Genetically modified foods: potential human health effects

This Review appeared in "as Chapter 16 in the book "Food Safety: Contaminants and Toxins" by by J P F D'Mello, Scottish Agricultural College, Edinburgh, UK, April 2003. We have reproduced it here with the kind permission of Dr Arpad Pusztai.

To make the review easier to navigate, you can go directly to any section by clicking on the title for that section in the Table of Contents. Then after reading that section, you can jump back to the Table of Content by clicking on the BACK button of your browser.. I have also highlighted the food names for easy reference.

Genetically modified foods: potential human health effects

Arpad Pusztai, Susan Bardocz and Stanley W.B. Ewen

Table of Contents


Arpad Pusztai (corresponding author) and Susan Bardocz private consultants; formerly Rowett Research Institute, Aberdeen, Scotland UK
(6 Ashley Park North, Aberdeen, AB10 6SF, Scotland UK);
phone/fax: 44-1224-594954; e-mail:

Stanley W.B. Ewen, University of Aberdeen, Department of Pathology, Forresterhill, Aberdeen, Scotland, UK;

1. Introduction

The scope of this review is restricted to data-based considerations about the safety of genetically modified (GM) foods of plant origin for health. No opinions, unless supported by experimental results, will be discussed. The emphasis will be on papers published in peer-reviewed journals. A few articles will be mentioned from non-peer-reviewed journals but only if they influenced the development of science-based ideas for the regulatory process. Environmental issues will not be dealt with.

Safety evaluation of whole foods derived from crops with considerable natural variability is more difficult than that of a single chemical, pharmaceutical, food additive or defined mixtures of them. Published results of tests for toxicity and nutritional wholesomeness of complex foodstuffs are therefore few and far between. A recent comment in Science described this in its title: "Health Risks of Genetically Modified Foods: Many Opinions but Few Data" (Domingo, 2000). Or even a cursory look at the list of references of a recent major review on food safety issues (Kuiper et al, 2001) showed that most of the the publications referred to were non-peer-reviewed institutional opinions or envisaged future scientific and methodological developments for safety assessments but were short on actual published scientific papers on which a reliable data base of safety could be founded. Judging from the absence of published data in peer-reviewed scientific literature, no human clinical trials with GM food have apparently ever been conducted. Most attempts to establish the safety of GM food have been indirect. At best, inferences have been drawn from animal trials but the preferred approach is to use compositional comparisons between the GM foodstuff and its traditional counterpart. If these results show no significant differences the two foodstuffs are "substantially equivalent", meaning that the GM is as safe as the non-GM food. Thus, as the regulation is almost exclusively dependent on "substantial equivalence", the published results of GM food analyses and inferences drawn from them for health will be critically examined in this review.

In genetic modification the intended gene is incorporated into the genome of a crop using a vector containing several other genes, including as a minimum, viral promoters, transcription terminators, antibiotic resistance marker genes and reporter genes. Although in GM food safety the role of the intended gene is very important, the potential effects of these other genes need also be taken into account because other parts of the construct or the insertion of the vector could substantially contribute to the overall effect (Ewen and Pusztai, 1999b). There is in fact some evidence that some of the other genes of the vector may have an effect on safety. This is particularly so as it is now known that DNA does not always break down in the alimentary tract (Schubbert et al, 1994; 1998; Hohlweg and Doerfler, 2001). This opens up the possibility that the antibiotic resistance marker gene, in addition to others, may be taken up by bacteria in the digestive tract and contribute to the spreading of antibiotic reistance via human gut bacteria. In this context one potentially important observation was that a substantial proportion (6 to 25%) of a genetically engineered plasmid survived a one hour exposure to human saliva (Mercer et al, 1999). Partially degraded plasmid DNA also successfully transformed Streptococcus gordonii, a bacteria that normally lives in the human mouth. Saliva also contains factors which increase the ability of bacteria to become transformed by naked DNA. Therefore the prospect of the uptake of undegraded or partially degraded vector genes, including the antibiotic resistance gene, will have to be seriously considered. However, the main concern in GM food safety is what are the direct effects of the expression of the main intended gene after its insertion into the plant genome via a gene construct. An additional concern is that this may also cause significant, indirect and unintended effects on the expression and functionality of the plant's own genes. The number of copies of the construct inserted and their location in the plant genome (pleiotropic effect) are of particular importance in this respect, with the possibility that many unexpected changes may occur. This possibility is in fact generally accepted and the inadequacy of the currently used methods to detect them are frequently acknowledged (Kuiper et al, 2001). Pleiotropic effects always occur with both conventional cross-breeding and genetic engineering and their unwanted consequences are usually eliminated by empirically selecting for the desired trait and discarding the potentially harmful ones. Some of these changes are unpredictable and therefore we can only compare the known properties and constituents but cannot look for, and even less analyse, unknown components. This imposes limitations on our selection criteria. Reliance based solely on chemical analysis of macro/micronutrients and known toxins is at best inadequate and, at worst, dangerous. More sophisticated analytical methods need to be devised, such as mRNA fingerprinting, proteomics and secondary metabolite profiling (Kuiper et al, 1999). However, and most importantly, there is an urgent need to develop comprehensive toxicological/nutritional methods to screen for the unintended potentially deleterious consequences for human/animal health of genetic manipulation to pinpoint the problems in advance of the incorporation of the GM foodstuff into the food chain (Ewen and Pusztai, 1999a). Although some limited animal testing has been done, only a few of these have been published. However, data from some of these studies have recently been placed on the internet. Although they were not peer-reviewed they were incorporated into this review because of their potential importance for other scientists.

2. Non-peer-reviewed safety tests on commercial GM crops in the public domain

2.1. FLAVR SAVRTM tomatoes
The first example of official safety evaluations of a GM crop, Calgene's FLAVR SAVRTM tomato, including a 28 day rat feeding trial, was commissioned by Calgene for the FDA (Food and Drug Administration, USA) before its general release. Although the details of this study have never been properly published, because this work had such an extraordinarily major effect and influenced GM food regulation in the USA and elsewhere, there is a compelling need to analyse the methods used and the conclusions reached. Fortunately, as a result of a Court case in the USA (Alliance for Bio-Integrity et al, vs Shalala et al) most data in the FDA's files are now on the internet in the public domain and can therefore be evaluated (Alliance for Biointegrity Home, 1998).

This GM tomato study shows most of the problems which may be encountered in GM food safety evaluation, particularly as if like the tomato, they are fruits rather than foodstuffs and their protein and energy contents are insufficient for supporting the growth of young animals. The methods used and results obtained in this study are important not only for their own sake but also for their influence on the process of regulation.

2.1.A. Substiantial equivalence

As "substantial equivalence" so prominently features in GM food regulation (Kuiper et al, 2001), including in this GM tomato study, there is a need to look more closely at this concept. This issue has been dealt with in some depth by a recent article (Millstone et al, 1999) in which the problems with this concept were highlighted, such as that "substantial equivalence" has never been properly defined and that there are no legally binding rules on how to establish it in practice.

Differences in growth conditions can have serious impact on composition and therefore in the absence of specification of the origin and the conditions of cultivation of the different GM and non-GM samples, strict scientific comparisons cannot be made. These are not valid unless the parent line is grown side-by-side with the GM line. Comparisons with historical or literary values have only limited scientific validity.

"Substantial equivalence" is a crude, non-scientific concept. It provides a loophole for the GM biotechnology companies not to carry out nutritional and toxicological animal tests to establish whether the biological effect of the GM crop-based foodstuff is substantially equivalent to its non-GM counterpart. It therefore allows them to claim that there is no need for biological testing because the GM crops are similar to their conventional counterpart, while on the other hand because they contain novel genes from other organism(s), they are patentable. However, unintentional and unpredictable changes can occur in plants because of the incorporation and positioning of the vector into the plant genome. It cannot therefore be known which of the hundreds of components of the GM crop may carry toxic or allergenic properties. As most of these are unknown, by definition they cannot be included in analytical comparisons. Determination of the amounts of protein, carbohydrates, fats and other nutrients can only be a starting point. The consumption of minor and unexpected constituents of potentially high biological activity may have considerable and disproportionally large effects on the digestive tract. Their presence, therefore, can only be revealed from animal studies and this makes it imperative that these are performed with a flawless design and experimentation.

The FLAVR SAVRTM tomato was produced by 'antisense' GM technology. As part of its safety evaluation it was subjected to compositional analysis for total protein, vitamins and minerals to establish whether any unexpected changes in gross fruit composition had arisen as a result of the integration of the FLAVR SAVRTM and kanr genes into the tomato genome. It was claimed that no significant changes were found and that the contents of potentially toxic glycoalkaloids, particularly tomatine, and to a lesser extent solanine and chaconine, were also similar (Redenbaugh et al, 1992) and therefore this GM tomato was substantially equivalent to other non-GM tomato lines. However, to supplement these, several feeding studies were also performed by commercial laboratories at the request of the FDA.

2.1.B. Acute toxicity

First, range-finding, limit acute oral toxicity tests of the processed tomatoes in rats were carried out by IIT Research Institute of Life Sciences Department (10 West 35th Street, Chicago, IL 60616, USA). A single dose of the homogenates prepared from about 80 g of various GM and control tomatoes respectively was administered (15 ml/kg) by gavage to groups of Harlan Sprague Dawley rats (5 male or female rats per group) ad lib fed on rat chow for 14 days to establish whether the GM tomatoes were toxic or not. As claimed, no test substance-related mortalities occurred and increases in mean body weights were not significantly different between GM and control groups. However, as the range of the rat starting weights was unacceptably wide; female rats weighed 131 to 186 g (±18%) and male rats were between 159 to 254 g (±23%), in such a short (14 day) study with 5 rats per group it would have been difficult for significant differences to develop. For comparison, only a few per cent variation in starting weights is permitted in papers published in high-quality nutritional journals. Thus, the poor design of this feeding study largely invalidated the conclusions that GM tomatoes were not toxic. To supplement these three more rat feedings studies of similar design were carried out by International Research and Development Corporation (Mattawan, MI 49071, USA).

2.1.C. Twenty eight day toxicology/histology study
Of the three studies the most complete set of data are available for the second. In this, four groups of rats (20 males and 20 females per group) fed standard rat chow for 28 days were twice daily gavaged with homogenised tomatoes (15 ml/kg). Two groups were given GM tomatoes, CR3-613 or CR3-623 (CR3-623 is the commercial FLAVR SAVRTM tomato). There were two control groups, one of which was gavaged with the parent CR3 tomato homogenates and a second control group in which the rats were gavaged with water even though the relevance of this group is somewhat questionable. At the request of Calgene, an expert panel was retained (ENVIRON Corp, Arlington, VA, USA) to evaluate the data. They concluded that gavaging rats with GM tomato puree resulted in no significant changes in body weight, food consumption and clinical chemistry or haematology parameters in comparison with control tomatoes. However, there was a possible treatment-related increase in glandular stomach erosion/necrosis in four out of twenty female rats but none in the controls or in male rats at the end of the 28 day feeding period. The number of four female rats was increased to seven when the histology slides were re-scored by PATHCO, an independent pathology working group. This prompted a repeat study in which the dose of the tomato puree was increased by two-fold. Unfortunately, in this study some of the CR3 control and CR3-623 GM tomato lines were grown at different locations and harvested different times from those in the second experiment. However, this was not regarded as important by the expert panel even though when the same tomatoes were used as in the second experiment, the results appeared to show similar tendencies; two out of the fifteen females developed stomach glandular erosions with the GM while none found in the control females. However, in a not clearly understandable way the ENVIRON panel concluded that the lesion of glandular erosion was not related to the administration of GM tomatoes. According to them, such lesions occur spontaneously in animals stressed, or given mucolytic agents, food restriction or when animals are restrained in cages even though these parameters have not been systematically investigated. Moreover, none of these circumstances applied, since tomatoes contain no mucolytic agents, food was provided ad libitum, and the rats were not restrained. It was also suggested that because the lesions were possibly of short duration they were incidental, not related to the test material and would have healed spontaneously. Unfortunately, none of these assumptions were confirmed by further experimentation as no samples other than those at the end of the 28 day experiment were taken to probe into the timing and reversibility of the incidence of the stomach lesions. Clearly, the results of these three studies should have prompted more experimentation to investigate in more details the effect of GM tomatoes on stomach histology and, what is even more important, these studies should have been extended to include the possible effects of GM tomatoes on both the small and large intestines.

The red or dark red pin-point lesions present in the stomach of female rats which were described as necrosis would be termed 'erosion' in human pathology which may have sequelae such as life-endangering haemorrhage. Erosions cannot be termed 'mild' as unpredictable haemorrhage can occur in the elderly human, particularly on low dosage aspirin to prevent thrombotic events and synergy with transgenic tomatoes may occur. The assumption that the lesions are related to stress does not explain the low incidence in other groups, particularly in the second study. The relevance and significance of gastric erosions in the human may be a matter of life and death in the older age groups. It has been implied that pathologists in general might not report such a lesion but, in the present era of vexatious litigation, mention would have been made in any human pathologist's report to avoid an accusation of negligence. This may not be required in veterinary pathology but these rat studies were done with humans in mind and therefore the pathology findings must be put in this human context. It is probably true to suggest that these lesions are of short duration but the serious nature of erosive lesions should not be trivialised. This is the more serious because seven out of forty rats eating GM tomatoes died within two weeks. The nature of these deaths was not specified and the evidence that they were not related to the ingestion of transgenic tomatoes was inconclusive.

In a further development, the Scientific Committee on Food of the European Commission, Directorate C (Opinion, 2000) concurred with the conclusion reached by the US Food and Drug Administration, Memorandum (1994). In their opinion although the results showed an unexplained disparity, they were not supportive of a substance-related effect of the FLAVR SAVRTM tomato. However, it is likely that the EU Committee may not have seen all the primary data and their opinion therefore been based on incomplete evidence. It is also regrettable that by ascribing the gastric erosions in rats "as an artefact of gavage studies" the EU Committee has in fact labeled the scientists carrying out the work as incompetent. As these erosions were found at the end of a 28 day study during which 160 rats were gavaged twice daily with tomatoes, it is unlikely that even poorly trained workers should not have become more competent as to commit such a blunder.

2.1.D. Effects on body weight, food intake and organ weights

The conclusion of the ENVIRON panel that feeding rats on GM tomatoes (CR3-623) for 28 days had no effect on weight gain, feed intake and organ weights could not be justified because the starting weights of the rats were so widely different, a range of 130 to 258 g (±33%) for males and 114 to 175 g (±21%) for females, that finding significant differences in weight gain, feed intake and organ weights was not likely. Indeed, weight gains varied between wide limits (102 to 230 g for males and 46 to 127 g for females) in 28 days. Even under these conditions although the average starting weight of the male rats gavaged with CR3-623 GM tomatoes was the highest (148.1 g) their final weight (316.5 g) was the lowest. Accordingly, the rats gavaged with GM tomatoes grew the least of the four groups of rats. The feed intake of the different groups also varied betwen wide limits; 133 to 203 g for males and 102 to 153 g for females. Not surprisingly, the feed conversion efficiency (weight gain/total feed intake) of female rats on GM tomatoes (0.152) was significantly (p<0.05) less than that (0.167) obtained for female rats on control non-GM CR3 tomatoes.

The large range of starting weight differences also excluded the possibility of finding significant differences in the organ weights of the four groups of rats. The standard deviations of mean values were very large, in some instances more than 20%. It is the more remarkable that even under these conditions some differences in organ weights were found, including the testes for males and the thyroid/parathyroid for females. Finding of no significant differences in biochemical, haematology and ophthalmology parameters between GM and non-GM tomatoes was not unexpected either because the large initial body weight differences.

Overall, it is regrettable that these rat toxicological feeding studies were poorly designed as a great deal of effort, work and money must have spent on them and because so much rested on the outcome. The FDA's conclusion that FLAVR SAVRTM presented no more dangers to consumers than ordinary tomatoes does not therefore appear to rest on good science and evidence which could stand up to critical examination. Rather tellingly, the results of these studies have never been published in peer-reviewed journals. The study as described not only raises question marks about the design, methods and conclusions for this study but also whether they could have any general validity for other GM foods. In this light it is the more surprising that after these studies the FDA has required no nutritional/toxicological testing of other GM foods.

2.2. Aventis' Chardon LL herbicide-resistant GM maize
Due to the UK government's attempt to place Chardon LL seed on the National List a part of the supporting evidence submitted by Aventis contained data on the composition of two lines of seed to establish their substantial equivalence to the conventional parent maize line. The evidence also included the results of a 14 day rat feeding study. All this is to be found in a file deposited by the Ministry of Agriculture Food and Fisheries (MAFF) with the British Library, 'The British Library File' (1997).

2.2.A. Compositional analysis

In the absence of specifying the origin and conditions of cultivation of the different GM and non-GM samples strict scientific comparisons could not be made between them. However, even under these conditions, the composition of T14 and T25 GM maize expressing Phosphinothricin Acetyltransferase Enzyme (PAT-PROTEIN) showed many statistically significant differences in fat and carbohydrate contents in comparison with non-GM grain samples, and fat, protein and fibre between silage samples from GM and non-GM maize. Thus, the conclusion that GM maize is not "materially different" from current commercial varieties cannot be regarded as valid.

2.2.B. Repeated dose oral toxicity (14 day feeding) study in rats

The rationale for this study was to assess the cumulative toxicity of PAT-PROTEIN given to rats in their diet for 14 days and to provide a rational basis for toxicologic risk assessment in man. Although testing of the PAT-PROTEIN can be commended, this study was no substitute for the nutritional testing of the entire GM plant, seeds, vegetative parts and silage in all target animal species. Without these the potentially harmful, unintended and unpredictable effects of the gene transfer, other components of the vector and gene insertion (positioning effect) cannot be established or excluded.

Unfortunately as the design of the experiment was faulty it is difficult to draw valid conclusions from a feeding study carried out with 5 rats per group in which the starting weight of the rats varied by more than ± 20% (53 to 82 g for males and 50 to 74 for females) rather than the usual ± 2%. For any differences to reach significance they needed to exceed ± 20% and to achieve this in a 14 day study should have required catastrophic experimental conditions. The five rats per group were not housed singly and therefore their individual feed intakes could not be monitored even though the huge differences in the starting weights should have led to major differences in the feed intakes of the individual rats. Moreover, the group feed intakes were not measured continuously.

There were four groups of rats (five male/female rats per group) in the experiment. However, rats in group one were fed a different diet (full rat chow) from the other three groups and therefore for (statistical) comparisons this group one was not appropriate. The diet of the second group contained 5 g/kg and the third group had 50 g/kg PAT-PROTEIN mixed in with 45 g/kg and 0 g/kg respectively of commercial (SOJAMIN, KLIBA Muhlen AG) low soya bean protein diet (approximately 11% raw protein). The diet of the fourth group contained 50 g/kg SOJAMIN but no PAT-Protein. Thus, for statistical analysis the second and third groups ought to have been compared to rats in the fourth group. Curiously, although the main target organ of the PAT-PROTEIN fed to rats was the digestive tract (and pancreas), the weights of these were not measured. This is a major experimental design fault.

The starting weight and the feed intake of the third group (high PAT-PROTEIN) was the highest but they ended up with the lowest final body weight. This indicated an elevated metabolic activity probably induced by the PAT-PROTEIN. Our ANOVA analysis shows that the weight gain for both male and female rats on the high PAT-PROTEIN diet (group three; 65.2 and 43.6 g for males and females, respectively) was significantly (p<0.05) less than that of either the fourth group (control; 72.8 and 48.8 g for males and females, respectively) or group two (low PAT-PROTEIN diet; 73.4 and 44.4 for males and females, respectively). As PAT-PROTEIN reduced feed conversion efficiency it is potentially harmful. The conclusion that "there were no differences which could be attributed to treatment with the test article" was therefore not valid. Similarly, that "there were no changes on ...., clinical biochemistry and urine analysis after 14 days" is not valid either as the authors' own results described differences between the groups in glucose, cholesterol, triglyceride and phospholipid levels indicating an increased metabolic functional load in the rats. It is unexplained why these differences were dismissed by the authors as incidental and unrelated to the treatment. Our ANOVA analysis revealed that the urine output in rats on the high PAT-PROTEIN diet was significantly (p<0.05) reduced, indicating treatment-related effects (urine output of 5.4 and 4.4 ml for males and females in group three vs 7.1 and 6.5 ml for males and females, respectively in control, group four).

The large differences in the starting weight of the rats probably prevented finding of significant differences in organ weights. However, even under these conditions rats fed the high level PAT-PROTEIN diet (third group) had the lowest liver, thymus and spleen weights of all other groups (even though the differences with controls were not significant). This is of particular importance because the macroscopical findings indicated thymus foci in 20-40% of the animals fed diets containing the PAT-PROTEIN.

In conclusion: The design and execution of this feeding study was poor and, contrary to the authors' conclusions, the results indicated treatment-related effects induced by PAT-PROTEIN (of unspecified origin). The results therefore could not be taken as evidence that the transfer of its gene into maize represented no risk for the rat and, by inference, for humans, particularly as no gut histology studies have been completed so far. Finally, a recent publication (Chiter et al, 2000) showed that DNA survived in intact form or slightly fragmented unless the GM maize was extensively heat-processed. Therefore the possibility exists that with underprocessed maize products humans and animals might be exposed to the DNA used in the genetic engineering.

3. Compositional studies published in peer-reviewed journals

3.1. Herbicide-resistant soybean.

Befitting its importance in both human and animal nutrition, a great deal of attention has been given to the compositional analysis of herbicide-resistant and other GM soyabeans. Several publications appeared in nutritional and other journals demonstrating the compositional "substantial equivalence" of GM and non-GM soya. Thus, it was claimed that the macronutrient composition of glyphosate-tolerant soybean (GTS) seeds resulting from the transformation of conventional soybean with a gene encoding for 5-enolpyruvylshikimate-3-phosphate synthase from Agrobacterium to make the soy herbicide-resistant, was equivalent to that of conventional soybeans. This equally applied to GTS unsprayed with glyphosate (Padgette et al, 1996) or sprayed with this herbicide (Taylor et al, 1999). It was claimed that the results of proximate chemical analyses of the contents of crude protein, oil, ash, fibre, carbohydrates and amino acids of solvent-extracted and toasted or untoasted soybean meals of unsprayed GTS and control soybean had shown that all these lines were substantially equivalent (Padgette et al, 1996). Similar findings were described for sprayed GTS (Taylor et al, 1999). Although this appeared to be true for most macronutrients, several significant differences between GM and control lines, such as in ash, fat and carbohydrate contents were also found (Table 2; Padgette et al, 1996). However, these were not regarded to have biological significance.

A closer inspection of the data in the papers, however, revealed that the statistical comparison of the macronutrients of GM and non-GM lines was not scientifically valid. Instead of comparing their amounts in a sufficiently large number of samples of each individual GTS with its appropriate individual parent line grown side-by-side at the same location and harvested at the same time to establish whether they were "substantially equivalent", what the authors compared was a large number of different samples from different locations and harvest times. As growth conditions have a major influence on seed composition, the range of the amounts of constituents in the different samples, regardless whether GM or non-GM, was so great (± 10% or more) that the chances of finding statistically significant differences were unreal. It is possible that from a practical point of view the variation in protein concentration of samples of the three lines between 36.8 to 45% would fall into the normal range of agronomic variability of soybeans and therefore may not be of major concern for agronomists. However, this comparison is not strict enough to establish whether the genetic modification introduced any unintended compositional changes. What is remarkable that even with this approach many significant changes in macronutrient levels were found. Thus, the claim of "substantial equivalence" of GTS lines with non-GM soybean is not supported by rigorous scientific evidence.

The potential importance in human health of natural isoflavones, such as genistein, daidzein and coumestrol present in soybeans is generally recognised. It was, therefore of considerable interest whether any changes occurred in these components as a result of genetic modification. Here the published evidence is controversial. Thus, while in some studies no meaningful differences were reported (Padgette et al, 1996; Taylor et al, 1999), an independent study claimed that GM soya samples had consistently contained significantly less isoflavones than the parent cultivars (Lappe et al, 1999). In one respect all authors agreed, i.e. that the isoflavone content of soybean seeds showed considerable variability between sites and was dependent on agronomic conditions. However, Lappe at al (1999) went further and claimed that while the variability of the GM samples was indeed considerable, conventional soybeans showed less variation in isoflavone content. As the isoflavone content of soybeans might affect human health there needs to be more awareness of potential health problems due to this variability. While the precise details of the changes in isoflavone content on genetic modification will have to be established in the future, to assure clinical consistency the origin and the actual phytoestrogen levels in soybean may need to be standardized.

In the study by Padgette et al (1996) no significant differences were found in the levels of antinutrients, such as trypsin inhibitors, lectin and oligosaccharide flatulance factors between solvent-extracted, toasted or untoasted GM and non-GM soybean seeds. However, the comparisons were made by the same method as for macronutrients and therefore the large range of natural variability excluded the possibility of finding significant differences. Interestingly, in single soybean meal samples of each of the two GTS and parent lines the trypsin inhibitor (also a major allergen in soybean) content was substantially higher, by almost 30%, in one of the two GTS lines, with a smaller increase in the other. No trypsin inhibitor analyses were performed on the protein isolate or protein concentrate samples originating from the meal samples. Although there were other compositional differences in these processed soybean products it is difficult to decide from single determinations whether these were significantly different or not.

In conclusion, there is insufficient evidence to date to decide whether the composition of GM and conventional soybeans is equivalent or not. In fact, some data, particularly those for phytoestrogens, were significantly different. Furthermore, because as not strictly comparable compositional data were used, the case for equivalence was not properly established. There is therefore an obvious need for further more critical studies.

3.2. GM-potatoes

Brief references to GM potatoes, particularly those expressing Bt toxin, can be found in non-peer-reviewed book chapters or other articles. In most instances these contain no data and are therefore of little scientific value. There are two exceptions, one of which is an article on the safety assessment of GM potatoes expressing the soybean glycinin gene (Hashimoto et al, 1999). However, it is not quite clear what the authors wanted to achieve because at the expression level of glycinin in potatoes of between 12-31 mg/g total soluble protein, no significant improvements in the protein content or amino acid profile could have been expected. Indeed, the results in the paper demonstrated that the total protein content of the GM potatoes appeared to be significantly less than that of the control line and that no improvement in the essential amino acid profile was achieved either. There appeared to be substantial differences in some vitamins between GM and control lines and the amounts of both solanine and chaconine increased in the GM lines. It is, therefore, not quite clear why was it claimed by the authors that their GM lines were equivalent to the parent line and could be utilised as safely. The other more recent study is a conventional compositional analysis of some macro and micro nutrients of tubers from insect and virus resistant potato plants (Rogan et al, 2000) performed by methods which are currently accepted by most novel food regulatory bodies. Although these showed some significant differences in a number of tuber constituents, in the absence of toxicological/nutritional animal studies it is difficult to ascertain whether these differences could have any biological effects on humans/animals, particularly as these conventional analyses could not have revealed the development of any unknown possible toxic/antinutritive components. Additionally, known antinutrients, such as lectins or enzyme inhibitors were not included in the analysis.

3.3. GM-rice

GM rice lines expressing the soybean glycinin gene have been developed (Momma et al, 1999) by a method similar to that used for GM potatoes. The glycinin expression level was between 40-50 mg glycinin/g total rice protein. The GM rice was claimed to contain 20% more protein but its moisture content was less than that of the parent line. However, from the paper it is not quite clear whether the increased protein content was due to the decreased moisture content of the seeds because it was not specified whether the values were expressed for air-dried or fully dried seeds. Thus, most of the arguments in the discussion whether the higher protein level was due to the positioning effect of gene insertion or metabolic interference will have to await clarification by further work.

3.4. GM-cotton

Several lines of GM cotton plants have been developed using the gene encoding an insecticidal protein from Bacillus thuringiensis subsp. kurstaki. These had increased protection against the major lepidopteran insect pests of cotton. As cottonseed is an important source of oil for human consumption and cottonsed and processed cottonseed meal for animal feed, extensive analytical work has been done to establish whether the GM lines were "substantially equivalent" to conventional lines (Berberich et al, 1996). The levels of protein, fat, carbohydrate, moisture, ash, amino acids and fatty acids in the insect-protected lines were claimed to be comparable to those found in commercial varieties. Moreover, the levels of antinutrients such as gossypol, cyclopropenoid fatty acids and aflatoxin were similar or less than those in conventional seeds. Thus, the GM varieties were suggested to be equivalent to conventional seeds and just as nutritious. However, the statistics used by the authors was identical to that used with glyphosate-resistant soya and therefore could be similarly criticised. Although the content of known constituents fell in between the wide range of values of commercial conventional lines this did not mean that they were compositionally equivalent, particularly as environmental stress could have major and unpredictable effects on antinutrient and toxin levels (Novak and Haslberger, 2000). Thus, without animal experimentation this approach could not reveal whether any new and unknown toxins/allergens had been created or not.

3.5. GM-maize (corn)

A glyphosate-tolerant (Roundup Ready) corn line GA21 has recently been developed. It was claimed (Sidhu et al, 2000) that, except for a few minor differences which the authors think are unlikely to be of biological significance, the results of compositional analyses of proximate, fibre, amino acid, fatty acid and mineral contents of the grain and proximate, fibre and mineral contents of forage, collected from 16 field sites over two growing seasons showed that control and GM lines were comparable. The comparison was carried out by a method similar to that described for GTS soya (Padgette et al, 1996) and this may therefore not be scientifically rigorous for the establishment of substantial equivalence.

4. Nutritional/toxicological studies published in peer-reviewed journals

4.1. Herbicide-resistant soybean

As part of a safety assessment of glyphosate-resistant soybean (GTS), the feeding value, wholesomeness (Hammond et al, 1996) and possible toxicity (Harrison et al, 1996) of two major GM lines of GTS was compared to that of the parent line. Processed GTS meal was included in the diets of rats, broiler chickens, catfish and dairy cows at the same concentrations as in commercial non-GM soybean rations. Rats and dairy cows were fed these diets for four, broilers for six and catfish for ten weeks. It was claimed that in rats, catfish and broilers the growth and feed conversion efficiency, in catfish the fillet composition, in broilers the breast muscle and fat pad weights and in dairy cows milk production and composition, rumen fermentation and digestibilities were similar for both GTS and parental lines. According to the authors these results confirmed that the GTS and parental lines had similar feeding values.

4.1.A. Rat studies

A critical evaluation of the rat study was hampered by the lack of adequate primary individual data in the paper. Thus, there was no full description of the rat diet. It appears that the total protein content of the diets was adjusted to 24.7 g protein/100 g diet to be isonitrogenous with Purina Laboratory Rat Chow by the addition of 24.8 g of GTS and parent soybean meals respectively (about 10% protein) to a base diet. All comparisons were made to rats fed commercial Purina Chow. The protein concentration in these diets was, however, appreciably higher than the usual 10-16% crude protein and exceeded the protein requirements of the rat. This extra protein could have potentially masked any possible transgene product effects, particularly with the raw unprocessed soybean diets in which the GM meals were incorporated only at the level of 5 or 10% of the diet. Thus, these meals only replaced 8.5 and 17% respectively of the totaql protein of 24.7 g/100 g diet. In other words the GM soybean protein in these meals was diluted by other dietary proteins by 12 and 6 fold, respectively, producing another possible masking effect. The composition of the control Purina Chow diet in the ground raw soybean feeding study was not described. This is important because the identity of the raw control soybeans included in the Purina Chow control diet was not specified.

In the feeding study 4 groups of rats (10 males and 10 females in each group) singly housed were fed diets containing the parental line or the GTS lines (40-3-2 or 61-67-1) for 28 days. No individual values (or their ranges) for feed intake or body weight were given. The bar diagrams of the combined body weight of rats at the end of each week of the four week experiment were rather uninformative. However, it was observed by the authors that the Purina Chow-fed male rats grew significantly better than the three experimental groups fed toasted soybeans (including the parental line). This was attributed to better commercial processing. However, the bar diagrams also indicated that the growth one of the GTS lines (61-67-1) was probably equal to that of the Purina Chow-fed control and therefore, by inference, these rats did also grow significantly better than the other two experimental lines (the GTS line 40-3-2 and the parental line). This again underlined the importance of giving individual data in papers without which it is difficult to assess the results. Similarly, there were no individual data for organ weights, such as liver, kidneys and testes. However, it was claimed that the kidney weights of the raw GTS line-fed (and parental control?) male rats were significantly higher than those of the controls, while the testes of the parental line-fed rats was significantly enlarged. According to the authors as these differences were neither dose-related or were only shown by the parental line, they were not caused by genetic modification. Rather curiously, the weights of the stomach and intestines, the main target organ in any nutritional testing, were not recorded. Observations were not recorded on other organs and no histology appears to have been done on these tissues either. The only tissue which was subjected to microscopy was the pancreas. However, the description of the findings was qualitative. Only minimal to mild lesions were found and these were claimed to be common to all groups. Under the conditions, however, it was not surprising because no pancreatic hypertrophy was found. This was probably due to the effect of the unusually high dietary protein concentration which, as the authors pointed out, masked and/or diluted the biological effect of the trypsin inhibitors. This is of particular concern because the trypsin inhibitor content of GTS lines in unprocessed soybean was significantly higher than in the control line (Padgette et al, 1996).

It is regrettable that the design of this important rat feeding study had such unfortunate omissions. It is of particular concern that no histology have apparently been carried out on gut tissue. Thus, more critical work is needed to decide whether the feeding value of GM and non-GM soybeans is equal or not.

4.1.B. Chicken study

The broiler chicken feeding study's experimental design closely followed that of commercial practice and therefore the results should only be indicative of the commercial feeding and production value of the various soybean lines. As the data were pooled from all birds fed on the same diet, it is not easy to see what, if anything, was the significance of the small differences in the study such as the slightly lower body weights, breast and fat pad weights obtained with the GTS lines (particularly with GTS 40-3-2) for the utilisation of GM soybean. It would have been preferrable to measure the nutritional performance of individual birds (or small groups) fed on different diets and then compare them after statistical analysis. In the absence of this we have to rely on the authors' conclusion that the design of the experiment gave the upper limit of differences in weight gain of 3.5% and gain/feed ratio of 2% and that the GTS lines vs parental line was within this limit. Thus, with this restriction, the feeding value of the GTS lines for broilers was practically equal to that of the parental soybean line.

4.1.C. Catfish experiment

Catfish are excellent and highly sensitive indicators for the feeding value of diets. It was obvious from the results that, similar to the findings with rats, one of the GTS lines, 61-67-1, was superior to the other lines (GTS 40-3-2 and the parental line) in most respects. Thus, fish on GTS line 61-67-1 ate more, had better weight gain and gain/feed ratio and weighed more at the end of the 10 weeks' study than the others even though the composition of the fillets from these fish were not significantly different. This significant difference in performance must, therefore indicate that genetic modification may not be as reproducible as it has been claimed and that the feeding value and metabolic effects of GM and parent lines are not always "substantially equivalent".

4.1.D. Study on lactating cows

Milk production and composition and performance data in the lactating cow study showed some significant differences between cows fed diets containing the different lines of soybean, indicating a lack of "substantial equivalence". In view of these differences, even though we may not at present know all their biological/nutritional consequences, it may be difficult to maintain the view that the feeding value of the GTS and parent lines is equal and further work is needed to establish whether the GTS lines are safe or not for humans/animals.

4.1.E. Testing of E. coli recombinant gene product

Extensive studies have been carried out to ascertain the safety of the gene product, 5-enolpyruvylshikimate-3-phosphate synthase (CP4 EPSPS) which renders the soybeans glyphosate-resistant (Harrison et al, 1996). Unfortunately, there are some flaws in these experiments, the most important of which is that in the acute gavage studies the authors did not use the enzyme isolated from GTS lines but E. coli. Although they were at pains to show that the EPSPS enzyme samples from the two sources were similar in lack of glycosylation, molecular size, reaction with a polyclonal anti-EPSPS antibody and enzyme assays, these methods have no sufficient power to show unequivocally whether they were identical. The authors themselves pointed out that post-translational modification of the completed polypeptide chains emerging from the ribosomes may be done differently in two such evolutionary distinct life forms as higher plants and prokaryotic bacteria. Amidation, acetylation and proteolytic processing can have such major effects on the conformation of the protein as to make these gene products to behave differently in the digestive system. Thus, the use of the E. coli recombinant protein for the acute mice gavage studies may invalidate the authors' conclusion that the gene product from soybean did not have any toxic effects. These studies must be re-done with the gene product isolated from the transgenic plant before the results could be accepted. In any case, in such gavage studies young, rapidly growing animals must be used to show any distinct effect on growth. As all animal weights were unchanged in the experiment the test system used could not have detected any effect unless the consequences of the gavaging had been disastrous. Feeding studies with the gene product in young rapidly growing rodents should be the preferred method for the demonstration of the deleterious effects.

The other flaw in the experimental design was the reliance on an in vitro simulated gastric/intestinal digestion assay which was also carried out with the E. coli recombinant gene product. To have physiologically valid results it would have been necessary to use the gene product isolated from GM soybean in an in vivo assay in the rat (or other suitable animals; see Rubio et al, 1994) or a full feeding trial. Thus, it has been shown before that the kidney bean (Phaseolus vulgaris) alpha-amylase inhibitor is fairly stable to proteolytic degradation in the rat gut (Pusztai et al, 1995, 1999), but when its gene was expressed in peas (Pisum sativum), it was rapidly digested and inactivated in the rat stomach/small intestine in vivo (Pusztai et al, 1999). This may have contributed to the safety of GM peas for rats and, by inference, possibly for other monogastric mammals. Thus, in vitro digestion assays may have little relevance to the safety of GM foodcrops.

In a separate feeding study (Teshima et al, 2000) the possible harmful effects of toasted glyphosate-resistant GM soybean were investigated at 30% inclusion level in the diet of rats and mice. After feeding these animals for 15 weeks no significant differences in nutritional performance, organ development, histopathology of the thymus, liver, spleen, mesenteric lymph nodes, Peyer's patches and small intestine and the production of IgE and IgG humoral antibodies between GM and non-GM line diets were found. However, as rats grew less than 30 g and mice none in 15 weeks, the conditions were so unphysiological that no valid conclusions could be drawn from these experiments.

4.2. GM corn

In a major commercial-scale broiler chicken feeding study with rations containing transgenic Event 176 derived Bt corn involving 1280 birds (Brake and Vlachos, 1998) it was claimed that no statistically significant differences in survival or bird weights between birds fed diets containing GM corn, Event 176, or an isogenic parent corn line, were found. Indeed, birds fed GM corn rations appeared to have significantly better feed conversion ratios and an improved yield of breast muscle. However, the authors cautioned against the conclusion that this enhanced performance could be attributed to the Bt corn per se. It is possible that the results might have been due to slight differences in the overall composition of the diets. This is reasonable considering the length of this study and possible problems of consistent diet preparation on a commercial scale. Minor differences in composition such as the slightly lower protein content of the GM corn and fat contents of the diets magnified to the scale of this trial make the results more relevant to commercial than to academic scientific studies.

In a poultry feeding study it was claimed that the GA21 Roundup Ready corn based diets gave similar performance data in growth, feed efficiency and fat pad weights to diets containing the parental control line (Sidhu et al, 2000). However this and a similar study carried out in Germany with a maize line expressing the PAT protein (Flachowsky and Aulrich, 2001) were commercial production experiments and made little contribution to scientific safety assessment.

In a separate study corn was genetically modified by the transfer of the gene of egg white avidin to make the seed resistant to storage insect pests (Kramer et al, 2000). It was also claimed that this GM corn was safe for mice as apparently when, instead of a balanced diet, they were solely fed on this crop, the mice suffered no ill effects. However, the mice used in the experiment were adults which did not grow at all and therefore the conclusion that the GM corn was safe is, at best, premature.

4.3. GM peas

Diets containing transgenic peas expressing the transgene for insecticidal bean alpha-amylase inhibitor (about 3 g/kg peas!) at two different inclusion levels in the diet, 30% or 65%, were subjected to nutritional evaluation with rats in a 10 day feeding trial (Pusztai et al, 1999). The nutritional performance of rats fed GM pea diets were compared to those obtained with rats pair-fed iso-proteinic and iso-energetic diets containing parent-line peas and also lactalbumin diets spiked with isolated bean and pea alpha-amylase inhibitors, respectively. At 30%, but not at 65% inclusion level, the nutritional value of diets containing transgenic or parent peas was not significantly different. Even at 65% level the difference was small mainly because the transgenically expressed pea recombinant alpha-amylase inhibitor was quickly (in less than 10 min) degraded in the rat digestive tract and therefore its antinutritive effect abolished. In contrast, spiking the parental line pea diet with the stable bean alpha-amylase inhibitor reduced its nutritional value (Pusztai et al, 1995; Pusztai et al, 1999).

In this study unfortunately no gut histology was done nor lymphocyte responsiveness measured and therefore one had to rely on the evaluation of nutritional parameters which are inherently less sensitive to find possible differences in metabolic responses between GM and conventional food components. Although there were significant differences in the development of some organs, mainly in the caecum and pancreas, most organ weights were remarkably similar. At the end of the study cautious optimism was expressed that GM peas may be used in the diets of farm animals, particularly at low/moderate levels recommended in commercial practice and if the progress of the animals will be carefully monitored. However, this relatively short feeding study with modest objectives cannot at this stage to be taken as proof of the safety of GM peas for human consumption. There is a need to carry out further and more specific risk assessment testing procedures which must be designed and developed with human consumers in mind. It also has to be kept in mind that only one particular line of GM peas was tested in which the endogenous antinutrient levels were selected to be similar to those of the parent peas. In some other GM lines, however, lectin levels could vary, up or down, by a factor of four. Moreover, in some field pea cultivars, such as Laura, the concentration of trypsin inhibitor increased by about 24%, the chymotrypsin by 100% while the haemagglutinating activity decreased by a factor of four in the GM line compared with its parent (Pusztai, unpublished). This strengthens the need that in the safety assessment of GM crops many lines should be included and that from the results of a single GM line no blanket approval should be given to other lines developed even if in the transformation the same vector was used and carried out at the same time.

4.4. GM potatoes: There have been four independent studies on different GM potatoes.

4.4.A. Glycinin-expressing potatoes
The safety of transgenic potatoes expressing the soybean glycinin gene was evaluated in a short (4 week) rat feeding study (Hashimoto et al, 1999). With an interesting experimental design, control rats and the experimental groups were fed the same control commercial diet. However, the rats were also daily force-fed by gavage with 2 g of respective potato lines/kg body weight. The potatoes used were a parental control line and two transformed lines, one with the glycinin gene and another one with a designed glycinin gene (coding for four additional methionines in the gene product), respectively. However, there were a number of problems with this study. Thus, although no difference in growth, feed intake, blood cell count and blood composition and internal organ weights between the groups was found, the uncertainty whether the animals were fed with raw or boiled/baked potatoes leaves a question mark over the interpretation of the results.

4.4.B. Bt toxin potatoes

An interesting, mainly histology study was carried out on the ileum of mice fed with potatoes transformed with a Bacillus thuringiensis var. kurstaki CryI toxin gene. As a control, the effect of the toxin itself was also investigated (Fares and El-Sayed, 1998). It was shown that both the delta endotoxin and, to a lesser extent the Bt-potato, caused villus epithelial cell hypertrophy and multinucleation, disrupted microvilli, mitochondrial degeneration and increased numbers of lysosomes and autophagic vacuoles and activation of crypt Paneth cells. As a result it was recommended that "thorough tests of these new types of genetically engineered crops must be made to avoid the risks before marketing". Unfortunately, some flaws in the experimental design detract from the strength of the conclusions. The most important of these was that, apart from indicating that the gene used in the transformation was the CryI gene from B. thuringiensis var. kurstaki, there was no description of the Bt potatoes. The gene expression level in the GM potato was not given and it was not clear whether the potatoes in the diet were cooked/baked or raw. Moreover, the amount of the Bt toxin used for supplementing the potatoes with in the control potato diet was not specified either. This made it impossible to make a quantitative comparison of the effects on the ileum of the Bt potato with the spiked control potato diets. The assumption that the ileum is the most important absorptive part of the rodent small intestine could also be argued because 90% of all nutrient absorption in fact occurs in the jejunum. As this was an electron microscopy study, the fixation of the ileal samples was not done on well-oriented sections but on chopped up fine tissue pieces and important detail of villus organisation was therefore lost. Finally, the delta-endotoxin induced hyperplastic changes on ileal villi should have been demonstrated by measuring cell proliferation and mitotic rates in ileal (and jejunal) crypts rather than on the villi. However, despite these shortcomings this study has once for all established that, in contrast to general belief, exposure of the mouse gut (ileum) to the CryI gene product has caused profound hypertrophic and hyperplastic changes in cells of the gut absorptive epithelium and can lead to mucosal sensitization (Vazquez Padron et al, 1999; 2000b). This shows up the fallacy of drawing comforting conclusions from in vitro simulated gut proteolysis tests. Clearly, concerns about the possible biological consequences of exposure to GM food should be addressed under in vivo conditions because even if an E. coli product breaks down in vitro, this does not necessarily mean that the same gene product expressed in the transgenic crop should also break down.

4.4.C. GNA-GM potatoes

Some of the results rat feeding studies with GM potatoes expressing the snowdrop (Galanthus nivalis) bulb lectin (GNA) gene were similar to the results of Fares and Al-Sayed (1998). A part of this work concerning the effect of GNA-GM potatoes on the histology of different compartments of the rat gut was published (Ewen and Pusztai, 1999b). Although this peer-reviewed scientific paper was criticised by some, most of the criticisms were unpublished personal opinions. Moreover, most of the published criticisms (e.g. Kuiper et al, 1999) were adequately answered (Ewen and Pusztai, 1999a). Some selected results of the nutritional/metabolic studies were, against the wishes of the authors, placed on the website of the Rowett Research Institute ( where most of the work was done (Bucksburn, Aberdeen, Scotland UK) . However, as not to jeopardize their eventual proper publication these results will only be briefly mentioned.

Young, rapidly growing rats (starting weight of 84 ± 1 g) were strictly pair-fed on iso-proteinic (60 g total protein/kg diet; most of which was from potatoes) and iso-caloric diets (in contrast to that described in Kuiper et al, 2001) supplemented with vitamins and minerals for 10 days. The test diets contained GM potatoes either raw or boiled. The control diets contained the same amount of parental line potatoes (raw or boiled) alone or supplemented with GNA at the same concentration as expressed in the GM potatoes. A positive control group of rats was also included in the experiment which were fed a lactalbumin-based high-quality control diet to check for any potential problems in rat behaviour and experimental conditions. As part of the nutritional/metabolic evaluation, samples of stomach, jejunum, ileum, caecum and colon have been taken, fixed and stained with haematoxylin and eosin for full quantitative histological evaluation (Plates 1,2 & 3) or reacted with GNA antibody and subsequently stained using a PAP (peroxidase - antiperoxidase) method to establish whether any GNA was bound to the epithelial surface (Plate 4). By measuring the mucosal thickness of the stomach and the crypt length of the intestines (Ewen and Pusztai, 1999b) it was shown that proliferation in the gastric mucosa was in part caused by GNA, the gene product. However, the growth-promoting stimulus on the small intestine of diets containing GM potatoes leading to crypt enlargement and a part of the stomach enlargement was not a GNA effect. As shown before and confirmed here, there was a slight binding of GNA to the small intestinal epithelium (Plate 4). However, GNA is not a mitotic lectin and therefore it did not induce hyperplastic growth in this tissue (Pusztai et al, 1990). Accordingly, the jejunal growth was probably due to some so far unknown effects of other parts of the genetic construct used for the transformation or the genetic transformation itself. Hyperplasia was previously shown by measuring the increase in crypt length (Ewen and Pusztai, 1999b). However, similar results were obtained by measuring the increase in crypt cell numbers (Table 1) and crypt mitotic figures (not fully significant) in the jejunum of GM potato-fed rats (Table 2). The results suggested that it is possible that crypt hyperplasia and increase in epithelial T lymphocyte infiltration observed with GM potatoes might also happen with other GM plants which had been developed using the same or similar genetic vectors and method of insertion. It is therefore imperative that the effects on the gut structure and metabolism of all GM crops should be thoroughly investigated as part of the regulatory process before their release into the human food chain.

4.4.D. Potatoes expressing cationic peptide chimeras

Desiree and Russet Burbank potatoes expressing N-terminus modified cecropin-melittin cationic peptide chimeras and control line potatoes fed to mice caused severe weight loss. As the animals did not grow even after supplementing these potatoes with Rodent Laboratory Chow. Apparently, mice fed with tubers from transgenic potatoes were as healthy and vital (sic) as those from the control group and their faecal pellets were comparable. The severe weight loss seriously questioned the value of the results of this poorly designed feeding experiment.

4.5. GM tomatoes

Finally, an important study will have to be described even though it was not published in a peer-reviewed journal but the ideas and experiments described had some influence on the development of GM regulation (Noteborn et al, 1995). Thus, a new laboratory GM tomato line was developed using the B. thuringiensis crystal protein CRYIA(b) gene but, instead of the cauliflower mosaic virus 35 s promoter (CaMV 35 s) which is used in practically all first generation GM crops, a potentially safer plant promoter was used. Although with this the expression level of the Bt toxin was only about 1/20th of that found with CaMV 35 s, this might be improved upon in future. In contrast to most other studies with GM crops, there was a commendable attempt to measure the binding of the gene product to the rat gut surface in vivo rather than using spurious arguments why the gene product should not bind. Although no in vivo binding was found, this should not detract from the significance of this initiative because, due to the lack of availability of sufficient quantities of Bt toxin isolated from GM tomatoes, an E. coli recombinant and potentially less stable form of the gene product was used in the experiment and its possible degradation in the gut may have accounted for the lack of binding. However, Bt toxin was shown by immunocytochemistry to bind to gut sections, including the caecum and colon, from humans and rhesus monkeys in vitro . Unfortunately their short term toxicity testing in mice (and rabbits) and the in vitro simulated proteolysis assasys were also carried out with the E. coli recombinant gene product and therefore their conclusions of finding no toxic effects may not be valid. Commendably, the authors carried out a 91 day feeding study with rats using freeze-dried GM vs parent line tomatoes which were included at a 10% level in the diets but no differences in food intake or body and organ weights were found. However, because the Bt toxin expression level in the tomatoes was low, the daily intake of the gene product(s) by the rats was also low. Moreover, the daily input of tomato proteins was only about 5-6% of the total dietary protein intake of the rats it was somewhat optimistic to expect any significant changes in these nutritional parameters. To have any reasonable chances to show up small differences in the nutritional value of GM vs parent line crops it would have been important to use as high a protein concentration as possible such as in the 110 day GM potato feeding study carried out at the Rowett ( in which the GM protein in the diet was only diluted twofold by other dietary proteins, and this allowed to show up the significant differences in the growth rates of rats fed on baked GM potato diets vs parent potato diets. In fact, to equalize the growth rates of the rats on the GM potatoes to that of the controls the GM diet had to be supplemented with extra 12 g lactalbumin/kg diet and this extra protein gave a quantitative measure of the difference of the nutritional value between GM and non-GM potatoes. Even at these similar growth rates the weights of some of the rats' vital organs, such as the gut and particularly the small intestine, the liver and kidneys were still significantly different.

There were other omissions in the Bt tomato study, the most important of which was that no Bt toxin survival was measured in the gut lumen and no gut gut histology was done to see if there was any Bt toxin-binding to or possible structural changes in the gut epithelium or whether lymphocyte infiltration occurred. This omission is particularly important because later studies had shown that the similar Bt toxin Cry1Ac could bind to gut epithelial cells in mice (Vazquez Padron et al, 2000) and induce mucosal antigenic sensitization (Vazquez Padron et al, 1999, 2000). The allergenic potential of Bt tomatoes was not investigated either. However, despite some of its shortcomings this study showed many novel and commendable features which after some improvements may, hopefully, be incorporated into the general GM food testing procedures.

5. Allergenicity

One of the major health concerns with GM food is its potential to increase allergies in the human population through the food chain. The possibility of fatal anaphylaxis in sensitised individuals after their unwitting exposure to allergenic proteins in unlabeled GM foodstuffs is a real danger. When a gene is transferred from a source of known allergenic potential, the assessment of the allergenicity of the GM crop is relatively straightforward. This can be done using in vitro tests with sera from individuals sensitised to the allergen from the original source. Similarly, it is relatively easy to assess the effect of genetic engineering on endogenous allergens in crops with some evidence of allergenicity. With tests, such as RAST, RAST inhibition, immunoblotting, the allergenic potential of the GM crop is easily measured. There are now several examples for these, such as the demonstration of the allergenicity of the brasil nut 2 S seed storage protein in transgenic soybean (Nordlee et al, 1996) or the codfish allergy in potatoes genetically engineered with cod protein genes to make the potatoes tolerate cold storage (Bindslev-Jensen and Poulsen, 1997). The claim that in glyphosate-tolerant soybean the introduction of the herbicide-resistance gene does not affect the allergenicity of the soy endogenous allergens is also a good example (Burks and Fuchs, 1995). Having been shown in a surveillance programme of farm workers before and after exposure to B. thuringiensis pesticide sprays that some developed skin sensitization and IgE antibodies to the Bt spore extract and that two of them had a positive skin-prick test, it may now be possible to test for the allergenicity of Bt toxins engineered into various crops (Bernstein et al, 1999). This is all the more important because the Cry1Ac toxin has now been shown to be a potent oral immunogen and adjuvant (Vazquez-Padron et al, 1999; 2000).

It is much more difficult to assess the allergenicity of GM foods when the gene is transferred from a plant whose allergenic potential is unknown. Moreover, it is also possible that as a result of the gene transfer or vector insertion a new allergen is developed or the expression level of a minor allergen is increased in the GM crop. The gene product can also have an allergenic adjuvant effect on a food component previously of low allergenic potential or some component in the GM food may have an adjuvant effect on the allergenicity of the transgene product. Unfortunately, while there are good animal models for nutritional/toxicological testing, no satisfactory animal models have so far been developed for allergenicity testing (Helm and Burks, 2000). For the time being only indirect methods are available for the assessment of the allergenic potential of GM foods derived from sources of unknown allergenicity. There are a number of recommended approaches to be followed. A useful preliminary step is to establish if there are any sequence homologies in the transgenic protein to any of the about 200 known allergens. If there are, in vitro tests for IgE reactivity need to be performed. It is thought that the peptide length in the transgenic protein which is optimally needed for binding B cell epitopes requires the presence of at least eight contiguous identical or similar amino acids. Unfortunately, the amino acids in the allergenic epitopes are rarely contiguous. Moreover, the absence of a positive reaction in in vitro testing does not guarantee that the transferred protein is not an allergen. In a decision-tree type of indirect approach the next step is to consider the molecular size, glycosylation, stability, solubility and isoelectric point of the transgenic protein and compare it with that of known allergens (O'Neil et al, 1998). Unfortunately, in most studies to date the all important stability of the transgenic protein to gut proteolysis is established in an in vitro simulated gastric/intestinal system (Astwood et al, 1996, Metcalfe et al, 1996) and this is fundamentally flawed. The results therefore at best are misleading and at worst erroneous. Reliance on the concept that most allergens are abundant proteins is probably also misleading because for example Gad c 1, the major allergen in codfish is not a predominant protein (Bindslev-Jensen and Poulsen, 1997).

When the gene responsible for the allergenicity of a crop is known, its cloning and sequencing opens the way for its reduction by antisense RNA strategy. Thus, in rice the low molecular weight alpha-amylase/trypsin inhibitors are major allergens. A part of the genomic sequence encoding this protein in an antisense direction was constructed between the promoter of the rice allergen gene and its waxy terminator and this was introduced into rice protoplasts. The allergenicity of the regenerated plants was significantly less than that of parental wild type rice (Nakamura and Matsuda, 1996).

In conclusion, allergenicity testing appears to be one of the Achilles heels of GM food safety. It is clear that if and when it is known that the protein gene is derived from a source with a history of allergenicity, there is a reasonable certainty that the GM crop will be allergenic. Unfortunately, the reverse is not true: the use of a gene from something that is not allergenic will not guarantee that the GM crop will not possess allergenicity. In the absence of new and reliable methods for allergenicity testing, particularly the lack of good animal models, at present it is almost impossible to definitely establish whether a new GM crop is allergenic or not in advance of its release into the human/animal food/feed chain.

6. Final conclusions One has to agree with the opinion expressed in Science (Domingo, 2000) that there are many opinions but very few data on the potential health risks of GM foods even though research to exclude such risks should have been carried out before the GM crops were introduced into the food chain. Our present data base is therefore woefully inadequate. This is clearly seen from a closer scrutiny of the reference lists of recent reviews which contain only a handful of toxicological/nutritional and immune studies of GM foodcrops published in peer-reviewed science journals (Kuiper et al, 2001; Betz et al, 2000; Ruibal-Mendieta and Lints, 1998; Pusztai, 2001). Moreover, the scientific quality of even what is published is, in most instances not up to the standards that ought to be expected. In this review data published in peer-reviewed and some non-peer-reviewed journals has been examined in detail. However, as our future is claimed to be dependent on the success or failure of the promise of genetic modification delivering GM foods which will be wholesome, plentiful and, most importantly, safe for us all, the emphasis was on strict but fair criticism. From the results the conclusion seems inescapable that the present crude method of genetic modification has not delivered GM crops that are predictably safe and wholesome. The promise of a superior second generation of GM crops is still in the future. It is possible that some of the first generation of GM crops may superficially satisfy some commercial endpoints, such as their use in broiler chicken production. However, we need to consider that these GM feed ration-fed animals will eventually be consumed by humans and there is absolutely nothing known about the potential hazards (if any) on human health of this indirect exposure to GM food. Furthermore, the examples in the papers highlighted some differences even between such crude things as macronutrient composition of GM and conventional lines. It is argued by some that these differences have little biological meaning. However, it was clear that most GM and parental line crops would arguably fall short of the definition of "substantial equivalence". This crude, poorly defined and unscientific concept outlived its possible previous usefulness. There is an urgent need to come up with novel scientific methodologies to probe into the compositional, nutritional/toxicological and metabolic differences between GM and conventional crops if we want to put this technology on a proper scientific foundation and also to allay the fears of the general public. We need more science and not less. For proper safety assessment our first concern ought to be to establish on a case-by-case basis the impact of components of GM foods on the digestive system, its structure and metabolism, because the way our body will respond to GM foods will be pre-determined at this level. According to The Royal Society (1999) we need "to refine the experimental design of the research done to date". New ideas were also advocated in the Lancet debate (Noteborn et al., 1999; Ewen and Pusztai, 1999a) and at the OECD Conference in Edinburgh in February 2000.

7. Recommendations

7.1. Main tasks and methods for safety assessment of GM crops

7.1.1. For compositional analysis and comparison the parent and transformed lines must be grown under identical conditions, treated and harvested the same way. In addition to proteins, starch, lipids, etc, of the parent and GM-lines, their contents of bioactive components should also be compared by novel methods (proteomics, fingerprinting, etc).

7.1.2. The stability to degradation by acid or pepsin or other proteases/hydrolases of GM-products, foreign DNA, including the gene construct, promoter, antibiotic resistance marker gene, etc, has to be established in the stomach and intestines of model animals in vivo. With GM-lectins, including Bt- (Bacillus thuringiensis) toxin the presence/absence of their epithelial binding should also be demonstrated by immunohistology.

7.1.3. The biological, immunological, hormonal properties and allergenicity of GM-products must be established with the GM-product isolated from the GM-crop and not with recombinanst from E. coli as these two may have substantially different properties.

7.1.4. As GM-food is unlikely to be highly poisonous, "toxicity" is an unhelpful concept and difficult to assay. In contrast, nutritional studies in which GM crop-based diets are fed to young growing animals should reveal their possible harmful effects on metabolism, organ development, immune/endocrine systems and gut flora which together determine the safety of the GM-crop and the development of the young into healthy adults.

7.1.5. For animal testing iso-proteinic and iso-energetic diets need to be formulated in which most of the dietary protein is derived from the GM-crop. The composition of the control diets should be the same as the GM-diet but containing the parent-line with or without supplementation with the isolated gene product at the same level as expressed in the GM-line. Groups of animals (5 or more per group) of similar weight, should be pairfed in short- and long-term experiments. Urine and faecal samples collected for the determination of Net Protein Utilization (NPU), Nitrogen Balance, and Feed Utilization Ratios. Blood-samples should be taken before, during and at the end of the experiments for immune studies (i.e. lymphocyte proliferation assay, Elispot), hormone assays (insulin, CCK, etc) and for the determination of other blood constituents. The animals are to be weighed daily and any abnormalities observed. After killing the animals, bodies are dissected, gut rinsed and its contents saved for further studies (enzymes, GM-products, DNA), sections taken for histology, wet and dry weights of organs recorded and analysed.

7.2. Evaluation

With suitable statistical analyses (ANOVA, multiple comparisons and/or multivariate analysis) the significance of differences, if any, in the parameters should be established:

If differences between animals fed GM- and parent-line diets indicate that the genetic modification must have had a significant effect on utilization and nutritional value, the GM-crop cannot be accepted for inclusion in the human/animal diet.

If similar to the GM-diet, the parent-line diet spiked with the gene product shows differences, the use of this gene in GM-food/feed is not acceptable.

If negative effects are not observed with the parent line diet containing the isolated gene product, it is likely that the harm is caused by the use of the particular construct or by an unwanted or unforeseen effect of the gene insertion on the genome.

Animal testing is but a first step and not a substitute for human studies. If there is no indication of harm to the animals the results will have to be validated with human volunteers in clinical double-blind, placebo-controlled drug-type tests. Such studies may have to go on for considerable lengths of time. It must also be kept in mind that any potential harm with GM food can be most acute in the young, elderly and sick, particularly those suffering from HIV, hepatitis or other viral diseases. Many people suffer from allergies and other disorders of the gastrointestinal tract and for these the consumption of GM food may have unforeseen consequences and some of these may be irreversible. Thus, for these the clear labeling of GM food must be made mandatory.

There is a compelling need to further develop the concepts of biological testing particularly for potential long-term effects. Since the GM-potato work with male rats showed abnormalities in the development of their sexual organs, it is imperative that similar experiments should be done with female rats to be followed by studies of the effects on reproductive performance of rats (or other animals) reared and maintained on GM vs non-GM diets for several generations.

If there is a general willingness to fund research along these or similar lines and the regulators will accept the concept of biological/toxicological testing transparently and inclusively, the methods are available for the work to start. Following this route, publishing the results and consulting the public will ensure that a technology which promised safe and plentiful food will deliver it for us all and, I am confident, that if people will see that everything has been done to establish its safety they will willingly accept it.


Table 1. Number of crypt cells in the jejunum of rats fed various potato diets
Diet Parent Parent vs parent + GNA (P) Parent + GNA Parent vs Transgenic (P) Transgenic Parent+GNA vs Transgenic (P)
Raw 15.9 (0.5) 0.037 17.0 (0.7) 0.000 20.3 (0.8) 0.006
Boiled 17.8 (1.1) 0.466 18.2 (0.2) 0.749 18.2 (1.2) 0.769
raw vs boiled 0.006   0.003   0.003

Number of nuclei were sequentially counted on well-oriented haematoxylin and eosin paraffin sections (4 um). Values represent means (SD) for 6 rats/treatment; 10 crypts per rats were counted. Differences between treatments are significant when p<0.05 (Student's t test). The effect of boiling (p=0.759) is not significant, while that of GNA added or as transgene product (p=0.019) and the effect of transformation (p=0.000) are highly significant. The interaction between GNA and cooking (p=0.043) and between transformation and cooking (p=0.018) are also significant (multivariant analysis with Tukey's test).

Table 2 Mitotic numbers per 100 crypts in the jejunum of rats fed potato diets
Diet Parent Parent + GNA Transgenic
Raw 48 49 75
Boiled 57 56 57

Mitotic cells were expressed per 100 crypts.

8. References

Alliance for Biointegrity (1998). Including Calgene FLAVR-SAVRTM tomato report, pp. 1-604. International Research and Development Corporation’s first test report, pp. 1736-1738; Conclusions of the expert panel regarding the safety of the FLAVR SAVRTM tomato, ENVIRON, Arlington VA, USA pp. 2355-2382; Four week oral (intubation) toxicity study in rats by IRDC, pp. 2895-3000.

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

Berberich, S.A., Ream, J.E, Jackson, T.L., Wood, R., Stipanovic, R., Harvey, P., Patzer, S. and Fuchs, R.L. (1996) The composition of insect-protected cottonseed is equivalent to that of conventional cottonseed. Journal of Agricultural Food Chemistry 44, 365-371.

Bernstein, I.L., Bernstein, J.A., Miller, M., Tierzieva, S., Bernstein, D.I., Lummus, Z., Selgrade, M.K., Doerfler, D.L. and Seligy, V.L. (1999) Immune responses in farm workers after exposure to Bacillus thuringiensis pesticides. Environmental Health Perspectives 107, 575-582.

Betz, F.S., Hammond, B.C. and Fuchs, R.L. (2000) Safety and advantages of Bacillus thuringiensis-protected plants to control pests. Regulatory Toxicology and Pharmacology 32, 156-173.

Bindslev-Jensen, C. and Poulsen, L.K. (1997) Hazards of unintentional/intentional introduction of allergens into foods. Allergy 52, 1184-1186.

Brake, J. and Vlachos, D. (1998) Evaluation of transgenic Event 176 "Bt" corn in broiler chicken. Poultry Science 77, 648-653.

Burks, A.W. and Fuchs, R.L. (1995) Assessment of the endogenous allergens in glyphosate-tolerant and commercial soybean varieties. Journal of Allergy and Clinical Immunology 96, 1008-1010.

Chiter, A., Forbes, J.M. and Blair, G.E. (2000). DNA stability in plant tissues: implications for the possible transfer of genes from genetically modified food. FEBS Letters 24098, 1-5.

Domingo, J.L. (2000) Health risks of genetically modified foods: Many opinions but few data. Science 288, 1748-1749.

European Commission Directorate C, (2000) Opinion of the Scientific Committee on Food on 'The Evaluation of Toxicological Information Related to the safety Assessment of Genetically Modified Tomatoes'. CS/NF/TOM/8 ADD 1 REV 3 Final.

Ewen, S.W.B. and Pusztai, A. (1999a) Authors' reply. The Lancet 354, 1727-1728.

Ewen, S.W.B. and Pusztai, A. (1999b) Effects of diets containing genetically modified potatoes expressing Galanthus nivalis lectin on rat small intestine. The Lancet 354, 1353-1354.

Fares, N.H. and El-Sayed, A.K. (1998) Fine structural changes in the ileum of mice fed on delta-endotoxin-treated potatoes and transgenic potatoes. Natural Toxins 6, 219-233.

Flachowsky, G. and Aulrich, K. (2001) Nutritional assessment of feeds from genetically modified organism. Journal of Animal and Feed Sciences 10, Suppl. 1., 181-194.

Hammond, B.G., Vicini, J.L., Hartnell, G.F., Naylor, M.W., Knight, C.D., Robinson, E.H., Fuchs, R.L. and Padgette, S.R. (1996) The feeding value of soybeans fed to rats, chickens, catfish and dairy cattle is not altered by genetic incorporation of glyphosate tolerance. Journal of Nutrition 126, 717-727.

Harrison, L.A., Bailey, M.R., Naylor, M.W., Ream, J.E., Hammond, B.G., Nida, D.L., Burnette, B.L., Nickson, T.E., Mitsky, T.A., Taylor, M.L, Fuchs, R.L. and Padgette, S.R. (1996) The expressed protein in glyphosate-tolerant soybean, 5-enolpyruvylshikimate-3-phosphate synthase from Agrobacterium sp. strain CP4, is rapidly digested in vitro and is not toxic to acutely gavaged mice. Journal of Nutrition 126, 728-740.

Hashimoto, W., Momma, K., Katsube, T., Ohkawa, Y., Ishige, T., Kito, M., Utsumi, S. and Murata, K. (1999) Safety assessment of genetically engineered potatoes with designed soybean glycinin: compositional analyses of the potato tubers and digestibility of the newly expressed protein in transgenic potatoes. Journal of Science of Food and Agriculture 79, 1607-1612.

Hashimoto, W., Momma, K., Yoon, H.J., Ozawa, S., Ohkawa, Y., Ishige, T., Kito, M., Utsumi, S. and Murata, K. (1999) Safety assessment of transgenic potatoes with soybean glycinin by feeding studies in rats. Bioscience Biotechnology Biochemistry 63, 1942-1946.

Helm, R.M. and Burks, A.W. (2000) Mechanisms of food allergy. Current Opinion in Immunology 12, 647-653.

Hohlweg, U. and Doerfler, W. (2001) On the fate of plant and other foreign genes upon the uptake in food or after intramuscular injection in mice. Molecules Genetics and Genomics 265, 225-233.

Kramer, K.J., Morgan, T.D., Throne, J.E., Dowell, F.E., Bailey, M. and Howard, J.A. (2000) Transgenic avidin maize is resistant to storage insect pests. Nature Biotechnology 18, 670-674.

Kuiper, H.A., Noteborn, H.P.J.M. and Peijnenburg, A.A.C.M. (1999) Adequacy of methods for testing the safety of genetically modified foods. The Lancet 354, 1315-1316.

Kuiper, A.H., Kleter, G.A., Noteborn, H.P.J.M. and Kok, E.J. (2001) Assessment of the food safety issues related to genetically modified foods. The Plant Journal 27, 503-528.

Lappe, M.A., Bailey, E.B., Childress, C. and Setchell, K.D.R. (1999) Alterations in clinically important phytoestrogens in genetically modified, herbicide-tolerant soybeans. Journal of Medical Food 1, 241-145.

Mercer, D.K., Scott, K.P., Bruce-Johnson, W.A., Glover, L.A. and Flint, H.J. (1999) Fate of free DNA and transformation of oral bacterium Streptococcus gordonii DL1 plasmid DNA in human saliva. Applied and Environmental Microbiology 65, 6-10.

Metcalf, D.D., Astwood, J.D., Townsend, R., Sampson, H.A., Taylor, S.L. and Fuchs, R.L. (1996) Assessment of the allergenic potential of foods derived from genetically engineered crop plants. In: Critical Reviews in Food Science and Nutrition 36(S):S165-186. CRC Press Inc. Boca Raton,


Millstone, E., Brunner, E. and Mayer, S. (1999) Beyond substantial equivalence. Nature 401, 525-526.

Momma, K., Hashimoto, W., Ozawa, S., Kawai, S., Katsube, T., Takaiwa, F., Kito, M, Utsumi, S. and Murata, K. (1999) Quality and safety evaluation of genetically engineered rice with soybean glycinin: Analyses of the grain composition and digestibility of glycinin in transgenic rice. Bioscience Biotechnology Biochemistry 63, 314-318.

Nakamura, R. and Matsuda, T. (1996) Rice allergenic protein and molecular-genetic approach for hypoallergenic rice. Bioscience Biotechnology Biochemistry 60, 1215-1221.

Nordlee, J.A., Taylor, S.L., Townsend, J.A. and Thomas, L.A. (1996) Identification of a Brazil nut allergen in transgenic soybean. New England Journal of Medicine 334, 688-692.

Noteborn, H.P.J.M., Bienenmann-Ploum, M.E., van den Berg, J.H.J., Alink, G.M., Zolla, L., Raynaerts, A., Pensa, M. and Kuiper, H.A. (1995) Safety assessment of the Bacillus thuringiensis insecticidal crystal protein CRYIA(b) expressed in transgenic tomatoes. In: ACS Symposium series 605 Genetically Modified Foods - Safety Issues, Eds. Engel, K.H, Takeoka, G.R. and Teranishi, R. Chapter 12, pp. 135-147. American Chemical Society, Washington, D.C.

Novak, W.K. and Haslberger, A.G. (2000) Substantial equivalence of antinutrients and inherent plant toxins in genetically modified novel foods. Food and Chemical Toxicology 38, 473-483.

O'Neil, C., Reese, G. and Lehrer, S.B. (1998) Allergenic potential of recombinant food proteins. Allergy and Clinical Immunology International 10, 5-9.

Osusky, M., Zhoul, G., Osuska, L., Hancock, R.E., Kay, W.W. and Misra, S. (2000) Transgenic plants expressing cationic peptide chimeras exhibit broad-spectrum resistance to phytopathogens. Nature Biotechnology 18, 1162-166.

Padgette, S.R. Taylor, N.B., Nida, D.L., Bailey, M.R., MacDonald, J., Holden, L.R. and Fuchs, R.L. (1996) The composition of glyphosate-tolerant soybean seeds is equivalent to that of conventional soybeans. Journal of Nutrition 126, 702-716.

Pusztai, A., Ewen, S.W.B., Grant. G., Peumans, W.J., van Damme, E.J.M., Rubio, L., Bardocz, S. (1990) Relationship between survival and binding of plant lectins during small intestinal passage and their effectiveness as growth factors. Digestion, 46(suppl. 2), 308-316.

Pusztai, A., Grant, G., Duguid, T., Brown, D.S., Peumans, W.J., Van Damme, E.J.M., Bardocz, S. (1995) Inhibition of starch digestion by α-amylase inhibitor reduces the efficiency of utilization of dietary proteins and lipids and retards the growth of rats. Journal of Nutrition, 125, 1554-1562.

Pusztai, A., Grant, G., Bardocz, S., Alonso, R., Chrispeels, M.J., Schroeder, H.E., Tabe, L.M. and Higgins, T.J.V. (1999) Expression of the insecticidal bean alpha-amylase inhibitor transgene has minimal detrimental effect on the nutritional value of peas fed to rats at v30% of the diet. Journal of Nutrition 129, 1597-1603.

Pusztai, A. (2001) Genetically Modified Foods: Are they a risk to Human/Animal Health?.

Redenbaugh, K., Hatt, W., Martineau, B, Kramer, M., Sheehy, R., Sanders, R., Houck, C. and Emlay, D. (1992) A case study of the FLAVR SAVRTM tomato. In: Safety Assessment of Genetically Engineered Fruits and Vegetables. CRC Press Inc. Boca Raton.

Rogan, G.J., Bookout, J.T., Duncan, D.R., Fuchs, R.L., Lavrik, P.B., Love, S.L., Mueth, M., Olson, T., Owens, E.D., Raymond, P.J. and Zalewski, J. (2000) Compositional analysis of tubers from insect and virus resistant potato plants. Journal of Agricultural and Food Chemistry 48, 5936-5945.

Rubial-Mendieta, N.L. and Lints, F.A. (1998) Novel and transgenic food crops: Overview of scientific versus public perception. Transgenic Research 7, 379-386.

Rubio, L.A., Grant, G., Caballé, C., Martinez-Aragon, A., Pusztai, A. (1994) High in vivo (rat) digestibility of faba bean (Vicia faba), lupin (Lupinus angustifolius) and soya bean (Glycine max) soluble globulins. Journal of Food Science and Agriculture 66, 289-292.

Schubbert, R., Lettmann, C. and Doerfler, W. (1994) Ingested foreign (phage M13) DNA survives transiently in the gastrointestinal tract and enters the blood stream of mice. Molecules, Genes and Genetics 242, 495-504.

Schubbert, R. Hohlweg, U., Renz, D. and Doerfler, W. (1998) On the fate of orally ingested foreign DNA in mice: chromosomal association and placental transmission in the fetus. Molecules, Genes and Genetics 259, 569-576.

Sidhu, R.S., Hammond, B.G., Fuchs, R.L., Mutz, J.N., Holden, L.R., George, B. and Olson, T. (2000) Glyphosate-tolerant corn: the composition and feeding value of grain from glyphosate tolerant corn is equivalent to that of conventional corn (Zea mays L.). Journal of Agricultural and Food Chemistry 48, 2305-2312.

Taylor, N.B., Fuchs, R.L., MacDonald, J.,Shariff, A.B. and Padgette, S.R. (1999) Compositional analysis of glyphosate-tolerant soybeans treated with glyphosate. Journal of Agriculture and Food Chemistry 47, 4469-4473.

Teshima, R., Akiyama, H., Okunuki, H., Sakushima, J-i, Goda, Y., Onodera, H., Sawada, J-i and Toyoda, M. (2000) Effect of GM and Non-GM soybeans on the immune system of BN rats and B10A mice. Journal of Food Hygiene Society of Japan 41, 188- 193.

Vazquez Padron, R.I., Moreno Fierros, L., Neri Bazan, L., De la Riva, G.A. and Lopez Revilla, R. (1999) Intragastric and intraperitoneal administration of Cry1Ac protoxin from Bacillus thuringiensis induces systemic and mucosal antibody responses in mice. Life Sciences 64, 1897-1912.

Vazquez-Padron, R.I., Moreno-Fierros, L., Neri-Bazan, L., Martinez-Gil, A.F., de la Riva, G.A. and Lopez-Revilla, R. (2000) Characterization of the mucosal and sytemic immune response induced by Cry1Ac protein from Bacillus thuringiensis HD 73 in mice. Brazilian Journal of Medical and Biological Research 33, 147-155.

Vazquez Padron, R.I., Gonzalez Cabrera, J., Garcia Tovar, C., Neri Bazan, L., Lopez Revilla, R., Hernandez, M., Morena Fierros, L. and De la Riva, G.A. (2000) Cry1Ac protoxin from Bacillus thuringiensis sp. kurstaki HD73 binds to surface proteins in the mouse small intestine. Biochemical and Biophysical Research Communications 271, 54-58.

The British Library File (1997) Public reference 'BL SUP 1113' Chardon Hearing Documents, No. 10;

The Royal Society (1999) Review of data on possible toxicity of GM potatoes. US Food and Drug Administration. Memorandum dated May 17 (1994). Summary of consultation with Calgene Inc, concerning FLAVR SAVRTM tomatoes'