The Use of Amino Acid Sequence Alignments to Assess Potential Allergenicity of Proteins Used in Genetically Modified Foods

Steven M. Gendel    Biotechnology Studies Branch Food and Drug Administration National Center for Food Safety and Technology Summit-Argo, Illinois 60501

I INTRODUCTION

Food allergies occur in approximately 5% of children and 2% of adults (Sampson, 1992; Hefle 1996). Although allergic reactions to foods can range from mild to life threatening, sensitive individuals experience extreme reactions when exposed to small amounts of allergen (Sampson, 1992). The only reliable way to deal with food allergy is to avoid the offending food; therefore, it is important that allergic individuals be aware of the content of all foods consumed. An allergic individual must avoid both whole foods that cause reactions (for example milk) and mixtures that contain components of the allergenic food (such as casein).

The production of transgenic foods raises two major concerns regarding allergenicity: the transfer of allergenic proteins to new hosts, and the potential for proteins from organisms that have not previously been part of the food supply to become allergens (FDA, 1992; Fuchs and Astwood, 1996). If a transferred protein is from a “commonly allergenic” donor, it may be possible to obtain some measure of the allergenicity of the protein in a new host by testing with sera from allergic individuals. This has been done in one case in which a Brazil nut protein was transferred to soybean and retained allergenicity (Nordlee et al., 1996). However, if a transferred protein is derived from a “less commonly allergenic” source, this approach to safety assessment is not practical because of the difficulty in obtaining a sufficient number of allergic sera. Similarly, there are no direct biochemical tests for potential allergenicity that can be used to assess new proteins in the food supply.

Food allergens, and allergens in general, are a diverse group of proteins. Food allergen proteins have been described as generally being between 10 and 70 kDa, highly expressed, possibly glycosolated, and resistant to degradation (Hefle, 1996). However, there are no data to show that any of these properties are necessary or sufficient to cause sensitization or an allergic reaction in a previously sensitized individual.

Several publications, including a recent report from the International Food Biotechnology Council, have suggested that the potential allergenicity of a transferred protein can be assessed by examining a set of physiochemical properties (including stability to digestion, prevalence, and stability to processing) and by comparing the sequence of the protein to those of known allergens. (Astwood and Fuchs, 1996; Fuchs and Astwood, 1996; Metcalfe et al., 1996). The sequence-based component of an allergenicity assessment can be carried out by aligning a query sequence with each member of a database of allergen sequences. A negative result, the failure to find significant sequence similarity between the query sequence and any known allergen, can be considered an indication of low probability of potential allergenicity. Such comparisons have, in fact, been used in the safety assessment process for several transgenic foods, although little specific information has been published on how these comparisons were performed (Astwood and Fuchs, 1996; Fuchs and Astwood, 1996).

Sequence alignments of this type can be carried out using programs that implement a number of different algorithms (Gribskov and Devereux, 1991). Most of these algorithms were developed primarily to detect evolutionary or functional relationships. However, allergenicity assessment involves the detection of short regions of structural similarity that are not evolutionarily or functionally related. Therefore, some assumptions that are built into the alignment programs may not be relevant in this context. In addition, these programs often have multiple user-definable input parameters that affect their functioning. The data sets used for alignment also affect the results obtained and the ease with which significant results can be recognized.

To determine the best method for utilizing sequence information in assessing the potential allergenicity of proteins used in new food varieties, I compared the results obtained by using different sequence alignment strategies with several test sequences and two allergen sequence databases. The test sequences included both synthetic control sequences and sequences for proteins that are currently being used in transgenic foods. The results of these tests showed that local alignment algorithms are more appropriate for use in this context than global allignment algorithms, use of the proper scoring matrix is necessary to reliably locate significant matches, and the lack of reliable criteria for defining an allergenic epitope makes it difficult to assess the biological significance of the matches that are identified.

II METHODS

All of the sequence analysis programs used were part of Version 8 of the GCG package (Genetics Computer Group, Inc., Madison, WI) running on an AXP 2100 computer (Digital Equipment Corp., Maynard, MA) under a VMS 6.1 operating system. The individual programs and parameters used are described in detail under Results.

Construction of the two allergen sequence databases has been described (Gendel, chapter 3 of this volume). Briefly, accessions for food and nonfood allergen sequences were identified in three large reference databases. These sequences were compared both within and between the reference databases to identify a complete set of accessions that includes all available allergen sequence variants. Because it is not known whether common sequence properties are involved in the allergenicity of food and nonfood allergens, each group of sequences was treated as a separate database. The overall composition of the allergen databases is described in Table I. The food allergen sequence database does not include wheat gluten proteins because it is not clear whether food allergies and gluten-associated enteropathies share a common etiology (O’Mahony and Ferguson, 1991; Metcalfe, 1992).

All known food allergens are proteins (Taylor, 1992; Hefle, 1996). Therefore, amino acid sequence comparisons should be used for assessing potential allergenicity. Direct comparison of amino acid sequences avoids three problems with nucleic sequence comparison that could obscure significant matches. First, because the genetic code is degenerate, proteins with identical amino acid sequences can have significantly different coding sequences. Second, because all known food allergen sequences originate from eukaryotes, the genomic sequences may contain introns. Although it may be possible in many cases to identify and use only the coding regions of the nucleic acid sequences, this can be much more complex than simply using the translated amino acid sequence. Third, some allergen sequences have been obtained from cDNA while others represent genomic clones. Again, this means that the possible presence of introns needs to be considered when making comparisons at the nucleic acid level.

Construction of three positive control sequences was carried out as follows. The sequence of a known food allergen, the 113-amino-acid cod parvalbumin protein known as allergen M or Gad cl (SwissProt accession A94236), was randomized by using the program SHUFFLE. This produces a random sequence with the same amino acid composition as the original sequence. The 10 amino acids numbered 51-60 in the original sequence were used to replace amino acids 11-20, 51-60, or 101-110 in the shuffled sequence. This produced sequences with regions located near the N-terminus (control sequence C1), the middle (control sequence C2), or the C-terminus (control sequence C3) that are identical to part of a known food allergen.

A set of transgene test sequences, proteins currently being used in transgenic plants, were identified from a variety of sources, including direct searching of database annotation, regulatory documents from both the Food and Drug Administration and the U.S. Department of Agriculture, and literature sources (Table II). Sequence testing was carried out using the accessions listed in Table II; the actual transgenic plants may express proteins with slightly different amino acid sequences if the gene used originated from a different strain or if it was modified during construction of the transgenic plant.