amino-acid sequence [23,24], hydrophilicity propensity [25] or surface-exposure [26] is studied. Actually, if epitopes are homogeneously distributed all along the allergen sequences, which remains to be verified, no specific structural feature or motif should emerge.
3.5. Analysis of epitope distribution in four examples The distribution of positive assays that led to epitope identifications in the literature was examined and reported in detail for four major allergens: Bet v 1, Der f 2, Ara h 2 and Bos d 5. These allergens have been highly studied in at least 10 references, and they displayed high FAE. As shown in Fig. 6, the positions of the positive assays were superimposed in the order of the publication year of their related reference. In the case of the Bos d 5 allergen, from one reference to another, the epitopes were homogeneously found along the entire sequence. These were identified in various cohort sizes (10, 30 or 40 individuals), and the studies were mainly performed in USA, China or Australia. The graph corresponding to Bos d 5 shows that epitopes may be any part of the allergen, but are less frequently found around position 80, although some residues of a discontinuous epitope were found in that region using X-ray crystallography. Thirteen references related to this allergen were found, but at least three placed all amino acids of the sequence in epitopes (PUBMED: 19577281, 11729348 and 22939793). However, Fig. 6 shows a high level of positive assays around positions 140–150, where information provided by multiple references was aggregated. Similarly, Ara h 2 displayed three areas that emerged as major epitopic portions of the allergen, in spite of a less homogeneous distribution of identification frequencies. However, the whole protein sequence was not systematically studied in each publication. For example, the reference identifier 21883278 [27] focused on an area based on its resemblance to the Jug r 2 allergen and its potential cross-reactivity to this allergen. The identifier 22042002 [28] also corresponds to a study of a limited area. These two references were the most recent works that were collected for Ara h 2 and were published in 2011 and 2012, which may suggest a focus of recent studies to restricted areas with
Meta-analysis of IgE-binding allergen epitopes
37
Figure 6 Superimposition of epitope identifications along the sequences of four major allergens, from oldest to the most recent references, labelled by their PUBMED identifier (PMID).
regards to previous results. However, a microarray immunoassay experiment (PUBMED: 18234310 [29]) displayed an FAE of approximately 70% of the whole sequence (excluding the 20 amino-acid signal peptide). Authors of this work have related the clinical sensitivity of the 24 studied patients to the epitope diversity. The identification of epitopes all along the Der f 2 respiratory allergen appeared more progressively with the
growing number of references than for the Bos d 5 food allergen. Like Ara h 2 and Bos d 5, this sequence contains a signal peptide. Signal peptides are generally excluded from the identification of B-cell epitopes, which should explain why the regressive curve of the FAE displayed a limitation of 85-90% of the protein sequence (Ymax = 0.87). Moreover, the number of positive assays that mapped on Bos d 5, Ara h 2, Der f 2 ranged differently, although the number of references, the
38 number of applied techniques and the FAE were similar. When the signal peptide is omitted, the number of positive assays ranged between 5 and 20 positive assays for Bos d 5, and between 1 and 10 for Der f 2. Because we computed the positive assay frequencies at each position, the height of any identification peak was partly linked to the overlap size of the assayed synthetic peptides, but also to the cohort size. The number of positive assays is thus a combination of multiple information, including various assay types and the epidemiology of the epitopes. As a consequence, the highest peaks may be considered key areas for IgE binding. In the case of Bet v 1, the epitope information was less abundant. This allergen has been described as displaying mostly conformational epitopes; however, the mapping of this epitope type remains a difficult task [30,31]. Indeed, this allergen seems to have been studied in a piecemeal fashion. The positive results displayed here were extracted from publications where data were not available along the entire protein sequence. Some negative results from other articles localised to the stretches of sequence at positions 1–22, 60–70 and 106–126. These last works aimed at developing immunotherapies by screening positive results on IgG isotypes and negative results on patient IgEs, while setting aside potential positive IgE reactions on the remaining sequence. The inclusion of negative results in this dataset might have weighted the positive amino-acid fraction that was computed on allergen sequences and would have highlighted potential major areas of focus. Furthermore, this could have also differentiated untested stretches of sequence from negative results in this type of graphical representation. However, negative results are often unpublished. In addition, such weighting required a homogeneous distribution of positive and negative data, and, as a consequence, the protein sequence should be entirely assessed and both results diffused. Unfortunately, some experiments are technically unable to satisfy these criteria. These four examples indicate that epitopes are distributed heterogeneously along a protein sequence, consequently resulting in potential major areas for immunoreactivity. Despite limitations linked to the current data on some allergens (negative results are unknown or analysis is limited to parts of sequences), these examples illustrated how to localise such major areas on a given allergen protein.
3.6. Major epitope localisation In our four detailed examples, a peak of positive assays indicated technical confirmation of an epitopic area, such as a confident result, rather than an immunodominant area. Indeed, the overlapping of identifications corresponds to a convergence of multiple technical approaches, including assays with synthetic peptides in solution vs. adsorbed peptides or epitopes of different lengths. Such peaks therefore locate areas that interact with antibodies in multiple configurations and eventually among different populations of patients. Therefore, these areas should be stable during processing of the allergen by the immune system. To provide a snapshot of the current knowledge of epitope collection for allergens and to locate potential major areas on their sequences, we have developed a web-based tool
V. Lollier et al. that is freely accessible at http://wwwappli.nantes.inra.fr/ LocAllEpi. It also includes computed surface features when the 3D structure is available. For localisation of immunodominant areas, our web tool displays an option for the number of positive sera in addition to binding assays. However, these data are often missing, especially when using pooled sera experiments. When available, this information is displayed as an absolute value and not as a relative value to the cohort size, which varies along the protein sequence according to references. We have noticed that the median size of cohorts was close to 10 patients, which might be used as a reference value. Some differences in identification frequencies were observed between molecules. These differences may be due to the diversity of experimental conditions or to the molecule itself, such as Bet v 1, where classical epitope mapping methods produced few positive results. In this situation, it is difficult to determine whether intrinsic characteristics of the protein intervene or to measure how the allergen is altered or protected by its environment during its processing by the immune system.
4. Conclusions When all epitopic areas are mapped independently of the experimental context, our meta-analysis confirms that all parts of an allergen sequence can be part of an epitope. This progressive sequence overlapping is not the result of a predominant identification technique. However, some parts of the allergen are more often involved than others. For this reason, we suggest that the integration of data from different studies could be of interest. This available data on epitope immunogenicity and epidemiology is highly heterogeneous and makes building learning data sets for prediction difficult. Nevertheless, the mapping of all positive assays along allergens provides a statistic view that would help to delineate major areas of interest for the development of diagnostic tools. Sharing information on a dedicated internet website should enlarge the dataset to more allergens and refine the location of major epitopes. For instance, the IEDB website provides any user the ability to submit data for curation through a Data Submission Tool (https://dst.liai.org). Studies on IgE epitopes are often dedicated to a precise application (vaccination or diagnosis on major allergens), and intermediary data, which could be valued in an extended context, are generally excluded during the publication process. However, these additional data could also lead to interesting analyses for characterising B-cell epitopes beyond the study of one allergen or protein family. Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.clim.2014.03.010.
Conflict of interest statement The authors declare that there are no conflicts of interest.
Acknowledgments This work was supported by the ANR (PREDEXPITOPE, ANR08-ALIA-14).
Meta-analysis of IgE-binding allergen epitopes
39
References [1] R. Vita, L. Zarebski, J.A. Greenbaum, H. Emami, I. Hoof, N. Salimi, R. Damle, A. Sette, B. Peters, The immune epitope database 2.0. Nucleic Acids Res. 38 (2010) D854–D862. [2] S. Saha, M. Bhasin, G.P.S. Raghava, Bcipep: a database of b-cell epitopes, BMC Genomics 6 (2005) 79. [3] O. Ivanciuc, C.H. Schein, W. Braun, SDAP: database and computational tools for allergenic proteins, Nucleic Acids Res. 31 (2003) 359–362. [4] V. Lollier, S. Denery-Papini, C. Larré, D. Tessier, A generic approach to evaluate how B-cell epitopes are surface-exposed on protein structures, Mol. Immunol. 48 (2011) 577–585. [5] M. Van Regenmortel, What is a B-cell epitope? in: Mike Schutkowski, Ulrich Reineke (Eds.), Epitope Mapping Protocols, Humana Press, New York, 2009, pp. 3–20. [6] R.C. Aalberse, R. Crameri, IgE-binding epitopes: a reappraisal, Allergy 66 (2011) 1261–1274. [7] J.A. Greenbaum, P.H. Andersen, M. Blythe, H. Bui, R.E. Cachau, J. Crowe, M. Davies, A.S. Kolaskar, O. Lund, S. Morrison, B. Mumey, Y. Ofran, J. Pellequer, C. Pinilla, J.V. Ponomarenko, G.P.S. Raghava, M.H.V. van Regenmortel, E.L. Roggen, A. Sette, A. Schlessinger, J. Sollner, M. Zand, B. Peters, Towards a consensus on datasets and evaluation metrics for developing B-cell epitope prediction tools, J. Mol. Recognit. 20 (2007) 75–82. [8] T.P. King, D. Hoffman, H. Lowenstein, D.G. Marsh, T.A. PlattsMills, W. Thomas, Allergen nomenclature. WHO/IUIS allergen nomenclature subcommittee, Int. Arch. Allergy Immunol. 105 (1994) 224–233. [9] A. Mari, E. Scala, P. Palazzo, S. Ridolfi, D. Zennaro, G. Carabella, Bioinformatics applied to allergy: allergen databases, from collecting sequence information to data integration. the allergome platform as a model. Cell. Immunol. 244 (2006) 97–100. [10] NCBI Resource Coordinators, Database resources of the national center for biotechnology information, Nucleic Acids Res. 41 (2013) D8–D20. [11] W. Li, A. Godzik, Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences, Bioinformatics 22 (2006) 1658–1659. [12] M.A. Larkin, G. Blackshields, N.P. Brown, R. Chenna, P.A. McGettigan, H. McWilliam, F. Valentin, I.M. Wallace, A. Wilm, R. Lopez, J.D. Thompson, T.J. Gibson, D.G. Higgins, Clustal W and clustal X version 2.0, Bioinformatics 23 (2007) 2947–2948. [13] K. Vaughan, J. Greenbaum, Y. Kim, R. Vita, J. Chung, B. Peters, D. Broide, R. Goodman, H. Grey, A. Sette, Towards defining molecular determinants recognized by adaptive immunity in allergic disease: an inventory of the available data, J. Allergy (Cairo) (2010) 628026. [14] D.Y. Son, S. Scheurer, A. Hoffmann, D. Haustein, S. Vieths, Pollen-related food allergy: cloning and immunological analysis of isoforms and mutants of Mal d 1, the major apple allergen, and Bet v 1, the major birch pollen allergen, Eur. J. Nutr. 38 (1999) 201–215. [15] Y. Ma, G. Gadermaier, B. Bohle, S. Bolhaar, A. Knulst, Z. Markovic-Housley, H. Breiteneder, P. Briza, K. HoffmannSommergruber, F. Ferreira, Mutational analysis of amino acid positions crucial for IgE-binding epitopes of the major apple (Malus domestica) allergen, Mal d 1, Int. Arch. Allergy Immunol. 139 (2006) 53–62. [16] M.D. Chapman, A. Pomés, H. Breiteneder, F. Ferreira, Nomenclature and structural biology of allergens, J. Allergy Clin. Immunol. 119 (2007) 414–420. [17] G. Schramm, A. Bufe, A. Petersen, H. Haas, R. Merget, M. Schlaak, W.M. Becker, Discontinuous IgE-binding epitopes
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
contain multiple continuous epitope regions: results of an epitope mapping on recombinant Hol l 5, a major allergen from velvet grass pollen, Clin. Exp. Allergy 31 (2001) 331–341. G.A. Bannon, T. Ogawa, Evaluation of available IgE-binding epitope data and its utility in bioinformatics, Mol. Nutr. Food Res. 50 (2006) 638–644. K.L. Bøgh, H. Nielsen, C.B. Madsen, E.N.C. Mills, N. Rigby, T. Eiwegger, Z. Szépfalusi, E.L. Roggen, IgE epitopes of intact and digested Ara h 1: a comparative study in humans and rats, Mol. Immunol. 51 (2012) 337–346. V. van Kampen, V. Liebers, I. Sander, Z. Chen, X. Baur, M. Raulf-Heimsoth, F.W. Falkenberg, B-cell epitopes of the allergen Chi t 1.01: peptide mapping of epitopes recognized by rabbit, murine, and human antibodies. Allergy 56 (2001) 118–125. S. Padavattan, S. Flicker, T. Schirmer, C. Madritsch, S. Randow, G. Reese, S. Vieths, C. Lupinek, C. Ebner, R. Valenta, Z. Markovic-Housley, High-affinity IgE recognition of a conformational epitope of the major respiratory allergen Phl p 2 as revealed by X-ray crystallography, J. Immunol. 182 (2009) 2141–2151. V. Davies, K. Vaughan, R. Damle, B. Peters, A. Sette, Classification of the universe of immune epitope literature: representation and knowledge gaps, PLoS One 4 (2009) e6948. B. Yao, L. Zhang, S. Liang, C. Zhang, Svmtrip: a method to predict antigenic epitopes using support vector machine to integrate tri-peptide similarity and propensity, PLoS ONE 7 (2012) e45152. J. Gao, E. Faraggi, Y. Zhou, J. Ruan, L. Kurgan, BEST: improved prediction of B-cell epitopes from antigen sequences, PLoS ONE 7 (2012) e40104. M. Odorico, J. Pellequer, BEPITOPE: predicting the location of continuous epitopes and patterns in proteins, J. Mol. Recognit. 16 (2003) 20–22. P. Haste Andersen, M. Nielsen, O. Lund, Prediction of residues in discontinuous B-cell epitopes using protein 3D structures, Protein Sci. 15 (2006) 2558–2567. S.J. Maleki, S.S. Teuber, H. Cheng, D. Chen, S.S. Comstock, S. Ruan, C.H. Schein, Computationally predicted ige epitopes of walnut allergens contribute to cross-reactivity with peanuts, Allergy 66 (2011) 1522–1529. L. Rosenfeld, W. Shreffler, L. Bardina, B. Niggemann, U. Wahn, H.A. Sampson, K. Beyer, Walnut allergy in peanut-allergic patients: significance of sequential epitopes of walnut homologous to linear epitopes of Ara h 1, 2 and 3 in relation to clinical reactivity, Int. Arch. Allergy Immunol. 157 (2012) 238–245. A.E. Flinterman, E.F. Knol, D.A. Lencer, L. Bardina, C.F. den Hartog Jager, J. Lin, S.G.M.A. Pasmans, C.A.F.M. BruijnzeelKoomen, H.A. Sampson, E. van Hoffen, Shreffler W.G., Peanut epitopes for IgE and IgG4 in peanut-sensitized children in relation to severity of peanut allergy, J. Allergy Clin. Immunol. 121 (2008) 737–743 (e10). M.D. Spangfort, O. Mirza, H. Ipsen, R.J.J. Van Neerven, M. Gajhede, J.N. Larsen, Dominating IgE-binding epitope of Bet v 1, the major allergen of birch pollen, characterized by X-ray crystallography and site-directed mutagenesis, J. Immunol. 171 (2003) 3084–3090. A. Gieras, P. Cejka, K. Blatt, M. Focke-Tejkl, B. Linhart, S. Flicker, A. Stoecklinger, K. Marth, A. Drescher, J. Thalhamer, P. Valent, O. Majdic, R. Valenta, Mapping of conformational IgE epitopes with peptide-specific monoclonal antibodies reveals simultaneous binding of different IgE antibodies to a surface patch on the major birch pollen allergen, Bet v 1, J. Immunol. 186 (2011) 5333–5344.
available at www.sciencedirect.com
Clinical Immunology www.elsevier.com/locate/yclim
Meta-analysis of IgE-binding allergen epitopes Virginie Lollier ⁎, Sandra Denery-Papini, Chantal Brossard, Dominique Tessier UR1268 BIA (Biopolymers Interactions Assemblies), INRA, 44300 Nantes, France
Received 10 March 2014; accepted with revision 19 March 2014 KEYWORDS IgE epitopes; Allergens; Bioinformatics; Database
Abstract IgE-binding epitopes are related to allergic symptoms by eliciting degranulation of special cells and release of molecules that trigger the hypersensitivity reaction. Little is known about what characterises allergen IgE-binding epitopes, although advances in analytical methods have led to the identification of a large number of them. To assess if a binary classification of allergen regions into epitopes or non-epitopes may accurately reflect biological reality, we computed the fraction of allergen amino acids that are involved in epitopes. A relationship between this fraction and the increasing number of literature references was modelled. Due to the wide variety of methods that are used in the literature, a peak in the number of matches between an allergen sequence and its epitopes confirms their validity. Accordingly, our graphical representation of positive assays along sequences provides an overview of epitope localisation, which should help to highlight major positions for IgE binding to allergens. © 2014 Elsevier Inc. All rights reserved.
1. Introduction Specific recognition of epitopes on an allergen by IgE antibodies that are bound to effector cells such as basophils is essential for the development of the allergic response. Consequently, B-cell epitope identifications play an important role in vaccine design, disease diagnosis and allergy research. Epidemiological studies at a molecular level that may highlight the parts of the protein most often involved in specific recognition by IgE antibodies are rare. Indeed, an
Abbreviations: FAE, fraction of amino acids in a protein sequence that are involved in epitopes. ⁎ Corresponding author at: UR1268 BIA (Biopolymers Interactions Assemblies), INRA, 44300 Nantes, France. E-mail address: [email protected] (V. Lollier).
http://dx.doi.org/10.1016/j.clim.2014.03.010 1521-6616/© 2014 Elsevier Inc. All rights reserved.
exhaustive study of the epitope repertoire of an allergen in multiple environments and in a large cohort is currently too expensive and may be technically impossible due to low quantities of the available sera. However, many experimental techniques have been widely used worldwide to identify and map epitopes on allergens. The resulting accumulated knowledge has been partially distributed among several databases [1–3]. However, those databases do not provide a condensed view of the epitope distribution on each allergen. In a previous study [4], we have shown that the cumulative epitope identifications mapped nearly everywhere along the protein sequence of the milk and latex major allergens (i.e., Bos d 5 and Hev b 6). By defining the B-cell epitope that is involved in a normal immunological response, Van Regenmortel [5] indicated that the number of identified B-cell epitopes of an antigen
32 corresponds to the size of the analysed immunological repertoire of a host that is immune to that protein. IgE antibodies tend to be more often cross-reactive than IgG antibodies [6]; therefore, they might be more numerous. In addition, Greenbaum et al. [7] suggested that the antigen surface might be considered a continuous landscape of epitopic regions. In the case of allergens, this surface undergoes unknown modifications during the crossing of the epithelial barrier. The aim of this article is to explore whether B-cell epitopes of allergens can be localised everywhere: not only on the overall protein surface but also all along its primary sequence. In this meta-analysis, we further examined whether the frequencies of amino-acid positions in the primary sequence, which are involved in epitopes could be used to overcome the lack of epidemiological data at the molecular level.
2. Methods 2.1. Selection of epitope data The data related to B-cell epitopes were extracted from the Immune Epitope DataBase (IEDB, August 2013). Only epitopes corresponding to positive assays for the IgE isotype were selected. The epitopes were also filtered based on their unique GenBank identifier, and only epitopes that were identified in allergens were maintained. The epitope data and data related to allergen names and clusters were stored in a relational database with the help of the ETL (Extract Transform and Load) software called Talend Open Studio. In the database, epitopes were described by their continuous or discontinuous sequence, their position within the protein, their bibliographic references and their assay types based on the IEDB ontology. One epitope longer than 100 amino acids was evaluated as too imprecise and removed from the dataset. Data stored in the relational database was used in the web application LocAllEpi.
2.2. Protein sequence selection The sequence identifiers of multiple public databases that corresponded to known allergen sources were collected from the IUIS [8] (allergen.org) and Allergome [9] web sites during August 2013. Based on these identifiers, the allergen sequences were extracted from the requests to the National Center for Biotechnology Information (NCBI) using Entrez Programming Utilities [10] and its cross-referencing system. Protein sequences were selected if they were referred to in IEDB, and their clustering was achieved with the help of the Cd-hit software (version 4.5.4) [11]. The clusters were labelled by the allergen name of the longest sequence. The epitope localisation within clusters was achieved by correcting the amino acid numbering from multiple sequence alignments of the variants using the clustalw software (version 2.1) [12]. Only clusters having at least one sequence member related to an identified IgE-binding epitope were used for further analysis.
V. Lollier et al.
2.3. Computation of the fraction of amino-acids involved in epitopes (FAE) For each allergen and cluster sequence, we computed the number of amino acids that were involved in at least one positive epitope assay and divided this number by the sequence length. We called the resulting value FAE (or fraction of amino-acids involved in epitopes). The FAE represents the fraction of overlap between a protein sequence and its set of epitope sequences. A complete overlap between a protein and its epitope repertoire is expressed by an FAE value close or equal to one. This implies that all amino acids of the allergen cluster or sequence are involved in epitopes. This fraction was graphically represented as a function of how many studies had been conducted in the literature or how many different techniques were performed. On the resulting scatter plots, each point represents a sequence or a cluster of sequences. To compare allergens and clusters regardless of their length, the number of references and of different assay types were also divided by the number of amino acids that constituted the sequence. The assay types were named with the Ontology of Immune Epitopes (ONTIE) used by IEDB to represent experiments that identify and characterize immune epitopes. A fitted curve was computed using the nls function within R for non-linear regression of data.
2.4. Display of the frequencies of overlaps between epitopes along sequences With the help of the start and end positions of the collected epitopes in the allergen sequence or in the cluster alignment, we computed the number of times an amino acid position was found in a positive assay among the references. Numbers obtained for each amino acid were graphically mapped as a function of their position within the different protein sequences. These frequencies of overlaps were associated with additional data in 3 graphics: with the number of positive sera in Appendix A, being labelled with their PubMed identifiers in Fig. 1 and with the assay type that was related to the studied epitope in Appendix B Computations and graphs were produced with the help of the R framework.
3. Results and discussion 3.1. Relationship between the abundance of literature and the FAE Our dataset was composed of 2971 IgE B-cell epitopes of allergens. Although the Immune Epitope DataBase [1] is not dedicated to allergy, we chose this database as a central hub for overall epitope information. It includes nearly 40% of the International Union of Immunological Societies (IUIS) allergens [13]. Compared to the existing databases on B-cell epitopes [2,3], the IEDB ontology provides a standardised description of the experimental conditions under which the epitopes were identified. This database is also frequently updated with continuous and discontinuous B-cell epitopes from many antigens. Because the experimental work on
Meta-analysis of IgE-binding allergen epitopes
33
Figure 1 Fraction of amino-acids involved in epitopes according to the increasing number of references. Proteins and clusters are coloured in blue, green, red and orange, respectively, for food, respiratory, skin and sting allergens.
protein sequences often implies the production of isoforms or variants, the collected sequences were clustered according to their similarity to obtain epitope knowledge along a single representative sequence of the allergen. Each cluster was labelled with the allergen name of the longest sequence. The allergens where further represented by 112 protein and cluster sequences. Assuming that the whole protein sequences have been scanned for epitope localisations, we analysed the FAE (see Methods section for the computation of Fraction of Amino-acids involved in Epitopes) in relation to how often the allergen molecules were studied. Fig. 1 shows that all residues of an allergen sequence were progressively involved in epitopes when the number of references increased. This result concerns food as well as respiratory and skin allergens, with the exception of the Pru a 1 cluster. Indeed, this allergen has been mostly studied with Mal d 1 using site-directed mutagenesis at a few limited, shared locations to explore cross-reactivity of Bet v 1 homologous allergens [14,15]. In this case, because a large part of the protein has not been scanned, the Pru a 1 cluster was considered as an outlier and was removed from the fitted curve computation. The resulting curve shows that the data follows an exponential model with significant parameters, where the number of amino acids constituting epitopes is a function of the number of references. In this model, the maximum FAE is close to 90%. Few references were sufficient to identify nearly 50% of the amino acids that took part in epitopes. By extrapolation from the fitted curve, this value could be obtained from two articles for a protein consisting of 225 amino acids, which was the median length of sequences in our dataset. Above this threshold, additional epitopic elements from further studies were less numerous, suggesting an overlapping of identifications between references.
3.2. Implications of allergen clustering We evaluated the impact of clustering on our results. As defined by Chapman et al. [16], variants could display greater than 70% sequence identity and share identical biochemical features. We chose this threshold because this similarity should result in potential cross-reactions in the context of allergy. Fig. 2 illustrates how the epitope data were spread over multiple isoforms of four major allergens (the complete list of allergen clusters and unclustered sequences are in the Appendix C). Most of the clusters contained isoforms of the same protein, although some clusters could be extended to allergens from the same conserved protein family. For example, the Der f 2 cluster also contained the Der p 2 sequences. However, the extension of sequences beyond isoforms of the same allergen had little influence on our measured FAE. In the example of Der f 2, the gathered sequence clustering data was related to an allergen protein from European and American house dust mites. As shown on Fig. 2, Der f 2 and Der p 2 allergens have similar sequences, with the exception of three residues (in bold face). Epitopes identified on the Der p 2 allergen were distributed nearly entirely along the representative sequence of the cluster (mauve, pink and light blue points, Fig. 3). Epitopes from Der f 2 overlapped with some Der p 2 epitopes, even at the less conserved positions 36 and 105. More generally, when we inspected the distribution of epitope references within the extended clusters shown in the table of Appendix C, few references were provided based on the extent of the clusters to additional allergens from the same protein family. Two clusters, named Bet v 2 and Cor a 9, were exceptions. The allergen sources within these last clusters, belonging, respectively to plant profilins and 11S globulins (legumins), are known to cross-react.
34
V. Lollier et al.
Figure 2 Sequence members of clusters representing four major allergens and multiple sequence alignment of Der f 2–Der p 2 proteins (where “*” means identity of sequences, “:” conserved residues and “.” semi-conserved residues). The identity percentage of a sequence is related to the representative sequence, which is indicated by an “x” within each cluster.
3.3. Influence of techniques Because the epitopes from the literature have been rapidly identified along whole protein sequences, we wondered
if we faced a complementarity of identification techniques or/and some imprecision of the most common of them. Most of the continuous and discontinuous epitopes were published after 1995. Since that time, analytical techniques
Figure 3 Epitope identifications for each sequence belonging to the Der f 2 cluster. The curves labelled 416898, 37958157 and 55859468 correspond to the GenBank identifiers of the Der f 2 variants. The labels 256095984, 157829757 and 1352237 are the GenBank identifiers of the Der p 2 variants. Positions where the amino acids differed in the multiple sequence alignment are indicated by black arrows, and letters correspond to amino acids of the 37958157 sequence id. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
Meta-analysis of IgE-binding allergen epitopes have rapidly evolved from hydrolysis fragmentation to X-ray 3D structure-solving, and the epitope sequence size has varied. In the present dataset, the epitopes were composed of peptides between 1 and 95 amino acids, and 70% of them were smaller than 20 amino acids. Epitopes of the continuous form appeared to predominate in datasets because the protein mutation and antibody/antigen structural resolution imply more complex steps than measuring IgE binding to protein-derived peptides. Nevertheless, the classification of epitopes into continuous and discontinuous forms has been recently revised, suggesting no clear-cut difference between these two types of epitopes [5,6]. For example, the weak binding of the smallest linear peptides could lead to the assumption that such peptides may be parts of larger, discontinuous epitopes [17]. Concerning the ideal epitope size, some results suggest that allergen epitopes are short stretches of the protein sequence [18,19], while other studies observed that different sized peptides display different profiles with reactive sera [17,20]. To assess if one particular technique predominates and explains the gradual overlap of protein sequence with epitopes, we examined the potential relationship between the assay types and the median FAE per sequence for each analysis reference (Fig. 4). The names of the assay types correspond to the IEDB ontology, which addresses either the technique that was used to identify B-cell epitopes (e.g., ELISA, X-ray crystallography) or an experimental scope (e.g., “Reduction of disease after treatment” or “Induction of tolerance”). The most frequent identification techniques were “Western/Immunoblot”, “ELISA” and “Antigen Competition of Antibody Binding”, which are similar because they focus on epitope mapping along the entire protein sequence. However, the FAE brought by these techniques did not exceed 30%. No assay type displayed an extremely high FAE, with the exception of “Phage Display/Immunopanning”.
Figure 4
35 However, this technique was related to only three references, which constitutes a weak sampling. In addition, the curve displayed in Fig. 5 further oriented to a complementarity of assay types. As an exception, the Phl p 2 allergen was plotted out of the regressive curve because a single reference was related to this allergen, whereas multiple analyses were conducted on a focused area (PUBMED: 19201867 [21]). In other cases, epitope matching with all amino acids occurred progressively according to the increasing number of different applied techniques. However, each of these techniques should have been applied on different populations, which should contribute to epitope diversity. The comparison of Figs. 1 and 6 gives some indications about this contribution. When points representing allergens move to the left from Figs. 1 to 6, the FAE is more explained by the number of references than the number of techniques. There is, therefore, an influence of the number of tested populations. This is the case, for example, for Bos d 9. Accordingly, when points move to the right, the FAE should then be linked to the number of different techniques that are involved in epitope identifications (e.g., Bet v 4). If one technique introduces imprecision, we should observe a fast and systematic overlap between epitopes and the protein sequence when this technique is used. This is not the case, as shown in Fig. 4. Moreover, Fig. 5 highlights that this overlap is at least partially due to the combination of several techniques. These results suggest more complementarity of the experimental conditions than an imprecision of some techniques.
3.4. Prediction concerns During our study, we noticed that some proteins were not entirely scanned for epitope localisation. However, two observations tend to confirm that B-cell epitopes of allergens
Frequencies of applied techniques and their corresponding median FAE.
36
V. Lollier et al.
Figure 5 FAE against the number of different assay types per allergen sequence. Proteins and clusters are coloured in blue, green, red and orange, respectively, for food, respiratory, skin and sting allergens. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
can be localised everywhere along primary sequences. The first is provided by the overall tendency that is shown in Fig. 1. The second observation relies on the fact that the FAE remains high for these proteins (e.g., more than 60% for Bet v 1). This localisation of epitopes along allergens raises further questions regarding the application of existing epitope prediction tools in the context of allergy. Their presence all along the sequence of allergens may be related to their ability to cross-react with antibodies. Epitopes of lesser affinity may then be involved in the allergic reaction without having induced the production of specific antibodies by B-cells. Considering the present level of information on epitopes, it is difficult to classify them according to their affinity. Moreover, only the BciPep database stores the immunogenicity of collected epitopes. Unfortunately, this database, which is not dedicated to allergy, gathers only 2% of IgE epitopes (December 2013). In addition, building datasets to classify epitopic amino acids against the non-epitopic amino acids of a sequence will be challenged when identifications all along the allergen are lacking. In the allergy context, the available data should be restricted to major allergens, and major-studied regions should be linked to symptom severity and, eventually, sequence conservation. A more extended meta-analysis of IEDB content has suggested a potential bias in dataset composition because a relationship between the number of epitopes and the morbidity of antigens was observed [22]. In the general context of antigenicity, prediction tools often consider epitopes as intrinsic features of the protein because they are based on their structure [7]. This is the case when the relationship between the antigenicity of peptides and their amino-acid sequence [23,24], hydrophilicity propensity [25] or surface-exposure [26] is studied. Actually, if epitopes are homogeneously distributed all along the allergen sequences, which remains to be verified, no specific structural feature or motif should emerge.
3.5. Analysis of epitope distribution in four examples The distribution of positive assays that led to epitope identifications in the literature was examined and reported in detail for four major allergens: Bet v 1, Der f 2, Ara h 2 and Bos d 5. These allergens have been highly studied in at least 10 references, and they displayed high FAE. As shown in Fig. 6, the positions of the positive assays were superimposed in the order of the publication year of their related reference. In the case of the Bos d 5 allergen, from one reference to another, the epitopes were homogeneously found along the entire sequence. These were identified in various cohort sizes (10, 30 or 40 individuals), and the studies were mainly performed in USA, China or Australia. The graph corresponding to Bos d 5 shows that epitopes may be any part of the allergen, but are less frequently found around position 80, although some residues of a discontinuous epitope were found in that region using X-ray crystallography. Thirteen references related to this allergen were found, but at least three placed all amino acids of the sequence in epitopes (PUBMED: 19577281, 11729348 and 22939793). However, Fig. 6 shows a high level of positive assays around positions 140–150, where information provided by multiple references was aggregated. Similarly, Ara h 2 displayed three areas that emerged as major epitopic portions of the allergen, in spite of a less homogeneous distribution of identification frequencies. However, the whole protein sequence was not systematically studied in each publication. For example, the reference identifier 21883278 [27] focused on an area based on its resemblance to the Jug r 2 allergen and its potential cross-reactivity to this allergen. The identifier 22042002 [28] also corresponds to a study of a limited area. These two references were the most recent works that were collected for Ara h 2 and were published in 2011 and 2012, which may suggest a focus of recent studies to restricted areas with
Meta-analysis of IgE-binding allergen epitopes
37
Figure 6 Superimposition of epitope identifications along the sequences of four major allergens, from oldest to the most recent references, labelled by their PUBMED identifier (PMID).
regards to previous results. However, a microarray immunoassay experiment (PUBMED: 18234310 [29]) displayed an FAE of approximately 70% of the whole sequence (excluding the 20 amino-acid signal peptide). Authors of this work have related the clinical sensitivity of the 24 studied patients to the epitope diversity. The identification of epitopes all along the Der f 2 respiratory allergen appeared more progressively with the
growing number of references than for the Bos d 5 food allergen. Like Ara h 2 and Bos d 5, this sequence contains a signal peptide. Signal peptides are generally excluded from the identification of B-cell epitopes, which should explain why the regressive curve of the FAE displayed a limitation of 85-90% of the protein sequence (Ymax = 0.87). Moreover, the number of positive assays that mapped on Bos d 5, Ara h 2, Der f 2 ranged differently, although the number of references, the
38 number of applied techniques and the FAE were similar. When the signal peptide is omitted, the number of positive assays ranged between 5 and 20 positive assays for Bos d 5, and between 1 and 10 for Der f 2. Because we computed the positive assay frequencies at each position, the height of any identification peak was partly linked to the overlap size of the assayed synthetic peptides, but also to the cohort size. The number of positive assays is thus a combination of multiple information, including various assay types and the epidemiology of the epitopes. As a consequence, the highest peaks may be considered key areas for IgE binding. In the case of Bet v 1, the epitope information was less abundant. This allergen has been described as displaying mostly conformational epitopes; however, the mapping of this epitope type remains a difficult task [30,31]. Indeed, this allergen seems to have been studied in a piecemeal fashion. The positive results displayed here were extracted from publications where data were not available along the entire protein sequence. Some negative results from other articles localised to the stretches of sequence at positions 1–22, 60–70 and 106–126. These last works aimed at developing immunotherapies by screening positive results on IgG isotypes and negative results on patient IgEs, while setting aside potential positive IgE reactions on the remaining sequence. The inclusion of negative results in this dataset might have weighted the positive amino-acid fraction that was computed on allergen sequences and would have highlighted potential major areas of focus. Furthermore, this could have also differentiated untested stretches of sequence from negative results in this type of graphical representation. However, negative results are often unpublished. In addition, such weighting required a homogeneous distribution of positive and negative data, and, as a consequence, the protein sequence should be entirely assessed and both results diffused. Unfortunately, some experiments are technically unable to satisfy these criteria. These four examples indicate that epitopes are distributed heterogeneously along a protein sequence, consequently resulting in potential major areas for immunoreactivity. Despite limitations linked to the current data on some allergens (negative results are unknown or analysis is limited to parts of sequences), these examples illustrated how to localise such major areas on a given allergen protein.
3.6. Major epitope localisation In our four detailed examples, a peak of positive assays indicated technical confirmation of an epitopic area, such as a confident result, rather than an immunodominant area. Indeed, the overlapping of identifications corresponds to a convergence of multiple technical approaches, including assays with synthetic peptides in solution vs. adsorbed peptides or epitopes of different lengths. Such peaks therefore locate areas that interact with antibodies in multiple configurations and eventually among different populations of patients. Therefore, these areas should be stable during processing of the allergen by the immune system. To provide a snapshot of the current knowledge of epitope collection for allergens and to locate potential major areas on their sequences, we have developed a web-based tool
V. Lollier et al. that is freely accessible at http://wwwappli.nantes.inra.fr/ LocAllEpi. It also includes computed surface features when the 3D structure is available. For localisation of immunodominant areas, our web tool displays an option for the number of positive sera in addition to binding assays. However, these data are often missing, especially when using pooled sera experiments. When available, this information is displayed as an absolute value and not as a relative value to the cohort size, which varies along the protein sequence according to references. We have noticed that the median size of cohorts was close to 10 patients, which might be used as a reference value. Some differences in identification frequencies were observed between molecules. These differences may be due to the diversity of experimental conditions or to the molecule itself, such as Bet v 1, where classical epitope mapping methods produced few positive results. In this situation, it is difficult to determine whether intrinsic characteristics of the protein intervene or to measure how the allergen is altered or protected by its environment during its processing by the immune system.
4. Conclusions When all epitopic areas are mapped independently of the experimental context, our meta-analysis confirms that all parts of an allergen sequence can be part of an epitope. This progressive sequence overlapping is not the result of a predominant identification technique. However, some parts of the allergen are more often involved than others. For this reason, we suggest that the integration of data from different studies could be of interest. This available data on epitope immunogenicity and epidemiology is highly heterogeneous and makes building learning data sets for prediction difficult. Nevertheless, the mapping of all positive assays along allergens provides a statistic view that would help to delineate major areas of interest for the development of diagnostic tools. Sharing information on a dedicated internet website should enlarge the dataset to more allergens and refine the location of major epitopes. For instance, the IEDB website provides any user the ability to submit data for curation through a Data Submission Tool (https://dst.liai.org). Studies on IgE epitopes are often dedicated to a precise application (vaccination or diagnosis on major allergens), and intermediary data, which could be valued in an extended context, are generally excluded during the publication process. However, these additional data could also lead to interesting analyses for characterising B-cell epitopes beyond the study of one allergen or protein family. Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.clim.2014.03.010.
Conflict of interest statement The authors declare that there are no conflicts of interest.
Acknowledgments This work was supported by the ANR (PREDEXPITOPE, ANR08-ALIA-14).
Meta-analysis of IgE-binding allergen epitopes
39
References [1] R. Vita, L. Zarebski, J.A. Greenbaum, H. Emami, I. Hoof, N. Salimi, R. Damle, A. Sette, B. Peters, The immune epitope database 2.0. Nucleic Acids Res. 38 (2010) D854–D862. [2] S. Saha, M. Bhasin, G.P.S. Raghava, Bcipep: a database of b-cell epitopes, BMC Genomics 6 (2005) 79. [3] O. Ivanciuc, C.H. Schein, W. Braun, SDAP: database and computational tools for allergenic proteins, Nucleic Acids Res. 31 (2003) 359–362. [4] V. Lollier, S. Denery-Papini, C. Larré, D. Tessier, A generic approach to evaluate how B-cell epitopes are surface-exposed on protein structures, Mol. Immunol. 48 (2011) 577–585. [5] M. Van Regenmortel, What is a B-cell epitope? in: Mike Schutkowski, Ulrich Reineke (Eds.), Epitope Mapping Protocols, Humana Press, New York, 2009, pp. 3–20. [6] R.C. Aalberse, R. Crameri, IgE-binding epitopes: a reappraisal, Allergy 66 (2011) 1261–1274. [7] J.A. Greenbaum, P.H. Andersen, M. Blythe, H. Bui, R.E. Cachau, J. Crowe, M. Davies, A.S. Kolaskar, O. Lund, S. Morrison, B. Mumey, Y. Ofran, J. Pellequer, C. Pinilla, J.V. Ponomarenko, G.P.S. Raghava, M.H.V. van Regenmortel, E.L. Roggen, A. Sette, A. Schlessinger, J. Sollner, M. Zand, B. Peters, Towards a consensus on datasets and evaluation metrics for developing B-cell epitope prediction tools, J. Mol. Recognit. 20 (2007) 75–82. [8] T.P. King, D. Hoffman, H. Lowenstein, D.G. Marsh, T.A. PlattsMills, W. Thomas, Allergen nomenclature. WHO/IUIS allergen nomenclature subcommittee, Int. Arch. Allergy Immunol. 105 (1994) 224–233. [9] A. Mari, E. Scala, P. Palazzo, S. Ridolfi, D. Zennaro, G. Carabella, Bioinformatics applied to allergy: allergen databases, from collecting sequence information to data integration. the allergome platform as a model. Cell. Immunol. 244 (2006) 97–100. [10] NCBI Resource Coordinators, Database resources of the national center for biotechnology information, Nucleic Acids Res. 41 (2013) D8–D20. [11] W. Li, A. Godzik, Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences, Bioinformatics 22 (2006) 1658–1659. [12] M.A. Larkin, G. Blackshields, N.P. Brown, R. Chenna, P.A. McGettigan, H. McWilliam, F. Valentin, I.M. Wallace, A. Wilm, R. Lopez, J.D. Thompson, T.J. Gibson, D.G. Higgins, Clustal W and clustal X version 2.0, Bioinformatics 23 (2007) 2947–2948. [13] K. Vaughan, J. Greenbaum, Y. Kim, R. Vita, J. Chung, B. Peters, D. Broide, R. Goodman, H. Grey, A. Sette, Towards defining molecular determinants recognized by adaptive immunity in allergic disease: an inventory of the available data, J. Allergy (Cairo) (2010) 628026. [14] D.Y. Son, S. Scheurer, A. Hoffmann, D. Haustein, S. Vieths, Pollen-related food allergy: cloning and immunological analysis of isoforms and mutants of Mal d 1, the major apple allergen, and Bet v 1, the major birch pollen allergen, Eur. J. Nutr. 38 (1999) 201–215. [15] Y. Ma, G. Gadermaier, B. Bohle, S. Bolhaar, A. Knulst, Z. Markovic-Housley, H. Breiteneder, P. Briza, K. HoffmannSommergruber, F. Ferreira, Mutational analysis of amino acid positions crucial for IgE-binding epitopes of the major apple (Malus domestica) allergen, Mal d 1, Int. Arch. Allergy Immunol. 139 (2006) 53–62. [16] M.D. Chapman, A. Pomés, H. Breiteneder, F. Ferreira, Nomenclature and structural biology of allergens, J. Allergy Clin. Immunol. 119 (2007) 414–420. [17] G. Schramm, A. Bufe, A. Petersen, H. Haas, R. Merget, M. Schlaak, W.M. Becker, Discontinuous IgE-binding epitopes
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
contain multiple continuous epitope regions: results of an epitope mapping on recombinant Hol l 5, a major allergen from velvet grass pollen, Clin. Exp. Allergy 31 (2001) 331–341. G.A. Bannon, T. Ogawa, Evaluation of available IgE-binding epitope data and its utility in bioinformatics, Mol. Nutr. Food Res. 50 (2006) 638–644. K.L. Bøgh, H. Nielsen, C.B. Madsen, E.N.C. Mills, N. Rigby, T. Eiwegger, Z. Szépfalusi, E.L. Roggen, IgE epitopes of intact and digested Ara h 1: a comparative study in humans and rats, Mol. Immunol. 51 (2012) 337–346. V. van Kampen, V. Liebers, I. Sander, Z. Chen, X. Baur, M. Raulf-Heimsoth, F.W. Falkenberg, B-cell epitopes of the allergen Chi t 1.01: peptide mapping of epitopes recognized by rabbit, murine, and human antibodies. Allergy 56 (2001) 118–125. S. Padavattan, S. Flicker, T. Schirmer, C. Madritsch, S. Randow, G. Reese, S. Vieths, C. Lupinek, C. Ebner, R. Valenta, Z. Markovic-Housley, High-affinity IgE recognition of a conformational epitope of the major respiratory allergen Phl p 2 as revealed by X-ray crystallography, J. Immunol. 182 (2009) 2141–2151. V. Davies, K. Vaughan, R. Damle, B. Peters, A. Sette, Classification of the universe of immune epitope literature: representation and knowledge gaps, PLoS One 4 (2009) e6948. B. Yao, L. Zhang, S. Liang, C. Zhang, Svmtrip: a method to predict antigenic epitopes using support vector machine to integrate tri-peptide similarity and propensity, PLoS ONE 7 (2012) e45152. J. Gao, E. Faraggi, Y. Zhou, J. Ruan, L. Kurgan, BEST: improved prediction of B-cell epitopes from antigen sequences, PLoS ONE 7 (2012) e40104. M. Odorico, J. Pellequer, BEPITOPE: predicting the location of continuous epitopes and patterns in proteins, J. Mol. Recognit. 16 (2003) 20–22. P. Haste Andersen, M. Nielsen, O. Lund, Prediction of residues in discontinuous B-cell epitopes using protein 3D structures, Protein Sci. 15 (2006) 2558–2567. S.J. Maleki, S.S. Teuber, H. Cheng, D. Chen, S.S. Comstock, S. Ruan, C.H. Schein, Computationally predicted ige epitopes of walnut allergens contribute to cross-reactivity with peanuts, Allergy 66 (2011) 1522–1529. L. Rosenfeld, W. Shreffler, L. Bardina, B. Niggemann, U. Wahn, H.A. Sampson, K. Beyer, Walnut allergy in peanut-allergic patients: significance of sequential epitopes of walnut homologous to linear epitopes of Ara h 1, 2 and 3 in relation to clinical reactivity, Int. Arch. Allergy Immunol. 157 (2012) 238–245. A.E. Flinterman, E.F. Knol, D.A. Lencer, L. Bardina, C.F. den Hartog Jager, J. Lin, S.G.M.A. Pasmans, C.A.F.M. BruijnzeelKoomen, H.A. Sampson, E. van Hoffen, Shreffler W.G., Peanut epitopes for IgE and IgG4 in peanut-sensitized children in relation to severity of peanut allergy, J. Allergy Clin. Immunol. 121 (2008) 737–743 (e10). M.D. Spangfort, O. Mirza, H. Ipsen, R.J.J. Van Neerven, M. Gajhede, J.N. Larsen, Dominating IgE-binding epitope of Bet v 1, the major allergen of birch pollen, characterized by X-ray crystallography and site-directed mutagenesis, J. Immunol. 171 (2003) 3084–3090. A. Gieras, P. Cejka, K. Blatt, M. Focke-Tejkl, B. Linhart, S. Flicker, A. Stoecklinger, K. Marth, A. Drescher, J. Thalhamer, P. Valent, O. Majdic, R. Valenta, Mapping of conformational IgE epitopes with peptide-specific monoclonal antibodies reveals simultaneous binding of different IgE antibodies to a surface patch on the major birch pollen allergen, Bet v 1, J. Immunol. 186 (2011) 5333–5344.
