The experimental protocol was conducted in an animal care unit authorized by the French Ministries of Agriculture and Research (Agreement Number A35-288-1). Animal experiments were performed according to ethical guidelines of animal experimentations (CEE 86/609 regulation), including the necessary observations of all tumors, in line with the requirements for a long-term toxicological study , up to a size where euthanasia on ethical grounds was necessary.
Concerning the cultivation of the maize used in this study, no specific permits were required. This is because the maize was grown (MON-00603-6 commonly named NK603) in Canada, where it is authorized for unconfined release into the environment and for use as a livestock feed by the Canadian Food Inspection Agency (Decision Document 2002-35). We confirm that the cultivation did not involve endangered or protected species. The GM maize was authorized for import and consumption into the European Union (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 same location, spaced at sufficient distance to avoid cross-contamination. The genetic nature, as well as the purity of the GM seeds and harvested material, was confirmed by qPCR analysis of DNA samples. One field of NK603 was treated with R at 3 L ha−1 (WeatherMAX, 540 g/L of G, EPA Reg. 524-537), and another field of NK603 was not treated with R. Corn cobs were harvested when the moisture content was less than 30% and were dried at a temperature below 30°C. From these three cultivations of maize, laboratory rat chow was made based on the standard diet A04 (Safe, France). The dry rat feed was made to contain 11%, 22%, or 33% of GM maize, cultivated either with or without R, or 33% of the non-transgenic control line. The concentrations of the transgene were confirmed in the three doses of each diet by qPCR. All feed formulations consisted of balanced diets, chemically measured as substantially equivalent except for the transgene, with no contaminating pesticides over standard limits. All secondary metabolites cannot be known and measured in the composition. However, we measured isoflavones and phenolic acids including ferulic acid by standard HPLC-UV. All reagents used were of analytical grade. The herbicide diluted in the drinking water was the commercial formulation of R (GT Plus, 450 g/L of G, approval 2020448, Monsanto, Belgium). Herbicide levels were assessed 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 (Toplit classic, 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 12 h-light/dark cycle, with free access to food and water. The location of each cage within the experimental room was regularly changed. This 2-year life-long experiment was conducted in a Good Laboratory Practice (GLP) accredited laboratory according to OECD guidelines. After 20 days of acclimatization, 100 male and 100 female animals were 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 closest isogenic non-transgenic maize control; six groups were fed with 11%, 22%, and 33% of GM NK603 maize either treated or not treated with R. The final three groups were fed with 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 level of some regular tap waters), 0.09% of R (400 mg/kg G, US MRL of 400 ppm G in some GM feed), 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 and palpation of animals, recording of clinical signs, measurement of any tumors, food and water consumption, and individual body weights.
Animals were sacrificed during the course of the study only if necessary because of suffering 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 by exsanguination under isoflurane anesthesia. In each case, detailed observations and anatomopathology 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, parathyroid glands, Peyer’s patches, pituitary, salivary glands, sciatic nerve, skin, spinal cord, stomach, thyroid, and trachea. The first 14 organs (at least ten per animal depending on the sex, Table 1) were weighted, plus any tumors that arose. The first nine were divided into two parts and one half was immediately frozen in liquid nitrogen/carbonic ice. The remaining parts including other organs were rinsed in PBS and stored in 4% formalin before anatomopathological study. These samples were used for further paraffin-embedding, slides, and HES histological staining. For transmission electron microscopy, the kidneys, livers, and tumors 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 and processed as previously described .
Blood samples were collected from the tail vein of each rat under short isoflurane anesthesia 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 first demonstrated that anesthesia did not impact animal health. Two aliquots of plasma and serum were prepared and stored at −80°C. Then, 31 parameters were assessed (Table 1) according to standard methods including hematology and coagulation parameters, albumin, globulin, total protein concentration, creatinine, urea, calcium, sodium, potassium, chloride, inorganic phosphorus, triglycerides, glucose, total cholesterol, alanine aminotransferase, aspartate aminotransferase, gamma glutamyl-transferase (GT), estradiol, and testosterone. 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 metabolic cages, and 16 parameters were quantified including creatinine, phosphorus, potassium, chloride, sodium, calcium, pH, and clearance. Liver samples taken at the end made it possible to perform assays of CYP1A1, 1A2, 3A4, 2C9 activities in S9 fractions, with glutathione S-transferase and gamma-GT.
In this study, multivariate analyses were more appropriate than pairwise comparisons between groups because the parameters were very numerous, with samples of ten individuals. Kaplan-Meyer comparisons, for instance, were not used because these are better adapted to epidemiological studies. Differences in the numbers of mammary tumors 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, and interpreting complex chemical and biological data. OPLS is a recent modification of the PLS method. PLS is a regression method used in order to find the relationship between two data tables referred to as X and Y. PLS regression  analysis consists in calculating by means of successive iterations, linear combinations of the measured X-variables (predictor variables). These linear combinations of X-variables give PLS components (score vectors t). A PLS component can be thought of as a new variable – a latent variable – reflecting the information in the original X-variables that is of relevance for modeling and predicting the response Y-variable by means of the maximization of the square of covariance (Max cov2(X,Y)). The number of components is determined by cross validation. SIMCA software uses the nonlinear iterative partial least squares algorithm (NIPALS) for the PLS regression. Orthogonal partial 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 of observations on the basis of their X variables. The X matrix consists of the biochemical data. The Y matrix contains dummy variables which describe the group membership of each observation. Binary variables are used in order to encode a group identity. Discriminant analysis finds a discriminant plan in which the projected observations are well separated according to each group. The objective of OPLS is to divide the systematic variation in the X-block into two model parts, one linearly related to Y (in the case of a discriminant analysis, the group membership), and the other one unrelated (orthogonal) to Y. Components related to Y are called predictive, and those unrelated to Y are called orthogonal. This partitioning of the X data results in improved model transparency and interpretability . Prior to analysis, variables were mean-centered and unit variance scaled.