Anthony J FitzGerald, *Robert A Goodlad, Nicholas A Wright
S. W. B. Ewen and A. Pusztai
Brian Fenton et al
D. C. Kilpatrick
Brian Fenton, Kiri Stanley, Steven Fenton, Caroline Bolton-Smith
Carl B Feldbaum
Roger A Fisken
Allan Mowat, Department of Immunology and Bacteriology, University of Glasgow, Western Infirmary, Glasgow, writes that Stanley Ewen and Arpad Pusztai (Oct 16, p 1353)1 raise intriguing questions about the potential of Galanthus nivalis agglutinin (GNA) to cause morphological alterations in the intestinal tract and suggest that the lectin's effects may be exacerbated in genetically modified potatoes. However, the ways in which Ewen and Pusztai assess the enteropathic properties of GNA mean that their findings must be interpreted with extreme caution.
The indices of crypt length and jejunal intraepithelial lymphocyte (IEL) count have been used to study intestinal immunopathology for 20 years or more2,3 but several aspects of Ewen and Pusztai's methodology warrant attention. First, writes Mowat, the study relies entirely on image analysis of formalin-fixed, paraffin-embedded sections, which are notoriously subject to shrinkage, distortion, and other fixation artifacts. These problems can be partly overcome by careful choice of well-oriented villus-crypt units combined with exhaustive measurement techniques, but Ewen and Pusztai do not indicate whether they did this. Errors created by measuring crypts in different planes of section on the sample could account for the high variation reported. The fact that the crypt lengths reported (60-90 5m in the jejunum) are much smaller than those normally found in the rat4 reinforces the view that the measurements may not have been accurate. More sensitive methods for processing and measuring intestinal tissues include non-formalin-based fixatives, microdissection, and direct morphometry of crypt and villus lengths.2-4 The enterocyte mitotic rate is probably the most sensitive index of intestinal pathology,2 and this can be measured easily by metaphase arrest or incorporation of bromodeoxyuridine.
Ewen and Pusztai use IEL counts to support their hypothesis that genetically modified potatoes cause jejunal lesions. They state that "IEL are known to increase when non-specific intestinal damage occurs", but an increase in IEL count is specifically a feature of enteropathies associated with activated T lymphocytes.3 Thus an increased IEL count in animals receiving lectins could be compelling evidence for these materials inducing immunologically mediated damage to the gut. However, Ewen and Pusztai have not shown this conclusively.. They do not seem to have counted IELs by a well-established method in which IEL are counted per 100 enterocyte nuclei or as an absolute number per length, volume or area of mucosa. That the technique they used is not ideal is underlined by the numbers of IEL they report, which are in the region of 7-11 per 48 villi. To reconcile these estimates, given that a single column of villus enterocytes in the rat jejunum contains 200-300 enterocytes and the density of IEL in the normal small intestine is 10-20 per 100 epithelial cells, is difficult. The low protein content of some of the diets, referred to by other commentators as a possible source of error,5 could account for some deviation of IEL numbers from normal but not for such a gross change. However, it is feasible that such a diet might have made the rats more susceptible to the intestinal infections known to cause the kind of changes in IEL and crypts3 noted here.
The speculation that the lectin caused jejunal crypt hyperplasia via a direct stimulatory effect on crypt cells cannot be substantiated by the data. Hyperplasia implies increased mitotic activity, which was not measured. Also, the time course for these changes is not described, and no parameters of villus pathology are provided. In the absence of this information, it is impossible to say whether the changes in crypt morphology are primary effects of the lectin or secondary to villus damage.
Interactions between lectins, intestinal epithelial cells, and the local immune apparatus is an important and poorly understood area. Appropriate methods for studying the enteropathic effects of lectins are available and are comparatively simple and inexpensive. Application of these techniques may help elucidate the issues raised by this provocative study.
Anthony J FitzGerald, *Robert A Goodlad, Nicholas A Wright of the Histopathology Unit, ICRF, 44 Lincoln's Inn Fields, London, write that while much of the debate has focused on the nature of the diets studied by Stanley Ewen and Arpad Pusztai 1 and possible differences between them, one central question is the effects of these on cell proliferation in the gut. Ewen and Pusztai talk about "proliferative effects" when they have not measured intestinal cell proliferation but merely crypt depth. Crypt depth might reflect hypoplasia and hyperplasia but this has yet to be shown. Various methods can be used to measure intestinal epithelial cell proliferation, such as the numbers of dividing cells in optimally sectioned crypts, but for definitive conclusions we need measurements related to the rate of crypt or gland cell production;2 the size of the epithelial population also needs to be assessed appropriately. Perhaps the best way of doing this is to use metaphase arrest and the micro-dissection method,3 in which not only the rate of crypt cell production but also good measurements of crypt and villus size can be captured simultaneously. Another point is that many such studies can be confounded by concomitant changes in the denominator,4,5 and the data on intraepithelial lymphocytes, with sectioned villus as the denominator, could be subject to the same criticism.
We hope these comments will help to ensure that if these studies are repeated (as they should be), robust, rapid, and reliable methods for assessment of cell proliferation are used.
Peter Lachmann, Academy of Medical Sciences, 10 Carlton House Terrace, London, UK writes that Stanley Ewen and Arpad Pusztai 's research letter1 describing measurements of intraepithelial lymphocyte counts and mucosal thickness in rats in a short-term feeding experiment of potatoes transgenic for snowdrop lectin is unacceptable for the following reasons.
The two measurements reported were not the test of a pre-existing hypothesis. They have been selected from an unstated number of comparisons.The probabilities need to be adjusted for the number of different comparisons made. This will almost certainly make them non-significant and the experiments therefore need to be repeated on a new group of rats.
S. W. B. Ewen and A. Pusztai reply that in their Oct 16 commentary Harry A Kuiper and colleagues discuss methods for testing the safety of GM foods.
Our research letter only addressed the biological effects of GNA-GM potato diets on the morphology of the rat gut. However, since the safety testing issue has been raised we would like to respond.
According to the Rowett Institute's audit report the parent and transgenic lines we used were "substantially equivalent". At least for protein content this was confirmed in the alternative report put on the internet by the Rowett (not by Pusztai, as stated by Kuiper et al). All experimental diets were isoproteinic and isocaloric and supplemented with vitamins and minerals for optimal growth, and the food intake of all rats was the same. Although 6% protein is suboptimal, the diet was not protein-free, as it was in the paper cited by Kuiper et al. Since the rats were growing, to use emotional terms such as "starvation" is misleading. Furthermore, starvation reduces gut size and crypt length, precisely the opposite what we found. Since trypsin and chymotrypsin inhibitors have no effect on gut morphology, reference to them is irrelevant. The potato lectin is antimitogenic 2 so its effect should be a reduction in crypt cell proliferation rate and crypt size, not an increase. Moreover, GNA was selected for GM insertion because its effect on the gut is minimal so differences in lectin content, as an explanation for the biological effects reported, are also irrelevant.
Kuiper et al refer to "no consistency" in the changes. Is there any reason why changes in the different gut compartments should be the same? The ingestion of potatoes may indeed be associated with several adaptive changes, including caecal hypertrophy, but all the rats were given essentially the same potato diet containing the same starch component. Our study was about the effects of GM potatoes on gut morphology, not on their toxicology. The number of rats used (six per treatment group) was more than sufficient and most peer-reviewed, published papers on the biological and nutritional effects of similar diets (including over 40 from our laboratory) use three to four animals per group. Our controls were also sufficient.
For both raw and cooked GM potatoes we used as controls both parent potato and parent potato supplemented with purified GNA, at the level expressed in GM potatoes. With this experimental design we could discriminate between the effects of the transgene product, the transformation, or the cooking and the interactions between these effects. The use of standard rodent diet of undefined composition as a control would have been inappropriate, and the effects of high-protein and low-protein diets would not have been comparable. We agree with Kuiper et al that attention should be paid to bioavailability and toxic effects. Kuiper et al cite their study done with a recombinant form of Bacillus thuringiensis toxin and not with toxin isolated from the GM-tomato. They themselves point out that since the stability, survival in the gut, and toxicity of a recombinant protein usually differ from those of the plant-gene product, the two forms can not be used for comparisons. Biotech companies, in their submissions to regulatory authorities, also rely on toxicity studies with E coli recombinants. We, by contrast, have been comparing the gene product from the GM plant with that of the non-transformed original.3
Indeed it would have been desirable for testing methods recommended by FAO/WHO and so on to have been used for GM food crops currently approved. Unfortunately, no such tests were required, carried out, or their results published. It is therefore surprising for Kuiper et al to state that "the data so far indicate GM-crops ... that have been introduced into the environment do not differ from traditionally grown crops except for the inserted traits". Several traits having nothing to do with the transgene save the insertion have been found to differ in GM lines. The Monsanto analyses of glyphosate-resistant soya showed that the GM-line contained about 28% more Kunitz trypsin inhibitor, a known antinutrient and allergen.4
GM-soya contained significantly less phyto-oestrogens than the parent lines. 5 Why is it that current GM crops need not be examined as thoroughly as the next generation?
We agree that "particular attention must be given to the detection and characterisation of unintended effects of genetic modification" but how will unknown toxins or allergens be found without first biologically testing the GM crop for toxicity? Writing of the sequence to be followed in future safety testing procedures, Kuiper et al say "depending on the outcome . . . further toxicological and nutritional studies may be needed". The admission by people close to the decision-making committees of the European Union that the biological testing of GM food is needed appears to be an acknowledg-ment that previous safety testing could not have been rigorous enough.
Allan Mowat comments on the tissue fixation. This has long been a contentious topic. There may not be an ideal fixative but it is mischievous to suggest that the fixative upon which the whole of human histopathology relies could be responsible for different crypt-length measurements. The histology was part of a large series of synchronous physiological measurements on the same animal, demanding rapid tissue handling. Our animals were young (about 85 g) and, at that stage of development, the gut can easily be dissected from the mesentery to provide an unwashed, undistended histological sample at a constant distance from the pylorus. Orientation and plane-of-section problems were minimised by "splinting" tissue samples on card during standardised fixation. The jejunal crypt length of our 85 g rats has not varied and the data reported are similar to those obtained from 5000 jejunal crypts over the past 10 years. Rat intestinal crypt length depends on diet, animal supplier, and housing conditions, all of which are standardised in our rat colony, and the chemical composition of the potato diet was subjected to the exhaustive chemical analysis.
Intraepithelial lymphocytes (IEL) are thought to provide a surveillance function for damaged or virally infected cells, and any increase in the IEL population need not be limited to hypersensitivity reactions. Our young animals could not have been exposed to GNA previously, and any difference between groups is likely to be due to luminal factors in the diet. Moreover, IEL numbers are unaffected by lectin treatment.1 The only difference in the diets is the presence of unidentified factors caused by genetic modification. Epithelial cell damage induced by the genetic modification cannot be excluded, and we would have studied it if our experiments had not been aborted. IELs averaged 7-11 per villus based on 48 villi counted, and we believe that our estimate is a useful indicator of unspecified immunological events of comparative value between groups. The suggestion that the low-protein content of the diet caused intestinal infection can be rejected; parasites were not evident in the unwashed histological preparations and the lamina propria did not contain excess eosinophils.
Hyperplasia refers specifically to an increased cell number, frequently, but not necessarily, accompanied by increased mitotic activity although increased crypt mitotic activity is indeed present in our GM-potato fed animals. If, after 10 days of ingestion of GM potatoes villus pathology had been evident we would have described the damage. We reject the notion that potato or GNA lectin could have produced the changes that caused us concern.
Peter Lachmann's points (1) and (2) are addressed in our reply to Mowat. To point (3) the answer is yes, the measurements were done double-blind. With regard to point (4), Lachmann does not understand that the significant changes in mucosal indices represent another nail in the coffin of "substantial equivalence" on which the GM regulatory system is based. The differences in gut metabolic responses demonstrate that GM potatoes were not "metabolically equivalent", and this is important whether the changes are pathological or not. Increased epithelial cell proliferation in the colon is not regarded desirable and the high energy cost of small-intestinal hyperplasia may compromise growth and development. In his point (5) Lachmann asserts that we had no prior hypothesis. Because the experimental design was obvious from the introductory part of our letter, to save space, a part of our sentence was omitted. Here it is: "It was thought that comparison of the histological parameters of the gut of rats fed potato diets containing either GM potatoes, or non-GM potatoes with or without being supplemented with GNA should give a clear indication whether GNA gene insertion had affected the nutritional and physiological impact of potatoes on the mammalian gut". Our table 1 clearly gives the statistical methods used and the number of comparisons. These methods were approved by independent statisticians. Lachmann says that the experiments need to be repeated. We would be happy to oblige. If our experiments are so poor why have they not been repeated in the past 16 months? It was not we who stopped the work on testing GM potatoes expressing GNA or other lectins or even potatoes transformed with the empty vector, which are now available.If Lachmann represents the view of the Academy of Medical Sciences on GM-food safety he should use his influence to make funds available for the continuation of this work in the UK.
Department of Pathology, University of Aberdeen, Aberdeen AB25 2ZD, UK
Sean Munro, MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK writes that Brian Fenton and colleagues' research letter (Oct 16, p 1354)1 on the insecticidal Galanthus nivalis lectin (GNA) makes several questionable assertions. They state that the "distribution, abundance, and micro-heterogeneity" of structures recognised by GNA are "largely unknown". They then provide results which they claim "show that human white blood cells have many proteins that strongly bind to GNA". Several previous studies have reported examination of GNA binding to human proteins and tissues (including breast, Peyer's patches, kidney, brain, photoreceptors, plasma proteins, placenta, gallbladder, and several human-derived cell lines). Moreover, the binding of GNA to proteins from human white blood cells has been already been reported by Benallal et al.2 The conclusions of the new work cannot be considered novel.
Fenton et al also argue that their results are relevant to the debate on genetically modified (GM) food, stating there needs to be "a greater under-standing of the interactions of plant lectins and human glycoproteins before they can be safely incorporated into the food chain". However, plant lectins are already incorporated into the human diet. Not only are lectins naturally present in potatoes, lentils, beans, peas, tomatoes, and other crop plants but also many of these lectins bind to human glycoproteins. The first plant lectins were identified over a 100 years ago, and their binding to human cells, and the potential toxicity of some of them, has long been known. Denaturation by cooking or digestion by enzymes in the gut can even allow consumption of lectins that might otherwise be harmful.
The interaction of lectins with glycoproteins in the digestive tract, and elsewhere in the body, has been extensively studied for decades. Although GNA is not incorporated into any of the GM food currently approved for human consumption in the UK, the need for testing of any food containing GNA is clearly apparent from past studies. The results of Fenton et al reveal nothing that was not already known and provide no new reasons to question the possible use of GNA in transgenic crops.
D. C. Kilpatrick, Academic Unit, Department of Transfusion Medicine, SE Scotland Blood Transfusion Service, Edinburgh EH1 2QN, UK writes that Brian Fenton and colleagues1 report that the lectin Galanthus nivalis (snowdrop) agglutinin (GNA) can bind glycoconjugates from human leucocytes, and they claim that the work "highlights the need for a much greater understanding of the interactions between plant lectins and human glycoproteins before they can be safely incorporated into the food chain". Binding of GNA to human leucocytes has been known for a decade, and plant lectins are already common components of foodstuffs.
Shortly after GNA was first isolated, we showed that it was weakly mitogenic for human mononuclear leucocytes,2 an activity that required binding to cell surfaces. Moreover, GNA acts synergistically with a submitogenic concentration of the tumour promoter, tetradecanoylphorbol-13-acetate,2 as Fenton et al conjecture it might. However, all this in itself is unremarkable. Plant lectins are present in active form in a wide variety of "healthy" foods.3 Lectins in general are resistant to the digestive process, and a small proportion of any ingested lectin may be transported into the circulation. Undoubtedly, some ingested lectins (eg, pokeweed mitogen) could be very harmful; equally, others, like the tomato lectin, seem to be harmless.4 GNA is not normally a dietary lectin for man so caution should be exercised before permitting its presence in foods. However, the data indicate that the GNA gene product itself is non-toxic; the point about the contentious work of Ewen and Pusztai5 is that some other component of the construct used to transfect the GNA gene may be responsible for unexplained (and potentially harmful) consequences.
Brian Fenton, Kiri Stanley, Steven Fenton, Caroline Bolton-Smith of the Scottish Crop Research Institute, Invergowrie, Dundee; and Nutrition Research Group, Cardiovascular Epidemiology Unit, University of Dundee, Dundee, UK reply that Sean Munro's first comments are misleading, for he quotes out of context. The distribution, abundance, and microheterogeneity of terminal [(alpha)]-1-3-mannose residues on human membrane-bound receptor proteins is largely unknown, whilst GNA (Galanthus nivalis agglutinin) binding to some plasma (secretory) proteins, such as [(alpha)]2-macroglobulin, is well documented. Benallal et al1 identified a 95 kDa leucocyte membrane protein which also bound GNA, and this probably corresponds to one of the molecules identified in our study. However, we have identified additional GNA binding proteins that also seem to vary between individualsQthe importance of which is unknown.
Both Munro and D C Kilpatrick cite the natural abundance of lectins in commonly consumed foods, and the known toxicity of some and apparently harmless nature of others. This variable toxicity emphasises the need for caution before lectins, which are not normally consumed, are incorporated into human-food plants. Snowdrop has never formed part of the human diet. GNA is a mannose binding lectin (MBL) and other MBLs are already in our diet, such as those from onion and leek. However, in these plants, their concentration (0a5-10 5g/g) is 100 times less than in non-edible plants (daffodil and snowdrop; 2-4 mg/g)2 and that required to control insects with GNA. The other dietary plant lectins that Munro mentions have differing specificities and are more digestible (after one hour only 30% of pea or bean lectin survives digestion whereas GNA is still 90% intact3). Daily and seasonal variation in human diets will result in differential intensity and duration of exposure to natural food plant lectins. By contrast, GNA could be incorporated into all main crop plants, resulting in chronic exposure from cooked and uncooked foods. Kilpatrick rightly advocates caution, and cites his work4 on which our own comments were based. This reference was omitted at the editing stage. Such recognition of the reported mitogenicity of GNA is not universal.5
Kilpatrick further highlights the importance of synergy. This synergy could occur between GNA and other dietary lectins or tumour promoters in man and could provide one explanation of Ewen and Pusztai's observationsQGM potatoes are likely to contain other changes as a consequence of tissue culture (an intrinsic part of the GM process). If change occurred in an endogenous mitogen it could act synergistically with the GNA gene product to produce the hyperplasia observed in the specific treatment group. A second possible explanation of their results is that the single isoform of GNA used in the transgenic potatoes is more highly mitogenic than the mixture of isoforms purified directly from snowdrops.2 These alternative explanations do not implicate the genetic manipulation process itself.
Carl B Feldbaum, Biotechnology Industry Organization, Washington, DC, writes that your decision to publish the research of Stanley Ewen and Arpad Pusztai 1 breaks unfortunate new ground for a scientific journal. Put simply, The Lancet has placed politics and tabloid sensationalism above its responsibility to report and assess new science. Most peer-reviewed journals are respected, and read, for the integrity of the research they publish and their dependability in weeding out irresponsible work.
Richard Horton (Oct 16, p 1314) 2 argues that The Lancet might have been criticised for suppressing information by not publishing Ewen and Pusztai's work, but I believe he has jeopardised the journal's credibility, especially among readers and contributors in the scientific community. This pandering to popular debate rather than promoting responsible scientific inquiry may appeal to some, but I believe that the editor's poor judgment will strengthen the resolve of other scientific journals to adhere to the publication standards The Lancet saw fit to abandon.
I doubt if The Lancet would have published Ewen and Pusztai's research if it had implied the safety of biotech foods. But that is fine. Those who work hard to apply biotechnology to agriculture have no interest in flawed data.
Horton is nave if he really believes that "publication of Ewen and Pusztai's findings is not, as some newspapers have reported, a 'vindication' of Pusztai's earlier claims". Anti-technology activists already have seized on The Lancet's publication of this work as precisely that. They claim that publication, in itself, is proof the research is valid because that is the standard scientific journals are supposed to apply.
In the USA, biotech crops and foods have been tested more than any other agricultural products in history. We have a regulatory system that applies science-based policies to guard the health of consumers and the environment. That system would have trashed Pusztai's potatoes if they had been submitted for approval. Too bad The Lancet failed to exercise such oversight on research submissions.
Roger A Fisken, Friarage Hospital, Northallerton DL6 1JG, UK writes that: I disagree with Richard Horton's view , expressed at the end of his commentary on the controversy surrounding genetically modified (GM) foods.1
It seems to be that we as scientists have not been nearly aggressive enough in attacking the scaremongering and sheer nonsense put out by the lay media on a variety of medical and scientific topics. Besides writing about these issues we should be lobbying the Press Complaints Commission and the government to try and ensure that journalists are taken to task and made to publish amendments if they grossly distort the facts in any kind of technical reporting, just as we would expect a retraction and apology for the libelling of an individual. If we do not do this we are in danger of sliding into a sort of mob rule when the media can with impunity whip up such a furore over supposed malpractice that protesters burn fields of experimental GM crops.
Consider the damage done to immunisation campaigns by the steady drip of adverse publicity in the media. This public hostility to one of the most effective forms of disease control in the history of medicine (a hostility to which, sadly, The Lancet has indirectly contributed) means that there is now a serious possibility of a measles epidemic because of the poor uptake of vaccine.
Let us be proud of the scientific method and of scientific rigour. We must try to stop the slide into alarmism and negativity to innovation. In the Middle Ages superstition led to the burning of witches; if we are not careful it could soon be scientists who are being burned.
Aaron Klug of The Royal Society, London, wrties that The Lancet criticised the Royal Society last May 1 for its "breathtaking impertinence" in reviewing Arpad Pusztai's work before it had been published. Richard Horton now repeats that criticism.2 We commented on Pusztai's unpublished work because he himself had commented on it, so extensively that it had become a matter of public interest. Since a one-sided debate was raging on the back of unvalidated experimental data, the Royal Society had a duty to examine such evidence as it could secure from all sources, including Pusztai himself. That is impertinence only if you endorse scientists flouting normal practice and rushing to the press with unvalidated data and invalid conclusions.
In introducing the Ewen and Pusztai research letter3 Horton helpfully describes the ambivalence of the referees and emphasises the value of having the data out in the open. He also states that the data are non-generalisable. It is therefore surprising that the journal allowed the paper to appear with two general conclusions in the final paragraph. In the circumstances, Horton's comments on the "failure to understand the . . . dialogue of accountability" are somewhat ironic.
Richard Horton, the editor, replies that Stanley Ewen and Arpad Pusztai 's research letter was published on grounds of scientific merit, as well as public interest. Four out of six invited reviewers recommended publication after revision on scientific grounds, one argued the public interest case, and one voted against publication. In the face of such clear support, the "irresponsible" action that Carl Feldbaum alludes to would have been to suppress publication. A debate about the science, rather than unsupported claims about that science, has now begunQa useful step forward.
Roger Fisken invites scientists to be more "aggressive". The Guardian newspaper has already reported one example of aggression, relating to The Lancet's decision to publish Ewen and Pusztai's work.1 That instance does not speak well of scientists' (in this case, a very senior scientist's) tolerance for open and reasoned debate.
Aaron Klug defends the Royal Society's wish to damn Ewen and Pusztai's work in the absence of both investigators. What he cannot defend is the reckless decision of the Royal Society to abandon the principle of due process in passing judgment on their work. To review and then publish criticism of these researchers' findings without publishing either their original data or their response was, at best, unfair and ill-judged.
** NOTICE: In accordance with Title 17 U.S.C. Section 107, this material is distributed for research and educational purposes only. **
Last Updated on 11/19/99
By Karen Lutz