J. Agric. Food Chem. 2004, 52, 5223−5232
5223
Detection and Quantification of Roundup Ready Soy in Foods by Conventional and Real-Time Polymerase Chain Reaction MICHAEL E. ROTT,* TRACY S. LAWRENCE, ERIKA M. WALL, MARGARET J. GREEN
AND
Centre for Plant Health, Canadian Food Inspection Agency, 8801 East Saanich Road, Sidney, British Columbia, Canada V8L 1H3
Transgenic soybean line GTS-40-3-2, marketed under the trade name Roundup Ready (RR) soy, was developed by Monsanto (USA) to allow for the use of glyphosate, the active ingredient of the herbicide Roundup, as a weed control agent. RR soy was first approved in Canada for environmental release and for feed products in 1995 and later for food products in 1996 and is widely grown in Canada. Consumer concern issues have resulted in proposed labeling regulations in Canada for foods derived from genetically engineered crops. One requirement for labeling is the ability to detect and accurately quantify the amount of transgenic material present in foods. Two assays were evaluated. A conventional qualitative Polymerase Chain Reaction (PCR) assay to detect the presence of soy and RR soy and a real-time PCR to quantify the amount of RR soy present in samples that tested positive in the first assay. PCR controls consisted of certified RR soy reference material, single transgenic soybeans, and a processed food sample containing a known amount of RR soy. To test real-world applicability, a number of common grocery store food items that contain soy-based products were tested. For some samples, significant differences in amplification efficiencies during the quantitative PCR assays were observed compared to the controls, resulting in potentially large errors in quantification. A correction factor was used to try to compensate for these differences. KEYWORDS: Soybean; Roundup Ready; quantification; real-time PCR; processed foods: genetically modified
INTRODUCTION
The European Union (EU) has had regulations for the labeling of novel foods since 1997. As a consequence, research on the most efficient, reliable, and accurate methods of detecting and quantifying genetically modified (GM) ingredients has been intense (1-12). Current EU regulations stipulate that products containing an ingredient of which g0.9% originates from a GM product must be labeled. For example, if a pure soy product contains g1% soy derived from a GM soy variety, it must be labeled, but if it contains only 0.5%, it would not have to be labeled. If the product consisted of only 10% soy and 90% other, but 1% of the soy present was derived from a genetically GM soy variety, it would also require labeling, even though on a weight per weight basis it contains less GM soy than the pure soy product containing 0.5% GM soy that does not require labeling. No mandatory labeling legislation has been implemented in Canada to date. However, the Canadian Food and Drug Act allows for the voluntary labeling of biotechnologyderived foods. Currently, the Canadian General Standards Board (CGSB) in conjunction with the Canadian Council of Grocery Distributors has developed a National Standard for Canada for * Corresponding author [telephone (250) 363-6650, ext. 263; fax (250) 363-6661; e-mail
[email protected]].
the voluntary labeling of foods and food ingredients obtained from biotechnology (13). The aim of the CGSB is to develop standards, based on recognized international protocols, that will provide a model for label declarations that are understandable and not misleading for Canadian consumers. Their goal is also to establish procedures to distinguish biotechnology-derived foods from conventionally produced foods from production to retail and to establish the use of testing and monitoring procedures. The CGSB does not legislate labeling requirements. In Canada, there are over 40 plant products containing unique transgenic DNA events approved for food (14). Of these, one of the most widely grown in Canada, in 2002, is herbicidetolerant (GTS-40-3-2) soybean (15), with over 100 varieties containing this event registered in Canada. GTS-40-3-2, also known as Roundup Ready (RR), was developed by Monsanto and confers tolerance to the glyphosate-based herbicide Roundup. The GTS-40-3-2 event consists of an enhanced cauliflower mosaic virus 35S promoter, a CTP4 leader sequence from Petunia hybrida, and the 5-enol-pyruvylshikimate-3-phosphate synthase (EPSPS) gene conferring the herbicide tolerance followed by an Agrobacterium tumefaciens nopaline synthase terminator (15, 16). A number of methods have been developed for the detection of RR soy. These include protein-based methods for the detection of the EPSPS gene product in
10.1021/jf030803g CCC: $27.50 Published 2004 by the American Chemical Society Published on Web 07/20/2004
5224
J. Agric. Food Chem., Vol. 52, No. 16, 2004
Rott et al.
Table 1. Soy Foods Used in This Survey food
degree of processing
form of soy ingredient
miso soy sauce natto TVPb meat alternative 1 meat alternative 2 cheesies cracker 1 gravy mix 1 frozen dessert soup soybean paˆte´ protein bar 1 soy beverage meal replacement beverage coffee whitener yogurt chocolate pudding meat alternative 3 tofu infant formula 1 infant formula 2
high high high high high high high high high moderate moderate moderate moderate moderate moderate moderate moderate moderate moderate moderate moderate moderate
fermented beans fermented beans fermented beans tspc tsp/soy sauce tsp hspd hsp oil/hsp soy powder/isp soy milk protein roasted soybean/ispa soy milk protein whole bean soy milk soy milk spce/isp soybean curd oil/isp/soy milk oil/isp/soy milk
a
food
degree of processing
form of soy ingredient
boiled soybean dried (organic) soybean roasted soy nuts soy nut snack soy nut spread simulated bacon bits protein bar 2 biscuit gravy mix 2 bread 1 bread 2 soup mix nutritional bar cheese crispbread cracker 2 flour
low low low low low low low low low low low low low low low low low
whole bean whole bean whole bean whole bean whole bean flour/soy protein/hsp flour/lecithin/isp flour flour/hsp flour/isp/lecithin defatted flour soy powder flour/isp/soybean/lecithin/soy butter flour/oil defatted flour flour/lecithin flour
isp, isolated soy protein. b TVP, textured vegetable protein. c tsp, textured soy protein. d hsp, hydrolyzed soy protein. e spc, soy protein concentrate.
transgenic raw or unprocessed soy products (15, 17-19) and PCR-based methods, both qualitative and quantitative, which can also be used for the more highly processed soy-based food products (2, 5, 6, 8-10, 15, 17, 20-22). Several of these methods have been officially approved or validated by other government agencies (2, 15, 18, 23). In this study a variety of foods containing soy ingredients that are available in the local supermarkets were sampled using two different PCR protocols. Our goal was to first establish that soy DNA and Roundup Ready (RR) DNA could be detected in the foods by qualitative PCR, targeting the endogenous soy lectin gene and the CTP4CTP EPSPS junction of the RR gene (15, 17, 24), and to determine from what type of foods it was possible to amplify soy DNA and to get an initial impression on the prevalence of genetically modified soy in common food products. Those products testing positive were then further analyzed by realtime quantitative PCR, targeting different regions but the same lectin and RR genes (15, 25) to determine the level of Roundup Ready soy in these foods. Few studies to date have attempted to quantify the amount of transgenic material in processed foods to determine the practical feasibility of such testing. MATERIALS AND METHODS Soy Food Samples. A selection of 39 soy food products including vegetarian foods, dry foods, snack foods, condiments, desserts, soups, baby formulas, and beverages representing the different levels of processing (Table 1), were chosen from local supermarkets. Soybean Reference Material. As reference material for qualitative and quantitative analysis, Roundup Ready soybean (variety S14M7) was obtained from a Canadian seed distributor. Because the purity of the seed lot was unknown, single-seed extractions were done to confirm that individual seeds were transgenic. One of these DNA extractions was then used as reference material. The single-seed extract was diluted to 40 ng/µL and serially diluted to 12 pg/µL. Six of these dilutions (5 µL per reaction) were chosen to represent the standard curve encompassing 200 ng-60 pg of DNA. Quantitative PCR Control. Soybean powder certified reference material (CRM) IRRM-410S (Fluka) 5% Roundup Ready soy used for quantitative real-time PCR was purchased from Sigma (Canada). A processed soy food sample, meat paˆte´, from the Genetically Modified
Material Analysis (GeMMA), Scheme, Report GeM18, with an assigned RR soy content of 8.5%, was also used as a control. Extraction of Genomic DNA. Each soy food product was homogenized in a Waring electric blender and then ground using an ice-cold mortar and pestle. Approximately 200 mg or 1 g samples from the ground material were used in the subsequent DNA extraction procedure. The single S14M7 soybean seed reference material (∼200 mg) was ground using a chilled mortar and pestle, whereas 50 mg of the CRM soybean powder was extracted directly. Modified Wizard. Briefly, 860 µL of extraction buffer [10 mM Tris (pH 8.0), 150 mM NaCl, 2 mM ethylenediaminetetraacetic acid (EDTA), 1% (w/v) sodium dodecyl sulfate], 100 µL of (5 M) guanidine hydrochloride, and 40 µL of (20 mg/mL) proteinase K were added to 200 mg of each ground sample and vortexed to mix thoroughly. For 1 g samples, the extraction buffer, guanidine hydrocloride, and proteinase K were scaled up proportionally. The samples were incubated at 60 °C for 30-60 min followed by a further incubation at 75 °C for 20 min to inactivate the proteinase K. The supernatant generated after centrifugation at full speed for 20 min (to pellet debris) was incubated for 10 min at 60 °C with RNase A added to a final concentration of 1 mg/mL. Following a 3 min centrifugation at full speed, 1 mL of the supernatant was mixed with 1 mL of Wizard resin (Promega) and then applied to a Wizard mini column using a 3 or 5 mL capacity disposable Luer-Lok syringe. For samples that had a dark supernatant (roasted soy nuts, biscuit, natto, and simulated bacon bits), the column was washed with 1 mL of CQW wash buffer (Nucleospin Food Kit, Macherey-Nagel). All samples were washed with 2 mL of 80% ethanol and then centrifuged at full speed for 10 min followed by a 10 min incubation at 37 °C to evaporate residual ethanol. The nucleic acids were eluted in 100 µL of 70 °C 10 mM Tris, pH 8.0. Qiagen Stool Kit. The soy sauce, chocolate pudding, and miso homogenized samples were dried at 37 °C for 24 h to remove excess moisture. To 1 g of dried sample, ASL buffer (supplied with the Stool Kit) was added until a consistency was reached such that the sample flowed freely in the tube (2-3 mL). The samples were vortexed until thoroughly homogenized. Following a 5 min incubation at 70 °C, the samples were centrifuged at full speed for 10 min to remove particulates. The supernatant was divided into multiple 2.0 mL microfuge tubes and 1 InhibitEX tablet added to each (1 tablet/1.2 mL supernatant). If the volume was 3 × 103 (200 ng-60 pg per reaction) of total genomic DNA extracted from the S14M7 single seed, amplified in separate reaction tubes using either the lectin primers/ probes or RR primers/probes, was used to generated the calibration curves for lectin and RR, respectively. The standard curves were the regression of crossing point (Cp) versus log of the nanograms of standard in each reaction, where the crossing point is the cycle at which the reaction fluorescence increases above a baseline level defined by the software (see Table 3). Each food sample was run in triplicate for each target. For the initial run, three different template concentrations for each sample were analyzed to determine which gave a Cp that best fit within the standard curve. In the subsequent run, this concentration was analyzed in duplicate, and the mean of all three Cp values was used in the analysis. Data were analyzed using LightCycler Data Analysis software version 3.5.5 and the “second derivative” algorithm. Determination of Percent RR Soy Relative to Total Soy DNA in Food Samples. Two standard curves were generated from seven replicates of the lectin and RR assays. The equations from the linear regressions of these curves were used to calculate the nanograms of lectin or RR DNA in the food samples on the basis of the mean of the crossing points. The percent RR soy was determined from the ratio of nanograms of RR divided by the nanograms of lectin multiplied by 100. Some foods required more template in the RR reaction compared to the lectin reaction in order to bring the crossing points within the range of the standard curve (4 × template, meat alternatives 1 and 2, soup mix; 8 × template, protein bars 1 and 2;× 20 × template, tofu and 5% Fluka CRM), and therefore the RR values had to be divided by these factors before the ratio to lectin could be made. RESULTS
Classification of Soy-Containing Foods. The 39 foods chosen for this study had soy listed as an ingredient on the label and are representative of a wide range of soy-containing foods available in the Canadian marketplace. The soy ingredients included whole bean, flour, defatted flour, powder, soy protein, soy milk, and lecithin. The proportion of soy relative to other food ingredients varied from product to product and could not be experimentally determined. Table 1 lists the soy foods and the degree of processing of the soy ingredients. Degree of processing was based on previously published descriptions of soybean products and their methods of preparation (15, 27, 28). Fermented, extruded (textured soy protein), pressed (lecithin and oil), and hydrolyzed soy protein soy ingredients were classified as highly processed. Mild alkaline extracted (isolated soy protein), water extracted (milk), precipitated (curd), and roasted ingredients were considered to be moderately processed. Whole beans and ground beans (flour) were considered to be the least processed. Foods with a number of different forms of soy in the ingredients were classified on the basis of the least processed form. DNA Extraction. All samples were initially extracted using the modified Wizard method. Those samples that gave no or poor lectin amplification by qualitative PCR were re-extracted with the Qiagen Stool Kit (soy sauce, chocolate pudding, miso, gravy mix 1, natto, cracker 1, and cheesies). To determine the degree of degradation, and therefore the quality of the extracted DNA, ∼200 ng of each sample was visualized by agarose gel electrophoresis (Figure 1). Many of the food extracts contained
5226
J. Agric. Food Chem., Vol. 52, No. 16, 2004
Rott et al.
Table 3. Quantitative Results for the Soy Foods
a Cp, crossing point (mean of nine replicates for standard curves and mean of three replicate for the food samples). b SD, standard deviation. c Four times more template used in the RR reaction compared to the lectin reaction. d Eight times more template used in the RR reaction compared to the lectin reaction. e Twenty times more template used in the RR reaction compared to the lectin reaction.
degraded DNA with an average fragment size of 2400%, and for one sample (meat alternative 3) the %RR soy changed significantly (from 0.39 to 5.0%) which could alter the classification of this food with respect to a proposed Canadian labeling threshold of 5%. DISCUSSION
In anticipation of food labeling for genetically modified products in Canada, a pilot project was undertaken to detect and quantify RR soy in a variety of soy-containing foods available in the Canadian marketplace. RR soy was chosen because the transgenes for this event are well characterized, a number of published tests for the detection and quantification are available, and it is widely grown and used in both Canada and the United States and was therefore expected to be detectable in a wide range of food products. From the survey, soy could be detected in 100% of the 39 food samples tested, and RR soy was detected in 28 (72%) of these foods. Of the foods testing positive, 11 foods (28%) contained quantifiable levels of RR soy, 8 of which (21%) contained levels of RR soy >5% and would be candidates for labeling. Assay Controls. To test the accuracy of quantitative tests, two known controls were analyzed. The 5% RR soybean powder (Fluka) IRMM certified reference standard was calculated to contain 4.6% RR soy, an error of only 8%. The second control, GeM18 meat paˆte´ sample, is more representative of the processed foods analyzed in this study. To summarize, the composition and preparation of this sample was as follows. A 6-7% RR soy (w/w) of total soy flour mix was added to other paˆte´ ingredients (pig liver, pork belly, water, skimmed milk powder, salt, sorbic acid, sodium polyphosphate, pepper, onion powder, and monosodium glutamate) at a level of 5%, mixed, and baked at 130 °C for 1 h and then at 150 °C until a core temperature of 80 °C was obtained. Homogeneity testing of the paˆte´ samples by GeMMA (10 subsamples, 2 replicates of each) gave a mean RR soy/total soy value of 5.5% with a range from 3.8 to 8.9%. The assigned %RR value for the paˆte´, based on the consensus mean of 81 laboratories participating in the GeM18 proficiency panel, was 8.5%, with acceptable individual results (based on z scores) ranging from 3.4 to 21.4%. In this study, the meat paˆte´ sample was calculated to contain 6.7% RR soy/total soy, which is close to both the mean value determined by homogeneity testing and the assigned value. Close agreement of the experimentally determined and accepted values for both the RR flour and processed food controls gives confidence that the values obtained for the unknown processed food samples are accurate. Food-Processing Effects. The food samples tested contained soy as an ingredient ranging from low (whole bean or flour) to highly processed (fermented bean). In general, it can be assumed that increasing levels of processing result in a decrease in the
5229
amount and quality of soy DNA that can be extracted due to degradation of the DNA caused by the different processing methods. This can have a significant effect on the ability to detect/quantify the amount of soy in these foods by PCR. Agarose gel electrophoresis of the DNA extracted from the food samples showed that many contained highly degraded DNA with average fragment sizes of 0.10, a correction factor was applied (eq 1). When this correction factor was applied to meat alternative 1 and protein bars 1 and 2, the amount of RR present was calculated to be below the LOQ. It has been demonstrated that low concentrations of target can lead to a calculation artifact with the LightCycler (34). The curve of Cp versus log of concentration, at low target concentrations, is no longer linear but parabolic, and calculations of E within this range would give artificially high values if a correction factor was not used. This appears to be the case for meat alternative 1 and protein bars 1 and 2. Applying the correction factor to bread 2 (∆∆E ) 0.20) resulted in an increase in %RR from 76 to 2477%, which clearly is not possible. This result is probably due to an error in the calculation of E for bread 2 due to an insufficient number of data points. Bread 2 contained the least amount of soy of all foods quantified (
