Empirical natural and social sciences produce knowledge (in German: Wissenschaftenschaffen Wissen) which should describe and explain past and present phenomena andestimate their future development. To this end quantitative methods are used.Progress in science needs controversial debates aiming at the best methods as basisfor objective, reliable and valid results approximating what could be the truth. Suchmethodological competition is the energy needed for scientific progress. In thissense, ESEU aims to enable rational discussions dealing with the article from G.-E.Séralini et al. (Food Chem. Toxicol. 2012, 50:4221–4231) by re-publishingit. By doing so, any kind of appraisal of the paper’s content should not beconnoted. The only aim is to enable scientific transparency and, based on this, adiscussion which does not hide but aims to focus methodological controversies.-Winfried Schröder, Editor of the Thematic Series “Implications forGMO-cultivation and monitoring” in Environmental Sciences Europe.


There is an ongoing international debate as to the necessary length of mammaliantoxicity studies, including metabolic analyses, in relation to the consumption ofgenetically modified (GM) plants [1]. Currently, no regulatory authority requires mandatory chronic animal feedingstudies to be performed for edible genetically modified organisms (GMOs), or evenshort-term studies with blood analyses for the full commercial formulations ofpesticides as sold and used, but only for the declared active principle alone. However,several 90-day rat feeding trials have been conducted by the agricultural biotechnologyindustry. These investigations mostly concern GM soy and maize that are engineeredeither to be herbicide-tolerant (to Roundup (R) in 80% of cases), or to produce amodified Bt toxin insecticide, or both. As a result, these GM crops contain newpesticide residues for which new maximum residue levels (MRL) have been established insome countries.

Though the petitioners conclude in general that no major physiological changes isattributable to the consumption of the GMO in subchronic toxicity studies [25], significant disturbances have been found and may be interpreted differently [6, 7]. A detailed analysis of the data in the subchronic toxicity studies [25] has revealed statistically significant alterations in kidney and liverfunction that may constitute signs of the early onset of chronic toxicity. This may beexplained at least in part by pesticide residues in the GM feed [6, 7]. Indeed, it has been demonstrated that R concentrations in the range of103 times below the MRL can induce endocrine disturbances in human cells [8] and toxic effects thereafter [9]. This may explain toxic effects seen in experiments in rats in vivo[10] as well as in farm animals [11]. After several months of consumption of an R-tolerant soy, the liver andpancreas of mice were affected, as highlighted by disturbances in sub-nuclear structure [1214]. Furthermore, this toxic effect was reproduced by the application of Rherbicide directly to hepatocytes in culture [15].

More recently, long-term and multi-generational animal feeding trials have beenperformed, with some possibly providing evidence of safety, while others conclude on thenecessity of further investigation because of metabolic modifications [16]. However, in contrast with the study we report here, none of these previousinvestigations have included a detailed follow-up of the animals, including multiple (upto 11) blood and urine sampling over 2 years, and none has investigated either the GMNK603 R-tolerant maize or Roundup.

Furthermore, evaluation of long-term toxicity of herbicides is generally performed onmammalian physiology employing only their active principle, rather than the completeformulations as used in agriculture. This was the case for glyphosate (G) [17], the declared active chemical constituent of R. It is important to note thatG is only able to efficiently penetrate target plant organisms with the help ofadjuvants present in the various commercially used R formulations [18]. Even if G has shown to interact directly with the active site of aromataseat high levels [19], at low contaminating levels, adjuvants may be better candidates than G toexplain the toxicity or endocrine disruptive side effects of R on human cells [8, 20] and also in vivo for acute toxicity [21]. In this regard, it is noteworthy that the far greater toxicity of fullagricultural formulations compared to declared supposed active principles alone hasrecently been demonstrated also for six other major pesticides tested in vitro[22]. When G residues are found in tap water, food, or feed, they arise from thetotal herbicide formulation although little data is available as to the levels of the Radjuvants in either the environment or food chain. Indeed, adjuvants are rarelymonitored in the environment, but some widely used adjuvants (surfactants) such asnonylphenol ethoxylates, another ethoxylated surfactant like POEA present in R, arewidely found in rivers in England and are linked with disruption of wildlife sexualreproduction [23]. Adjuvants are found in groundwater [24]. The half-life of POEA (21 to 42 days) is even longer than for G (7 to 14days) in aquatic environments [25]. As a result, the necessity of studying the potential toxic effects of totalchemical mixtures rather than single components has been strongly emphasized [2628]. On this basis, the regular measurement of only G or other supposed activeingredients of pesticides in the environment constitute at best markers of fullformulation residues. Thus, in the study of health effects, exposure to the dilutedwhole formulation may be more representative of environmental pollution than exposure toG alone.

With a view to address this lack of information, we performed a 2-year detailed ratfeeding study. Our study was designed as a chronic toxicity study and as a directfollow-up to a previous investigation on the same NK603 GM maize conducted by thedeveloper company, Monsanto [3]. A detailed critical analysis of the raw data of this subchronic 90-day ratfeeding study revealed statistically significant differences in multiple organ functionparameters, especially pertaining to the liver and kidneys, between the GM and non-GMmaize-fed group [3, 7]. However, Monsanto's authors dismissed the findings as not‘biologically meaningful’ [3], as was also the case with another GM corn [29]. The European Food Safety Authority (EFSA) accepted Monsanto's interpretationon NK603 maize [30], like in all other cases.

Our study is the first and to date the only attempt to follow up Monsanto'sinvestigation and to determine whether the differences found in the NK603 GM maize-fedrats, especially with respect to liver and kidney function, were not biologicallymeaningful, as claimed, or whether they developed into serious diseases over an extendedperiod of time.

The Monsanto authors adapted Guideline 408 of the Organization for Economic Co-operationand Development (OECD) for their experimental design [3]. Our study design was based on that of the Monsanto investigation in order tomake the two experiments comparable, but we extended the period of observation fromMonsanto's 90 days to 2 years. We also used three doses of GMOs (instead of Monsanto'stwo) and Roundup to determine treatment dose response, including any possible non-linearas well as linear effects. This allowed us to follow in detail the potential healtheffects and their possible origins due to the direct or indirect consequences of thegenetic modification itself in the NK603 GM maize, or due to the R herbicide formulationused on the GM maize (and not G alone), or both. Because of recent reviews on GM foodsindicating no specific risk of cancer [2, 16], but indicating signs of hepatorenal dysfunction within 3 months [1, 7], we had no reason to adopt a carcinogenesis protocol using 50 rats per group.However, we prolonged to 2 years the biochemical and hematological measurements andmeasurements of disease status, as allowed, for example, in OECD protocols 453 (combinedchronic toxicity and carcinogenicity) and 452 (chronic toxicity). Both OECD 452 and 453specify 20 rats per sex per group but require only 50% (ten per sex per group, the samenumber that we used in total) to be analyzed for biochemical and hematologicalparameters. Thus, these protocols yield data from the same number of rats as ourexperiment. This remains the highest number of rats regularly measured in a standard GMdiet study, as well as for a full formulated pesticide at very low environmentallyrelevant levels.

We used the Sprague-Dawley strain of rat, as recommended for chronic toxicology tests bythe National Toxicology Program in the USA [31], and as used by Monsanto in its 90-day study [3]. This choice is also consistent with the recommendation of the OECD that fora chronic toxicity test, rats of the same strain should be used as in studies on thesame substance but of shorter duration [32]. We then also tested for the first time three doses (rather than the twousually employed in 90-day protocols) of the R-tolerant NK603 GM maize alone, the GMmaize treated with R, and R alone at very low environmentally relevant doses, startingbelow the range of levels permitted by regulatory authorities in drinking water and inGM feed.

Overall, our study is the first in-depth life-long toxicology study on the fullcommercial Roundup formulation and NK603 GM maize, with observations on 34 organs andmeasurement of 56 parameters analyzed at 11 time points for most organs, and utilizing 3doses. We report here the major toxicological findings on multiple organ systems. Asthere was no evidence in the literature on GM food safety evaluation to indicateanything to the contrary, this initial investigation was designed as a full chronictoxicity and not a carcinogenicity study. Thus, we monitored in details chronologicallyall behavioral and anatomical abnormalities including tumors. A full carcinogenicitystudy, which usually focuses only on observing incidence and type of cancers (not alwaysall tumors), would be a rational follow-up investigation to a chronic toxicity study inwhich there is a serious suspicion of carcinogenicity. Such indications had not beenpreviously reported for GM foods.

Our findings show that the differences in multiple organ functional parameters seen fromthe consumption of NK603 GM maize for 90 days [3, 7] escalated over 2 years into severe organ damage in all types of testdiets. This included the lowest dose of R administered (0.1 ppb,50 ng/L G equivalent) of R formulation administered, which is well belowpermitted MRLs in both the USA (0.7 mg/L) [33] and European Union (100 ng/L) [34]. Surprisingly, there was also a clear trend in increased tumor incidence,especially mammary tumors in female animals, in a number of the treatment groups. Ourdata highlight the inadequacy of 90-day feeding studies and the need to conductlong-term (2 years) investigations to evaluate the life-long impact of GM foodconsumption and exposure to complete pesticide formulations.


Biochemical analyses of the maize feed

Standard biochemical compositional analysis revealed no particular differencesbetween the different maize types and diets, the GM and non-GM maize being classifiedas substantially equivalent, except for transgene DNA quantification. For example,there was no difference in total isoflavones. In addition, we also assayed for otherspecific compounds, which are not always requested for establishing substantialequivalence. This analysis revealed a consistent and statistically significant (p< 0.01) decrease in certain phenolic acids in treatment diets, namelyferulic and caffeic acids. Ferulic acid was decreased in both GM maize and GM maize +R diets by 16% to 30% in comparison to the control diet (889 ± 107, 735 ±89, respectively, vs. control 1,057 ± 127 mg/kg) and caffeic acid in the samegroups by 21% to 53% (17.5 ± 2.1, 10.3 ± 1.3 vs. control 22.1 ± 2.6mg/kg).

Anatomopathological observations and liver parameters

All rats were carefully monitored during the experiment for behavior, appearance,palpable tumors, and infections. At least ten organs per animal were weighed and upto 34 analyzed postmortem, at the macroscopic and/or microscopic levels(Table 1). Due to the large quantity of data collected,it cannot all be shown in one report, but we present here the most importantfindings. There was no rejection by the animals of the diet with or without GM maize,nor any major difference in body weight (data not shown).

Table 1 Protocol used and comparison to existing assessment and to non-mandatoryregulatory tests

The most affected organs in males were the liver, hepatodigestive tract, and kidneys(Table 2; Figure 1A,B,C,D,E,F,G,H,I). Liver abnormalities such as hepatic congestions andmacroscopic and microscopic necrotic foci were 2.5 to 5.5 times more frequent in alltreatments than in control groups, where only two rats out of ten were affected withone abnormality each. For instance, there were 5 abnormalities in total in the GMO11% group (2.5 times higher than controls) and 11 in the GMO 22% group (5.5 timesgreater). In addition, by the end of the experiment, Gamma GT hepatic activity wasincreased, particularly in the GMO + R groups (up to 5.4 times higher), this probablybeing reflective of liver dysfunction. Furthermore, cytochrome P450 activitygenerally increased in the presence of R (either in drinking water or in the GMmaize-containing diet) according to the dose and up to 5.7 times greater at thehighest dose.

Table 2 Summary of the most frequent anatomical pathologies observed
Figure 1
figure 1

Anatomopathological observations in rats fed GMO treated or not by Roundupand effects of Roundup alone. Macroscopic (A to D) andmicroscopic (A' and C') photographs show male left kidneys andlivers (E to I) and female pituitaries (J to M), inaccordance to Table 2. The number of each animal andits treatment is specified. Macroscopic pale spots (I) and microscopicnecrotic foci in liver (G clear-cell focus, H basophilic focuswith atypia), and marked or severe chronic progressive nephropathies, areillustrated. In females, pituitary adenomas (K to M) are shownand compared to control (J, rat number and C for control).Apostrophes after letters indicate organs from the same rat.

Transmission electron microscopic observations of liver samples confirmed changes forall treated groups in relation to glycogen dispersion or appearance in lakes,increase of residual bodies and enlargement of cristae in mitochondria(Figure 2, panels 2 to 4). The GM maize-fed groupseither with or without R application showed a higher heterochromatin content anddecreased nucleolar dense fibrillar components, implying a reduced level of mRNA andrRNA transcription. In the GMO + R group (at the highest dose), the smoothendoplasmic reticulum was drastically increased and nucleoli decreased in size,becoming more compact. In the R alone treatment groups, similar trends were observed,with a partial resumption of nucleolar activity at the highest dose.

Figure 2
figure 2

Ultrastructure of hepatocytes in male rats from groups presenting thegreatest degree of liver pathology. (1) Typical control rat hepatocyte(bar 2 μm except in 4). (2) Effects with Roundup at the lowestdose. Glycogen (G) is dispersed in the cytoplasm. L, lipid droplet; N, nucleus;R, rough endoplasmic reticulum. (3) Details of treatment effects with22% dietary GMO (bar 1 μm). a, cluster of residual bodies (asterisks); b,mitochondria show many enlarged cristae (arrows). (4) Hepatocytes ofanimal fed GM maize (GMO) at 22% of total diet. Large lakes of glycogen occurin the cytoplasm. M, mitochondria.

Degenerating kidneys with turgid inflammatory areas demonstrated the increasedincidence of marked and severe chronic progressive nephropathies, which were up totwo fold higher in the 33% GM maize or lowest dose R treatment groups(Table 2; Figure 1, firstline).

Biochemical analyses of blood and urine samples

Biochemical measurements of blood and urine were focused on samples taken at the 15thmonth time point, as this was the last sampling time when most animals were stillalive (in treated groups 90% males, 94% females, and 100% controls). Statisticalanalysis of results employed OPLS-DA 2-class models built between each treated groupper sex and controls. Only models with an explained variance R2(Y) ≥80%, and a cross-validated predictive ability Q2(Y) ≥ 60%, were usedfor selection of the discriminant variables (Figure 3),when their regression coefficients were significant at a 99% confidence level. Thus,in treated females, kidney failures appeared at the biochemical level (82% of thetotal disrupted parameters). Levels of Na and Cl or urea increased in urine with aconcomitant decrease of the same ions in serum, as did the levels of P, K, and Ca.Creatinine and creatinine clearance decreased in urine for all treatment groups incomparison to female controls (Table 3). In GMmaize-treated males (with or without R), 87% of discriminant variables werekidney-related, but the disrupted profiles were less obvious because of advancedchronic nephropathies and deaths. In summary, for all treatments and both sexes, 76%of the discriminant variables versus controls were kidney-related.

Figure 3
figure 3

Orthogonal partial least squares-discriminant analysis (OPLS-DA) forbiochemical data (females fed 33% GMO versus controls). (A) First,detailed examples of significant discriminant variables distribution betweenfemales fed 33% GMO (bold line) and controls (dotted line). On X axis,animals; on Y axis, serum or urine biochemical values for Na, Cl,estradiol, testosterone. (B) Wider view of OPLS-DA regressioncoefficients for predictive component, with jack-knifed confidence intervals at99% confidence level, indicating discriminant parameters versus controls atmonth 15. U, urinary; UEx, excreted in urine during 24 h; APPT, activatedpartial thromboplastin time; MCV, mean corpuscular volume; PT, prothrombinetime; RBC, red blood cells; ALT, alanine aminotransferase; MCHC, meancorpuscular hemoglobin concentration; A/G, albumin/globulin ratio; WBC, whiteblood cells; AST, aspartate aminotransferase. Profiles evidence kidney ionleakages and sex hormonal imbalance versus controls.

Table 3 Percentage variation of parameters indicating kidney failures of femaleanimals

Furthermore, in females (Table 3), the androgen/estrogenbalance in serum was modified by GM maize and R treatments (at least 95% confidencelevel, Figure 3). For male animals at the highest Rtreatment dose, levels of estrogens were more than doubled.

Tumor incidence

Tumors are reported in line with the requirements of OECD chronic toxicity protocols452 and 453, which require all ‘lesions’ (which by definition includetumors) to be reported. These findings are summarized in Figure 4. The results are presented in the form of real-time cumulative curves(each step corresponds to an additional tumor in the group). Only the growing largestpalpable growths (above a diameter of 17.5 mm in females and 20 mm in males) arepresented (for example, see Figure 5A,B,C). These werefound to be in 95% of cases non-regressive tumors (Figure 5D,E,F,G,H,I,J) and were not infectious nodules. These arose from time totime; then, most often disappeared and were not different from controls afterbacterial analyses. The real tumors were recorded independently of their grade, butdependent on their morbidity, since non-cancerous tumors can be more lethal thanthose of cancerous nature, due to internal hemorrhaging or compression andobstruction of function of vital organs, or toxins or hormone secretions. Thesetumors progressively increased in size and number, but not proportionally to thetreatment dose, over the course of the experiment (Figure 4). As in the case of rates of mortality (Figure 6), this suggests that a threshold in effect was reached at the lowerdoses. Tumor numbers were rarely equal but almost always more than in controls forall treated groups, often with a two- to threefold increase for both sexes. Tumorsbegan to reach a large size on average 94 days before controls in treated females andup to 600 days earlier in two male groups fed with GM maize (11 and 22% with orwithout R).

Figure 4
figure 4

Largest non-regressive tumors in rats fed GMO treated or not by Roundup andeffects of Roundup alone. Rats were fed with NK603 GM maize (with orwithout application of Roundup) at three different doses (11%, 22%, and 33% intheir diet; thin, medium, and bold lines, respectively) compared to thesubstantially equivalent closest isogenic non-GM maize (control, dotted line).Roundup was administered in drinking water at three increasing doses, samesymbols, environmental (A), MRL in some agricultural GMOs (B),and half of minimal agricultural levels (C), see ‘Methods’).The largest tumors were palpable during the experiment and numbered from 20 mmin diameter for males and 17.5 mm for females. Above this size, 95% of growthswere non-regressive tumors. Summary of all tumors are shown in the barhistograms: black, non-regressive large tumors; white, small internal tumors;grey, metastases.

Figure 5
figure 5

Examples of female mammary tumors observed. Mammary tumors are evidenced(A, D, H, representative adenocarcinoma, from the samerat in a GMO group) and in Roundup and GMO + Roundup groups, two representativerats (B, C, E, F, I, J fibroadenomas)are compared to controls. A normal representative rat in controls is not shown,only a minority of them having tumors up to 700 days, in contrast with themajority affected in all treated groups. (G) The histologicalcontrol.

Figure 6
figure 6

Mortality of rats fed GMO treated or not with Roundup and effects of Roundupalone. The symbols of curves and treatments are explained in the captionof Figure 4. Lifespan during the experiment for thecontrol group is represented by the vertical bar ± SEM (grey area). In barhistograms, the causes of mortality before the grey area are detailed incomparison to the controls (0). In black are the necessary euthanasia becauseof suffering in accordance with ethical rules (tumors over 25% body weight,more than 25% weight loss, hemorrhagic bleeding, etc.); and in hatched areas,spontaneous mortality.

In female animals, the largest tumors were in total five times more frequent than inmales after 2 years, with 93% of these being mammary tumors. Adenomas, fibroadenomas,and carcinomas were deleterious to health due to their very large size(Figure 5A,B,C) rather than the grade of the tumoritself. Large tumor size caused impediments to either breathing or digestion andnutrition because of their thoracic or abdominal location and also resulted inhemorrhaging (Figure 5A,B,C). In addition, one metastaticovarian cystadenocarcinoma and two skin tumors were identified. Metastases wereobserved in only two cases; one in a group fed with 11% GM maize and another in thehighest dose of R treatment group.

Up to 14 months, no animals in the control groups showed any signs of palpabletumors, whilst 10% to 30% of treated females per group developed tumors, with theexception of one group (33% GMO + R). By the beginning of the 24th month, 50% to 80%of female animals had developed tumors in all treatment groups, with up to threetumors per animal, whereas only 30% of controls were affected. A summary of allmammary tumors at the end of the experiment, independent of size, is presented inTable 2. The same trend was observed in the groupsreceiving R in their drinking water (Figure 4, R treatmentpanels). The R treatment groups showed the greatest rates of tumor incidence, with80% of animals affected (with up to three tumors for one female), in each group.Using a non-parametric multiple comparison analysis, mammary tumor incidence wassignificantly increased at the lowest dose of R compared to controls (p <0.05, Kruskal-Wallis test with post hoc Dunn's test). All females except one(with metastatic ovarian carcinoma) presented in addition mammary hypertrophies andin some cases hyperplasia with atypia (Table 2).

The second most affected organ in females was the pituitary gland, in general aroundtwo times more than in controls for most treatments (Table 2; Figure 1J,K,L,M). Again, at this level ofexamination, adenomas and/or hyperplasias and hypertrophies were noticed. For all Rtreatment groups, 70% to 80% of animals presented 1.4 to 2.4 times more abnormalitiesin this organ than controls.

The large palpable tumors in males (in kidney and mostly skin) were by the end of theexperimental period on average twice as frequent as in controls, in which only oneskin fibroma appeared during the 23rd month. At the end of the experiment, internalnon-palpable tumors were added, and their sums were lower in males than in females.They were not significantly different from controls, although slightly increased infemales (Figure 4, histogram insets).


The rates of mortality in the various control and treatment groups are shown as rawdata in Figure 6. Control male animals survived on average624 ± 21 days, whilst females lived for 701 ± 20 days during theexperiment, plus in each case, a 5-week starting age at reception of animals and a3-week housing stabilization period. After mean survival time had elapsed, any deathsthat occurred were considered to be largely due to aging. Before this period, 30%control males (three in total) and 20% females (only two) died spontaneously, whileup to 50% males and 70% females died in some groups on diets containing the GM maize(Figure 6, panels GMO, GMO + R). However, the rate ofmortality was not proportional to the treatment dose, reaching a threshold at thelowest (11%) or intermediate (22%) amounts of GM maize in the equilibrated diet, withor without the R application on the crop. It is noteworthy that the first two malerats that died in both GM maize-treated groups had to be euthanized due to Wilms'kidney tumors that had grown by this time to over 25% of body weight. This wasapproximately a year before the first control animal died. The first female deathoccurred in the 22% GM maize feeding group and resulted from a mammary fibroadenoma246 days before the first control female death. The maximum difference in males wasfive times more deaths occurring by the 17th month in the group consuming 11% GMmaize and in females six times greater mortality by the 21st month on the 22% GMmaize diet with and without R. In the female cohorts, there were two to three timesmore deaths in all treated groups compared with controls by the end of the experimentand deaths occurred earlier in general. Females were more sensitive to the presenceof R in drinking water than males, as evidenced by a shorter lifespan(Figure 6, panels R). The general causes of deathrepresented in histogram format within each of the panels in Figure 6, are linked mostly to mammary tumors in females and to problemsin other organ systems in males.


This report describes the first long-term (2-year) rodent (rat) feeding studyinvestigating possible toxic effects arising from consumption of an R-tolerant GM maize(NK603) and a complete commercial formulation of R herbicide. The aims of thisinvestigation were essentially twofold. First, to evaluate whether the signs oftoxicity, especially with respect to liver and kidney functions, seen after 90 days'consumption of a diet containing NK603 R-tolerant GM maize [3, 7] escalated into serious ill health or dissipated over an extended period oftime. Second, to determine if low doses of full commercial R formulation at permittedlevels were still toxic, as indicated by our previous in vitro studies [8, 9]. The previous toxicity study with NK603 maize employed only this GM crop thathad been sprayed with R during cultivation [3]. However, in our study presented here, in addition to extending the treatmentperiod from 90 days to 2 years and in order to better ascertain the source of any illhealth observed, we included additional test feeding groups. These consisted of NK603maize grown without as well as with R application and R alone administered via drinkingwater. Furthermore, we used three levels of dosing in all cases rather than the twopreviously used [3], in order to highlight any dose response effects of a given treatment. It isalso important to note that our study is the first to conduct blood, urine, and organanalyses from animals treated with the complete agricultural formulation of R and notjust G alone, as measured by the manufacturer [35].

Our data show that the signs of liver and kidney toxicity seen at 90 days from theconsumption of NK603 GM maize [3, 7] do indeed escalate into severe disease over an extended period. Furthermore,similar negative health effects were observed in all treatment groups (NK603 GM maizewith or without R application and R alone).

What is also evident from our data is that ill effects were not proportional to the doseof either the NK603 GM maize ± R or R alone. This suggests that the observeddisease may result from endocrine disruptive effects, which are known to benon-monotonic. Similar degrees of pathological symptoms occurred from the lowest to thehighest doses, suggesting a threshold effect [36]. This corresponds to levels likely to arise from consumption or environmentalexposure, such as either 11% GM maize in food, or 50 ng/L G equivalent of R-formulation,a level which can be found in some contaminated drinking tap waters and which fallswithin authorized limits.

Death in male rats was mostly due to the development of severe hepatorenalinsufficiencies, confirming the first signs of toxicity observed in 90-day feedingtrials with NK603 GM maize [7]. In females, kidney ion leakage was evident at a biochemical level at month15, when severe nephropathies were observed in dead male animals at postmortem, at theanatomopathological level. Early signs of toxicity at month 3 in kidney and liver werealso observed for 19 edible GM crops containing pesticide residues [1]. It is known that only elderly male rats are sensitive to chronic progressivenephropathies [37]. Therefore, the disturbed kidney functional parameters may have been inducedby the reduced levels of phenolic acids in the GM maize feed used in our study, sincecaffeic and ferulic acids are beneficial to the kidney as they prevent oxidative stress [38, 39]. This possibility is consistent with our previous observation that plantextracts containing ferulic and caffeic acids were able to promote detoxification ofhuman embryonic kidney cells after culture in the presence of R [40]. It is thus possible that NK603 GM maize consumption, with its reduced levelsof these compounds, may have provoked the early aging of the kidney physiology,similarly to R exposure causing oxidative stress [41]. Disturbances in global patterns of gene expression leading to disease viaepigenetic effects cannot be excluded, since it has been demonstrated that numerouspesticides can cause changes in DNA methylation and histone modification, therebyaltering chromatin compaction and thus gene expression profiles [42].

Disturbances that we found to occur in the male liver are characteristic of chronictoxicity, confirmed by alterations in biochemical liver and kidney function parameters.The observation that liver function in female animals was less negatively affected maybe due to the known protection from oxidative stress conferred by estrogen [43]. Estrogen can induce expression of genes such as superoxide dismutase andglutathione peroxidase via the MAP kinase-NF-kB signaling pathway, thus providing anantioxidant effect [43]. Furthermore, liver enzymes have been clearly demonstrated as sex-specific intheir expression patterns, including in a 90-day rat feeding trial of NK603 GM maize [7]. However, in a long-term study, evidence of early liver aging was observed infemale mice fed with R-tolerant GM soy [12]. In the present investigation, deeper analysis at an ultrastructural levelrevealed evidence of impediments in transcription and other defects in cell nuclearstructure that were comparable in both sexes and dose-dependent in hepatocytes in alltreatments. This is consistent with the well-documented toxic effect of very lowdilutions of R on apoptosis, mitochondrial function, and cell membrane degradation,inducing necrosis of hepatocytes, and in other cell lines [8, 9, 44, 45].

The disruptions of at least the estrogen-related pathways and/or enhancement ofoxidative stress by all treatments need further confirmation. This can be addressedthrough the application of transcriptomic, proteomic, and metabolomic methods to analyzethe molecular profile of kidneys and livers, as well as the GM NK603 maize [4648]. Other possible causes of observed pathogenic effects may be due to disturbedgene expression resulting from the transgene insertional, general mutagenic, ormetabolic effects [49, 50] as has been shown for MON810 GM maize [51, 52]. A consequent disruption of general metabolism in the GMO cannot be excluded,which could lead, for example, to the production of other potentially active compoundssuch as miRNAs [53] or leukotoxin diols [54].

The lifespan of the control group of animals corresponded to the mean for the strain ofrat used (Harlan Sprague-Dawley), but as is frequently the case with most mammals,including humans [55], males on average died before females, except for some female treatmentgroups. All treatments in both sexes enhanced large tumor incidence by two- to threefoldin comparison to our controls and also the number of mammary tumors in comparison to theHarlan Sprague-Dawley strain [56] and overall around threefold in comparison to the largest study with 1,329Sprague-Dawley female rats [57]. This indicates that the use of historical data to compare our tumor numbersis not relevant, first, since we studied the difference with concurrent controlschronologically (and not only at the end of the experiment, as is the case in historicaldata), and second, since the diets of historical reference animals may have beencontaminated with several non-monitored compounds including GMOs and pesticides atlevels used in our treatments. In our study, the tumors also developed considerablyfaster than in controls, even though the majority of tumors were observed after 18months. The first large detectable tumors occurred at 4 and 7 months into the study inmales and females, respectively, further underlining the inadequacy of the standard90-day feeding trials for evaluating GM crop and food toxicity [1]. Future studies employing larger cohorts of animals providing appropriatestatistical power are required to confirm or refute the clear trend in increased tumorincidence and mortality rates seen with some of the treatments tested in this study. Asalready stated, our study was not designed as a carcinogenicity study that would haverequired according to OECD the use of 50 rats per sex per group. However, we wish toemphasize that the need for more rats to provide sufficient statistical power may bebiased by the presence of contaminants in the diets used in gathering historical controldata, increasing artificially the background of tumors, which would inappropriately becalled in this case ‘spontaneous’ or due to the genetic strain. Forinstance, toxic, hormonal disrupting or carcinogenic levels of pesticides, PCBs,plasticizers, dioxins, or heavy metals may contaminate the diets or drinking water usedfor the establishment of ‘spontaneous’ tumors in historical data [5862].

In females, induced euthanasia due to suffering and deaths corresponded mostly to thedevelopment of large mammary tumors. This was observed independently of the cancer gradebut according to impact on morbidity. These appeared to be related to the varioustreatments when compared to the control groups. These tumors are generally known to bemostly estrogen-dependent [63]. We observed a strikingly marked induction of mammary tumors in groupsadministered R alone, even at the very lowest dose (50 ng/L G equivalent dilution inadjuvants). At this concentration in vitro, G alone is known to induce humanbreast cancer cell growth via estrogen receptors [64]. In addition, R with adjuvants has been shown to disrupt aromatase, whichsynthesizes estrogen [19], and to interfere with estrogen and androgen receptors in cells [8]. Furthermore, R appears to be a sex endocrine disruptor in vivo inmales [10]. Sex steroid levels were also modified in treated rats in our study. Thesehormone-dependent phenomena are confirmed by enhanced pituitary dysfunction in treatedfemales. An estrogen-modified feedback mechanism may act at this level [65, 66]. The similar pathological profiles provoked by the GM maize + R diet may thusbe explained at least in part by R residues present in this feed. In this regard, it isnoteworthy that the medium dose of the R treatment tested (400 mg/Kg G equivalent)corresponds to acceptable residue levels of this pesticide in some edible GMOs.

Interestingly and perhaps surprisingly, in the groups of animals fed with the NK603 GMmaize without R application, similar effects with respect to enhanced tumor incidenceand mortality rates were observed. For instance, comparing the 11% GMO-treated femalegroup to the controls, the assumption that the tumors are equally distributed isrejected with a level of significance of 0.54% with the Westlake exceedance test [67]. The classical tests of Kolmogorov-Smirnov (one-sided) andWilcoxon-Mann-Whitney reach α values of significance, which are respectively of1.40% and 2.62%.

A possible explanation for this finding is the production of specific compound(s) in theGM feed that are either directly toxic and/or cause the inhibition of pathways, which inturn generates toxic effects. This is despite the fact that the variety of GM maize usedin this study was judged by industry and regulators as being substantially equivalent tothe corresponding non-GM closest isogenic line [3, 30]. As the total chemical composition of the GM maize has not been measured indetail, the use of substantial equivalence as a concept in risk assessment isinsufficient to highlight potential unknown toxins and therefore cannot replacelong-term animal feeding trials for GMOs.

A cause of the ill effects resulting from NK603 GM maize alone observed in this studycould be the fact that it is engineered to overexpress a modified version of theAgrobacterium tumefaciens 5-enolpyruvylshikimate-3-phosphate synthase(EPSPS-CP4) [3], which confers R tolerance. The modified EPSPS is not inhibited by G, incontrast to the wild-type enzyme in the crop. This enzyme is known to drive the firststep of aromatic amino acid biosynthesis in the plant shikimate pathway. In addition,estrogenic isoflavones and their glycosides are also products of this pathway [68]. A limited compositional analysis showed that these biochemical pathways werenot disturbed in the GM maize used in our study. However, our analysis did reveal thatthe levels of caffeic and ferulic acids in the GM diet, which are also secondarymetabolites of the plant shikimate pathway, but not always measured in regulatory tests,were significantly reduced. This may lower their protective effects againstcarcinogenesis and mammalian tumor formation [69, 70]. Moreover, these phenolic acids, and in particular ferulic acid, may modulateestrogen receptors or the estrogenic pathway in mammalian cells [71]. This does not exclude the possibility of the action of other unknownmetabolites. This explanation also corresponds to the fact that the observed effects ofNK603 GM maize and R were not additive but reached a threshold. This implies that boththe NK603 maize and R may cause hormonal disturbances in the same biochemical andphysiological pathways.


In conclusion, the consumption of NK603 GM maize with or without R application or Ralone gave similar pathologies in male and female rats fed over a 2-year period. It waspreviously known that G consumption in water above authorized limits may provoke hepaticand kidney failure [33]. The results of the study presented here clearly indicate that lower levelsof complete agricultural G herbicide formulations, at concentrations well belowofficially set safety limits, can induce severe hormone-dependent mammary, hepatic, andkidney disturbances. Similarly, disruption of biosynthetic pathways that may result fromoverexpression of the EPSPS transgene in the GM NK603 maize can give rise to comparablepathologies that may be linked to abnormal or unbalanced phenolic acid metabolites orrelated compounds. Other mutagenic and metabolic effects of the edible GMO cannot beexcluded. This will be the subject of future studies, including analyses of transgene, Gand other R residue presence in rat tissues. Reproductive and multigenerational studieswill also provide novel insight into these problems. This study represents the firstdetailed documentation of long-term deleterious effects arising from consumption of aGMO, specifically a R-tolerant maize, and of R, the most widely used herbicideworldwide.

Taken together, the significant biochemical disturbances and physiological failuresdocumented in this work reveal the pathological effects of these GMO and R treatments inboth sexes, with different amplitudes. They also show that the conclusion of theMonsanto authors [3] that the initial indications of organ toxicity found in their 90-dayexperiment were not ‘biologically meaningful’ is not justifiable.

We propose that agricultural edible GMOs and complete pesticide formulations must beevaluated thoroughly in long-term studies to measure their potential toxic effects.



The experimental protocol was conducted in an animal care unit authorized by theFrench Ministries of Agriculture and Research (Agreement Number A35-288-1). Animalexperiments were performed according to ethical guidelines of animal experimentations(CEE 86/609 regulation), including the necessary observations of all tumors, in linewith the requirements for a long-term toxicological study [32], up to a size where euthanasia on ethical grounds was necessary.

Concerning the cultivation of the maize used in this study, no specific permits wererequired. This is because the maize was grown (MON-00603-6 commonly named NK603) inCanada, where it is authorized for unconfined release into the environment and foruse as a livestock feed by the Canadian Food Inspection Agency (Decision Document2002-35). We confirm that the cultivation did not involve endangered or protectedspecies. The GM maize was authorized for import and consumption into the EuropeanUnion (CE 258/97 regulation).

Plants, diets, and chemicals

The varieties of maize used in this study were the DKC 2678 R-tolerant NK603(Monsanto Corp., USA), and its nearest isogenic non-transgenic control DKC 2675.These two types of maize were grown under similar normal conditions, in the samelocation, spaced at sufficient distance to avoid cross-contamination. The geneticnature, as well as the purity of the GM seeds and harvested material, was confirmedby qPCR analysis of DNA samples. One field of NK603 was treated with R at 3 Lha−1 (WeatherMAX, 540 g/L of G, EPA Reg. 524-537), and anotherfield of NK603 was not treated with R. Corn cobs were harvested when the moisturecontent was less than 30% and were dried at a temperature below 30°C. From thesethree cultivations of maize, laboratory rat chow was made based on the standard dietA04 (Safe, France). The dry rat feed was made to contain 11%, 22%, or 33% of GMmaize, cultivated either with or without R, or 33% of the non-transgenic controlline. The concentrations of the transgene were confirmed in the three doses of eachdiet by qPCR. All feed formulations consisted of balanced diets, chemically measuredas substantially equivalent except for the transgene, with no contaminatingpesticides over standard limits. All secondary metabolites cannot be known andmeasured in the composition. However, we measured isoflavones and phenolic acidsincluding ferulic acid by standard HPLC-UV. All reagents used were of analyticalgrade. The herbicide diluted in the drinking water was the commercial formulation ofR (GT Plus, 450 g/L of G, approval 2020448, Monsanto, Belgium). Herbicide levels wereassessed by G measurements in the different dilutions by mass spectrometry.

Animals and treatments

Virgin albino Sprague-Dawley rats at 5 weeks of age were obtained from Harlan(Gannat, France). All animals were kept in polycarbonate cages (820 cm2,Genestil, France) with two animals of the same sex per cage. The litter (Toplitclassic, Safe, France) was replaced twice weekly. The animals were maintained at 22± 3°C under controlled humidity (45% to 65%) and air purity with a 12h-light/dark cycle, with free access to food and water. The location of each cagewithin the experimental room was regularly changed. This 2-year life-long experimentwas conducted in a Good Laboratory Practice (GLP) accredited laboratory according toOECD guidelines. After 20 days of acclimatization, 100 male and 100 female animalswere randomly assigned on a weight basis into ten equivalent groups. For each sex,one control group had access to plain water and standard diet from the closestisogenic non-transgenic maize control; six groups were fed with 11%, 22%, and 33% ofGM NK603 maize either treated or not treated with R. The final three groups were fedwith the control diet and had access to water supplemented with respectively 1.1× 10−8% of R (0.1 ppb or 50 ng/L of G, the contaminating levelof some regular tap waters), 0.09% of R (400 mg/kg G, US MRL of 400 ppm G in some GMfeed), and 0.5% of R (2.25 g/L G, half of the minimal agricultural working dilution).This was changed weekly. Twice-weekly monitoring allowed careful observation andpalpation of animals, recording of clinical signs, measurement of any tumors, foodand water consumption, and individual body weights.


Animals were sacrificed during the course of the study only if necessary because ofsuffering according to ethical rules (such as 25% body weight loss, tumors over 25%body weight, hemorrhagic bleeding, or prostration) and at the end of the study byexsanguination under isoflurane anesthesia. In each case, detailed observations andanatomopathology was performed and the following organs were collected: brain, colon,heart, kidneys, liver, lungs, ovaries, spleen, testes, adrenals, epididymis,prostate, thymus, uterus, aorta, bladder, bone, duodenum, esophagus, eyes, ileum,jejunum, lymph nodes, lymphoreticular system, mammary glands, pancreas, parathyroidglands, Peyer's patches, pituitary, salivary glands, sciatic nerve, skin, spinalcord, stomach, thyroid, and trachea. The first 14 organs (at least ten per animaldepending on the sex, Table 1) were weighted, plus anytumors that arose. The first nine were divided into two parts and one half wasimmediately frozen in liquid nitrogen/carbonic ice. The remaining parts includingother organs were rinsed in PBS and stored in 4% formalin before anatomopathologicalstudy. These samples were used for further paraffin-embedding, slides, and HEShistological staining. For transmission electron microscopy, the kidneys, livers, andtumors were cut into 1 mm3 fragments. Samples were fixed in pre-chilled 2%paraformaldehyde/2.5% glutaraldehyde in 0.1 M PBS pH 7.4 at 4°C for 3 h andprocessed as previously described [13].

Biochemical analyses

Blood samples were collected from the tail vein of each rat under short isofluraneanesthesia before treatment and after 1, 2, 3, 6, 9, 12, 15, 18, 21, and 24 months:11 measurements were obtained for each animal alive at 2 years. It was firstdemonstrated that anesthesia did not impact animal health. Two aliquots of plasma andserum were prepared and stored at −80°C. Then, 31 parameters were assessed(Table 1) according to standard methods includinghematology and coagulation parameters, albumin, globulin, total proteinconcentration, creatinine, urea, calcium, sodium, potassium, chloride, inorganicphosphorus, triglycerides, glucose, total cholesterol, alanine aminotransferase,aspartate aminotransferase, gamma glutamyl-transferase (GT), estradiol, andtestosterone. In addition, at months 12 and 24, the C-reactive protein was assayed.Urine samples were collected similarly 11 times, over 24 h in individual metaboliccages, and 16 parameters were quantified including creatinine, phosphorus, potassium,chloride, sodium, calcium, pH, and clearance. Liver samples taken at the end made itpossible to perform assays of CYP1A1, 1A2, 3A4, 2C9 activities in S9 fractions, withglutathione S-transferase and gamma-GT.

Statistical analysis

In this study, multivariate analyses were more appropriate than pairwise comparisonsbetween groups because the parameters were very numerous, with samples of tenindividuals. Kaplan-Meyer comparisons, for instance, were not used because these arebetter adapted to epidemiological studies. Differences in the numbers of mammarytumors were studied by a non-parametric multiple comparisons Kruskal-Wallis test,followed by a post hoc Dunn's test with the GraphPad Prism 5 software.

Biochemical data were treated by multivariate analysis with the SIMCA-P (V12)software (UMETRICS AB Umea, Sweden). The use of chemometrics tools, for example,principal component analysis (PCA), partial least squares to latent structures (PLS),and orthogonal PLS (OPLS), are robust methods for modeling, analyzing, andinterpreting complex chemical and biological data. OPLS is a recent modification ofthe PLS method. PLS is a regression method used in order to find the relationshipbetween two data tables referred to as X and Y. PLS regression [72] analysis consists in calculating by means of successive iterations, linearcombinations of the measured X-variables (predictor variables). These linearcombinations of X-variables give PLS components (score vectors t).A PLS component can be thought of as a new variable - a latent variable - reflectingthe information in the original X-variables that is of relevance formodeling and predicting the response Y-variable by means of the maximizationof the square of covariance (Max cov2(X, Y)). The numberof components is determined by cross validation. SIMCA software uses the nonlineariterative partial least squares algorithm (NIPALS) for the PLS regression. Orthogonalpartial least squares discriminant analysis (OPLS-DA) was used in this study [73, 74].

The purpose of discriminant analysis is to find a model that separates groups ofobservations on the basis of their X variables. The X matrixconsists of the biochemical data. The Y matrix contains dummy variableswhich describe the group membership of each observation. Binary variables are used inorder to encode a group identity. Discriminant analysis finds a discriminant plan inwhich the projected observations are well separated according to each group. Theobjective of OPLS is to divide the systematic variation in the X-block intotwo model parts, one linearly related to Y (in the case of a discriminantanalysis, the group membership), and the other one unrelated (orthogonal) toY. Components related to Y are called predictive, and thoseunrelated to Y are called orthogonal. This partitioning of the Xdata results in improved model transparency and interpretability [75]. Prior to analysis, variables were mean-centered and unit variancescaled.