Although the debate surrounding food containing genetically modified organisms (GMOs) is still far from a consensus in Europe, a decision to label GM foods to allow consumers an informed choice has been made by some countries and governments. The enforcement of the Swiss Food Regulation1 and of the EU Novel Food Regulation2,3 is based on a PCR detection system specific for the 35S promoter originating from cauliflower mosaic virus (see Fig. 1). This 35S promoter element is often used in GM plants, including those currently approved in Switzerland (Roundup Ready soybean (RRS), Bt-176, and Bt-11 corn).
The qualitative 35S-PCR detection system has been thoroughly tested in collaborative studies; however, its detection limit was found to vary up to a factor of 20 (from 100 pg to 2 ng RRS-DNA) between different analytical laboratories. Additionally, due to its high sensitivity, the 35S-PCR detection system is not suited to distinguish between intended GMO mixtures and GMO comingling due to unseparated channels for conventional and GMO raw materials during harvest, transport, and stocking.
In practice, the 35S-PCR test allows the detection of GMO contents of foods and raw materials in the range of 0.01–0.1%. Attempts have been made to diminish the interlaboratory variations by the standardization of the PCR protocol4, by the use of certified external reference material with well-defined amounts of GMOs5, and by the development of quantitative PCR detection systems6,7.
Recently, the Swiss government revised its food regulations, introducing a threshold value of 1% GMO content as the basis for food labeling. The enforcement of such threshold values clearly requires quantitative detection systems such as quantitative competitive PCR (QC-PCR6,7), real-time PCR8, or immunochemical detection of modified proteins using ELISA systems9. Whereas ELISA is widely acknowledged to be of practical use at the earliest stages of manufacture, there is the disadvantage of protein denaturation as a consequence of processing. Thus, the detection of DNA by quantitative PCR techniques has a number of advantages, including the survival of DNA in many, albeit not all manufacturing processes.
The suitability of QC-PCR for the enforcement of threshold values was successfully ring tested in Switzerland in 1998 by 12 analytical laboratories. Eight food samples—including soy flour, grist, and protein—were analyzed by different QC-PCR systems. By comparing the RRS content of the samples to external reference material5, GMO content could be classified as negative, below 0.5%, between 0.5% and 2%, and above 2%. No false negative results among 246 determinations for the six samples containing RRS and no false positive results for the negative control sample 6 were reported (see Fig. 2). The performance of the RRS-specific and of the 35S promoter–specific QC-PCR were compared by χ2 testing, yielding a contingency coefficient of 0.81. Interlaboratory differences were found to be smaller with QC-PCR compared to qualitative PCR, and were mainly due to insufficient sample homogenization. Furthermore, the precision of the RRS-specific QC-PCR detection system was assessed by four independent determinations of the GMO content of certified reference material containing 0.5% and 2% RRS, respectively. The relative standard errors of the mean were 9% and 2%, respectively, and the discrimination between the two GMO contents was highly significant (p < 0.001). Thus, QC-PCR allows the survey of threshold values for GMO labeling, as mandated in the revised Swiss regulations.
Despite the encouraging results of this ring trial, a number of analytical problems remain to be solved. The application of a soya-specific QC-PCR system revealed that the amount of amplifiable DNA can be strongly affected by food processing, and can vary up to fivefold (see Fig. 3). Thus, for the determination of the relative GMO content of processed food, the results of GMO-specific QC-PCR have to be normalized by the results of a plant-specific QC-PCR system (e.g., specific for soya in the case of RRS).
Such double QC-PCR test systems will also allow the determination of the GMO content of major food ingredients in composed foods. Clearly, the presence of unamplifiable DNA will affect all quantitative PCR detection systems, including real-time PCR. Quantitative PCR systems will be ring tested in Switzerland and in the EU in the near future to provide answers concerning the robustness and precision of quantitative PCR methods. It is important to mention that, in this context, such ring trials should not be based on known certified GMO reference material, since this will unequivocally influence the final results.
The introduction of threshold values for GMO labeling will have an impact on sample sizes and sampling plans that needs to be addressed urgently. Besides statistical considerations, the DNA content of a processed food ingredient might influence the sample size. For example, soya lecithin contains little DNA. One question to be answered is how much lecithin should be used for DNA extraction—one gram, one kilogram, or one ton?
It is doubtful whether the compilation of a so-called negative list, which at the moment includes highly refined oils and certain products derived from starch10, will clarify the situation for the food industry, since consumers are more interested in food authenticity than food analysis. On the other hand, GMO testing has to remain feasible for the control laboratories. The recent moratorium on new GMOs by the EU ministers of environment raises another question: What rules should be applied for the testing of nonapproved GMOs?
We would like to thank the members of the Swiss Food Manual subcommission 29a (Albert Eugster, Peter Brodmann, Ulrich Vögeli, Jürg Ruf, Urs Pauli, Wolfram Hemmer, Jean-Yves Deru, Rolf Meyer, Jürg Rentsch, and Anita Pfefferkorn) for discussions, suggestions, and their participation in the ring test together with Andreas Wurz (Gene-Scan GmbH, Freiburg, Germany) and Frederik Janssen (Food Inspection Service, Zutphen, The Netherlands).
About this article
Nature Biotechnology (2001)