The events studied are shown in Table 1, the samples are shown in Table 2. After homogenization and grinding the sample, two hundred milligrams were used for DNA extraction by the method of bromide Cetyltrimethyl ammonium bromide (CTAB) as in [2] . The integrity and quantification of extracted DNA were observed using electrophoresis.

The purity of DNA was checked with the inhibition test performed with standard curves from a sample called “undiluted” using endogenous primers, i.e., the values of Ct (Threshold cycle: is the cycle in which each amplification curve crosses the threshold line, serving as a basis for comparison between samples; threshold is the detection threshold set by the user to analyze results at the end of a real-time PCR) of the endogenous gene amplification, were compared with the data extrapolated from the calibration curve. The criteria accepted by the Community Reference Laboratory for Genetically Modified Food and Feed (CRL-GMFF) [3] for the absence of PCR inhibitors is when the average difference (ΔCt) between the measured value and the extrapolated Ct value for the “undiluted” sample is <0.5 cycles and the “slope” between −3.6 and −3.1.

Specific primers designed from the region 5' or 3' end of the genome/insert interaction, with the Primer 3 program were optimized. The initiators of endogenous reference genes for corn, soybean and cotton, respectively, adh1 (ADH, maize alcohol dehydrogenase), lec (LEC, lecithin) and adhC (ADH, cotton alcohol dehydrogenase C gene) were used. The reactions and conditions were optimized for the 7500 Fast Real Time Applied Biosystems (APPLIED BIOSYSTEMS) to a volume of 15.0 µl with 7.5 µl of the mix BRYT^TM Green (Go Taq qPCR Master Mix of PROMEGA).

The efficiency of PCR standard curve was calculated from the value of “slope” being:

^*TL means lepidopteran tolerance and TG glyphosate tolerance.

Standard curves were performed for event and endogenous reference. For each sample, the amount of the event was determined from curves standard, and reference.

The standard curve was taken with 20%, 2.86%, 0.41% and 0.057% of DNA event, mixed with none event, for reaction of 100 ng DNA.

The absolute number of copies of the standard curve was determined by dividing the weight of DNA (nano-

^*Powercore was provided by Down Agroscience, Viptera3 by Syngenta Seeds andIntactaRR2PRO by Nidera Seeds.

grams) by the published average IC value as in [4] of genome DNA, for corn (2725 picograms), soybean (1.13 pg) and cotton (2.33 pg). Table 3 shows the values of copy number of the event in points of standard curve for samples of corn, soybean and cotton

For normalization of quantification of an event in a sample, the event copy number was divided by the copy number of the endogenous gene and multiplied by 100 to yield the percentage value:

The evaluation of linearity, working range, sensitivity, and limit of detection, limit of quantification, repetitiveness, precision, reproducibility, accuracy, and robustness was made according to the parameters defined by CRL- GMFF [3] and DOQ-CGCRE-008-INMETRO [5] . CRL-GMFF [3] gives recommendations to evaluate and validate analytical methods of GMO, according with the Commission regulation (EC) No. 1829/2003 in Europe.

The linear range of work established when the method is linear with an acceptable level of accuracy and precision, is accepted to be 1/10 and at least 5 times the concentration required by legislation (JRRC 56609-Mon810). In Brazil, the GMO limit is 1%, determined by Decree 4680 [1] . Therefore, working range must be from 0.1% to at least 5%.

The parameter used for the sensitivity is the slope, being the acceptance criterion for the standard curve, the average value in the range of −3.1 to −3.6.

By law, the LOQ (limit of quantification) is less than 1/10^th and the LOD (limit of detection), at least 1/20^th of the threshold value; as in Brazil, the limit of GMOs is 1% the limits correspond to 0.1% and 0.05%.

The precision has been achieved by the repeatability and reproducibility. The repeatability of identification by three replicates for each measurement performed in the same analysis, determined by the coefficient of linearity R² (correlation coefficient of a standard curve obtained by linear regression analysis) which should be ≥0.98, and the limit of repeatability (). To a 95% significance level: r = 2.8∙S_r, where S_r is the standard deviation associated with the Ct readings for the same analysis.

The reproducibility was verified by analyses with two different technicians in different days, using the same apparatus under the same conditions of temperature and time previously optimized for each primer; the differences in the percentage of quantitation between the analyses must not be greater than the reproducibility limit R (); or, for a 95% significance level: (where = variance of reproducibility of quantification percentage detected by the two technicians).

The accuracy criterion (agreement between the result of the laboratory and the reference value) defined as ±25% as in [3] , requires a reference value. The value used as a reference was the analysis of Eurofins in Brazil (from Eurofins Agroscience Services), although it was not possible to have all events analyzed. The same samples used for quantification studies were analyzed by Eurofins for validation and comparison of studied methodology.

The relative error (RE) was expressed as a percentage by means of the expression:

where: Xlab = value obtained experimentally or arithmetical average of obtained values; Xv = value accepted as true.

The robustness by the measures of reproducibility and accuracy was inferred within the limits stipulated by CRL-GMFF [3] that shall not deviate more than ±30%.

The performance of the extraction of DNA, which is essential for the success of PCR analysis, was tested for the presence of inhibitors. By the inhibitor test, no samples of corn, soybean and cotton, showed for the average difference (ΔCt) between the measured value and the extrapolated Ct value, ΔCt > 0.5 cycle, indicating the significant absence of inhibitors.

The specificity of the primers developed in the region of genome/insert interaction, was tested by using a reaction of 15.0 µl, with 20 ng of DNA on “FAST” method (initial heating at 95˚C/2min followed by 40 cycles of denaturation, annealing and extension at 95˚C/10s and 60˚C/30s). All analysis showed amplification when performed with specific primer of Table 5 and were completely specific in relation to all events of the other samples studied. Also the dissociation peak showed practically only the correspondent peak for the studied event.

The concentration of primers (forward and reverse) used were the same or almost the one used in the quantification analysis (Table 4).

The linearity measured by PCR efficiency was from 0.9 to 1.01 indicating that the PCR product doubled every

cycle when all reagents are also available, demonstrating the linear response.

The criterion working range 0.1% to until at least 5% was observed for the working range of 0.057% to 20% for the standard curve reaction using 100 ng DNA.

The values of LOD and LOQ (0.1% and 0.05%) were practically followed, as the minimum used was 0.057%.

For all three analysis repetitions, the differences in the percentage of quantitation between the analyses were not greater than the repeatability limit and the R² was ≥0.98.

The variance of reproducibility of quantification for all events (0.33 to 3.0) originated by the variance of the average results of repeating in three days or by different technicians, was lower than the limits of reproducibility and within the parameter set by CRL-GMFF [3] which is ≤30% showing the reproducibility.

The precision was observed for all analysis, as well the repeatability and reproducibility.

Table 6 shows data for validation by comparison the quantification average results (some analyses could not be made by Eurofins) obtained by Eurofins and our laboratory, with the average variance, the coefficient of linearity R² and the relative error (RE). The variance is related to comparison of at least three laboratory analyses on different days.

The average quantification of laboratory tests carried at least three times on different days (at least two repetition tests with the same technician, and one with different technician) ranged from 92 to 115. By CRL-GMFF [3] , the accuracy criterion is ±25%. The results of the relative errors (RE) methodologies between our laboratory and Eurofins (except for PRO, NK603 and GA21, which clearly have had problems because all samples were pure event), were from 2 to 18, below the stipulated by the CRL-GMFF [3] , showing the accuracy of the laboratory

method. As it was found according to the measurement of accuracy and reproducibility that the method was within the limits set by the CRL-GMFF [3] , it may infer the robustness of the methodology.