《生物技术与人类》课程教学资源（经典阅读）Assessment of the safety of foods derived from genetically modified

1052 A.Konig et al.Food and Chemical Toxicology 42 (2004)1047-1088 variation between species and individuals are both ten. structurally the same and as the transfer of genetic The NOAEL is therefore often divided by 100 to estab- material across species barriers occurs not only in lish the margin of safety.This information then con- laboratories,but also has been a major driving force in stitutes the basis for a determination of a reference dose, evolution.Hence,concepts were developed to focus exposure to which is deemed safe.The amount of a the safety assessment of GM crops on any functional substance exposure to which over a life time is deemed and chemical changes that result from the genetic safe is the Acceptable Daily Intake (ADI).Risk assess- modification. ment requires judgements on what data are considered Foods prepared and used in traditional ways,and sufficient and what uncertainties need to be taken into consumed under anticipated conditions are generally account. regarded as safe based on their history of human con- Information on the quantity and distribution of a sumption,even though they may contain natural toxins potentially hazardous substance in the environment is or anti-nutritional substances,such as neurotoxic then required in order to determine where populations glycoalkaloids in potatoes,or carcinogenic coumarins are expected to come into contact with the substance; in courgettes.The assessment of novel foods,including for foods,dietary intake assessments of populations are foods derived from a GM crop,relies on the use of a required for exposure estimates.Particular attention is food generally recognised as safe as a comparator paid to expected average and worst-case exposure levels (FAO/WHO,1991).The term 'assessment of substantial of the most sensitive subgroups of a population.This equivalence'describes this comparative assessment information is used to determine the population groups approach.This term was first coined by the Office of that may be at risk and the distribution of such risks. Device Evaluation (ODE)of US FDA in the context of These are the elements that allow estimation of the the evaluation of new medical devices that have a com- probability that harm will occur.Exposure assessment parable function to existing medical devices (Miller, often needs to take into account important societal fac- 1999).Authorities and agencies involved in food safety tors necessary to anticipate behaviour of a wide range of assessment in most countries have based their safety individuals that might affect their exposure. assessment strategies and guidelines on this approach Risk characterisation then combines information on (UK Department of Health,1991;US FDA,1992; hazard and exposure.This includes the probable extent, Health Canada,1994;Japan MHLW,2000;European nature,and duration of exposure with considerations of Commission 1997b). hazard characteristics and relevance of those hazards Application of the concept of substantial equivalence for humans in order to estimate the likely risk to human requires the comparison of the GM crop and an appro- health.Any uncertainties inherent in the risk assessment priate 'safe'comparator according to the agronomical should be highlighted.If the expected exposure exceeds and morphological characteristics.and the chemical the established reference dose that was deemed to be composition,including macro-and micro-nutrients,key safe,this has implications for risk management deci- toxins,and key anti-nutrients.This allows identification sions:risk mitigation measures for chemicals can of significant differences between the GM crop and the include measures such as prescribing use of personal comparator,usually the traditionally-bred parent crop protective equipment or restrictions on conditions of (OECD,1993a).Compositional parameters are then use of the chemical:risk management measures for selected that are typical for the crop that is assessed and foods can include labelling,as is the case for allergenic representative of the main metabolic pathways.Sig- foods.The general principles of risk analysis as descri- nificant changes in these parameters are expected to be bed above apply to the safety assessment of foods indicative of any more fundamental changes in the crop derived from GM crops. that need to be evaluated for their potential to have adverse consequence to human health. 2.3.The concept of safety assessment of foods derived The hazard identification and characterisation of GM from GM crops crops therefore is conducted in four steps:(i)Char- acterisation of the parent crop and any hazards asso- The techniques of molecular biology allow the trans- ciated with it:(ii)characterisation of the transformation fer of genes from one organism to another without process and of inserted recombinant DNA (the poten- sexual reproduction and across species.This process tial consequences of any gene transfer event of the allows desirable alterations to be introduced into plant recombinant DNA to microbes or humans should also genomes in a more specific and controlled manner than be assessed);(iii)characterisation of the introduced can be achieved through conventional breeding and proteins (their potential toxicity and allergenicity)and selection of crops.International panels of experts metabolites;and(iv)identification of any other targeted deemed that there are no risks inherent in the use of and unexpected alterations in the GM crop,including recombinant DNA technologies (OECD,1986;Royal changes in the plant metabolism resulting in composi- Society,1998,2002),as all DNA is chemically and tional changes and assessment of their toxicological

variation between species and individuals are both ten. The NOAEL is therefore often divided by 100 to estab￾lish the margin of safety. This information then con￾stitutes the basis for a determination of a reference dose, exposure to which is deemed safe. The amount of a substance exposure to which over a life time is deemed safe is the Acceptable Daily Intake (ADI). Risk assess￾ment requires judgements on what data are considered sufficient and what uncertainties need to be taken into account. Information on the quantity and distribution of a potentially hazardous substance in the environment is then required in order to determine where populations are expected to come into contact with the substance; for foods, dietary intake assessments of populations are required for exposure estimates. Particular attention is paid to expected average and worst-case exposure levels of the most sensitive subgroups of a population. This information is used to determine the population groups that may be at risk and the distribution of such risks. These are the elements that allow estimation of the probability that harm will occur. Exposure assessment often needs to take into account important societal fac￾tors necessary to anticipate behaviour of a wide range of individuals that might affect their exposure. Risk characterisation then combines information on hazard and exposure. This includes the probable extent, nature, and duration of exposure with considerations of hazard characteristics and relevance of those hazards for humans in order to estimate the likely risk to human health. Any uncertainties inherent in the risk assessment should be highlighted. If the expected exposure exceeds the established reference dose that was deemed to be safe, this has implications for risk management deci￾sions: risk mitigation measures for chemicals can include measures such as prescribing use of personal protective equipment or restrictions on conditions of use of the chemical; risk management measures for foods can include labelling, as is the case for allergenic foods. The general principles of risk analysis as descri￾bed above apply to the safety assessment of foods derived from GM crops. 2.3. The concept of safety assessment of foods derived from GMcrops The techniques of molecular biology allow the trans￾fer of genes from one organism to another without sexual reproduction and across species. This process allows desirable alterations to be introduced into plant genomes in a more specific and controlled manner than can be achieved through conventional breeding and selection of crops. International panels of experts deemed that there are no risks inherent in the use of recombinant DNA technologies (OECD, 1986; Royal Society, 1998, 2002), as all DNA is chemically and structurally the same and as the transfer of genetic material across species barriers occurs not only in laboratories, but also has been a major driving force in evolution. Hence, concepts were developed to focus the safety assessment of GM crops on any functional and chemical changes that result from the genetic modification. Foods prepared and used in traditional ways, and consumed under anticipated conditions are generally regarded as safe based on their history of human con￾sumption, even though they may contain natural toxins or anti-nutritional substances, such as neurotoxic glycoalkaloids in potatoes, or carcinogenic coumarins in courgettes. The assessment of novel foods, including foods derived from a GM crop, relies on the use of a food generally recognised as safe as a comparator (FAO/WHO, 1991). The term ‘assessment of substantial equivalence’ describes this comparative assessment approach. This term was first coined by the Office of Device Evaluation (ODE) of US FDA in the context of the evaluation of new medical devices that have a com￾parable function to existing medical devices (Miller, 1999). Authorities and agencies involved in food safety assessment in most countries have based their safety assessment strategies and guidelines on this approach (UK Department of Health, 1991; US FDA, 1992; Health Canada, 1994; Japan MHLW, 2000; European Commission 1997b). Application of the concept of substantial equivalence requires the comparison of the GM crop and an appro￾priate ‘safe’ comparator according to the agronomical and morphological characteristics, and the chemical composition, including macro- and micro-nutrients, key toxins, and key anti-nutrients. This allows identification of significant differences between the GM crop and the comparator, usually the traditionally-bred parent crop (OECD, 1993a). Compositional parameters are then selected that are typical for the crop that is assessed and representative of the main metabolic pathways. Sig￾nificant changes in these parameters are expected to be indicative of any more fundamental changes in the crop that need to be evaluated for their potential to have adverse consequence to human health. The hazard identification and characterisation of GM crops therefore is conducted in four steps: (i) Char￾acterisation of the parent crop and any hazards asso￾ciated with it; (ii) characterisation of the transformation process and of inserted recombinant DNA (the poten￾tial consequences of any gene transfer event of the recombinant DNA to microbes or humans should also be assessed); (iii) characterisation of the introduced proteins (their potential toxicity and allergenicity) and metabolites; and (iv) identification of any other targeted and unexpected alterations in the GM crop, including changes in the plant metabolism resulting in composi￾tional changes and assessment of their toxicological, 1052 A. Ko¨nig et al. / Food and Chemical Toxicology 42 (2004) 1047–1088

A.Konig et al.Food and Chemical Toxicology 42 (2004)1047-1088 1053 allergenic,or nutritional impact.The exposure assess- represents the best available assessment paradigm;no ment includes estimating the dietary intake of the alternative approaches for the safety assessment of new food derived from GM crops and anticipating foods derived from GM crops have been proposed the effect of food processing on any of the introduced (FAO/WHO,2000;Codex Alimentarius Commission, changes. 2003b).Detailed international guidelines for choice of The successful application of the concept of sub- comparators and for best practices in statistical analysis stantial equivalence largely depends on three critical have been and are being established under the auspices elements:the availability of an appropriate comparator of the Organisation for Economic Coordination and and an understanding of the range of variation to be Development (see also Section 4.1). expected within the measured characteristics of that In summary,the concept of substantial equivalence is comparator;the choice of parameters in the single con- widely accepted by international and national agencies stituent compound analyses,the number and type of as the best available guidance for the safety assessment which will strongly influence the validity of any conclu- of new GM crops.The approach recognises that foods sions on comparative safety;and the ability to dis- are complex matrices containing tens of thousands of criminate between differences in the GM crop and the individual constituents,and that their safety assessment comparator that result from the genetic modification therefore requires a comparative approach focusing on and those differences in the plant's germplasm,some of those parameters deemed indicative of the normal which may be attributed to soma-clonal variation functioning of the plant and its metabolism (including introduced during tissue culture,and environmental or biosynthesis of any compound that might affect human cultivation conditions.All such changes that might have health).As with all scientific concepts,the concept of health implications warrant further investigation,even substantial equivalence is evolving and,together with if they can not only be attributed to the genetic modifi- guidelines,making its application more systematic.The cation,unless they are not manifested in subsequent assessment helps to determine whether the GM crop is generations foreseen for commercial cultivation. 'as safe as'its conventional counterpart.Dialogue Any identified differences are then further assessed as between experts and civil society will contribute over to whether they might have adverse implications for time to further clarify and structure risk analysis strate- human health in the range of exposure scenarios.The gies to improve the salience of assessments to address concept of substantial equivalence is thus the starting concerns of policy makers and the public (Tait,2001; point and guiding concept for the safety assessment,not Jasanoff,2000;Schauzu,2000). its conclusion (FAO/WHO.2000:Codex Alimentarius Commission.2003b).If there are no significant differ- ences between the GM crop and the comparator or if 3.Methods for toxicity testing there are differences that will,with reasonable certainty, not adversely affect health,the GM product is con- Regulatory requirements for chemicals such as food sidered 'as safe as'its counterpart.This approach also additives and pesticides,many of which were first insti- applies to GM crops with more complex metabolic tuted in the 1970s.have led to the development of a modifications,where no single parent crop might be a battery of tests to assess the safety of chemicals in foods. suitable comparator,but where single widely consumed Strategies for assessing the food safety of chemicals substances,food constituents.ingredients.or other often combine three approaches:investigation of the whole foods that are deemed safe under representative structure/function relationship for indications of poten- conditions of use may serve as comparators. tial toxicity and allergenicity;in vitro assays with Critics of the concept of substantial equivalence claim enzymes,receptor proteins,or cultured cell lines:and in that current testing approaches do not sufficiently vivo animal studies.The selection of animal studies is address putative unintended and unexpected effects and based on considerations including a molecule's struc- can not rule out the occurrence of potential long-term ture,function and in vitro toxicity results,as well as effects that result from sustained human exposure to ethical criteria.Of these three distinct approaches,evi- such crops that might have subtle compositional chan- dence from animal tests is usually most indicative of ges that may be difficult to detect (Millstone et al., potential toxic effects of a test substance in humans. 1999).Furthermore,some critics maintain that there is a However,the extrapolation of results from animal tests lack of detailed international standards guiding the choice to humans is uncertain.unpredictable differences can of parameters to be measured in the comparative ana- include inter-species and inter-individual differences in lysis and in the application of rigorous statistical analy- metabolism,physiological processes,and lifestyle.These sis,reducing the quality of individual assessments(SBC, uncertainties are usually addressed through the use of 2001).Groups of international experts have reviewed uncertainty factors (see Section 2.2).Individual tox- the concept of substantial equivalence in the light of icological tests can be designed to be specific to test an these criticisms.It was concluded that the concept still hypothesis of a molecule's toxic effect on one particular

allergenic, or nutritional impact. The exposure assess￾ment includes estimating the dietary intake of the new food derived from GM crops and anticipating the effect of food processing on any of the introduced changes. The successful application of the concept of sub￾stantial equivalence largely depends on three critical elements: the availability of an appropriate comparator and an understanding of the range of variation to be expected within the measured characteristics of that comparator; the choice of parameters in the single con￾stituent compound analyses, the number and type of which will strongly influence the validity of any conclu￾sions on comparative safety; and the ability to dis￾criminate between differences in the GM crop and the comparator that result from the genetic modification and those differences in the plant’s germplasm, some of which may be attributed to soma-clonal variation introduced during tissue culture, and environmental or cultivation conditions. All such changes that might have health implications warrant further investigation, even if they can not only be attributed to the genetic modifi- cation, unless they are not manifested in subsequent generations foreseen for commercial cultivation. Any identified differences are then further assessed as to whether they might have adverse implications for human health in the range of exposure scenarios. The concept of substantial equivalence is thus the starting point and guiding concept for the safety assessment, not its conclusion (FAO/WHO, 2000; Codex Alimentarius Commission, 2003b). If there are no significant differ￾ences between the GM crop and the comparator or if there are differences that will, with reasonable certainty, not adversely affect health, the GM product is con￾sidered ‘as safe as’ its counterpart. This approach also applies to GM crops with more complex metabolic modifications, where no single parent crop might be a suitable comparator, but where single widely consumed substances, food constituents, ingredients, or other whole foods that are deemed safe under representative conditions of use may serve as comparators. Critics of the concept of substantial equivalence claim that current testing approaches do not sufficiently address putative unintended and unexpected effects and can not rule out the occurrence of potential long-term effects that result from sustained human exposure to such crops that might have subtle compositional chan￾ges that may be difficult to detect (Millstone et al., 1999). Furthermore, some critics maintain that there is a lack of detailed international standards guiding the choice of parameters to be measured in the comparative ana￾lysis and in the application of rigorous statistical analy￾sis, reducing the quality of individual assessments (SBC, 2001). Groups of international experts have reviewed the concept of substantial equivalence in the light of these criticisms. It was concluded that the concept still represents the best available assessment paradigm; no alternative approaches for the safety assessment of foods derived from GM crops have been proposed (FAO/WHO, 2000; Codex Alimentarius Commission, 2003b). Detailed international guidelines for choice of comparators and for best practices in statistical analysis have been and are being established under the auspices of the Organisation for Economic Coordination and Development (see also Section 4.1). In summary, the concept of substantial equivalence is widely accepted by international and national agencies as the best available guidance for the safety assessment of new GM crops. The approach recognises that foods are complex matrices containing tens of thousands of individual constituents, and that their safety assessment therefore requires a comparative approach focusing on those parameters deemed indicative of the normal functioning of the plant and its metabolism (including biosynthesis of any compound that might affect human health). As with all scientific concepts, the concept of substantial equivalence is evolving and, together with guidelines, making its application more systematic. The assessment helps to determine whether the GM crop is ‘as safe as’ its conventional counterpart. Dialogue between experts and civil society will contribute over time to further clarify and structure risk analysis strate￾gies to improve the salience of assessments to address concerns of policy makers and the public (Tait, 2001; Jasanoff, 2000; Schauzu, 2000). 3. Methods for toxicity testing Regulatory requirements for chemicals such as food additives and pesticides, many of which were first insti￾tuted in the 1970s, have led to the development of a battery of tests to assess the safety of chemicals in foods. Strategies for assessing the food safety of chemicals often combine three approaches: investigation of the structure/function relationship for indications of poten￾tial toxicity and allergenicity; in vitro assays with enzymes, receptor proteins, or cultured cell lines; and in vivo animal studies. The selection of animal studies is based on considerations including a molecule’s struc￾ture, function and in vitro toxicity results, as well as ethical criteria. Of these three distinct approaches, evi￾dence from animal tests is usually most indicative of potential toxic effects of a test substance in humans. However, the extrapolation of results from animal tests to humans is uncertain, unpredictable differences can include inter-species and inter-individual differences in metabolism, physiological processes, and lifestyle. These uncertainties are usually addressed through the use of uncertainty factors (see Section 2.2). Individual tox￾icological tests can be designed to be specific to test an hypothesis of a molecule’s toxic effect on one particular A. Ko¨nig et al. / Food and Chemical Toxicology 42 (2004) 1047–1088 1053

1054 A.Konig et al.Food and Chemical Toxicology 42 (2004)1047-1088 organ,a combination of specific endpoints,or they can for assessing the likely allergenicity of a novel protein be broad and non-targeted. (FAO/WHO,2001;Codex Alimentarius Commission, A combination of database screening,in vitro,and in 2003c;Kleter and Peijnenburg,2002).The merits and vivo testing approaches is therefore deployed in most limitations of this approach for assessing the safety of food safety testing strategies and provides assurance novel proteins introduced into GM crops are described that the tested food will be as safe as other foods that in more detail in Section 4.3.4. are routinely consumed.The characterisation of the test substance's physico-chemical properties and structure- 3.2.In vitro methods activity relationship that might be indicative of the pre- sence or absence of potential adverse health effects helps In vitro methods contribute to the safety assessment to frame and focus the testing approach.Several tests of chemicals,including food additives:they can in some established for chemicals are applicable to,or have been cases serve as indicators for specific toxic effects of dis- adapted for testing purified recombinant proteins or crete molecules and substances.There are several dif- other substances contained in foods derived from GM ferent types of in vitro methods;these include the in crops.Some animal methods have also been adapted to vitro simulation of digestive systems to assess stability; gain information pertaining to the safety of whole foods bioassays of the activity of purified enzymes;immorta- derived from GM crops.The following section will lised cultured cell lines:or in vitro reconstructions of provide a brief overview on the three distinct approa- receptor or membrane systems.The methods can serve ches developed for chemicals,highlighting which of either as screening systems to assess potential toxicity of these have been deployed in strategies to assess the a compound.or for investigations of a toxicological safety of foods derived from GM crops. mechanism underlying a specific effect observed in vivo or predicted from the structure of a molecule.Such 3.1.Investigating the relationship between structure and assays can serve for instance to assess whether mole- activity for indications of potential adverse effects cules bind to,inhibit,or stimulate proteins with specific functions.In vitro tests include tests that are claimed to The investigation of a substance's structure-activity be indicative of specific organ toxicity.An overview on relationship starts with a description of the defining the status of many in vitro methods in relation to the physico-chemical properties.Computers help to assess assessment of acute toxicity endpoints has been pub- whether a molecule shares characteristics of known lished (Walum.1998).A databank of in vitro techniques toxicants through screening databases with information in toxicology,called INVITTOX,has also been estab- on structural and physico-chemical properties of known lished (INVITTOX.2003).A critical discussion of the toxicants.Some toxicants show a clear structure-activity merits and limitations of in vitro methods is provided in relationship and their mechanism of toxicity is fully Eisenbrand et al.(2002). understood;other classes of toxicants just share com- Few in vitro tests are however validated formally,and mon structural elements or physico-chemical properties extrapolation of results from in vitro tests to in vivo that may be indicative of toxicity.Some physico- situations is often challenging,often their predictive chemical or structural characteristics of molecules are value has not been systematically assessed;uncertainties therefore indicative of a potential toxic effect.A have to be clearly stated.Advantages of the use of in molecule's physico-chemical properties help for instance vitro methods include that they are relatively cheap and the prediction of its propensity to intercalate in high-throughput.In vitro test results may be indicative DNA and interfere with its replication (Barratt,1998). of toxic effects;test systems relying on reconstituted Databases also exist listing proteins with (food) purified protein or cell components,or immortalised allergenic properties (see Table 1).Computer-based laboratory cultures of cell lines are,however,not methods have also been developed for the comparison representative of the functioning of such cell compo- of the primary amino acid sequences of proteins allow- nents or cells in living organisms.In vivo tests are ing identification of contiguous epitopes that might therefore required to confirm observations on toxicity mediate an allergenic effect.Limitations of the method from in vitro tests.The use of in vitro methods may include that non-continuous epitopes cannot be identi- become of greater importance with developments in the fied,and false positive matches through epitopes that area of genomic research and microarray systems to are similar but do not mediate allergenic effects.Any monitor changes in gene expression.Future develop- indication of toxicity or allergenicity obtained through ments in this area are discussed in more detail in Section computer screening has to be confirmed through other studies. The US Pharmacopeia describes simulated gastric and A computer-assisted search for similarity to proteins intestinal fluid preparations that have been used to known to cause adverse effects,such as protein toxins or assess the stability of proteins(US Pharmacopeia,1990; allergens,forms part of the current recommendations Astwood et al..1996).The test is used as one indicator

organ, a combination of specific endpoints, or they can be broad and non-targeted. A combination of database screening, in vitro, and in vivo testing approaches is therefore deployed in most food safety testing strategies and provides assurance that the tested food will be as safe as other foods that are routinely consumed. The characterisation of the test substance’s physico-chemical properties and structure￾activity relationship that might be indicative of the pre￾sence or absence of potential adverse health effects helps to frame and focus the testing approach. Several tests established for chemicals are applicable to, or have been adapted for testing purified recombinant proteins or other substances contained in foods derived from GM crops. Some animal methods have also been adapted to gain information pertaining to the safety of whole foods derived from GM crops. The following section will provide a brief overview on the three distinct approa￾ches developed for chemicals, highlighting which of these have been deployed in strategies to assess the safety of foods derived from GM crops. 3.1. Investigating the relationship between structure and activity for indications of potential adverse effects The investigation of a substance’s structure-activity relationship starts with a description of the defining physico-chemical properties. Computers help to assess whether a molecule shares characteristics of known toxicants through screening databases with information on structural and physico-chemical properties of known toxicants. Some toxicants show a clear structure–activity relationship and their mechanism of toxicity is fully understood; other classes of toxicants just share com￾mon structural elements or physico-chemical properties that may be indicative of toxicity. Some physico￾chemical or structural characteristics of molecules are therefore indicative of a potential toxic effect. A molecule’s physico-chemical properties help for instance the prediction of its propensity to intercalate in DNA and interfere with its replication (Barratt, 1998). Databases also exist listing proteins with (food) allergenic properties (see Table 1). Computer-based methods have also been developed for the comparison of the primary amino acid sequences of proteins allow￾ing identification of contiguous epitopes that might mediate an allergenic effect. Limitations of the method include that non-continuous epitopes cannot be identi- fied, and false positive matches through epitopes that are similar but do not mediate allergenic effects. Any indication of toxicity or allergenicity obtained through computer screening has to be confirmed through other studies. A computer-assisted search for similarity to proteins known to cause adverse effects, such as protein toxins or allergens, forms part of the current recommendations for assessing the likely allergenicity of a novel protein (FAO/WHO, 2001; Codex Alimentarius Commission, 2003c; Kleter and Peijnenburg, 2002). The merits and limitations of this approach for assessing the safety of novel proteins introduced into GM crops are described in more detail in Section 4.3.4. 3.2. In vitro methods In vitro methods contribute to the safety assessment of chemicals, including food additives; they can in some cases serve as indicators for specific toxic effects of dis￾crete molecules and substances. There are several dif￾ferent types of in vitro methods; these include the in vitro simulation of digestive systems to assess stability; bioassays of the activity of purified enzymes; immorta￾lised cultured cell lines; or in vitro reconstructions of receptor or membrane systems. The methods can serve either as screening systems to assess potential toxicity of a compound, or for investigations of a toxicological mechanism underlying a specific effect observed in vivo or predicted from the structure of a molecule. Such assays can serve for instance to assess whether mole￾cules bind to, inhibit, or stimulate proteins with specific functions. In vitro tests include tests that are claimed to be indicative of specific organ toxicity. An overview on the status of many in vitro methods in relation to the assessment of acute toxicity endpoints has been pub￾lished (Walum, 1998). A databank of in vitro techniques in toxicology, called INVITTOX, has also been estab￾lished (INVITTOX, 2003). A critical discussion of the merits and limitations of in vitro methods is provided in Eisenbrand et al. (2002). Few in vitro tests are however validated formally, and extrapolation of results from in vitro tests to in vivo situations is often challenging, often their predictive value has not been systematically assessed; uncertainties have to be clearly stated. Advantages of the use of in vitro methods include that they are relatively cheap and high-throughput. In vitro test results may be indicative of toxic effects; test systems relying on reconstituted purified protein or cell components, or immortalised laboratory cultures of cell lines are, however, not representative of the functioning of such cell compo￾nents or cells in living organisms. In vivo tests are therefore required to confirm observations on toxicity from in vitro tests. The use of in vitro methods may become of greater importance with developments in the area of genomic research and microarray systems to monitor changes in gene expression. Future develop￾ments in this area are discussed in more detail in Section 5. The US Pharmacopeia describes simulated gastric and intestinal fluid preparations that have been used to assess the stability of proteins (US Pharmacopeia, 1990; Astwood et al., 1996). The test is used as one indicator 1054 A. Ko¨nig et al. / Food and Chemical Toxicology 42 (2004) 1047–1088

A.Konig et al.Food and Chemical Toxicology 42 (2004)1047-1088 1055 as to whether the recombinant protein shares the whole foods to animals at doses that are large multiples characteristic of stability to digestion under these con- of the expected human exposure.Frequently,in feeding ditions that is common to many allergens (see Section trials with whole foods at the highest administered dose. 4.3.4) no adverse effect is observed.In animal studies where no adverse effect is observed.and where administered doses 3.3.Animal tests do not significantly exceed expected human exposure levels,it is not possible to account for uncertainties A variety of tests with laboratory animals have been regarding variations of susceptibility between species and developed for identifying and characterizing health between individuals within a population.Furthermore, hazards associated with exposure to single well-defined the food matrix in which the test material is adminis- chemicals and food additives.Tolerance studies are tered needs careful consideration:it may be difficult to sometimes used to confirm the absence of adverse effects ensure nutritional balance when diets contain high pro- at expected levels of exposure.Ethical considerations portions of novel foods or food extracts.Inadvertent should be an important driver in decisions on whether changes in nutritional status may result in adaptive chan- to conduct in vivo studies,in the choice of test species, ges that may mistakenly be attributed to adverse effects and in the design of the study protocol.Both tox- from intake from the GM crop and can be attributed to icological animal tests and tolerance studies are dis- changes resulting from the genetic modification. cussed in more detail below. Animal feeding studies with whole foods that take Animals can be used for the determination of acute account of these difficulties through careful design and toxicity of a substance,usually involving administration description of conditions under which the tests are car- of a large single dose followed by a short period of ried out and any remaining uncertainties have,however, observation.Sub-chronic toxicity,chronic toxicity,or been used to complement other tests for the safety carcinogenicity is tested in animals over prolonged per- assessments of foods derived from GM crops.Such tests iods of one or several months,or the lifetime of an ani- may provide useful information,in particular if doses mal.The use of these methods has recently been that are multiples of anticipated human exposure can be reviewed in detail by the European Concerted Action administered to animals (Hammond et al.,1996:Kuiper Food Safety in Europe (FOSIE)(Barlow et al.,2002). etal.,2001). They form the basis of toxicity testing of chemicals,and Studies in which animals (or,all be it rarely,human have provided valuable information for this purpose volunteers)are administered the expected level of intake (Kroes and Kozianowski,2002).In addition to the or low multiples of that level,so-called tolerance stud- standard tests listed in Appendix A,other studies could ies,can be used to complement other safety testing be carried out where deemed necessary to address approaches.Well-established protocols for tolerance immuno-toxicity;endocrine toxicity;individual organ studies of pharmaceuticals are available and can be toxicity;and toxico-kinetic investigations on absorp- adapted for this purpose.The highest dose used must be tion,distribution,metabolism,and excretion (ADME) without adverse nutritional effects and must respect the of a substance.Most of these methods have been stan- appropriate balance of nutrients required by the test dardised and OECD guidelines are available for their species.A classical range of parameters such as growth conduct and interpretation (see Appendix A;OECD, rate and feed efficiency are measured to identify possible 1993b). adverse effects on specific endpoints.The use of appro- These animal methods,if deemed necessary,can be priate controls is an important determinant of the adapted to test for potential adverse effects of recombi- validity of the results.For instance,if the novel com- nant proteins or novel compounds or metabolites ponent is normally present in a particular matrix,then introduced into crops through genetic modification (see that matrix would be an appropriate control.For the Section 4.3).It can however be difficult to obtain suffi- tolerance assessment of GM crops controls should cient quantities of purified recombinant proteins for include animal groups using the corresponding com- testing in animals over prolonged periods of time. parator.Similar methods can be used for animal and Adapting OECD guidelines and principles for testing human studies(see also Section 4.4.3). specific well defined chemicals to the assessment of whole foods,including whole foods derived from GM 3.4.Post-market monitoring crops,presents,however,several challenges.First,ani- mal tests for chemicals are usually designed to identify a Post-market monitoring systems have been estab- dose response relationship from which the consequences lished by several food companies for certain food pro- of exposure to low doses,typical of human exposure, ducts to act as early warning systems and to facilitate can be extrapolated.The highest administered dose is product recall in the event where health concerns might normally expected to produce some observable adverse be associated with a specific food.The organisation of effect.It is,however,often not possible to administer post-market monitoring is primarily the responsibility

as to whether the recombinant protein shares the characteristic of stability to digestion under these con￾ditions that is common to many allergens (see Section 4.3.4). 3.3. Animal tests A variety of tests with laboratory animals have been developed for identifying and characterizing health hazards associated with exposure to single well-defined chemicals and food additives. Tolerance studies are sometimes used to confirm the absence of adverse effects at expected levels of exposure. Ethical considerations should be an important driver in decisions on whether to conduct in vivo studies, in the choice of test species, and in the design of the study protocol. Both tox￾icological animal tests and tolerance studies are dis￾cussed in more detail below. Animals can be used for the determination of acute toxicity of a substance, usually involving administration of a large single dose followed by a short period of observation. Sub-chronic toxicity, chronic toxicity, or carcinogenicity is tested in animals over prolonged per￾iods of one or several months, or the lifetime of an ani￾mal. The use of these methods has recently been reviewed in detail by the European Concerted Action Food Safety in Europe (FOSIE) (Barlow et al., 2002). They form the basis of toxicity testing of chemicals, and have provided valuable information for this purpose (Kroes and Kozianowski, 2002). In addition to the standard tests listed in Appendix A, other studies could be carried out where deemed necessary to address immuno-toxicity; endocrine toxicity; individual organ toxicity; and toxico-kinetic investigations on absorp￾tion, distribution, metabolism, and excretion (ADME) of a substance. Most of these methods have been stan￾dardised and OECD guidelines are available for their conduct and interpretation (see Appendix A; OECD, 1993b). These animal methods, if deemed necessary, can be adapted to test for potential adverse effects of recombi￾nant proteins or novel compounds or metabolites introduced into crops through genetic modification (see Section 4.3). It can however be difficult to obtain suffi- cient quantities of purified recombinant proteins for testing in animals over prolonged periods of time. Adapting OECD guidelines and principles for testing specific well defined chemicals to the assessment of whole foods, including whole foods derived from GM crops, presents, however, several challenges. First, ani￾mal tests for chemicals are usually designed to identify a dose response relationship from which the consequences of exposure to low doses, typical of human exposure, can be extrapolated. The highest administered dose is normally expected to produce some observable adverse effect. It is, however, often not possible to administer whole foods to animals at doses that are large multiples of the expected human exposure. Frequently, in feeding trials with whole foods at the highest administered dose, no adverse effect is observed. In animal studies where no adverse effect is observed, and where administered doses do not significantly exceed expected human exposure levels, it is not possible to account for uncertainties regarding variations of susceptibility between species and between individuals within a population. Furthermore, the food matrix in which the test material is adminis￾tered needs careful consideration: it may be difficult to ensure nutritional balance when diets contain high pro￾portions of novel foods or food extracts. Inadvertent changes in nutritional status may result in adaptive chan￾ges that may mistakenly be attributed to adverse effects from intake from the GM crop and can be attributed to changes resulting from the genetic modification. Animal feeding studies with whole foods that take account of these difficulties through careful design and description of conditions under which the tests are car￾ried out and any remaining uncertainties have, however, been used to complement other tests for the safety assessments of foods derived from GM crops. Such tests may provide useful information, in particular if doses that are multiples of anticipated human exposure can be administered to animals (Hammond et al., 1996; Kuiper et al., 2001). Studies in which animals (or, all be it rarely, human volunteers) are administered the expected level of intake or low multiples of that level, so-called tolerance stud￾ies, can be used to complement other safety testing approaches. Well-established protocols for tolerance studies of pharmaceuticals are available and can be adapted for this purpose. The highest dose used must be without adverse nutritional effects and must respect the appropriate balance of nutrients required by the test species. A classical range of parameters such as growth rate and feed efficiency are measured to identify possible adverse effects on specific endpoints. The use of appro￾priate controls is an important determinant of the validity of the results. For instance, if the novel com￾ponent is normally present in a particular matrix, then that matrix would be an appropriate control. For the tolerance assessment of GM crops controls should include animal groups using the corresponding com￾parator. Similar methods can be used for animal and human studies (see also Section 4.4.3). 3.4. Post-market monitoring Post-market monitoring systems have been estab￾lished by several food companies for certain food pro￾ducts to act as early warning systems and to facilitate product recall in the event where health concerns might be associated with a specific food. The organisation of post-market monitoring is primarily the responsibility A. Ko¨nig et al. / Food and Chemical Toxicology 42 (2004) 1047–1088 1055

1056 A.Konig et al.Food and Chemical Toxicology 42 (2004)1047-1088 of the manufacturer of the food.Methods vary from and to link such variations to health outcomes (FSA, establishing channels of communication in the firm to 2002).Possible requirements for post-market monitor- receive direct consumer feedback on the product,to the ing of GM crops with more complex modifications, repurchase of products to determine the quality of the including altered nutritional value,are discussed in product on the supermarket shelf. Section 4.6. Post-market monitoring programmes may serve to confirm the absence of specific adverse health effects of certain products after they have been marketed.The 4.The safety assessment of foods derived from GM feasibility and validity of post-market monitoring crops depends on the health endpoint of interest and on the way the product is marketed.First,post-market mon- Safety considerations for foods derived from GM itoring of manifestation of adverse effects from intake of crops are fundamentally the same as those for conven- specific foods is most feasible if it is driven by an tional foods or other types of novel foods (Cockburn, hypothesis on a potential and specific adverse effect that 2002).Since the genetic improvement of crops has is acute and has distinctive symptoms.Acute adverse always been the aim of plant selection and breeding, effects generally associated with relatively high intakes 'traditional'approaches are appropriate in order to of a substance,or allergic reactions,are likely to become assess the safety of foods derived from all GM crops, apparent by post-market monitoring for spontaneous regardless of the crop species or the trait introduced by events.However,long-term or rare effects require genetic modification. generally a more targeted and intrusive study design. This section sets out a systematic stepwise approach Randomised controlled human trials could be used to on how to select appropriate combinations of test investigate possible medium/long term effects,but the methods to assess the safety of foods derived from GM wide variation in diets and dietary components from crops(Figs.I and 2).The objective of the assessment is day to day and year to year should be recognised.While to determine whether these new foods are at least as safe clinical studies in humans may provide wider assurance as foods produced from conventional crops;this of safety of whole foods.they cannot fully reproduce approach hence offers a high level of safety assurance. the diversity of the populations who will consume the The outlined approach serves to structure case-by- marketed product.The possibility therefore remains case assessments of specific products;it provides gui- that unpredicted side effects may occur in some sectors dance on how to design a test programme for the safety of the population,such as those with certain disease assessment of a GM crop that is tailored to the specific conditions or with particular genetic characteristics.In characteristics of the parent crop and the introduced addition,risk assessment relies on an estimate of expo- trait(s).The safety assessment focuses on the new gene sure to the food,which is variable and subject to product(s)and of whole foods derived from the GM uncertainty before the food is marketed.Critical issues crop.Both intended and potential unintended effects to the success of a post-market monitoring programme from the genetic modification are taken into account. relating to the marketing approach of the foods are the The assessment involves the following steps:(i)char- ability to estimate exposure with reasonable accuracy acterisation of the parent crop;(ii)characterisation of (traceability)and to match any reported effects to the the donor organism(s)from which any recombinant consumption of the material (Wal et al.,2003).For DNA sequences are derived,the transformation pro- identity-preserved products this may be feasible, cess,and the introduced recombinant DNA sequences; whereas for commodities it is much more difficult,if not (iii)safety assessment of the introduced gene products impossible. (proteins and metabolites);and (iv)food safety assess- To date,no GM crop has been placed on the market ment of whole food derived from,or edible part of,the for which post-market monitoring was deemed neces- GM crop. sary.The current safety testing strategy has been con- The assessment focuses on any changes introduced sidered sufficiently predictive for these approvals. through the genetic modification,including introduced Problems limiting the interpretation of post-market genes and gene products,and potentially altered levels monitoring of foods derived from commodity crops of endogenous compounds or the formation of new have been highlighted(FAO/WHO,2000).The United metabolites.Methods for the detection of unexpected Kingdom Food Standards Agency has commissioned a changes in the composition due to the genetic modifi- study to examine the feasibility of using supermarket cation process are discussed and evaluated in the paper and household survey data for post-market monitoring by Cellini et al.(2004).Possible consequences of trans- of novel foods.This study assesses the government's fer of the recombinant sequences to gastrointestinal ability to detect variations of food purchasing and microflora or to humans should also be assessed and are consumption at the district level in Great Britain,as evaluated in the paper by Van den Eede et al.(2004). this is seen as an indicator for the feasibility to detect The assembled information may also help to identify

of the manufacturer of the food. Methods vary from establishing channels of communication in the firm to receive direct consumer feedback on the product, to the repurchase of products to determine the quality of the product on the supermarket shelf. Post-market monitoring programmes may serve to confirm the absence of specific adverse health effects of certain products after they have been marketed. The feasibility and validity of post-market monitoring depends on the health endpoint of interest and on the way the product is marketed. First, post-market mon￾itoring of manifestation of adverse effects from intake of specific foods is most feasible if it is driven by an hypothesis on a potential and specific adverse effect that is acute and has distinctive symptoms. Acute adverse effects generally associated with relatively high intakes of a substance, or allergic reactions, are likely to become apparent by post-market monitoring for spontaneous events. However, long-term or rare effects require generally a more targeted and intrusive study design. Randomised controlled human trials could be used to investigate possible medium/long term effects, but the wide variation in diets and dietary components from day to day and year to year should be recognised. While clinical studies in humans may provide wider assurance of safety of whole foods, they cannot fully reproduce the diversity of the populations who will consume the marketed product. The possibility therefore remains that unpredicted side effects may occur in some sectors of the population, such as those with certain disease conditions or with particular genetic characteristics. In addition, risk assessment relies on an estimate of expo￾sure to the food, which is variable and subject to uncertainty before the food is marketed. Critical issues to the success of a post-market monitoring programme relating to the marketing approach of the foods are the ability to estimate exposure with reasonable accuracy (traceability) and to match any reported effects to the consumption of the material (Wal et al., 2003). For identity-preserved products this may be feasible, whereas for commodities it is much more difficult, if not impossible. To date, no GM crop has been placed on the market for which post-market monitoring was deemed neces￾sary. The current safety testing strategy has been con￾sidered sufficiently predictive for these approvals. Problems limiting the interpretation of post-market monitoring of foods derived from commodity crops have been highlighted (FAO/WHO, 2000). The United Kingdom Food Standards Agency has commissioned a study to examine the feasibility of using supermarket and household survey data for post-market monitoring of novel foods. This study assesses the government’s ability to detect variations of food purchasing and consumption at the district level in Great Britain, as this is seen as an indicator for the feasibility to detect and to link such variations to health outcomes (FSA, 2002). Possible requirements for post-market monitor￾ing of GM crops with more complex modifications, including altered nutritional value, are discussed in Section 4.6. 4. The safety assessment of foods derived from GM crops Safety considerations for foods derived from GM crops are fundamentally the same as those for conven￾tional foods or other types of novel foods (Cockburn, 2002). Since the genetic improvement of crops has always been the aim of plant selection and breeding, ‘traditional’ approaches are appropriate in order to assess the safety of foods derived from all GM crops, regardless of the crop species or the trait introduced by genetic modification. This section sets out a systematic stepwise approach on how to select appropriate combinations of test methods to assess the safety of foods derived from GM crops (Figs. 1 and 2). The objective of the assessment is to determine whether these new foods are at least as safe as foods produced from conventional crops; this approach hence offers a high level of safety assurance. The outlined approach serves to structure case-by￾case assessments of specific products; it provides gui￾dance on how to design a test programme for the safety assessment of a GM crop that is tailored to the specific characteristics of the parent crop and the introduced trait(s). The safety assessment focuses on the new gene product(s) and of whole foods derived from the GM crop. Both intended and potential unintended effects from the genetic modification are taken into account. The assessment involves the following steps: (i) char￾acterisation of the parent crop; (ii) characterisation of the donor organism(s) from which any recombinant DNA sequences are derived, the transformation pro￾cess, and the introduced recombinant DNA sequences; (iii) safety assessment of the introduced gene products (proteins and metabolites); and (iv) food safety assess￾ment of whole food derived from, or edible part of, the GM crop. The assessment focuses on any changes introduced through the genetic modification, including introduced genes and gene products, and potentially altered levels of endogenous compounds or the formation of new metabolites. Methods for the detection of unexpected changes in the composition due to the genetic modifi- cation process are discussed and evaluated in the paper by Cellini et al. (2004). Possible consequences of trans￾fer of the recombinant sequences to gastrointestinal microflora or to humans should also be assessed and are evaluated in the paper by Van den Eede et al. (2004). The assembled information may also help to identify 1056 A. Ko¨nig et al. / Food and Chemical Toxicology 42 (2004) 1047–1088