Science-based Approaches to Assessing Allergenicity of New Proteins in GE Foods

Science-based Approaches to Assessing Allergenicity of New Proteins in Genetically Engineered Foods
Presentation to FDA Food Biotechnology Subcommittee, Food Advisory Committee.
College Park, MD
August 14, 2002
Michael Hansen, Ph.D., Research Associate, Consumers Union


Thank you for the chance to present the views of Consumers Union, publisher of Consumer Reports, to this Subcommittee. The Food and Drug Administration is taking a very positive, important and much-needed step by undertaking an effort to develop a protocol for assessing the potential allergenicity of genetically engineered food. Food allergies can be life threatening for the estimated 2% of adults and 8% of children who suffer from them. The potential to inadvertently transfer a gene that codes for an allergen through genetic engineering was demonstrated when Brazil nut genes were transferred into soybeans by Pioneer Seeds. Pioneer fortunately conducted tests that determined that an allergen had been inadvertently transferred, and voluntarily stopped developing the product. However this case, and the subsequent case of Starlink corn, whose potential allergenicity was much more difficult to predict, underline the need to have a sound, consistent, and comprehensive assessment protocol. The FDA protocol should be one which when scientific data is incomplete errs on the side of protecting consumer health, and is used by all companies developing products and by all agencies regulating them.

We think this guidance should be incorporated in the rule on Pre-Market Notification which FDA has under development. Our comments will focus primarily on the specifics of what the assessment should contain and how it should be conducted.

The FDA can profitably draw on the work of several excellent bodies that have already given considerable thought to the difficult question of allergenicity assessment. We would like to draw special attention to the 2001 report of the Joint FAO/WHO Joint Expert Consultation on Allergenicity of Foods Derived from Biotechnology, chaired by Dr. Dean Metcalfe of the National Institute of Health (FAO/WHO, 2001), to the Annex on Allergenicity to the Guidelines for Assessment of the Safety of Recombinant DNA Plants, agreed to last March by the Codex Alimentarius Ad Hoc Committee on Foods Derived from Biotechnology, and to the work of the Environmental Protection Agency’s (EPA) FIFRA Scientific Advisory Panel (SAP) which was charged with developing mammalian toxicity assessment guidelines for protein plant pesticides and with assessing the human safety of Starlink corn (SAP, 2001a,b).

Key Points
As an overview, we urge FDA to:

–make this protocol a rule not guidance; it needs to be mandatory and not voluntary

–include all allergens, dermal and inhalant as well as food when determining amino acid sequence homology/similarity between new proteins and known allergens

–consider all the assessment criteria used by the EPA SAP and/or FAO/WHO 2001 Expert Consultation and/or EPA: amino acid sequence homology, digestive stability, heat stability, animal models, physical characteristics (size/molecular weight, probable glycosylation)

— integrate all criteria into a decision tree since no single criterion is absolutely predictive as to allergenicity; we suggest use of the FAO/WHO 2001 decision tree, modified where necessary to take account of the fact that some assessment techniques are much better developed than others. In general, FDA should require standardized procedures/methodologies for the use of individual assessment criteria used in the decision tree with a view to a harmonized application of the decision tree.

–conduct tests for all “newly expressed proteins” (language from Annex 1 of draft safety assessment guidelines for rDNA plants from the Codex Alimentarius Ad Hoc Task Force on Foods Derived from Modern Biotechnology); this means not just the intended transgene product (e.g. protein), but also includes all the unintended newly expressed proteins (e.g. the process of GE may turn on genes in a plant/animal that had been previously turned off, or the transgene protein could interact with the complex metabolic pathway in the organisms to create a new protein).

–require proteins be tested in purified form and as they exist in the food

–require purified proteins be extracted from the plant and/or animal from which the food will be derived; FDA should not allow a company to test the protein as it is expressed in a bacterial or other microbial source if that is not the form that will be consumed.

We will now comment on several key assessment techniques that we think must be part of an assessment protocol.

Amino Acid sequence homology/similarity

Although no single criterion has been shown to be absolutely accurate in predicting the allergenicity of a (novel) protein, perhaps the most basic criterion that has been employed is the notion that proteins which are similar in structure (e.g. homologous) to a know allergen will have a greater likelihood of being an allergen than a protein which has little or no structural similarity to known allergens. Thus, virtually all protocols that have been developed to test the allergenicity of genetically engineered proteins include the comparison of amino acid sequence of novel (engineered) proteins with those of known allergens (Metcalfe et al., 1996; NRC, 2000; SAP, 2000a; FAO/WHO, 2001).

Perhaps the first protocol developed to help test for the possible structural similarity between a novel protein (of unknown allergenic potential) and a known allergen was contained in decision tree developed by the industry-funded International Food Biotechnology Council (IFBC) in conjunction with the Allergy and Immunological Institute of the International Life Sciences Institute (ILSI) [Metcalfe et al., 1996]. Given that the 3-dimensional structure of most allergenic epitopes is not known, the IFBC/ILSI decision tree focused on the amino acid sequence homology of the newly introduced protein and a data base of known allergens and recommended that any sequence of eight contiguous amino acids in the test protein that exactly matches a corresponding sequence in a known allergen, using a global algorithm that optimizes alignments/matches across the entire full-length of the protein, should be a cause for concern and should trigger further investigation. This has been termed the “eight amino acid match approach” (EAAM-approach). Sequence of identity of less than 8 amino acids is not considered to raise concerns about potential allergenicity. A slightly modified version of the IFBC/ILSI decision tree can be seen in Figure 1.

In the six years since the IFBC/ILSI decision tree approach, a number of changes or refinements to the approach, based on accumulating scientific knowledge, have been suggested for detecting structural similarity between novel proteins and known allergens. Some of the suggested refinements include: i) allowing for substitution of chemically similar amino acids in the 8-amino acid sequence (Fuchs and Astwood, 1996; Gendel, 1998b; SAP, 2000a); ii) using identity of 6 or 4 identical contiguous amino acids rather than 8 (SAP, 200a; Becker, 2001; FAO/WHO 2001); iii) using local alignments (regions with a high degree of similarity) rather than the entire protein (e.g. a global alignment) when comparing unrelated proteins (Gendel 1998b; Becker, 2001; FAO/WHO, 2001); iv) using 35% overall amino acid homology to a known allergen as an additional criterion (FAO/WHO 2001); and v) developing databases and methods to test for conformational or discontinuous epitopes (defined by 3-D structure rather than simple amino acid sequence) including those caused by changed glycolysation patterns (SAP 2000a; FAO/WHO, 2001; Becker, 2001).

Most of the above problems/suggested modifications of the IFBC/ILSI decision tree approach to sequence homology have been succinctly described by Dr. Wolf-Mienhard Becker in his paper, “Sequence homology and allergen structure,” written for the 2001 Joint WHO/FAO Expert Consultation (Becker, 2001). Dr. Becker notes that the use of the EAAM-approach “leads to the insight that conformational epitopes and contiguous parts of these epitopes after denaturation, and epitopes made up by sugar residues, are not identifiable by this procedure. Apart from the result [that] identified linear epitopes of peanut or cod fish only consist of 6 or 4 contiguous amino acid residues which are essential for IgE binding. Thus the EAAM-approach would fail. The conclusion from that is that the EAAM-approach even including only six contiguous amino acids can only identify potential allergenic components but not rule them out. Since predicting or excluding allergenicity is a matter of immunology the epitope, the interface between chemical structure and the immune system, should come into focus. . . . chemical structure is suitable but the most convincing tools are epitope receptors such as patients’ IgE or monoclonal antibodies to test the allergenicity of the protein in question in the genetically engineered food. Since the maturation of the immune system cannot be predicted monitoring studies of immune responses in consumers should be undertaken after the genetically engineered food has reached the market” (Becker, 2001: 1).

The focus on epitopes is a crucial one since the immune system cannot recognize the whole structure of a macromolecule, such as a protein or glycoprotein, but can only smaller sections called determinants or epitopes. The caveat to this is that the immunological behavior of an epitope can be affected by the whole structure of the macromolecule. In principal, two types of epitopes exist: linear (or continuous) epitopes based directly on the primary protein structure (e.g. amino acid sequence) and conformational (or discontinuous) epitopes based on the (3-dimensional) surface area of a molecule formed by discontinuous sections of the primary protein structure. Two compartments of the immune system that deal with epitopes are the B-cells and T-cells. T-cell epitopes are exclusively linear in nature while B-cells respond to both conformation and linear epitopes. Many (but not all) classical food allergens tend to contain linear epitopes while aeroallergens and pollen-related food allergens (those responsible for “oral allergy syndrome”) often contain conformational epitopes.

The EAAM-approach codified in the IFBC/ILSI decision tree focuses on T-cell epitopes, where 8 amino acids is the minimal size for such epitopes. However, B-cell epitopes can be smaller and can occur in food allergens, as Becker notes with the case of certain peanut and cod allergens (Becker, 2001).
While epitopes are clearly more important than the general amino acid sequence of a known allergen, very few epitopes have been determined. Only a small-to-moderate percentage of food allergens have even been identified. Various protein data bases contain the amino acid sequence of 180 major allergens of which 30 are food allergens of plant origin (Metcalfe et al., 1996). At the same time, a literature review found more than 150 foods associated with sporadic allergic reactions (Hefle et al., 1996). It should be noted, though, that roughly 90% of all moderate to severe allergic reactions to food come from eight types of food sources: peanuts, soybeans, milk, eggs, fish, crustacea, wheat and tree nuts. The number of identified epitope sequences for the various food allergens is miniscule compared to the probable number of epitopes that exist. Indeed, one of the main suggestions for further work is that “Research is needed to map all the epitopes of known allergens and to develop monoclonal antibodies against them” (Becker, 2001: 4). We concur whole-heartedly.

Becker also notes that glycosylation patterns can affect allergenicity and immunogenicity of a protein. He cites the example of “a-amylase [where it is known] that this allergen and protein is glycosylated, when expressed in eucaryotic plants and immunologically active but not in E. coli” (Becker, 2001: 3). As further noted in the final report of the FAO/WHO Expert Consultation, “Glycosylation may alter the epitope structure, either by shielding part of the protein surface (particularly if the glycosylation is extensive), or by introducing glycan epitopes. Glycan epitopes are known to be highly cross-reactive” (FAO/WHO, 2001: 11). Since E. coli does not glycosylate proteins, while many plants and animals do, we feel that all allergy testing of novel proteins be based on the protein as it is expressed in the organism destined for food and not on the protein as expressed in a bacterial host such as E. coli, as has routinely been permitted by the EPA and FDA.

The FAO/WHO Expert Consultation developed a standardized methodology for determining sequence homology between and introduced protein and known allergens. It started with the IFBC/ILSI decision tree and updated that tree on the basis of evolving scientific knowledge in the area. In contrast to the IFBC/ILSI decision tree, FAO/WHO suggested using identity of 6 rather than 8 identical contiguous amino acids as a criterion for further concern and using local alignments rather than global alignments when comparing unrelated proteins. They also suggested additional criteria such as a 35% overall amino acid sequence homology as a cause for further concern and the development of databases and methods to test for discontinuous epitopes including those changed by glycosylation patterns. FAO/WHO recommended the following standardized methodology for determining sequence homology:

“6.1. Sequence Homology as Derived from Allergen Databases
The commonly used protein databases (PIR, SwissProt and TrEMBL) contain the amino acid sequence of most allergens for which this information is known. However, these databases are currently not fully up-to-date. A specialized allergen database is under construction.
Suggested procedure on how to determine the percent amino acid identity between the expressed protein and known allergens.

Step 1: obtain the amino acid sequence of all allergens in the protein databases . . . in FASTA-format (using the amino acids from the mature protein only, disregarding the leader sequences, if any). Let this be data set (1).

Step2: prepare a complete set of 80-amino acid length sequences derived from the expressed protein (again disregarding the leader sequence, if any). Let this be data set (2).

Step 3: go to EMBL internet address: and compare each of the sequences of the data set (2) with all sequences of data set (1), using the FASTA program on the web site for alignment with the default settings for gap penalty and width.

Cross-reactivity between the expressed protein and a known allergen (as can be found in the protein databases) has to be considered where there is: 1) more than 35% identity in the amino acid sequence of the expressed protein (i.e. without the leader sequence, if any), using a window of 80 amino acids and a suitable gap penalty (using Clustal-type alignment programs or equivalent alignment programs) or: 2) identity of 6 contiguous amino acids.
If any of the identity scores equals or exceeds 35%, this is considered to indicate significant homology within the context of this assessment approach. The use of amino acid sequence homologies to identify prospective cross-reacting allergens in genetically-modified foods has been discussed in more detail elsewhere (Gendel, 1998a, Gendel, 1998b).

Structural similarity with known allergens may still be important if significant amino acid identity is found, but it is below 35%. In this case significant cross-reactivity is unlikely. However, some families of structurally related proteins are known to contain several allergens. Some examples are: lipocalins, non-specific lipid transfer proteins, napins (2S albumins from seeds), parvalbumins.

If the expressed protein belongs to such a family, it may be considered to have a higher probability to be an allergenic protein. . . . Since identity of 6 contiguous amino acids has an appreciable risk of occurring by chance, verification of potential crossreactivity is warranted when criterion (1) is negative, but criterion (2) is positive. In this situation suitable antibodies (from human or animal source) have to be tested to substantiate the potential for crossreactivity” (FAO/WHO, 2001: 10-11).

The report of the FAO/WHO Expert Consultation makes a reference to the work of Dr. Steven Gendel, chief of FDA’s Biotechnology Studies Branch. In a pair of papers Dr. Gendel discusses the various databases of allergens and how to use them to determine sequence similarity between an expressed protein and known allergen (Gendel, 1998a, b). Dr. Gendel argues persuasively for use of local algorithms rather than global algorithms when assessing allergenicity of novel proteins because most novel proteins are not evolutionarily related. As he points out, “sequence algorithms can be divided into global algorithms that optimize alignments across the entire length of the sequences involved and local algorithms that attempt to optimize alignments only with regions of high similarity. Global alignment algorithms are of greatest utility when the sequences involved are related. Allergenicity assessment involves sequence alignments between proteins that are not evolutionarily related. Therefore, it is likely that local alignment will be more useful” (Gendel, 1998b: 50). Gendel tests this assumption with known allergens and finds that the local alignment works best. The original IFBC/ILSI decision tree used a global alignment algorithm. Local alignment algorithms include the FASTA and BLAST program, which give similar results (Gendel, 1998b); FAO/WHO recommends use of the FASTA program. Gendel notes that “Although it is likely that immunological cross-reactivity requires extensive sequence similarity, absolute identity may not be necessary (for example, see Elsayed et al., 1982)” (Gendel, 1998b: 57). He then goes on to develop a “biochemical similarity matrix” which “divides the amino acids into six classes based on biochemical characteristics (i.e., hydrophilic acid amino acids, hydrophilic basic amino acids, etc.). . . Alignment of members of the same class is scored as a mismatch. The realignment was confined to a region of 15 to 20 amino acids in each case to preserve the previously located identities” (Gendel, 1998b: 58).

Using this methodology, Gendel finds significant sequence homology between b-lactoglobulin (major milk allergen) and Cry3A (found in Bt potatoes) and between Cry1Ab or Cry1Ac and vitellogenin (egg allergen). He concludes, “although it is clear that some amino acid residues are critical for specific binding, some conservative substitutions may not affect allegenicity. Therefore, it may be prudent to treat sequence matches with a high degree of identity that occur within regions of similarity as significant even if the identity does not extend for eight or more amino acids. For example, the similarity between Cry1A(b) and vitellogenin might be sufficient to warrant additional evaluation” (Gendel, 1998b: 60).

In sum, we urge FDA to follow the protocol laid out by FAO/WHO as slightly modified by Dr. Gendel (e.g. allow chemically similar amino acid residues to be used when determining short sequence similarity/identity for the contiguous amino acid sequences). We also agree with the EPA SAP, FAO/WHO and Dr. Becker that developing databases and methods (such as monoclonal antibodies using animal and/or human materials) to test for conformational or discontinuous epitopes including those caused by changed glycolysation patterns is of key importance and urge FDA to try and encourage studies in these areas.
Digestive Stability (enzymatic digestion)

A number of scientific and other sources-including the Environmental Protection Agency, FIFRA’s Science Advisory Panel (SAP), the International Life Sciences Institute (ILSI), the FAO/WHO Expert Consultation on Allergenicity of Foods Derived from Biotechnology and Codex Alimentarius’ Ad Hoc Task Force on Foods Derived from Modern Biotechnology-have agreed that digestive stability (or enzymatic digestion) of new protein produced in foods developed via bioengineering should be a criterion that is assessed. A number of these sources agree that in order for the criterion of digestive stability to be used, standardized methods need to be developed so that any laboratory can repeat them. As Drs. Steve Taylor and Samuel Lehrer pointed out in an early paper in this area, “Although the assessment of the resistance to hydrolysis of proteins could offer valuable information regarding the potential allergenicity of specific proteins, a rigorous protocol for such experiments has not been established. Ideally, this protocol would mimic digestive proteolysis and included tests on the isolated protein and the protein in the appropriate food matrix” (Taylor and Lehrer, 1996: ).

All sources quoted above agree that assessing digestive stability should involve simulating the environment of the human digestive system. One can either simulate the environment of the stomach, via simulated gastic fluid (SGF), or simulate the environment of the intestine, via simulated intestinal fluid (SIF). Most of the authors prefer the use of SGF. However, some note that if significant amounts of the undegraded or protein fragments survive SGF, then SIF testing should ensue (Helms, 2001). There has also been debate about the protocol for developing SGF. One of the first studies that demonstrated a link between allergenicity of a protein and resistance to digestion used the United States Pharmacopiea (USP) protocol for SGF (Astwood et al., 1996). However, the USP protocol for SGF has been criticized for not being sufficiently physiological in nature (Helms, 2001). Since the publication of the Astwood et al. paper in 1996, there have been a number of scientific meetings, symposia and papers that have further discussed protocols (or the need for them) for testing digestive stability; these are reviewed by Dr. Ricki Helm, of the Arkansas Children’s Hospital Reseach Institute, in his paper “Stability of Known Allergens (Digestive and Heat Stability)” written for the FAO/WHO expert consultation. In this paper, Dr. Helms, after reviewing the scientific work in this area, makes the following recommendations for protocols for SGF and SIF:
“Simulated gastric fluid (SGF)
1-Standardized source materials and pH ranges.
a. Pepsin should be from a reliable source and enzymatic activity should be expressed in arbitrary units prior to assessment of novel protein degradation. For this, the method used by Ryle (6) could be applied, i.e., enzymatic activity based upon measuring TCA precipitable hemoglogin after hydrolysis for 10 min.

b. A standardized enzyme/protein ratio should be established.

c. Bovine serum albumin should be used as a digestible protein.

d. Peanut allergens (and/or a stable protein readily available in pure form) should be used as a non-digestible protein.

e. The novel protein should be assessed in enriched or pure form, both recombinant and natural sources. If the matrix is to be assessed,
assessment should be from both the natural and transgenic form.

f. The effects of pH determinations should be made at 1.0, 1.5, 2.0, 4.0 and 6.0 due to the pH variation in the stomach following a meal.

g. Sampling of digestion should be taken at the following time points, 0, 15, and 30 seconds; and 1, 2, 4, 8, 15 and 60 minutes.

h. A scale in arbitrary units should be established using the digestible and non-digestible proteins to characterize the novel protein.

i. Reasonable criteria of digestibility for acceptance should be determined. (This could be based upon the data being collected by members testing the protocol recommended by the ILSI/HESI working subgroup).

j. All analyses should be made at 37°C.

2-Standardized analytical methods for determining degree of degradation.
a. Column chromatography (e.g., HPLC) should be used to assess the degree of degradation.

b. B. SDS-PAGE analysis, both denaturing and non-denaturing conditions, should be standardized according to the following criteria.
i. A common gel system should be used, e.g., Novex system.
ii. 10-20% acrylamide gradient gels
iii. A sensitive staining method should be used (Silver stain or colloidal gold).
c. Immunoblot analysis.
i. A standardized blotting system should be used, e.g., Novex sytem.
ii. Both polyclonal and monoclonal antibody assessments should be used to determine degree of degradation.
d. Data should be provided in publishable format.

Simulated intestinal fluid (SGF)[sic; should be SIF] This assay should only be used if there are considerable amounts of undegraded or protein fragments identified in the SGF. A gastroenterologist should be consulted for best physiologic conditions. Pancreatin sources are too variable, therefore a standardized mixture of enzymes should be used.

1-A minimal composition to that of physiological state, i.e., pancreatic drainage fluid of animal to enzyme mixtures in test sample, should be used.
a. Homogenous sources of trypsin/amylase/lipase/elastase/chymotrypsin are recommended from reliable sources (Worthington). (This will be difficult to manage, as sources may be limited and purity questionable).

2-Standards and conditions for SGF should be applied” (Helms, 2001: pp. 9-10).
The paper by Dr. Helm (Helm, 2001) served as a starting point for discussion of the Joint FAO/WHO Expert Consultation on Allergenicity of Foods Derived from Biotechnology. The final report of the Expert Consultation recommended a slightly modified version of Dr. Helm’s protocol (for example, rather than test the protein at a range of pHs to simulate the stomach at various times after feeding, the FAO/WHO Expert Consultation recommends testing only at pH 2.0), but it contained far more specific details about what the protocol should contain.

Their recommendation follows:
“6.4. Pepsin Resistance
Purified of enriched expressed protein (non-heated and non-processed) should be subjected to pepsin degradation conditions using Standard Operating Procedures and Good Laboratory Practices (SOP/GLP). In addition, the expressed protein should be assessed in its principle edible form under identical pepsin degradation conditions to those used to examine the expressed protein. Both known non-allergenic (soybean lipoxygenase, potato acid phosphatase or equivalent) and allergenic (milk beta lactoglobulin, soybean trypsin inhibitor or equivalent) food proteins should be included as comparators to determine the relative degree of the expressed pepsin resistance. The protein concentrations should be assessed using a colorimetric assay (e.g., Bicinchoninic acid assay (BCA), Bradford Protein Assay, or equivalent protein assay) with bovine serum albumin (BSA) as a standard. Pepsin proteolytic activity should be assessed (Ryle). Enzyme/protein mixtures should be prepared using 500mg of protein in 200mL of 0.32% pepsin (w/v) in 30mM/L NaCl, pH 2.0, and maintained in a shaking 37°C water bath for 60 minutes. Individual 500 microgram aliquots of pepsin/protein solution should be exposed for periods of 0, 15, 30 seconds and 1, 2, 4, 8, 15, and 60 minutes, at which time each aliquot should be neutralized with an appropriate buffer. Neutralized protein solutions should be mixed with SDS-PAGE sample loading buffer with and without reducing agent (DTT or 2-ME) and heated for 5 minutes at 90°C. Samples containing 5mg/cm gel of protein should be evaluated using 10-20% gradient Tricine SDS-PAGE gels or equivalent gel system under both non-reducing and reducing electrophoretic conditions. Protein in the gels should be visualized by silver or colloidal gold staining procedures. Evidence of intact expressed protein and/or intact fragments greater than 3.5 kDa would suggest a potential allergenic protein. Evidence of protein fragments less than 3.5 kDa would not necessarily raise issues of protein allergenicity and the data should be taken into consideration with other decision tree criteria. For detection of expressed protein in an edible food source, a polyclonal IgG immunoblot analysis should be performed according to the laboratory procedures. The immunoblot analysis should be compared to the silver or colloidal gold stained SDS-PAGE gel and reflect the stained pattern of the expressed protein run under identical conditions” (FAO/WHO, 2001: 12-13).

One significant extension of Dr. Helm’s protocol that the FAO/WHO Expert Consultation included was the notion that “the expressed protein should be assessed in its principle edible form under identical pepsin degradation conditions to those used to examine the expressed protein” (FAO/WHO, 2001: 12).

CU absolutely agrees that the expressed form of the protein should be assessed both in purified form and as part of the food that it occurs in. The reason for this is that the food matrix can act as a buffer allowing the expressed protein to survive digestion. There are many examples of this. For instance, a number of growth hormones in milk, such as insulin-like growth factor-1 (IGF-1) or epidermal growth factor (EGF), are protected from digestion by the presence of casein (Kimura et al., 1997; Playford et al., 1993; Xian et al., 1995). One study with IGF-1 found that 9% survived digestion when fed in pure form to rats; in the presence of casein, 67% survived digestion (Kimura et al., 1997). More recently, a study involving transgenic soy or corn DNA found that while 80% of the naked DNA was degraded in gastric simulations, none of the transgene DNA was digested when it was part of the food stuff: “The data showed that 80% of the transgene in naked soya DNA was degraded in the gastric simulations, while no degradation of the transgene contained within GM soya and maize were observed in these acidic conditions” (Martin-Orue et al., 2002: 533). While we realize that DNA is not a protein, the general phenomenon-partial survival of substance when part of a food compared to testing the pure substance-we feel is applicable. Furthermore, as Dr. Helm pointed out in his paper for the FAO/WHO Expert

Consultation, recent industry and scientific thinking in this area concur: “The working committee on the ‘Characteristics of Protein Food Allergens’ held by ISLI/HESI following the symposium established the following criteria be taken into consideration. . . . 3-Deliver: Consideration should be given to how the material will be introduced into the diet. Assessment of allergenicity should be based on the matrix/matrices that the novel protein would be introduced into the diet” (Helm, 2001: 6).

In conclusion, we urge that FDA require companies to follow the protocol as laid out in FAO/WHO Expert Consultation, which we described above. If there are to be deviations from this protocol, companies should be required to give a scientific justification for such deviations. In particular, we feel the FDA should not allow the companies to simply use USP protocol for SGF. Furthermore, FDA should not allow the companies to simply test the protein at pH 1.2 (as per the USP protocol). If a company wants to test the protein at pH 1.2, the FDA should also require higher pHs as well, including, at least, pH 2.0.

Second, we feel the FDA should require the company to test the protein in both the purified expressed form as well as in the form in which it occurs in food, e.g. as part of the food matrix. For the purified expressed form, we feel that the company should extract the protein from the transgenic material that is intended to be commercialized and not use a form of the protein that is extracted from a bacterial or other microbial source.

Finally, if a significant portion of the expressed protein does survive digestion in SGF, we recommend that it be tested further in SIF, using the protocol laid out by Dr. Helm.

Heat stability

Both allergy scientists as well as the Environmental Protection Agency (EPA) consider stability of a protein to heat to be a characteristic property of food allergens (Sampson, 1999; EPA, 2001; Helm, 2001; and Taylor and Hefle, 2001). During the Bt crop reregistration process, EPA vaguely adopted heat stability as a criterion for potentially allergenicity for the Bt Cry endotoxins, stating that a characteristic “considered as an indication of possible relation to a food allergen are [is] a protein’s ability to withstand heat or the conditions of food processing” (EPA, 2001b: IIB2). However, EPA has neither strictly required nor even suggested a test protocol for such data. Indeed, for a couple of Bt crops-Novartis’ Bt corn (Cry1Ab) and Monsanto’s Bt cotton (hybrid Cry1Ac/Ab)-the EPA accepted data that processed corn or cottonseed meal were inactive in an insect bioassay. Monsanto submitted a more formal heat stability study for a relatively new Bt corn variety (containing Cry1F rather than the usual Cry1Ab), but the methodology was flawed. The study’s main methodological flaw consisted of the sole end-point (e.g., measure of degradation) being “growth inhibition of neonate tobacco budworm larvae” following “application of treated Cry1F to the surface of an insect diet” (EPA 2001b: 10). Such a study implicitly assumes that the insecticidal mode of action correlates with allergenicity and that loss of insecticidal action means no allergenicity. There is no scientific justification for such an assumption. Theoretically, a protein could be allergenic and have insecticidal activity; loss of that activity does not imply loss of allergenicity. As has been noted by a number of scientists, degraded proteins or protein fragments can still elicit an allergenic response even though the protein is functionally inactive; a perfect example is the major milk allergen b-lactoglobulin (Haddad et al., 1979).

In contrast to the EPA’s lack of a consistent protocol, Dr. Helm has developed a science-based protocol as part of the paper on the topic that he wrote for the 2001 FAO/WHO Expert Consultation: “Heat Stability: The definition of heat stability should be standardized using the following criteria. 1-Heat treatment of the novel protein, native and recombinant, should be for 5 minutes at 90°C. 2-Assessment of stability by a combination of molecular sieving using HPLC and standardized SDS-PAGE analysis (both native and denaturing/reducing gels). See SDS-PAGE protocol below” [see the section on digestive stability, above for this protocol] (Helm, 2001: 8-9).

We urge that the FDA require data on heat stability and use the science-based protocol as outline by Dr. Helm (Helm, 2001). We would suggest the following additions/explanations to the protocol. The recombinant protein should be tested in both purified form and as part of the food in which it occurs. The purified form of the protein should be extracted from the engineered organism (usually plant) that will make up the food; the company should not be permitted to use a bacterial or other microbial source to produce the recombinant protein. Also, the engineered protein should be added to a food matrix/matrices, preferably to the matrix in which it will occur.

Animal models

Both the EPA’s FIFRA Scientific Advisory Panel (SAP), which looked at StarLink corn, and the FAO/WHO Expert Consultation recommended the use of animal models. The FIFRA SAP investigating StarLink considered immunological response in the brown Norway rat (BNR) and bioavailability of the protein in bloodstream of BNR as criteria suggesting of allergic potential of the Cry9C protein although they stated that these two assays had not used a standardized methodology (SAP, 2000b).

The FAO/WHO Expert Consultation had this to say about animal models:
“6.5. Animal Models
For additional assessment of the potential allergenicity of expressed proteins, informative data can be generated using animal models in development. A number of animal models may be considered to assess on a relative scale the potential allergenicity using oral sensitization routes with the Brown Norway rat model (Knippels et al., 1998) or intraperitoneal administration in murine models (Dearman et al., 2000) or other relevant animal models. Results should be presented in characteristic Th1/Th2 antibody (isotype) profiles for assessing the potential immunogenic/allergenic activity. The different routes of administration in animal models (oral versus intraperitoneal) may not give the same results. Therefore, selection of one route of administration is not meant to exclude other routes of sensitization. It is recommended to consider the results from two sensitisation routes in the same or different animal species.

It is recommended that the potential allergenicity be ranked against well known strong and weak food allergens and non-allergenic proteins in the animal model. As additional information becomes available with respect to animal models, protocols may need to be modified to give optimal conditions for assessing protein allergenicity.

Although the present animal models provide additional information on potential allergenicity of novel proteins, they do not reflect all aspects of IgE-mediated food allergies in humans” (FAO/WHO, 2001: 13).

While there is some question as to the reliability/applicability of use of animal models for predicting food allergy, animal models have been used quite successfully in predicting/evaluating inhalant allergens. One of the speakers at a conference titled “Assessment of the Allergic Potential of Genetically Modified Foods,”-sponsored by the National Toxicology Program, EPA, FDA and NIH and held in December 2001 in Chapel Hill, North Carolina-was Katherine Sarlo, principal scientist at Proctor & Gamble. Dr. Sarlo gave a talk at the meeting about how useful rodent models have been over the years in testing for allergic reactions to enzymes used in their detergents. According to Dr. Sarlo, when P&G first started using enzymes in their detergents in the mid-1960s, many workers in their plants developed allergies to the enzymes. In the intervening decades, P&G developed accurate rodent models using certain strains of guinea pigs and mice. Certain strains of guinea pigs developed an IgG response to the enzymes that caused allergic reactions in some workers, while certain strains of mice showed both IgG and IgE responses to the enzymes. The strains of mice and guinea pigs used were ones in which there was a correlation between the responses of the animals and the responses of the workers. Over the years, the used of these particular animal models, combined with medical surveillance of the workers and modification of the environment to dramatically reduce the problem. In sum, the experience of P&G definitely demonstrates that animal models are both useful and provide predictable responses as to how humans respond to allergens. She suggested that a similar approach be used to investigate potential allergenicity of genetically engineered foods (Anonymous, 2002).

We feel that the experience of P&G definitely shows that animal models can be successfully used to predict allergenicity of proteins. We therefore recommend that FDA urge companies to conduct animal studies, utilizing either the protocol as laid out by FAO/WHO or the protocol developed by P&G. In this regard, perhaps the same strains of guinea pigs and mice that were successful surrogates for humans when prediciting inhalant allergenicity of proteins may be successfully used to predict food allergenicity. Indeed, we suggest that FDA begin such research with these strains of guinea pigs and mice.


Anonymous. 2002. EPA, NIH, and FDA seek allergenicity protocols to test and regulate genetically modified foods. Chemical and Engineering News. January 7, 2002.

Astwood, J.D., Leach, J.N. and R.L. Fuchs. 1996. Stability of food allergens to digestion in vitro. Nature Biotechnology 14 (10): 1269-1273.

Becker, W. M. 2001. Topic 4: Sequence Homology and Allergen Structure. Working Paper Biotech 01/06 for the Joint FAO/WHO Expert Consultation on Allergenicity of Foods Derived from Biotechnology, January 22-25, 2001. Rome, Italy.

Dearman, R.J., Caddick, H., Basketter, D.A. and I. Kimber. 2000. Divergent antibody isotope responses induced in mice by systemic exposure to proteins: a comparison of ovalbumin with bovine serum albumin. Food Chemistry Toxicology, 38: 351-360.

Elsayed, S., Apold, J., Hiolen, E., Vik, H. and E. Florvagg. 1982. The immunological reactivity of the three homologous repetitive tetrapeptides in the region 41-64 of allergen M from cod. Scandanavian Journal of Immunology, 16: 77-82.

EPA. 2001a. Biopesticides Registration Action Document: Revised Risks and Benefits Sections – Bacillus thuringiensis Plant-Pesticides. July 16, 2001, EPA.

EPA. 2001b. Biopesticide Registration Action Document: Bacillus thuringiensis Cry1F Corn, August 2001, EPA.

FAO/WHO. 2001. Evaluation of Allergenicity of Genetically Modified Foods. Report of a Joint FAO/WHO Expert Consultation on Allergenicity of Foods Derived from Biotechnology, January 22-25, 2001. Rome, Italy.
Fuchs, R. and J. Astwood. 1996. Allergenicity assessment of food derived from genetically modified plants. Food Technology, 50(2): 83-88.

Gendel, S.M. 1998a. Sequence databases for assessing the potential allergenicity of proteins used in transgenic foods. Advances in Food and Nutrition Research, 42: 63-92.

Gendel, S.M. 1998b. The use of amino acid sequence alignments to assess potential allergenicity of proteins used in genetically modified foods. Advances in Food and Nutrition Research, 42: 44-61.
Haddad et al. 1979. IgE antibodies to peptic and peptic-tryptic digests of beta-lactoglobulin: significance in food hypersensitivity, Annals of Allergy, 42: 368- .

Hefle, S.L., Nordlee, J.A. and S.L. Taylor. 1996. Allergic foods. Critical Reviews in Food Science and Nutrition, 36: S69-S89.

Helm, R.M. 2001. Topic 5: Stability of Known Allergens (Digestive and Heat Stability). Working Paper Biotech 01/07 for the Joint FAO/WHO Expert Consultation on Allergenicity of Foods Derived from Biotechnology, January 22-25, 2001. Rome, Italy.

Kimura, T., Murakawa, Y., Ohno, M., Ohtani, S. and K. Higaki. 1997. Gastrointestinal absorption of recombinant human insulin-like growth factor-I in rats. The Journal of Pharmacology and Experimental Therapeutics, 283: 611-618.

Knippels et al. 1998. Oral sensitization to food proteins: a Brown Norway rat model. Clinical and Experimental Allergy, 28: 368-375.

Martin-Orue, S.M., O’Donnell, A.G., Arino, J., Netherwood, T., Gilbert, H.J. and J.C. Mathers. 2002. Degradation of transgenic DNA from genetically modified soya and maize in human intestinal simulations. British Journal of Nutrition, 87: 533-542.

Metcalfe, D.D., Astwood, J.D., Townsend, R., Sampson, H.A., Taylor, S.L. and R.L. Fuchs. 1996. Assessment of the allergenic potential of foods derived from genetically altered crop plants. Critical Reviews in Food Science and Nutrition, 36: S165-S186.

Playford, R.J., et. al. 1993. Effect of luminal growth factor preservation on intestinal growth. Lancet 2: 843-848.

Sampson, H. A. 1999. Food allergy. Part 1: Immunopathogenesis and clinical disorders. Journal of Allergy and Clinical Immunology, 103(5): 717-28.

SAP 2000a. Session II – Mammalian Toxicity Assessment Guidelines for Protein Plant Pesticides. FIFRA Scientific Advisory Panel Report No. 2000-03B, September 28, 2000.

SAP 2000b. Assessment of Scientific Information Concerning StarLink Corn. FIFRA Scientific Advisory Panel, SAP Report No. 2000-06, December 1, 2000.

Taylor, S.L. and S.B. Lehrer. 1996. Principles and characteristics of food allergens. Critical Reviews in Food Science and Nutrition, 36: S91-S118.

Taylor, S.L. and S.L. Hefle. 2001. Will genetically modified foods be allergenic? Current Reviews of Allergy and Clinical Immunology, 107(5): 765-71.

Xian, C.J., C.A. Shoubridge and L.C. Read. 1995. Degradation of IGF-I in the adult rat gastrointestinal tract is limited by a specific antiserum or the dietary protein casein. Journal of Endocrinology, 146: 215-225.

articles post