Allergenicity Assessment of Novel Food Proteins: What Should Be Improved?

Abstract

Allergenicity prediction is one of the most challenging aspects in the safety assessment of foods derived from either biotechnology or novel food proteins. Here we present a bottom-up strategy that defines a priori the specific risk assessment (RA) needs based on a database appropriately built for such purposes. Allergenicity prediction is one of the most challenging aspects in the safety assessment of foods derived from either biotechnology or novel food proteins. Here we present a bottom-up strategy that defines a priori the specific risk assessment (RA) needs based on a database appropriately built for such purposes. Obtaining alternative proteins from genetic engineering approaches and novel food sources is a priority for Food 2030i, the research and innovation policy framework from the European Commission to future-proof our food system. This is because there is an urgent need for plants with improved resistance to abiotic stress and pathogens and for dietary proteins from novel sources that will support reductions in greenhouse gas emissions from a more sustainable food system. One of the most difficult aspects of assessing the safety of foods derived from biotechnology or novel food proteins is allergenicity RA. Central to the current weight-of-evidence approach applied to assess potential allergenicity is the use of bioinformatics to compare the sequence of a novel protein with those of allergen proteins, which cause allergic reactions [1.EFSA Panel on Genetically Modified Organisms Scientific opinion on guidance for risk assessment of food and feed from genetically modified plants.EFSA J. 2011; 9: 2150Crossref Google Scholar, 2.EFSA Panel on Dietetic Products, Nutrition, and Allergies Guidance on the preparation and presentation of an application for authorisation of a novel food in the context of Regulation (EU) 2015/2283.EFSA J. 2016; 14: 4594Google Scholar, 3.EFSA Panel on Genetically Modified Organisms Guidance on allergenicity assessment of genetically modified plants.EFSA J. 2017; 15: 4862Google Scholar]ii. Traditionally, the sequence of a novel protein is compared with those of known allergens using a local alignment algorithm such as FASTA and a threshold value against amino acid sequence alignments (i.e., 35% sequence similarity over a sliding window of 80 amino acids) is used to indicate potential allergenic risks requiring further assessmentii. The scientific community has also used this approach to compare allergens and non-allergens and to develop more advanced in silico tools (e.g., using machine learning) for improved allergenicity prediction for novel food proteins [4.Dimitrov I. et al.AllergenFP: allergenicity prediction by descriptor fingerprints.Bioinformatics. 2014; 30: 846-851Crossref PubMed Scopus (245) Google Scholar, 5.Maurer-Stroh S. et al.AllerCatPro – prediction of protein allergenicity potential from the protein sequence.Bioinformatics. 2019; 35: 3020-3027Crossref PubMed Scopus (53) Google Scholar, 6.Westerhout J. et al.Allergenicity prediction of novel and modified proteins: not a mission impossible! Development of a random forest allergenicity prediction model.Regul. Toxicol. Pharmacol. 2019; 107104422Crossref PubMed Scopus (10) Google Scholar]. However, poor understanding of the specific characteristics of a protein that confer potential allergenicity limits the usefulness of bioinformatics for RA. Here we present a bottom-up strategy, which, contrary to the current paradigm, defines a priori the specific RA needs for the investigation of any given novel protein’s cross-reactive allergenic potential. A bottom-up strategy would place greater emphasis on the development of allergen sequence databases, where curation allows additional criteria to be applied to rank the clinical relevance of the allergens including, for example, their proven ability to act as triggers of allergic disease. Such a robust, reliable, and verifiable database will allow risk assessors to calibrate and frame the RA around the defined public health objectives. For example, similarity of a novel protein to a potent allergen affecting many individuals will be of greater concern and of higher regulatory burden than if similarity is shown to an allergen that affects only a few allergic individuals and has low potency to elicit a reaction. Currently, the allergen sequence databases used for allergenicity RA do not provide systematic data about their allergenic potential and often employ different inclusion criteria. For example, an allergen from peanuts called Ara h 5 was identified using human sera by screening an expression library [7.Kleber-Janke T. et al.Selective cloning of peanut allergens, including profilin and 2S albumins, by phage display technology.Int. Arch. Allergy Immunol. 1999; 119: 265-274Crossref PubMed Scopus (268) Google Scholar]. However, this protein is of very low abundance in peanut seed but is more likely to be expressed in pollen and is thought to be a problem for inhalant allergen cross-reactivity [8.Becker W.M. et al.Peanut allergens: new consolidated findings on structure, characteristics, and allergome.Allergol. Select. 2018; 2: 67-79Crossref PubMed Google Scholar]. This allergen therefore poses a low risk to individuals with allergies to peanut seed but might be important in considering pollen allergy. Currently, such metadata are missing from allergen sequence databases, reducing the usefulness of bioinformatic analysis and making interpretation of the sequence similarities identified more difficult. Their lack also undermines efforts to develop advanced bioinformatics tools that might better predict the risk of a novel protein triggering allergic reactions by providing higher sensitivity, specificity, and accuracy than the classical FASTA algorithm. Such a bottom-up approach could benefit RA strategies for novel proteins and their potential capacity to trigger adverse immune reactions related to both celiac disease (CD) and IgE cross-reactivity. In the case of CD, there have been several attempts to build a database containing many celiac epitopesiii,iv. However, no agreement has been reached on the inclusion criteria – based on precise evidence requirements – necessary to identify CD-relevant gluten epitopes [9.Sollid L.M. et al.Update 2020: nomenclature and listing of celiac disease-relevant gluten epitopes recognized by CD4+ T cells.Immunogenetics. 2020; 72: 85-88Crossref PubMed Scopus (68) Google Scholar]. While Propepperiii has more than 400 celiac epitopes, AllergenOnlineiv comprises more than 1000 celiac epitopes, and Sollid and coauthors [9.Sollid L.M. et al.Update 2020: nomenclature and listing of celiac disease-relevant gluten epitopes recognized by CD4+ T cells.Immunogenetics. 2020; 72: 85-88Crossref PubMed Scopus (68) Google Scholar] identified only around 40 epitopes, showing large inconsistencies between databases. These discrepancies are the documented evidence of the lack of consensus on the inclusion criteria for building an appropriate/reliable database. Thus, while Sollid and coauthors [9.Sollid L.M. et al.Update 2020: nomenclature and listing of celiac disease-relevant gluten epitopes recognized by CD4+ T cells.Immunogenetics. 2020; 72: 85-88Crossref PubMed Scopus (68) Google Scholar] have compiled a reviewed list of the most important and immunodominant epitopes mainly based on the recognition of gluten peptides by CD4+ T cells from one or more CD patients, the Propepperiii and AllergenOnlineiv databases comprehensively cover the full scale of currently available peptides that have been tested for T cell activation potential, even if the induction of celiac enteropathy has not been clinically demonstrated for the vast majority of these peptides. Likewise, it is important to stress that several additional relevant epitopes are as yet likely to be identified [9.Sollid L.M. et al.Update 2020: nomenclature and listing of celiac disease-relevant gluten epitopes recognized by CD4+ T cells.Immunogenetics. 2020; 72: 85-88Crossref PubMed Scopus (68) Google Scholar]. Our proposal can resolve the issue. Practically, we propose to define clear inclusion criteria that can serve to launch an initial screening phase designed to identify any potential protein fulfilling such a priori established criteria. The criteria defined by Sollid and coauthors [9.Sollid L.M. et al.Update 2020: nomenclature and listing of celiac disease-relevant gluten epitopes recognized by CD4+ T cells.Immunogenetics. 2020; 72: 85-88Crossref PubMed Scopus (68) Google Scholar] could serve such purpose, although other criteria may also be fit for purpose. Since no univocal list of criteria exists, an in-depth discussion and possibly consensus within the scientific community is indispensable. In a following phase, we propose to rank T cell epitopes according to their clinical relevance and related features to further boost the delivery of more transparent and robust RA to the public (Figure 1). Attempts in this direction have already been initiated [3.EFSA Panel on Genetically Modified Organisms Guidance on allergenicity assessment of genetically modified plants.EFSA J. 2017; 15: 4862Google Scholar,10.Fernandez A. et al.Safety assessment of immune-mediated adverse reactions to novel food proteins.Trends Biotechnol. 2019; 37: 796-800Abstract Full Text Full Text PDF PubMed Scopus (11) Google Scholar], and additional efforts are imperative since new findings show that proteins from origins other than cereals might inherit hazardous potential for individuals with CD [11.Petersen J. et al.T cell receptor cross-reactivity between gliadin and bacterial peptides in celiac disease.Nat. Struct. Mol. Biol. 2020; 27: 49-61Crossref PubMed Scopus (45) Google Scholar]. In the case of IgE food allergy, several databases have been developed and are currently in use for RA purposesiv–vii, offering diverse possibilities. The criteria for inclusion of allergens and the number of entries varies between databases [12.Radauer C. Breiteneder H. Allergen databases – a critical evaluation.Allergy. 2019; 74: 2057-2060Crossref PubMed Scopus (11) Google Scholar], while those relying on the reviewed UniProtKBviii may be affected by the curation strategies used in that database, which relies on the identification of canonical sequences to reduce redundancy [13.The UniProt Consortium UniProt: a worldwide hub of protein knowledge.Nucleic Acids Res. 2019; 47: D506-D515Crossref PubMed Scopus (3489) Google Scholar]. These differences can lead to misunderstandings and different RA outcomes; for example, bioinformatic analysis results may vary depending on the database used to search for relevant hits. Under the current paradigm, whenever a relevant hit is identified, the follow-up RA strategy analyses the quality of the pairwise sequence alignment and the specific similarity regions between the novel protein and the allergen. Clinical relevance is usually considered only as an additional element in the overall picture. Preliminary attempts to differentiate allergens (allergens and putative allergens versus proteins with insufficient evidence of allergenicity) within a database were initiated by AllergenOnline. However, there are no clear common views on how such a grading of allergenic potency should be assigned/interpreted and the weight-of-evidence that a risk assessor should attribute to such a classification. Attempts in such directions have been proposed for allergenic tree nuts [14.Javed B. et al.A protocol for a systematic review to identify allergenic tree nuts and the molecules responsible for their allergenic properties.Food Chem. Toxicol. 2017; 106: 411-416Crossref PubMed Scopus (4) Google Scholar]. Nevertheless, the current approach heavily relies on expert judgement to interpret a posteriori the outcome of the bioinformatic analysis of the assessment, which can lead to a lack of harmonization, reproducibility, and transparency of the RAs. An alternative approach would be to define a priori the characteristic and clinical relevance that any potential allergen has and the specific follow-up actions to be undertaken if ‘hits’ are identified (Figure 2). Building fit-for-purpose RA databases for food allergy is an urgent priority. To this end, the ranking of CD-relevant gluten T cell epitopes and IgE allergens within any given database is of major importance as it provides more precise information on the clinical relevance of any given allergen to the RA process. This will translate into a sounder RA capturing specific needs considering the input from the scientific community and stakeholders. These two relevant actors should interact and eventually collaborate following principles previously described [15.Fernandez A. Paoletti C. Unintended effects in genetically modified food/feed safety: a way forward.Trends Biotechnol. 2018; 36: 872-875Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar]. This approach will also allow refinement of the current, oversimplistic, RA view where proteins are categorized as allergens or non-allergens according to their inclusion or exclusion in a specific allergen database. Resources devoted to the development of sophisticated and refined bioinformatics tools will be better used once additional relevant protein features are defined and considered in the assessment. Ranking allergens according to allergenic potential and the subsequent development of targeted bioinformatics tools founded on enhanced algorithms will streamline RA approaches, fostering more transparent and credible information for the public. Improving the quality of the bioinformatic assessment will also make it more straightforward to identify when further testing is warranted using more expensive in vitro and in vivo tests, which also have their shortcomings and are often not validated. They can also be difficult to execute, especially when relying on the availability of serum panels from food-allergic subjects. Developing an international consensus on a more robust approach to allergen-sequence database curation will be essential to improve the quality of allergenicity RA of foods produced by biotechnology and novel foods, which will be urgently needed in an era of climate change and transition towards more sustainable food systems. The authors thank Claudia Paoletti and Elisabeth Waigmann for the inspiring comments. A.F. is employed by the European Food Safety Authority (EFSA). The positions and opinions presented in this article are those of the authors alone and do not necessarily represent the views or scientific works of the EFSA. ihttps://ec.europa.eu/knowledge4policy/publication/food-2030-innovative-eu-research-ensures-food-system-future-ready_en iiwww.fao.org/3/a-a1554e.pdf iiihttps://propepper.net ivwww.allergenonline.org/ vhttps://comparedatabase.org/ viwww.allergome.org/ viiwww.allergen.org/ viiiwww.uniprot.org/

Topics & Concepts