Comments to ANZFA about

Applications A346, A362 and A363

from the Food Legislation and Regulation Advisory Group (FLRAG)

of the Public Health Association of Australia (PHAA)

on behalf of the PHAA.

 

(Note: Comments about Applications A355 and A387 will be contained in a subsequent document.)

 

 

General Comments

 

These applications are supported by experimental evidence provided by Monsanto, the applicant company.  This literature is quoted in the text of ANZFA's risk draft analysis reports.  However none of these documents were peer reviewed.  That is, all of them were internal publications by Monsanto or Optimum Quality Grains and none of them were published in scientific journals.  Consequently, the accuracy and veracity of the findings, and the methods employed have not come under scrutiny by other scientists.  There is also no information about whether some of these documents were submitted to peer-reviewed journals and were rejected.  The studies therefore appear to be reports of unchecked, unverified (and possibly peer-review rejected) experiments from the applicant company.

 

In science, the results of the first round of experiments or observations are published in the scientific literature.  Other scientist then repeat the experiments, sometimes with minor modifications that equate to a type of sensitivity analysis of the original results.  Others further explore the results by undertaking different experiments from the original.  Thus a body of knowledge is built-up in the area, and a picture is obtained. This does not appear to have happened at all with these foods.  It is often the case in science that a new concept or treatment, that looks good upon first report, often falls into disrepute after further investigation.  It is therefore of concern that this scientific process does not seem to have occurred with these foods and that the applicant company's reports do not seem to have been publicly published, peer-reviewed and verified by independent scientists. It is therefore of concern that ANZFA is recommending the safety of these foods for 18.7 million Australians to eat on the basis of such apparently minimal, unverified experimentation.

 

Another concern is the clear conflict of interest in an applicant company doing its own safety assessments.  The company will clearly benefit financially if the food is given a clean safety assessment by regulatory authorities, yet ANZFA appear to be accepting their safety evidence without discount and in the complete absence of produced independent assessments. 

 

 

Statistical Analyses on Compositional Studies

 

The applicant company provides data that often compares the genetically engineered (GE) food to a similar non-GE control food.  In the draft risk analysis reports, however, it is clear that the sample sizes are most inadequate to find statistical significance.  For example, n=2 in the glyphosate-tolerant canola line GT73 (Application A363).  With such low numbers it is almost a foregone conclusion that a statistically significant difference will NOT be found between the GE food and the non-GE food for most analyses, even if one exists in nature.  In fact, if a statistically-significant difference is found with such small numbers it indicates that the difference may be substantial indeed.  The FLRAG contends that a much more suitable sample size would have been at least n=50 to obtain an accurate picture of the compositional analysis of these foods, and that on the basis of these results, sample size calculations should have been performed to determine the number of plants that would be required to find statistical significance at an alpha of 0.05 and a power of 80% and 90%.  Monsanto should then have undertaken these analyses at the indicated sample sizes, and all of the results should have been published in peer-reviewed journals for scientific discussion before submission to ANZFA.

 

Moreover, in the statistical analyses presented in the draft risk analysis reports, often only a mean and a range are given.  Peer-reviewed scientific journals would require most or all of the number (n), mean, standard deviation, 95% confidence interval of the mean, the nature of the statistical test (eg t-test) and a p-value for each measurement, or the paper would be rejected for publication.  Similar non-parametric statistics would be required if the data are not normally distributed.  Why have these not been given?  Their omission prevents a full assessment of the data by others.  For example, as no standard deviations are given, sample size calculations cannot be done by others.  The FLRAG requests that it be given the raw data, means, standard deviations, confidence intervals, nature of statistical test and p values for all means or ranges given in these documents.

 

The FLRAG is also seeking official clarification from Monsanto, through ANZFA, as to why such small numbers were used, given that at the time, Monsanto had field trials in operation that must have totalled hundreds to thousands of plants.  Why were only such small numbers of plants sampled?  Furthermore, the FLRAG is seeking clarification as to the methodology of how plants were chosen, with the methodology of how they were randomly chosen for analysis, if this is the case. 

 

The FLRAG is also seeking official clarification of Monsanto, through ANZFA, as to why, when statistically significant differences in compositional analyses were determined between the GE plant and its control, Monsanto did not follow-up these results with substantial further experimentation to determine why these differences occurred.  Instead, such differences in composition tend to be dismissed as being within the natural variation of the plant. Such a statement ignores the evidence from the controls and the reason for having controls in the first place.  Controls are used in order to provide a proper comparison with the experimental group under the conditions present at the time of the experiment.  For example, the GE canola GT73 (Application A363) was compared to a control that was its parental line in field trials, presumably so that any differences between these groups could be attributed to the genetic engineering, and not to differences in soil, air temperature, rainfall, fertiliser etc.  However, when this comparison yielded statistically significant differences, the GE canola was then compared to the 'Westar range', a database housed in Canada, presumably from canola grown over a number of years from a number of sites.  Doing this permits differences due to soil, climate, etc to creep in again.  Moreover, proof should be provided for the assertion that the differences were due to the natural variation of the plant.  It is normally the role of such things as the statistical test, the standard error and the 95% confidence interval of the mean to provide a measure of this.  It is also important that the sample size be big enough to obtain an accurate measure of these.  We therefore contend that it is unacceptable to dismiss significant differences between the GE plant and its control and instead to compare the GE results to a broader pool of data as if this pool is the new control.  Similarly, we believe that it is unacceptable to compare the mean concentration of something (eg proline in GE canola GT73) with 'previously reported values' because the concentration of proline lies outside the ranges of both the control and the 'Westar range'.

 

Essentially, we contend that Monsanto should have compared the compositional differences between the GE plant and its control by setting-up field trials using methods such as a randomised block design or a completely randomised design so that the effects of things such as soil type, fertiliser and macro and microclimate were effectively removed. Then when sufficient sample sizes are used, the differences in composition between the plants, as a result of the genetic changes alone, can be determined.  The statistical tests on the results, by their very nature, should take into account any variability.  If Monsanto have done this properly then the composition of the GE plant can be properly compared to its control and any compositional differences found can be regarded as valid, in which case a difference in the composition of the GE plant may be due to the production of a novel substance, and full animal and human health feeding studies should be instigated to determine the safety of the food before it is assessed as safe.  If they have not done these experiments as described, then we contend that the experiments should be repeated, as described above, so that the true compositional differences due to the genetic changes can be given to ANZFA so that ANZFA can make a proper assessment of the food's safety. We argue that, simply on the basis of the sample sizes used, these experiments should be repeated.

 

 

Animal and Human Testing

 

It seems that one of the major areas of contention between ANZFA on the one hand and consumers and bodies such as the FLRAG on the other hand is over whether animal testing is required.  The ANZFA state in their draft risk analysis reports (eg page 48 of A346) that animal experimentation using the whole GE food will not yield accurate information compared to testing the 'test substance' directly in dose-response experiments.  ANZFA's argument appears to based on the assertion that the only new substance that can be found in GE foods is the new substance that has been genetically engineered to appear.  This is an untested hypothesis for these foods and a core matter of disagreement.  It should be noted that unexpected substances have previously been known to appear in GE plants.  For example, a tobacco plant, genetically engineered to produce gamma linoleic acid also unexpectedly produced a substance never before seen in tobacco plants: octodecatetraenoic acid.

 

The null hypothesis (HO) and the alternative hypothesis (H1) for this can therefore be stated as:

HO:  The only new substance to appear in the GE plant will be the one genetically engineered to appear.

H1:  Another, unexpected substance will appear in the genetically engineered plant.

 

For HO to be accepted, H1 must be rejected at p<0.05 on the basis of experimental evidence. This experimental evidence has not been produced for these foods, yet ANZFA has accepted Ho.

 

Given the complexity of food, we contend that  the only way to directly test H1 would be to undertake feeding experiments of the whole food, using methods similar to those widely used for decades in medical science, such as those used in nutrition and cancer experimentation.  That is, animals would be fed the food at varying concentrations of that food, for a long period of time, with body weights, food and water intakes regularly recorded, the animals would be sacrificed and their health would be assessed during autopsy.  Given the unknown nature of any new, unexpected substances, we contend that experimental animals should be fed for many months on a given GE food, that 3 experimental groups be used; (1) a control group fed a 'substantially-equivalent' non-GE food, (2) a group fed the GE food and (3) a group fed the GE food that has come out of farmers' fields, with any herbicides, etc applied to it.  We contended that given the unknown nature of any new, unexpected substance produced by the plant, the assessment should involve at least 50 experimental animals in each experimental group and should be as comprehensive as possible and therefore at least biochemistry, immune function, neurology, liver function, kidney function and gut function should be performed during a full autopsy that would also look for tumours and morphological changes.  In addition, experimental animals should be allowed to breed to check for teratological and other effects in off-spring. 

 

ANZFA's own documents indicate that new, unexpected substances may be appearing in the GE food.  For example, in Application A346 for insect-protected corn line MON 810, the amino acid profiles indicate that out of the 18 amino acids tested, 8 were significantly different from the control corn.  That is, despite very small sample sizes, about half of the tested amino acids were significantly different between the GE corn and its control corn.  As amino acids are the 'building-blocks' of proteins, such a major difference in the building blocks of proteins indicates that the proteins themselves may be markedly different between the GE corn and its control.  This indicates that the concentrations of one or more existing proteins may have significantly changed, and/or that one or more new proteins may be being produced. 

 

Given this, we contend that it is of the utmost importance that these foods undertake at least thorough animal testing, and at least the first phase of the four phases of a clinical trial (testing in small numbers of healthy human volunteers), before being released to 18.7 million Australians to eat. 

 

The ANZFA provide evidence that certain DNA sequences are transferred to the GE foods and that others are not.  ANZFA are reminded that the applicant company for these GE foods, Monsanto, have recently admitted that two further DNA sequences are present in their Roundup Ready soybeans that were not previously believed to be there.  How confident is ANZFA that this situation will not happen with these GE foods?

 

The ANZFA argues that 'as DNA from all living organisms is structurally similar, the presence of transferred DNA in food products, in itself, poses no health risk to consumers'.  However, there is also 'considerable structural similarity' between the DNA of food-borne diseases and their hosts, yet these diseases clearly create a problem for their hosts.  Moreover, results from Schubbert et al (Proc Natl Acad Sci 94:961-966) clearly indicate that foreign DNA ingested by mice can reach peripheral leukocytes, spleen and liver via the intestinal wall mucosa and can be found covalently linked to mouse DNA. In addition, viral DNA can cross not only the gut wall, when fed to mice, but the placenta of pregnant mice and can be found in the foetus and in new-born animals (Doerfler and Schubbert (1998) Wien. Klin. Wochenschr 110:40-44).  Furthermore, Professor Hans-Hinrich Kaatz in Germany has recently shown that the gene conferring resistance to glufosinate in canola had transferred to some bacteria and yeast in the gut of bees.

 

 

 

Specific comments about Application A346

Food produced from insect-protected corn line MON 810

 

 

This GE corn has been genetically engineered to contain a gene cassette containing a Cryl(A)b gene from B. Thuringiensis, the E35S promoter from cauliflower mosaic virus, the intron from the maize hsp70 gene and the 3' untranslated region of the nopaline synthase gene (NOS 3') from the Ti plasmid of Agrobacterium tumefaciens. 

 

The expression of the novel protein, Cryl(A)b, designed to rupture the gut of certain grubs, has not been fed to so many people before, nor in such concentrations before, nor in a form before that cannot be removed from the corn by washing or peeling-back the protective 'leaves' of the corn to expose the kernels.   We note that the only reference quoted to support the assertion that 'the toxicity of these proteins is very specific to Lepidoptera and there is no evidence that they are active against non-target insects, birds, fish or mammals' is not a peer-reviewed scientific paper, but only the proceedings from a workshop.  Consequently, we believe that ANZFA need to supply further evidence to support the contention that CrylA(b) is safe for regular, considerable human ingestion.

 

The document states that the 'products from insect-protected corn are largely consumed as processed corn products and the processing is likely to destroy the function of any DNA present in the food.'  However, corn kernels and corn plant waste are likely be fed to animals such as cows, chickens and pigs in its raw form, and we will consume the milk, eggs and meat (often undercooked) from these animals.  This has not been considered nor assessed.

 

In the toxicity studies of the Cry1A(b) protein on mice, the nature of the 'gross pathology' needs to be described to determine whether it was adequately done.  Similarly, it is noted that whilst the CryA(B) B.t.k HD-1 core protein was determined to be rapidly degraded in a simulated in vitro gastric system (by 90% in 2 minutes), the protein was extremely long-lived in simulated intestinal fluid.  Furthermore, although the protein works on the gut wall of insects, these simulated systems appear not to have contained human gastric or intestinal mucosa, so that the activity of this protein on these mucosa appears not to have been determined in these studies. It is noted that there is no description of how these simulated systems match true in vivo conditions.  The ANZFA are asked to provide a description of how similar they are.

 

Although Monsanto considered the compositional analyses of silage, these results do not appear to be presented here, as only results from fields trials appear to be presented.  The results presented for proximate analyses, amino acids, fatty acids, carbohydrates, tocopherols and inorganics all appear to have been done on a sample size of n=6.  If significant differences existed, they would be unlikely to be found with such inadequate sample sizes.  The experiments need to be repeated with larger sample sizes, such as n=50.  Even so, statistically significant differences were found in approx half of the amino acids tested (Table 4 of the document).  Furthermore, the text associated with Table 4 does not match the figures given in Table 4.  The text states that 'the values were within the values reported in the literature'.  However, it is clear in Table 4 that this is not the case for cysteine and histidine, which are both higher than the quoted literature range. Moreover, phenylalanine, proline, serine and tyrosine are all at the very high end of the literature range, while tryptophan is at the very low end of that range.  Better sample sizes and the provision of confidence intervals would have made it clearer how these amino acids correspond with the literature range.  Moreover, it is stated in the caption of Table 4 that the level of tryptophan was significantly different in line MON 810 compared to its control, yet both have a mean of 0.6 in the table.  It is highly unlikely that 2 samples with the same mean would be significantly different, particularly with such low sample sizes.  It is suggested that ANZFA should check the figures in Table 4 to determine if errors are present.

 

We request the raw data, mean, standard deviation, 95% confidence intervals, statistical test used and p value for all 18 amino acids measured for MON 810, Control (818) and Lines 800/801 as well as the literature ranges for these.   Also, please provide reasons why Monsanto did not measure all amino acids, or if they did, why the results are not given.

 

It is of concern to us that the building-blocks of proteins have been significantly altered in the GE corn, indicating that the protein composition of the corn has also been altered.  Given that the text states that the expected new protein constitutes less than 0.001% of the total protein in each tissue of leaf, kernel, whole plant tissue and pollen, the change in amino acid profile cannot be attributed to the presence of this new, expected protein in the plant.  It indicates that other proteins may have been produced, which may be potentially toxic.  It is therefore of concern to us that these results have not been followed-up with animal experiments to determine if any new, unexpected proteins may cause disease.  Similarly, it is noted that the concentrations of calcium and beta tocopherols were also significantly different in the GE corn compared to its control.  Given all of these results, it is unlikely that the corn could be regarded as 'substantially equivalent'. 

 

It is also of interest that data on trypsin inhibitors were not obtained from the same plants as the other analyses.  Instead, it appears that the GE corn MON 810 was not used, but that 'seven hybrid MON 810 corn corn (sic) lines' were used, ie that hybrids of MON 810 with another corn were used, significance tests appear not to have been done and results were only expressed as a range.  Why were means, standard deviations, confidence intervals, the type of statistical test and p values not given for trypsin inhibitor levels, what was MON 810 hybridised with and why were hybrid plants sampled for this test?  This is of particular interest, as the produced ranges indicate that the hybrid of MON 810 may have a significantly higher level of trypsin inhibitor than the control upon significance testing.  If this is the case, then the non-hybrid MON 810 may have even higher levels. 

 

Therefore, FLRAG contends that insect-resistant corn line MON 810 should not be permitted to be sold in Australia on the basis that:

·        It has undergone insufficient safety testing

·        The testing that has been done has been inadequately done

·        The results obtained even from inadequate testing indicate a difference in the composition of the plant which may result in the production of toxic and other substances such as anti-nutrients, such as trypsin inhibitors.

Instead, we contend that comprehensive testing by independent researchers on the whole food as described under 'Animal and Human Testing' above, should be undertaken and the corn line MON 810 only released for consumption if the results from these tests show the corn to be safe.

 

 

 

Specific comments about Application A362

Food produced from glyphosate-tolerant corn line GA21

 

 

This corn has been genetically engineered to contain a gene cassette containing a modified corn EPSPS gene fused to a chloroplast transit peptide (CTP) sequence which has been derived from sequences obtained from corn and sunflower, the actin 1 promoter and first intron from rice and the 3' untranslated region of the nopaline synthase gene (NOS 3') from the Ti plasmid of Agrobacterium. 

 

The modified EPSPS (mEPSPS) was produced by taking the wildtype EPSPS gene from corn (Zea mays) and introducing 2 changes using in vitro techniques, which resulted in the production of a modified EPSPS protein with a lower affinity for the herbicide glyphosate.  This allowed the enzyme, and therefore the plant, to continue to function normally in the presence of the herbicide.  The specific amino acid changes to the EPSPS protein are not available to the public, because they have been classified as 'confidential commercial information as defined by Section 3 of the ANZFA Act (1991) as amended'.  We have concerns about this.  One reason for concern is that the text states that only 2 out of 445 amino acids are different in the modified EPSPS protein.  However, the text also states that the modified EPSPS protein 'shows more than 99.3% homology' with the conventional corn.  These two statements are clearly contradictory.  The true situation cannot be determined without looking at the amino acid structure of the modified EPSPS protein.  Furthermore, we believe that it is inappropriate to withhold the known structure of a new protein in the diet from medical and public health experts, and the general public who are consuming it.  As a result, the FLRAG wishes to lodge a formal protest about the 'confidential commercial information' classification of the modified EPSPS protein, as it considers health and safety considerations and the public's right to know to be more important than any commercial considerations of the applicant company.  Consequently, the FLRAG requests the public release of the structure of the modified EPSPS protein. 

 

The FLRAG seeks assurances that the gene cassette has not been incorporated into a coding region of corn DNA, particularly as "these studies did not provide molecular detail about the ends of the inserted DNA, particularly where these are adjacent to flanking corn DNA".

 

In addition, the new DNA sequences were incorporated into corn DNA using a 'particle acceleration transformation system'.  This system incorporates new DNA into the plant's DNA in a random position.  It also provides for the possibility that partial segments of introduced DNA could be incorporated into positions other than the single position described above.  Given Monsanto's recent admission that two sections of DNA have been found in their Roundup Ready soybeans that were not previously expected to be there, how confident is the ANZFA that the multiple copies of DNA described by Monsanto are the only pieces of DNA that have been introduced into the genome of the corn and that no other unexpected DNA sequences are to be found in the gene cassette?

 

The ANZFA document concludes that 'there is no biological potential for the transfer of novel genetic material from corn line GA21 to intestinal microorganisms, as a result of the genetic modification'.  However, recent results from Professor Hans-Hinrich Kaatz in Germany demonstrate this to be incorrect.  That is, he found that the gene conferring resistance to glufosinate in canola had transferred to some bacteria and yeast in the gut of bees.  The gene producing the novel mEPSPS protein contains novel genetic material, by definition.

 

The document states that 'although several methods of analysis showed that the mEPSPS is expressed in the edible grain from the plant at levels approximately ten times higher than endogenous EPSPS expression levels, the safety assessment concluded that the higher levels did not raise any health concerns.  The high prevalence of this family of plant and microbiol proteins in the human diet supported this conclusion.'  The document provided supporting data from acute toxicity studies on mice.  However, there are many examples in medicine where the effects of acute exposure on human health have been different to chronic exposure.  In addition, the effects of chronic human exposure at low concentrations may not be the same as chronic exposure at ten times the concentration. Essentially, the effects of feeding people high concentrations of the new protein over tens of years cannot be determined by feeding 20 mice a single oral gavage of a given high concentration of the protein and taking very basic data for 13-14 days, particularly when the protein fed to the mice came from partially-purified mEPSPS protein produced in E. Coli in a laboratory, rather than from corn in a field.

 

The FLRAG wishes to be told the following about the toxicity testing in mice, either because they are not clear from the document, or are not given in the document:

·        Whether there were 20 mice for each of the three target doses or whether there were a total of 20 mice being tested with mEPSPS. 

·        The ages of the mice at the beginning of the study

·        At what stages through the study, clinical observations were performed

·        What items were measured for the clinical observations and observations for toxicity, and the data for these.

·        Body weights and food consumptions over time, and the full results of statistical analysis on these, such as the raw data, means, standard deviations, 95% confidence intervals of the means, the name of the statistical tests used and the p values obtained.  Just providing us with final, cumulative values are not sufficient.  We wish to see trends over time.

·        What was involved with the gross pathology examination and a list of all irregularities that were observed, as well as which group they were observed in.

·        Why no organs were weighed and no tissues were examined microscopically

·        Why the unilateral corneal opacity, noted in one male mouse at the high dose level of the test material was considered not to be treatment related and why further experimentation was not undertaken to determine the proportion of treated mice afflicted.

·        Why Bovine Serum Albumin (BSA) was also used as a control, and why, although it was a control it appears to have been compared to the 'carrier control group' in the statistical analyses.

·        The carrier that was used.

 

The ANZFA document states that while approx. 30% of corn grown in Australia is manufactured into foods for human consumption, the vast majority of the remainder is used as stockfeed or exported.  It appears therefore, that a significant proportion of corn grown in Australia may be fed to stock.  It is expected that the corn will not be 'cooked' before feeding, so that any DNA or protein will not be inactivated before feeding.  We will be consuming the milk, eggs and meat (often undercooked) from these animals.  This has not been considered nor assessed.

 

The composition of the glyphosate-tolerant GE corn was compared to control plants that consisted of 'the population of non-transgenic negative segregants (that is, plants lacking the mEPSPS gene addition) present in untreated plots of transgenic GA21 corn'.  It would therefore appear that the control plants were those plants that had undergone the 'particle acceleration transformation system' to try to make them into GA21 plants, but where the process appears to have failed.  Can the ANZFA please provide details as to why these plants were chosen as controls as it is contended that a more suitable control would have been plants that had not been treated at all in this manner.  This is because of the possibility that minor sequences of DNA may have been incorporated into parts of the genome of these plants which may not have been picked-up, but which may nevertheless alter the concentrations of amino acids or fatty acids.

 

For reasons described under 'Statistical Analyses on Compositional Studies', we contend that the compositional analyses require much larger sample sizes than n=5.  At these very small sample sizes, it would be very hard indeed to pick-up statistically-significant differences even if they existed.  Moreover, the mean, standard deviation, 95% confidence interval of the mean, the nature of the statistical test (eg t-test) and p-values are all required for each amino acid and fatty acid measurement.  Also, please provide information as to why Monsanto did not measure all amino acids, and only these few fatty acids.  Also, please describe what the 'least squares mean' is as it does not seem to appear in statistics text books and a number of statisticians consulted were unaware of its existence, and please provide reasons why this type of mean was used rather than a standard measure of central tendency.

 

In spite of the very low sample sizes, statistically-significant differences were found for 5 amino acids in the GE corn treated with glyphosphate.  These changes in amino acid profile cannot be explained by the action of glyphosate, as the action of glyphosate is to 'specifically bind to and block the activity of 5-enolpyruvylshikimate-3-phosphate synthetase (EPSPS), an essential enzyme involved in the biosynthesis of aromatic amino acids in all plants, bacteria and fungi'.  That is, the action of glyphosate would be to reduce the concentration of the amino acids phenylalanine, tyrosine and tryptophan.  Yet these are not the amino acids that have been significantly changed in the glyphosate-treated corn line GA21 (Table 2 of ANZFA's document).  The ANZFA document also states that 'there is only one new protein, namely the mEPSPS enzyme, produced by the genetic modification to the corn plants.'  The document also clearly states that this is different from the wildtype EPSPS by only two amino acids.  If this is the case, then only 2 amino acids at the most would be expected to be significantly different in the GE corn compared to the wildtype corn.  Yet this is clearly not the case.  That is, even with the tiny sample sizes used, five amino acids are significantly different.  The concentration of 2 of these, being lysine and arginine, have been changed by more than 8%.  We are unable to determine if changes in these 2 amino acids reflect the 2 amino acids that have been changes in the mEPSPS protein, as this is 'confidential commercial information'.  However, as the mEPSPS protein is only present at 'approximately 0.01% of the total protein in grain [in] transgenic corn line GA21', and these 2 amino acids have changed by over 8%, the structure of the mEPSPS protein is unlikely to be the cause of these changes in amino acid composition.  It is probable that even more differences in amino acid composition may have been found with larger sample sizes.  Consequently, we argue that the different amino acid concentrations found in this GE corn cannot be explained either by the expression of the only expected protein, nor by the action of the herbicide.  As amino acids are the building blocks of proteins, the results indicate that other proteins may have been produced.  These may be potentially toxic.  It is therefore of concern that these results have not been followed-up with experiments of the whole food to determine if any new, unexpected substances are present which may cause disease.  The experiments that we believe should be done are listed under "Animal and Human Testing", above.

 

In Table 2, the concentration of serine is statistically different between the control (untreated) corn line and the untreated corn line GA21.  However, both have a mean of 5.3%.  It is highly unlikely that 2 samples with the same mean would be statistically different, particularly with such low sample sizes.  It is suggested that ANZFA should check the figures in Table 2 to determine if errors are present.  We contend that these figures should also have been given to 2 decimal places.

 

The document also states that only one fatty acid was found to be statistically different, stearic acid.  If significant differences existed in other fatty acids, they would be unlikely to be picked-up with such low sample sizes.  We believe that the experiment should be repeated with much larger sample sizes.  The fact that stearic acid was nevertheless found to be different indicates that changes to the structure of the corn genome may also have resulted in changes to the fatty acid composition of the corn as well as to the amino acid composition.  Did Monsanto look for any unexpected fatty acids as a result?  If not, why not?

 

Therefore, the FLRAG contends that glyphosate-tolerant corn line GA21 should not be permitted to be sold in Australia on the basis that:

·        It has undergone insufficient safety testing

·        The testing that has been done has been inadequately done

·        The results obtained even from inadequate testing indicate a difference in the composition of the plant which may result in the production of toxic and other substances.

Instead, we contend that comprehensive testing by independent researchers on the whole food as described under 'Animal and Human Testing' above, should be undertaken and the corn line GA21 only released for consumption if the results from these tests show the corn to be safe.

 

 

 

Specific comments about Application A363

Food produced from glyphosate-tolerant canola line GT73

 

 

The document states that the principle food product of canola is oil, which can be used in a variety of manufactured foods such as salad and cooking oil, margarine, shortening, mayonnaise, sandwich spreads, creamers and coffee whiteners.

 

This GE canola (GT73) has been genetically engineered from the parental canola line (Westar) to contain two genes that confer tolerance to the herbicide glyphosate (ie Roundup).  One is the 5-enolpyruvylshikimate-3-phosphate synthetase (CP4-EPSPS) gene from Agrobacterium fused with a modified figwort mosaic virus 35S promoter, a chloroplast transit peptide (CTP) sequence from Arabidopsis thaliana and a terminator region from the 3' end of the pea rbcS E9 gene.  Glyphosate normally inhibits the synthesis of aromatic amino acids.  Together, these sequences permit the plant to continue to synthesise these amino acids when sprayed with glyphosate. 

 

The other gene sequence is the gox (glyphosate oxidoreductase) gene from Ochromobactrum anthropii, a soil bacterium, fused with a modified figwort mosaic virus 35S promoter, a chloroplast transit peptide sequence from Arabidopsis thaliana and a terminator region from the 3' end of the pea rbcS E9 gene.  The gox gene has been modified to improve the affinity of the enzyme for glyphosate (the gox variant GOXv247).  The resultant protein is a novel protein that differs by 3 amino acids to the normal gox protein.  Together, these sequences permit the plant to breakdown glyphosate to aminomethylphosphonic acid and glyoxylate, thereby inactivating the herbicide.

 

These were transferred using Agrobacterium-mediated transformation.  The ANZFA document states that only one complete copy of each of these 2 sequences was transferred.

 

The FLRAG seeks assurances that the gene cassette has not been incorporated into a coding region of the canola DNA.  Furthermore, given Monsanto's recent admission that two sections of DNA were unexpectedly found in their Roundup Ready soy, how confident is the ANZFA that only one copy of each of these sequences has been incorporated into the plant's DNA and that no other unexpected DNA sequences are present anywhere in the canola DNA?

 

The document states in several places the point that:  'processing of canola seed to oil involves the removal of all DNA and protein which effectively results in the removal of the CP4 EPSPS and GOX proteins from the food fraction'.  Yet in other places, it is stated that protein (and presumably DNA) is present in canola oil.  Moreover, a concentration is even given for protein in refined oil, at 0.290 ppm for GT73 and 0.327 for the parent line, Westar.  This is clearly contradictory.  Moreover, no allowance is given for cold-pressed oil which may contain a much higher proportion of protein and DNA, as it may not go through the extent of processing as described in the document.  Clearly then, the CP4 EPSPS protein and the GOX novel protein would in fact be present in canola oil, albeit at very low levels.

 

The document states that 'it is extremely unlikely that novel genetic material will transfer from GM foods to bacteria in the human digestive tract because of the number of complex and unlikely steps that would need to take place consecutively'.  However, recent results from Professor Hans-Hinrich Kaatz in Germany demonstrated that this could in fact happen.  He found that the gene conferrring resistance to glufosinate in canola had transferred to some bacteria and yeast in the gut of bees.

 

The document then goes on to state that 'it is equally unlikely that novel genetic material will transfer from GM foods to human cells via the digestive tract'.  However, results from Schubbert et al (Proc Natl Acad Sci 94:961-966) clearly indicate that foreign DNA ingested by mice can reach peripheral leukocytes, spleen and liver via the intestinal wall mucosa and could be found covalently linked to mouse DNA.  In addition, viral DNA can cross not only the gut wall but the placenta of pregnant mice, to be found in the foetus and in new-born animals (Doerfler and Schubbert (1998), Wien. Klin. Wochenschr. 110:40-44).

 

The document also states that 'current scientific knowledge has not revealed any DNA sequences from ingested foods that have been incorporated into human DNA'.  We are not aware of any GE food feeding trials done on people, so this result is not surprising, as it simply reflects a lack of experimentation.

 

Consequently, we argue that ANZFA's conclusion that 'the horizontal gene transfer of any genetic material from the glyphosate tolerant canola, whether novel DNA or not, is not considered to pose any risk to public health and safety' may need to be revised.

 

We contend that the reporting of the results of the compositional analyses are most inadequate.  In particular, we contend that they require much larger sample sizes.  Some sample sizes are as low as 2!  It is therefore little wonder that statistically significant differences were not found for so many comparisons between the GE canola and it parent line.  Moreover, the mean standard deviation, 95% confidence interval of the mean, the nature of the statistical test (eg t-test) and p-values are all required for each amino acid and fatty acid measured.  Also, please provide information why Monsanto didn't measure all amino acids, and only these fatty acids.

 

The document spends some length on the concentrations of two naturally-occurring toxins, erucic acid and the glucosinolates.  The document states that erucic acid has cardiopathic potential and glucosinolates have goitrogenic properties which make rapeseed unsuitable for human consumption.  Canola has been bred from rapeseed to have lower levels of these and must conform to a CODEX (1993) standard of <2% erucic acid in oil and < 30mmol/g (12g/kg) glucosinolates in defatted, toasted canola meal.  The document also states that canola meal is not regarded as a food for human consumption, but is used in animal feeds.

 

Compositional analyses on oil from GE canola, GT73,  purport to show that this GE canola meets the standards for erucic acid based on sample sizes of between 2 and 7, depending on the year the samples were taken.  We contend that these sample sizes are most inadequate to indicate the true concentration of erucic acid, particularly as it may change with year, soil and climatic conditions around the world.  Moreover, when the concentrations of erucic acid are compared between GT73 and the control (Westar), a mean for GT73 is compared to a range for Westar.  We contend that means, standard deviations, 95% confidence intervals of the means, the statistical test used and p-values are all required for analyses such as these, as are much larger sample sizes.

 

The document states that of the glucosinolates, 5 compounds, referred to as alkyl glucosinolates are thought to have most of the anti-nutritional properties, and that the sum of four of these, which were listed, must be less than 30 mmol/g.  Some other glucosinolates also tend to be monitored.  The document describes the processing steps used to remove the oil from the canola seed and concludes that 'these processing steps as well as the final refinement effectively remove glucosinolates from the refined bleached deoderised oil'.  However, oil can also be sold as cold-pressed oil, and there is the risk that glucosinolates may be present in high concentrations in any cold-pressed canola oil.

 

The glucosinolate concentrations are presented in Table 5 of the document.  The alkyl glucosinolate and thioalkyl glucosinlates from 1992 field trials were about twice as high in the GE canola as its parental control.  Alkyl glucosinolates in the GE canola remain high in the 1993 field trials.  The document notes that they are well below the industry limit of 30 mmol/g.  However, this conclusion is based on sample sizes of between 2 and 7, depending on the year. We contend that these sample sizes are most inadequate to indicate the true concentration of glucosinolates and whether they conform to CODEX standards, particularly as the concentration of this family of compounds may change with year, soil and climatic conditions around the world.  Moreover, we contend that means, standard deviations, 95% confidence intervals of the means, the statistical test used and p-values are all required for analyses such as these, as are much larger sample sizes. 

 

The document replies to the Consumers' Federation of Australia Incorporated's concern that anti-nutrient levels in canola are safe and will not rise over time, by stating that 'the anti-nutrient levels of the genetically modified canola have not increased over the three years of field trials conducted'.   However, the document also states that 'in the 1994 field trial, Cargill used an alternative technique to determine the glucosinolates content, which makes direct comparison to previous years' values invalid'.  That is, the result from the third year's trials cannot be compared to the results from the other years' trials.  Therefore, any trend should be determined from only 2 years' data, a situation which is inadequate to indicate trend.

 

We seek to know why the method to measure glucosinolates was changed.  If both methods are measuring glucosinolates, but the results are not comparable, then this indicates that one method must be returning inaccurate results.  Which one is doing this?  Was the error an over or under estimate?  This adds to the importance of independent assessments of this GE food and the need for larger sample sizes, because if it is assumed that Monsanto moved to a more (rather than less) accurate measurement of glucosinolates with the later Cargill method, then the 1992 and 1993 measurements would be inaccurate and the decision that glucosinolates in the GE canola are below the CODEX standards lies with just the 1994 analyses, which use a mere sample size of 2!

 

The document provided data from acute toxicity studies of various proteins on mice.  However, there are many examples in medicine where the effects of acute exposure on human health have been different to chronic exposure.  In addition, the effects of chronic human exposure at low concentrations may not be the same as chronic exposure at many times the concentration. Essentially, we contend that the effects of feeding people high concentrations of the new protein over tens of years cannot be determined by feeding 20 mice a single oral gavage of a given high concentration of the protein and taking very basic data for 8-9 days.  This is particularly so when the CP4 EPSPS protein used in these studies was not produced by canola in vivo, but produced by bacteria in vitro and not bound to the CTP.

 

The FLRAG whishes to be told the following about the toxicity testing in mice, either because they are not clear from the document, or are not given in the document:

·        The number of mice in each dosage group of each protein tested. 

·        The ages of the mice at the beginning of the study

·        At what stages through the study, clinical observations were performed

·        What items were measured for the clinical observations and observations for toxicity, and the data for these.

·        Body weights and food consumptions over time, and the full results of statistical analysis on these, such as the raw data, means, standard deviations, 95% confidence intervals of the means, the name of the statistical tests used and the p values obtained.  Just providing us with final, cumulative values are not sufficient.  We wish to see trends over time.

·        The number of mice that died, when they died and in which group they were in.

·        What was involved with the gross pathology examination and a list of all irregularities that were observed, as well as which group they were observed in.

·        Organ weights if they were done.  If not done, why it was not done. 

·        The results of tissue microscopy, if done.  If not done, why this was not done.

 

It is noted that the CP4 EPSPS protein, while it had a short half life of less than 15 seconds in the simulated gastric system had a much longer half-life of 10 minutes in the simulated intestinal system.  It is also noted that while the gox protein had a short half life of less than 30 seconds in the simulated intestinal system, no half life was given for the simulated gastric system.  It is also noted that there is no description of how these simulated systems match true in vivo conditions.  The ANZFA are asked to provide a description of how similar they are.

 

In Table 6, the profile of 4 fatty acids that lie outside the CODEX specifications for canola for both the GE canola and its non GE parent are given for refined, bleached and deodorised oil.  The unit of measurement has not been given, and needs to be, as does the sample size, whether the figures are means or not, standard deviations, 95% confidence interval of the means, the nature of the statistical test and the p-value.  It appears that in fact a statistical comparison may not have been made between the fatty acid profiles of the GE and non GE canola in this table.  We believe that it needs to be done.  Also, the figures given in this table seem to be at odds with the figures for fatty acids given in other tables.  For example, the concentration of arachidic acid (20:0) for Westar is given as 1.02 in Table 6.  However, in Table 9.1, the range for this is given as 0.6-0.8 for 1992 and 0.6-0.7 for 1993.  Clearly, the 2 tables contradict.  Which one is correct?

 

In Table 6, the document states that there is no nutritional or toxicological significance associated with these fatty acids.  If this is the case, can the ANZFA please inform us as to why these fatty acids need to be kept within certain low levels in order for the oil to meet CODEX specifications?  The document also states that these raised levels are considered to 'reflect the natural variation within canola rather than any affect of the genetic modification on the canola line'.  Please provide reasons for this assertion, particularly in light of the section: 'Statistical Analyses on Compositional Studies', above. Furthermore, evidence for this assertion could have been provided if a greater sample size had been taken and the results expressed as the mean, standard deviation and 95% confidence interval of the mean, as the 'natural variation' would, by definition, have then been measured directly.  However, this does not appear to have been done.

 

In Table 7, means and ranges are given for proximate analyses of canola.  The GE canola was not found to be significantly different to the parent line.  However, with sample sizes as low as 2, this is not surprising . We contend that the experiment needs to be repeated with much larger sample sizes, such as n=50, and the results need to be given as the mean, standard deviation, 95% confidence interval of the mean, the name of the statistical test and the p value.  There is a very similar problem in Table 8, which gives the mean values for percent protein and percent fat in canola.  Here, in spite of the very small sample sizes, the percent fat content was found to be significantly higher in the GE canola compared to the control.  It is noted that the fat content was also higher in the 1993 and 1994 field trials, but was not found to be significantly different, probably due to the very small sample sizes.  Therefore, the mean fat values may not be due to 'the natural range of variation that occurs in canola', as stated.  Similarly, on the basis of very small sample sizes, the document states that 'there were no differences in fatty acid profiles between mean values for the treated or untreated GT73 and Westar seed'.  The same criticisms hold for these analyses, as for the other tables.  Also, in Table 9.1, while a mean is given for the GE canola (GT73) only a range is given for Westar and the Co-op values.  Once again, these need to be given as the mean, standard deviation, 95% confidence interval of the mean, the nature of the statistical test and the p value.  Also, in this document, why are some figures given to 2 decimal places when others are given to 1?  For values of this size, we believe that the results should be given to 2 decimal places. 

 

The text states that 'Table 10 lists the amino acids that were found to be slightly lower in the genetically modified canola plants'.  In fact, this table lists amino acids that were different to the controls - some are lower and some are higher.  Why are these results given as a g/100g seed weight when other ANZFA safety assessments for GE foods tend to give the results of a given amino acid concentration as a percentage of the total amino acid pool? 

 

It is noted that the amino acid profile is significantly different between the GE canola and its parental line.  In particular, the concentration of proline is different in the GE canola for all three years.  We argue that the results cannot be explained by the production of the new, expected proteins, EPSPS and gox, as they 'constitute less than 0.02% of the seed on a fresh weight basis'. These results indicate that the building blocks of proteins, the amino acids, have been changed in the GE canola compared to the parental line.  It indicates that other proteins may have been produced, which may be potentially toxic.  It is therefore of concern to us that these results have not been followed-up with animal experiments to determine if any new, unexpected proteins may cause disease.  Instead, the differences were dismissed as being 'within the natural variation range known for canola'. See the section 'Statistical Analyses on Compositional Studies', above for comment of this assertion.

 

In Table 10, no value is given for the mean tryptophan concentration for Westar.  Why has this been omitted?  Also, no sample sizes have been given for this table.

 

In Section 5.2, analyses for sinapines, phytic acid and certain minerals were discussed.  It was concluded that the results were 'the same' or were 'comparable' between the GE canola and its parental control.  More information is required, such as the sample size, means, standard deviations, 95% confidence interval of the mean, statistical test used and p value for each analysis.

 

Finally, the document also presents the results from studies used to assess the 'ability to support typical growth and well-being', using rats.  In the first study, the rats (20 per group) were fed various concentrations (0, 5 or 15% w/w) of ground but otherwise unprocessed canola as well as processed (toasted and defatted) canola meal for 4 weeks.  However, in this first study, the rats fed GE canola were not fed the GE food under consideration in this document (GT73), but a mixture of it and another GE canola (GT200).  Significantly decreased weight gains were observed in male rats fed the highest concentration of the GE canola mix, whether they were fed the processed or unprocessed canola, when compared to those fed the parental Westar control.  The document stated that there were no differences in food consumption that could account for this result. The absolute and/or liver weights for both GE and non GE canola was increased approx 5-20% compared to controls.