Table of Contents

1. Introduction
2. Background - Food Allergens
3. Background - Assessing Food Allergenicity - Known and Potential
4. Transgenic Crops and Known Allergens - Left Side of the Decision Tree in Figure 1
5. Transgenic Crops and Unknown Allergens - Right Side of the Decision Tree in Figure 1
6. Protein Expression and Unknown Allergens
7. Screening Methods to Identify New Allergens in the Food Supply
8. Summary

Introduction and Background Information

Crops developed through biotechnology have been marketed and used by farmers since the mid-1990s. Farmers in the United States have adopted these improved crops rapidly because they increase yields, reduce losses from damaging crop pests, increase income and improve the quality of the crops and food produced from them. Some examples of these crops are; insect resistant Bt corn, Bt cotton and herbicide tolerant canola, corn and soybeans. There are also numerous transgenic crops, under development by plant biotechnology companies and research institutions, that will enter the market during the next several years. Some of these will have direct consumer health benefits. At the same time farmers are increasing their usage of transgenic crops, questions have been raised about the food and environmental safety of these crops.

This is an overview of the current scientific knowledge about food allergies related to biotechnology food safety questions. Specific topics related to allergenicity receiving attention from industry, government regulators, consumers, activists, and scientists include:

• The possibility that genes from known allergens may be inserted into crops not typically associated with allergenicity;
• The possibility of creating new, unknown allergens by either inserting novel genes into crops or changing the expression level of endogenous proteins;
• The adequacy of screening methods to detect the creation of new allergens in transgenic crops.

All parties agree that even though the incidence of food allergies is very low, the potential consequences for some people can be serious. Therefore, introduction of possible allergens into transgenic crops is an issue that is addressed before the crop enters the food supply.

Food allergies are relevant to products developed through biotechnology because the improvements often involve adding or changing genes. Genes make proteins, and though very few proteins are allergens, biotechnology products are tested for potential allergenicity before they are marketed as a key component of the regulatory consultation and review process.

Return to Table of Contents

The medical community distinguishes food allergies from other adverse reactions to food, such as lactose intolerance. A food allergy involves an idiosyncratic reaction of the immune system to a normally harmless food or food component. Most food allergies are mediated by type-I IgE (immunoglobulin E) reactions, and are the subject of this overview. Food allergens are molecules that induce the production of IgE that bind to the surface of mast cells and basophils distributed throughout the body. Subsequent exposure causes an allergic response (histamine release from basophils and mast cells) in a sensitized individual within minutes or, at most, a few hours. The molecules that typically cause an allergic response are proteins or glycoproteins that range in molecular weight from ~10,000 to 70,000 daltons and are stable to proteases and heat (Taylor & Lehrer, 1996).

The percentage of proteins that are allergenic is very small. Only approximately 200 of the hundreds of thousands of proteins humans consume in food are food allergens (Day, 1992). Hefle et al. (1996) surveyed the literature and identified approximately 180 foods eported to be allergenic. Of those, eight food groups (peanuts, soybeans, crustacea, ish, cow's milk, eggs, tree nuts and wheat) account for 90% of all reported food llergies worldwide (Metcalfe, et al., 1996).

In addition, relatively few people have food allergies. In the U.S., approximately 1-2.5% of the adults and 6-8% of the children are affected by food allergies. Many children outgrow these allergies by three years of age (Sampson, 1992, 1995, 2000). People with food allergies are usually allergic to a few specific proteins within one or two foods (Metcalfe et al., 1996). Within a group that is allergic to a specific food, there is variation in the amount of an allergen required to trigger the reaction as well as the severity of the response in each individual. It often is not possible to separate the protein(s) causing the allergic reaction from the thousands of proteins present in a given food that is identified as causing allergies. In addition, the exact protein responsible can vary among different people. Finally, the allergenicity of a food or protein may vary across different societies and cultures due to differences in exposure. For example, peanut allergies are more common in North America, and rice allergies are more common in Japan.

Return to Table of Contents

Recognizing that traditional food safety assessments based on toxicological testing of small amounts of food components or food additives would rarely, if ever, apply to whole foods produced through biotechnology, the International Food Biotechnology Council (IFBC) was founded in 1988 to develop criteria and procedures to evaluate the safety of genetically modified foods. In its 1990 report Biotechnologies and Food: Assuring the Safety of Foods Produced by Genetic Modification, the IFBC adopted a decision tree approach to food safety assessment that has been widely recommended and adopted by numerous regulatory agencies (WHO/FAO, 1991; FDA, 1992; OECD, 1993).

Potential allergenicity of the inserted protein is one of the food safety assessments included in the 1990 IFBC document. For the food allergy assessment, further consideration esulted in another decision-tree approach that was developed by a panel of food allergy xperts in the Allergy and Immunology Institute (AII) of the International Life Sciences nstitute (ILSI) in collaboration with the IFBC. This decision tree process (Figure 1) was ublished (Metcalfe et al., 1996) and has been accepted and used by industry and egulators. This same decision tree was adopted by the FAO/WHO in June 2000, but in March 2001, a new consultation recommended significant additions.

This narrative focuses on the original 1996 IFBC/ILSI decision tree process in Figure 1. The updated FAO/WHO 2001 decision tree is included below (Figure 2) and lists the major differences (A-D) from the 1996 decision tree. While the differences at first appear additive, none of them have been completely evaluated and validated to determine their predictive value against known allergens and non-allergens.

Figure 1

Figure 1 illustrates the decision tree approach developed by IFBC/ILSI (1996) and has been used to assess the potential allergenicity of proteins encoded by introduced genes. "Commonly Allergenic" refers to the group of eight major allergens responsible for 90% of food allergies. "Less Commonly Allergenic" refers to the minor allergens responsible for the remaining 10% of food allergies. DBPCFC= Double blind placebo-controlled food challenge; IRB= Institutional Review Board.

In Figure 1 the first decision point depends on whether the gene being introduced to the crop plant is derived from a known allergen or from a source without a history of allergenicity. The split in the decision tree reflects this difference and results in testing methodologies that would be applied in each case to assess allergenicity and is not a presumption of probable allergenicity or lack of allergenicity. In regard to allergic potential for a gene under consideration, three possible sources exist: from commonly allergenic food, less commonly allergenic and unknown allergenic potential.

• If the gene is derived from a commonly allergenic food, by definition, a population of allergic individuals exists and can be used as a source of serum for immunoassays and clinical tests [skin prick tests; double-blind, placebo-controlled, food challenge (DBPCFC)].
  
  Such definitive tests for allergenicity can be used to test for the specific protein under investigation to determine if it is the cause of the known allergic response. This is possible since it is expected that a sizeable group of individuals has an allergy to the food that is the source of the gene. Therefore, these standardized tests can be carried out if a commonly allergenic food (peanuts, soybeans, tree nuts, milk, eggs, fish, crustacea, and wheat) is the source of the introduced gene.
  
  
• If the protein is from a food source that is less commonly allergenic and less than five individuals can be identified for the immunoassays, then the testing procedure shifts to the right side of Figure 1, and the protein is treated as if the source was not known to be allergenic (see below). These less commonly allergenic foods comprise 10% of the documented food allergies. Many food groups have been involved, including, for example; apples, celery, coffee, chocolate, maize, lemon, pork, beef, rice and turkey. Allergic responses may have occurred in only one patient. The severity of the symptoms has varied and, in some cases, has been difficult to evaluate (Bush et al., 1996). Because allergic reactions to these foods are sporadic, uncommon and, in some cases, unconfirmed, it is difficult to find a sizable population of allergic individuals for developing standardized allergenicity tests or identifying the allergenic protein. Thus in these cases, the IFBC-ILSI decision tree provides a stepwise testing procedure based on the known physiochemical characteristics of typical food allergens and is identical to that used for an unknown.
  
  
• Those genes that are from sources not known to be associated with allergic reactions do not have a population of individuals identified who can serve as a resource for positive sera for immunoassays or as test subjects for clinical tests. Instead, tests based on the physiochemical characteristics of a typical food allergen are conducted. These tests, outlined in the decision tree in Figure 1 on the right side, include a comparison of amino acid sequences to known allergens, stability to gastric fluids (pepsin resistance) and stability to heat. Proteins readily degraded by gastric fluids are not likely to trigger an allergic reaction because they are not available long enough to be recognized as antigens by the immune system. On the other hand, stability to gastric juices and heat are not absolute predictors of allergenicity. Many indigestible proteins in food have no history of allergenicity, and a few rapidly digestible proteins, such as patatin from potatoes, are allergens (Astwood et al., 2001). Tests of stability to gastric fluids and heat indicate only whether the intact protein will be available to interact with the immune system. Details on the testing procedures for unknown allergens follow.
  

Figure 2

Figure 2 differences between 2001FAO/WHO Decision Tree and 1996 ILSI/IFBC Decision Tree.

• Sequence Homology
  Conduct sequence homology tests on all genes and use six contiguous amino acids instead of eight amino acids as used in the 1996 decision tree.
• Increased Serum Sample Size
  Use 25 (2001 tree) rather than the 14 for commonly allergenic foods (1996 tree) or 5 for less commonly allergenic foods.
• Targeted Serum Concept
  Screen sera from individuals positive for "targeted" environmental allergens such as food allergens, grass, trees, weed pollens, molds, animals, insects, etc. (2001 tree) and not just specific sera food allergens (1996 tree).
• Pepsin Resistance and Animal Models
  Standardize the protocol for pepsin resistance and interpret the results in context with the results from an animal model, neither of which has been validated with known allergens and non-allergens.

Return to Table of Contents

Discussion

The ability to move genes from crops known to be allergens in certain individuals to "non-allergenic" crops raises concerns for people affected by food allergies. People with food allergies learn to avoid the allergenic food and its products. Therefore, if the allergenic protein occurs outside of its "native" crop, people allergic to that protein could inadvertently consume it and presumably could have an allergic response.

The tests outlined on the left side of the IFBC-ILSI decision tree (Figure 1 ) reliably prevent known food allergens, such as peanuts, from being inadvertently transferred to a new food source. This was demonstrated by Pioneer and reported in the scientific literature (Nordlee et al., 1996). In this example, the gene for a protein high in methionine, the 2S albumin gene, was being considered as a means to increase the nutritional quality of soybeans for animal feed. At that time, it was known that Brazil nuts caused an allergenic response in certain individuals when consumed, but the specific allergenic proteins had not been identified. Nordlee's work demonstrated for the first time that the 2S albumin protein from the Brazil nut was an allergen. Based on this work, the Brazil nut 2S albumin gene was abandoned by Pioneer as a possible gene candidate. Additional testing of 2S albumin proteins from sunflower with sera from people allergic to foods containing sunflower products demonstrated that the 2S albumin was reactive, and therefore may be an allergen in that food as well. Based on this information, all commercialization work with 2Salbumin genes was abandoned.

The 2001 FAO/WHO decision tree process maintains the same testing scheme when the source of the transferred gene is from a source known to be allergenic; that is, screening the protein of the gene in immunoassays using sera from individuals known to be allergic to the source. The difference is that the 2001 FAO/WHO decision tree recommends the use of 24 sera for testing compared to 14 sera outlined in the 1996 decision tree. However, the FAO/WHO document also recognizes that the use of a smaller number of very well documented, high quality sera may be preferable to the use of larger numbers of lesser-quality sera. In addition, it adds the concept of "Targeted Serum Screen" when a protein is not positive in the Specific Serum tests (moves to right side of decision tree).

The concept of "Targeted Serum Screen" is not a new idea and is based on some observed cross-reactions between allergens and IgE antibodies from individuals with primarily airborne allergies. The groupings outlined in the 2001 FAO/WHO document are as follows:

While this inclusion may at first appear additive, it is important to note that in a report from a panel of food allergy experts convened by the USEPA (SAP, 2000), these experts stressed that "IgE binding does not necessarily correlate with a clinical response and that all food allergens have to be adequately characterized to make definitive statements on cross-reactivity." It is also worth noting that the Bt proteins (from a bacteria source) would not have any targeted screening (see table above).

Return to Table of Contents

Most genes that have been introduced into biotechnology crops are derived from plant and microbial sources with no known history of allergenicity. This issue of introducing new proteins into the food supply is not unique to transgenic crops. In traditional crop breeding techniques, hundreds of proteins are transferred to new crops. Globalization has resulted in the introduction of entirely new foods to different cultures. For example, the Kiwi fruit that was marketed in the early 1980's in North America and Europe triggered allergic reactions in some individuals. Unlike the Kiwi fruit or traditionally bred crops, transgenic crops are tested for allergenicity potential before reaching the market.

The right side of the 1996 decision tree (Figure 1) contains two main evaluation procedures to help assess the potential allergenicity of a protein from a non-allergenic source. These steps evaluate sequence homology to known allergens and physiochemical characteristics such as stability to digestion and processing. These are described below.

5.1   Sequence Similarity to Known Allergenic Proteins (Sequence Homology)

Computer-aided sequence analysis has become an essential molecular biological tool for the estimation of the structures, functions, and evolutionary relationships of proteins (Saier, 1994). Sequence comparisons have been useful in the identification of common motifs that can represent a family of proteins (i.e. DNA binding motif). Application of this technology is useful for assessing whether a protein contains a short amino acid sequence that is common to known allergens. There do not appear to be any "typical" patterns of sequences associated with allergenic proteins nor any particular conservation of secondary and tertiary structure among the allergenic epitopes. Comparisons of allergenic sequences against the global database of proteins reveals sequence similarities with a number of proteins, indicating that there may be more subtle structural features that define an allergenic epitope. Therefore, it has been suggested that sequence comparisons for potential allergens should be made against a comprehensive collection of all allergen sequences rather than all sequences (Gendel, 1998a).

With this in mind, two non-redundant databases of allergen sequences and their variants (food and non-food allergens) have been established using information from three large reference protein sequence databases (Gendel, 1998b). As primary sequences of new allergens are published, updated versions of these databases are made available on-line at http://www.iit.edu/~sgendel. The number of known allergens (not just food allergens) in databases is roughly 300 (Schmidt, 1998). Thus a protein under consideration for insertion into a crop can be compared against these databases. The strategy for comparison is to take overlapping short amino acid sequences (8 in 1996/ILSI; 6 in 2001 WHO/FAO) through the entire length of the protein in question. For a protein to be allergenic, it must contain at least two IgE antibody reactive sites (B-cell epitopes) to cause mediator release from mast cells and basophils. Because discontinuous epitopes are determined by the three dimensional structure of the protein, they are hard to identify.

Analyses suggest those exact matches of eight or more amino acids occur infrequently by chance and a negative result can usually eliminate a potential allergen. Matches of six contiguous amino acids, as proposed by FAO/WHO, are likely to occur more frequently, by chance. Indeed, there may be too many random matches for this to be a useful screen. The sequence analysis must be carried out with the proper algorithms and scoring parameters. The preferred and recommended procedure for identifying a possible epitope is to initially use local alignment FASTA (Fast A or Fast Alignment) or BLAST (Basic Local Alignment Search Tool) with a scoring matrix based on identity (identity matrix) rather than similarity (Gendel, 1998b). It is especially important that the results are then used to do a follow-up sequence comparison using an evolutionary or biochemical similarity-scoring matrix. Some conservative substitutions may not affect allergenicity and a potential allergenic protein may go undetected based on absolute identity alone. As an additional screen for allergenic protential, the FAO?SHO document also proposes that an overall sequence similarity of 35% between an introduced protein and a known allergen should trigger concerns about potential cross reactivity.

The approach described above does not take into account immunologically significant conformational epitopes that are dictated by the three dimensional structure of the protein. Sequences amino acids, but in the proper structural context, may suffice for binding to IgE (Elsayed et al., 1982; Miller et al., 1996). There is not enough data to make a more generalized correlation. The research and regulatory community is addressing this issue by generating data on the relationship that exists, if any, between amino acid sequence, protein structure and allergenicity. A workshop (ECVAM, 2001) sponsored by the Joint Research Centre of the European Commission reviewed the criteria for such an analysis. A recommendation of that workshop was to establish a single global allergen database incorporating both linear and conformational epitopes and include all allergens.

5.2   Physiochemical Characteristics - Stability to Processing/Digestion

Food allergens are resistant to typical food processing and preparation conditions, in particular, high temperatures. Since a large number of allergens are heat stable, the susceptibility of a protein to heat denaturation is used as a characteristic to assess potential allergenicity. A protein that is unstable at high temperatures (>90(C) will probably not remain intact during processing and, is not likely to be an allergen. However, if the food were consumed unprocessed, then heat stability data would not be useful for evaluating the potential of a protein to be a food allergen.

Another likely physiochemical determinant of potential food allergen is the ability of the intact protein to reach the intestinal mucosa where it could stimulate an immune response locally, or be absorbed into the circulatory system to cause an immune response. The probability of reaching the intestinal mucosa is higher if the allergenic protein maintains its integrity in the pepsin rich, highly acidic environment of the stomach. Thus, stability to gastric digestion (pepsin resistance) has been a key property to assess when testing a protein for potential allergenicity (Astwood et al., 1996).

To assess stability of a protein to gastric digestion, simulated gastric fluid (SGF) is sed as a model of stomach digestion (The United States Pharmacoepeia 23, 1995). In the SGF model system, major food allergens such as ovalbumin, phosphitin, (-lactoglobulin, conglycinin and other peanut/soybean allergens are stable for periods up to 60 minutes compared with the rapid digestion (<15 sec) of the common non-allergenic proteins such as RUBISCO or sucrose synthetase. The digestion mixture is then evaluated by SDS-PAGE using tricine buffers (Shagger and Von Jagow, 1987) to separate proteins that are then visualized by Coomassie Blue staining. The tricine system has the capacity to resolve proteins/peptides in the 1 to 100 kilodalton range and is the recommended system for detecting peptide fragments>10 amino acid residues long that may be created which are then stable.

Stability to gastric digestion does not prove a protein is an allergen, nor does susceptibility to gastric digestion prove a lack of allergenicity. Based on current knowledge, it is assumed and consistent with proteins tested that enzymatic breakdown to small peptide fragments will destroy the antigenic epitope that is essential to sensitization and an allergic reaction. However, as noted in a previous section, peptides of <8 amino acids may still retain IgE binding capacity.

Another question regarding the SGF model system is the degree to which this in vitro system reflects an in vivo situation. Food is seldom ingested as a purified protein and other food matrices can influence protein stability and gut pH. In such cases, looking at additional digestive proteases (i.e. pancreatin, trypsin, chymotrypsin etc.) rather than just pepsin may be useful when assessing the biological stability of an unknown protein.

Despite the limitations noted on stability to SGF (pepsin resistance), the data provide some predictive value of whether a given protein may be a food allergen when it is combined with other parameters such as the source of the gene, the homology sequence similarity, and stability to heat/processing. This testing scheme has been used for evaluation of the potential allergenicity of all the proteins inserted into the transgenic products (Astwood et al., 1996; Rasche, 1998; Harrison et al., 1996).

The 2001 FAO/WHO decision tree process maintains the sequence homology searches and digestibility studies (pepsin resistance), but suggests some changes to methodology and interpretation of these studies.

Return to Table of Contents

The allergenic proteins in the group of eight commonly allergenic foods usually constitute a major portion of the proteins in their respective foods. Thus, an important question is do the levels of exposure to a protein affect the sensitization phase of food allergy? Studies in animals (Kataoka et al., 1998; Yoshikazu et al., 1998; Kumar et al., 1995; Pfeiffer et al., 1995; Kaz'mina et al., 1984) have shown that the induction of sensitization versus the induction of tolerance (no adverse immune reaction) to proteins is influenced by multiple factors which include the:

• the dose of antigen administered,
• the age of the animal at the time of administration of the antigen,
• the type of adjuvant used at the time of dosing
• the nature of the antigen administered

In plants the levels of proteins already vary considerably depending on genetics, growth site, weather conditions and many other factors. Any examination of the significance of changes of normal levels of protein expression must take into account the natural variation in protein levels. However, an increase in the level of a known allergen may cause additional, previously non-allergic, individuals to be sensitized. When the levels of a normal protein are increased beyond the normal variation, this protein would be treated as an unknown allergen source and the right side of the same decision tree is used.

In cases where a gene is inserted into commonly allergenic foods, methods are available to determine if the insertion caused a change in the endogenous levels of the known allergenic proteins in that plant. These assays, immunoblot and RAST inhibition tests, are quantifiable. For example, soybeans are one of the eight commonly allergenic foods. During product development of two different transgenic soybeans, academic scientists compared the endogenous allergens in each transgenic soybean line to their respective conventional commercial varieties. The RAST inhibition tests demonstrated that the insertion of the genes did not change the level of expression of allergenic proteins in either line (Burks et al., 1995; Lehrer et al., 1997). This methodology could also be used to detect a reduction in the level of a known allergen that might have some value.

Minor changes in the increase in level of a traditionally trace protein, by the insertion of a gene, can be detected by two-dimensional gel separation procedures that compare non-transgenic and transgenic protein levels. Any significant alteration can be evaluated.

Return to Table of Contents

As outlined in both the 1996 IFBC/ILSI and the 2001 FAO/WHO decision trees, transfer of a gene known to be an allergen is unlikely to occur when the testing procedures outlined on the left side of the tree are followed. For proteins from less commonly allergenic foods or sources without any history of allergy or consumption, the results from sequence homology, gastric digestibility, processing stability and quantity of protein expressed in the plant are interpreted in context with each other to evaluate the risk of transferring an allergen. This weight-of-evidence approach has proved reliable for the rapidly digestible proteins and was endorsed by both the National Academy of Sciences (NAS, 2000) and two Joint Food and Agriculture Organization/World Health Expert Consultations (FAO/WHO 2000, 2001).

While the 1996 tree does not include animal models in the decision tree, the FAO/WHO 2001 tree does. Data on several proposed food allergy models in the rat (Knippels et al., 1998, 1999a, 1999b, 2000; Atkinson et al., 1996; Dearman et al., 2000), mouse, (Li et al., 1999; Van Halteren et al., 1997; Hilton et al., 1997) and dog (Del Val et al., 1999; Ermel et al., 1997) as well as in vitro models (Yamanishi et al., 1997; Zafiropoulow et al., 1997; Akesson et al., 2000,stickler et al., 2000) have been published. None of these proposed models have been validated with both known food allergens and non-allergens despite their inclusion in the 2001 FAO/WHO decision tree. In the text of that report, it was acknowledged that the animal models are not validated, and that "they do not reflect all aspects of IgE-mediated food allergies in humans." In a report by another panel of scientific experts convened by the USEPA stated:

"Although available, rodent models are not yet regarded as valid models for extrapolation to sensitization in human food allergy. It is also unlikely that any animal model will reliably predict the equivalent of human clinical reactivity." (SAP, 2000).

The need for more research aimed at the development of validated and relevant models for food allergy in humans is acknowledged and supported by many scientists, including those from industry, government and academia.

Return to Table of Contents

All foods, not only those developed using genetic engineering techniques, are potential sources of allergens for certain sensitive individuals. A food allergy is an adverse reaction to an otherwise harmless food or food component that activates the body's immune system. Although the risk of an allergic reaction to food is low, questions are being asked about whether foods derived from biotechnology are more likely to cause food allergies than foods from conventionally-bred crops. These questions were recognized early on by scientists and medical experts. As a result, scientists and medical experts from the Allergy and Immunology Institute of the International Life Sciences Institute (ILSI) in collaboration with the International Food Biotechnology Council (IFBC) developed a decision tree procedure for testing such products that was formalized and published in 1996 (Metcalfe et al., 1996).

Developers of biotechnology-derived crops and government regulators evaluate these crops for food safety in many ways before they are marketed, especially focusing on the potential for allergenicity. There are numerous methods used to determine whether or not a new protein from a transgenic crop is likely to be an allergen. These testing methods for allergenicity are described in this document. Public and private scientists continue to work cooperatively on developing additional tests and models to improve the accuracy of the methods used to determine the potential allergenicity of foods, including those derived from transgenic crops.

Return to Table of Contents