Heat processing compromises GMO detection in soybean-enriched biscuits

Abstract

The widespread adoption of genetically modified (GM) crops necessitates robust detection methods, particularly for processed foods, where DNA degradation compromises analytical reliability. This study investigated the effect of industrial baking temperatures (190–210 °C) on the detectability of Roundup Ready® soybean (GTS 40-3-2) DNA in a biscuit matrix containing 0–100% GM soybean flour. Real-time PCR analyses targeted the soybean-specific lectin gene, the CaMV 35 S promoter, and the cp4 epsps transgene. The results demonstrate that thermal processing induces sequence-specific DNA degradation, which is more pronounced for the 35 S promoter and cp4 epsps sequences than for the lectin reference gene. This observation indicates that the assumption of equivalent amplification behavior between target and reference sequences—a fundamental premise of comparative qPCR approaches such as the ΔΔCq method—may not consistently hold in thermally processed matrices. Near the European Union’s 0.9% labeling threshold, such differential degradation patterns may increase the risk of false-negative results. Overall, this study demonstrates that the principal challenge in the analysis of processed foods lies not in absolute GMO quantification per se, but in the reliable interpretation of detection results under conditions of sequence-specific DNA degradation. These findings underscore the need for matrix-optimized analytical protocols tailored to processed foods and for a critical re-evaluation of current qPCR-based quantification frameworks in this context.

Introduction

The development and application of genetically modified organisms (GMOs) have attracted increasing attention in the context of plant genetic engineering while also raising concerns regarding potential health and environmental risks. Although consumer attitudes toward GMOs vary, there is a common demand for greater transparency about their presence and proportion in food products. Recent advances in genome-editing technologies have further blurred the regulatory boundaries surrounding genetically modified organisms (GMOs)¹.

The European Union (EU) introduced the first labeling regulations for GMOs in 1997², and marketing authorizations for such organisms are currently governed by the Novel Food Regulation (EC) No. 2003/29³. Since then, many countries have implemented their own regulatory frameworks for labeling genetically modified foods and feeds^4,5. However, labeling thresholds differ significantly across regions. For instance, the EU requires labeling if the GMO content exceeds 0.9% (w/w), whereas Japan, South Korea, and Australia/New Zealand set thresholds of 5%, 3%, and 1%, respectively^6,7,8. Moreover, the EU enforces a zero-tolerance policy for unauthorized GMOs in food products, although up to 0.1% is tolerated in animal feed⁹.

To maintain such regulations, highly sensitive and reliable molecular detection methods are essential. Various techniques have been developed for GMO monitoring^7,8,9,10 with polymerase chain reaction (PCR)—including conventional and real-time PCR—being the most widely used technique owing to its robust sensitivity and specificity¹⁰. In general, GMO analysis comprises multiple steps, including DNA extraction, screening, qualitative and quantitative PCR, and event-specific assays^5,14,15,16.

Commercial cultivation of genetically modified crops began in 1996 and has expanded to approximately 206.3 million hectares in 27 countries by 2023¹⁷. Globally, eleven genetically modified crop species are cultivated, among which soybean accounts for approximately 71% of the total GMO-planted area^5,13,14,15. Roundup ready soybean (RRS), event GTS 40-3-2, is cultivated primarily in the United States, Argentina, and Brazil and is exported to the EU and other countries for food and feed use¹⁷.

This soybean variety was genetically engineered to tolerate glyphosate, the active ingredient in the herbicide Roundup^®, by introducing the epsps gene from Agrobacterium tumefaciens strain CP4. The resulting enzyme, 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS), has reduced glyphosate-binding affinity, enabling the plant to survive herbicide application²⁰.

While PCR-based methods are highly effective for detecting GMOs in raw materials, their performance can be compromised during food processing. In general, a widely used thermal process in food production can fragment or degrade DNA, thus affecting PCR sensitivity and accuracy. Despite this concern, few studies have systematically investigated how baking conditions—particularly temperature and duration—impact the detectability of GMOs in complex food matrices such as biscuits²¹.

To address this gap, the present study aims to evaluate the impact of heat treatment on the quantification of GMOs via real-time PCR, with a particular focus on the integrity and amplifiability of event-specific target sequences. Biscuit samples were prepared by baking dough containing genetically modified soybean flour (event GTS 40-3-2) at seven concentrations (0%, 0.1%, 0.5%, 0.9%, 2%, 50%, and 100%) and three baking temperatures (190 °C, 200 °C, and 210 °C). Both the dough and the baked samples were subjected to qualitative and quantitative real-time PCR analysis. This experimental setup allowed a direct comparison of measured GMO levels with regulatory thresholds and enabled the evaluation of DNA degradation under different thermal conditions.

This study represents a novel approach to evaluating the impact of heat treatment on GMO detection in biscuits produced with varying proportions of Roundup Ready^® soybean flour. While most previous GMO detection studies have focused on raw or minimally processed foods, the present work simulated real industrial baking conditions, providing valuable insights into the effects of high-temperature processing on DNA integrity, purity, and qPCR performance within a complex food matrix.

Materials and methods

The samples and reference materials

The ingredients used for biscuit production, including salt, vegetable oil (shortening Mustanoğlu Gıda, Isparta, Turkey), sugar, baker’s special sugar, and baking powder (sodium bicarbonate), were purchased from local markets. High-fructose corn syrup (HFCS 42%) was obtained from Sunar Mısır Entegre Tesisleri San. ve Tic. A.Ş. in Adana, Türkiye. Biscuit-grade wheat flour was sourced from the Türkmenler Flour Factory in Gaziantep, Türkiye. Roundup Ready™ soybean flour (RRSF) was supplied by the Turkish Feed Industrialists’ Association.

Certified reference materials (CRMs) produced by ERM (European Reference Material) at the IRRM—Institute for Reference Materials and Measurements—in Geel, Belgium, were used as negative and positive controls at 0% and 0.1% GMO levels: non-GMO Soy ERM®-BF410ap and 0.1% GMO Soy ERM®-BF410cp. Additionally, 1% GMO-positive soy controls were provided by Roche.

Biscuit preparation

Biscuit production was carried out according to the AACC Method 10-50.05, with slight modifications, as shown in Table 1.²²

Table 1 Formulations used in biscuit production²².

For biscuit production, shorteners, baker’s special sugar, sodium bicarbonate, and salt were placed in a Hobart N-50 mixer (Germany) and mixed at low speed for 3 min. After each minute, the mixer was paused, and the material adhering to the bowl walls was scraped down with a spatula. Subsequently, corn syrup and deionized water were added, and the mixture was mixed for 1 min at low speed. Wheat flour or a wheat–RRSF blend was then incorporated and mixed for an additional 2 min at low speed.

RRSF was added at six different concentrations (0.1%, 0.5%, 0.9%, 2%, 50%, and 100%) by substituting an equivalent amount of wheat flour. After mixing, the dough was rolled to a thickness of 6 mm via a rolling pin and cut into round shapes (60 mm diameter) with a metal cutter. The biscuits were placed on greaseproof paper-lined trays and baked for 10 min at three different temperatures (190 °C, 200 °C, and 210 °C) in a commercial oven. A fixed baking time of 10 min was selected for all temperatures based on preliminary tests to ensure a uniformly well-baked product quality, thereby isolating the effects of temperature and GMO concentration on DNA degradation.

After baking, the biscuits were allowed to cool at room temperature for 30 min and then stored in hermetically sealed plastic bags until further analysis. The control group consisted of biscuits without the addition of RRSF. All formulations were prepared in triplicate.

DNA isolation

Two different DNA extraction methods were used: Qiagen DNeasy Plant Kit (Cat. No. 69106, Germany) and the High Pure DNA Isolation Kit (Cat. No. 11796828001, Roche, USA). Each sample was extracted in at least duplicate.

The quality and concentration of the extracted DNA were measured spectrophotometrically via a BioTek Microplate Reader (Epoch, USA). Samples were briefly centrifuged, and 1 µL of each DNA sample was applied to the pedestal of a BioTek Microplate Reader (Epoch, USA). A blank was prepared by adding 1 µL of the DNA rehydration solution alone to the pedestal.

DNA purity was assessed by measuring absorbance at 260 nm and 280 nm, and the A260/A280 ratio was calculated. An ideal ratio of 1.8 was considered indicative of pure DNA, whereas ratios greater than 2.0 suggested RNA contamination, and ratios below 1.6 indicated protein contamination. DNA concentration was determined in ng/µL using the respective spectrophotometric device. These procedures ensured the isolation of high-quality, amplifiable DNA from various processed food matrices²³.

Screening of plant genes and the 35 S promoter region

A screen for plant genomic DNA, a soya-specific gene (lectin), the 35 S promoter, and the Roundup Ready (RR) soybean transgene was performed via real-time PCR via a Light Cycler 480II (Roche Diagnostics). Region-specific primers, probes, positive control DNA, and sterile deionized water were supplied by Roche. The PCR mixture was prepared in reaction tubes on the basis of the number of samples analyzed. Certified reference materials (CRMs) containing 0%, 0.1%, and 1% RR soybean DNA were used as positive controls, whereas sterile deionized water served as the negative control.

The components used in each well of the 96-well PCR plate are listed in Tables 2 and 3.

After a brief centrifugation (1500–2000 rpm for 30–60 s), 15 µL of the prepared PCR mixture and 5 µL of isolated DNA were added to each well. The plate was then briefly mixed at 3000 rpm for 10 s in a plate centrifuge and loaded into the Light Cycler 480II. The PCR program described in Table 4 was applied.

Table 2 PCR components per well for screening and quantification.

Table 3 PPmix composition for each specific region (35 S promoter, plant gene, and lectin).

Table 4 Real-time PCR thermal cycling conditions for the 35 S promoter, plant gene, and lectin.

Plant gene screening was initially performed to confirm the presence of amplifiable DNA. The PCR mixture included primers and probes specific to the plant gene. The 0% RR soybean CRM (ERM^®-BF410ap) was used as the positive control; sterile deionized water was used as the negative control.

Lectin gene screening was conducted to determine the presence of soybean in both dough and baked biscuit samples. Primers and probes targeting the lectin gene were used. The 1% RR soybean CRM (ERM^®-BF410dp) was used as a positive control, and a soy-free sample served as the negative control.

The 35 S promoter region was detected to identify the presence of genetic modifications. The 1% RR soybean CRM (ERM^®-BF410dp) served as the positive control, whereas the 0% RR CRM (ERM^®-BF410ap) was used as the negative control.

All the samples were analyzed in triplicate, and the procedures were conducted according to the manufacturer’s instructions to ensure reliability.

Relative quantification of roundup ready soy by real-time PCR

Relative quantification of Roundup Ready (RR) soybean–derived transgenic DNA in dough and biscuit samples was performed using a LightCycler^® 480 II real-time PCR system (Roche Diagnostics, Mannheim, Germany)²⁴. All qPCR assays were conducted using DNA isolated with the Qiagen DNeasy Plant Pro kit due to its superior and more consistent purity yields. Amplification data were analyzed using the Fit Point algorithm implemented in the LightCycler^® 480 software (Version 1.5), which determines crossing point (Cq/Cp) values based on a fixed fluorescence threshold within the log-linear phase of amplification²⁵.

For each reaction, a GMO-specific target gene (cp4 epsps) and a soybean-specific endogenous reference gene (lectin) were analyzed simultaneously, in accordance with ISO 21569:2005 guidelines for GMO detection¹⁶. This dual-target strategy enables normalization of the transgenic signal to soybean-derived DNA within each sample.

Relative quantification was performed using the comparative quantification (ΔΔCq) module integrated into the LightCycler^® 480 software, following the mathematical model described by Livak and Schmittgen²⁶. Certified reference material ERM^®-BF410d (GTS 40-3-2 soybean; Joint Research Centre, European Commission) was used as the calibrator sample²⁷. For each unknown sample, ΔΔCq values were calculated according to the following equation:

$\begin{gathered} \Delta \Delta {\text{Cq }}={\text{ }}[{\text{Cq}}\_{\text{Target }}\left( {{\text{Unknown}}} \right)\, - \,{\text{Cq}}\_{\text{Reference }}\left( {{\text{Unknown}}} \right)] \hfill \\ - [{\text{Cq}}\_{\text{Target }}\left( {{\text{Calibrator}}} \right)\, - \,{\text{Cq}}\_{\text{Reference }}\left( {{\text{Calibrator}}} \right)] \hfill \\ \end{gathered}$

Relative quantity (RQ) values were then derived as:

${\text{RQ}}={2^ \wedge }( - \Delta \Delta {\text{Cq}})$

and expressed as percentages relative to the calibrator.

Importantly, this ΔΔCq-based approach does not quantify the absolute proportion of genetically modified soybean within the total flour matrix of the biscuits. Both cp4 epsps and lectin signals originate exclusively from soybean-derived DNA and therefore co-vary with the soybean flour content. Consequently, the calculated values are best interpreted as relative detectability indices reflecting changes in transgenic DNA integrity and PCR detectability following processing²⁵.

The normalization principle underlying this approach is based on the ΔCt method:

${\text{Relative}}\;{\text{Quantity}}={2^ \wedge }\left( { - \Delta Ct} \right)100$

where ΔCt represents the difference between target and reference Cq values. This equation is provided solely to illustrate the normalization concept underlying the ΔΔCq method; all quantitative results in this study were generated exclusively using the ΔΔCq module of the software^{26,27,28,29,30}.

This analytical framework assumes comparable amplification efficiencies for target and reference assays and is consistent with the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines²⁸. Accordingly, all PCR reactions were performed in triplicate, and no-template controls (NTCs) containing sterile deionized water were included in each run to monitor potential contamination^28,29. Quantitative analysis reports were generated using the LightCycler^® 480 software and provided by Roche Diagnostics Turkey²⁹.

Oligonucleotides

All oligonucleotides (primers and probes) and reagents used in the PCR analyses were obtained from Roche Diagnostics Turkey A.Ş. The primer and probe sequences used for plant gene, lectin gene, and 35 S promoter detection are shown in Table 5.

Table 5 Primers and probes used in the detection of plant genes, the lectin gene, and the 35 S promoter.

Statistical analyses

The data obtained from the GMO analyses were evaluated using SPSS version 22.0. The statistical analysis employed a two-way grouping strategy to address distinct research questions. First, dough samples were analyzed as a separate group to establish baseline detection and quantification parameters in the absence of thermal processing. Second, all biscuit samples were analyzed together to enable direct comparison of baking temperature effects across different RRSF concentrations while maintaining adequate statistical power. Multiple comparisons among means were performed using Duncan’s multiple range test at a significance level of p < 0.05.

Results and discussion

DNA isolation

We evaluated the performance of two commercial DNA isolation kits (Qiagen and Roche) in both raw dough and baked biscuit samples. The performance of two commercial DNA extraction kits (Roche High Pure DNA Kit and Qiagen SureFood Kit) was evaluated in both raw dough and baked biscuit samples containing various concentrations of Roundup Ready soybean flour (RRSF; 0–100%). DNA concentrations obtained with the Roche kit ranged from approximately 10 to 18 ng/µL, with A260/A280 ratios varying between 0.62 and 1.99, indicating considerable variability in DNA purity across the samples (Table 6).

Table 6 Representive DNA concentration and purity (A260/A280) of biscuit samples extracted using Roche high pure DNA Kit.

Both Qiagen and Roche kits successfully recovered sufficient DNA from raw dough samples; however, thermal processing during baking significantly affected DNA quality. DNA concentrations obtained with the Qiagen kit ranged from approximately 10 to 19.3 ng/µL. However Spectrophotometric monitoring indicated that DNA isolated from heat-processed samples using the Qiagen kit generally exhibited more consistent and ideal purity ratios (A260/A280 1.7–1.9) compared to the Roche kit (A260/A280 0.6–19). Therefore, to ensure greater reproducibility, DNA isolated with the Qiagen DNeasy Plant Pro kit was used for all subsequent qPCR analyses.

Thermal processing, particularly above 200 °C, contributes to the degradation of high-molecular-weight DNA fragments, reducing the size of recoverable DNA and thereby affecting amplification efficiency^21,31. Our findings align with³², who demonstrated that high-temperature processing in bakery products reduces PCR detectability, particularly at low GMO levels, while the endogenous lectin gene remains more consistently amplified. The observed increase in Cp values during qPCR analysis confirmed that DNA fragmentation occurs despite apparently acceptable purity ratios, emphasizing that spectrophotometric purity alone is insufficient to predict PCR suitability.

These findings are in agreement with previous studies that evaluated DNA extraction efficiency from processed food matrices. Türkeç et al.³³ reported variable DNA yield and quality across different extraction methods for detecting genetically modified materials in soy-based foods and feeds. Similarly, Ashrafi-Dehkordi and Hemmati³⁴ demonstrated that a modified CTAB method outperformed conventional methods for soy products.

Consistent with our findings, Vahdani et al.³⁵ comparatively evaluated several DNA extraction protocols for soy-based products and concluded that the most efficient method was the one yielding DNA with an average A260/A280 ratio of approximately 1.8, indicating high purity and suitability for PCR amplification. This observation aligns well with our results, where the Sure Food kit consistently produced DNA of comparable purity and integrity, particularly in heat-processed matrices. The convergence between these independent studies reinforces that achieving an optimal purity ratio around 1.8 is a key indicator of reliable DNA recovery and amplifiability in GMO analysis of processed soy-containing foods.

While Ballari and Martin³⁶ observed that physical DNA fragmentation did not affect relative quantification in purified maize flour, our study demonstrates that in complex baked matrices, DNA quality—specifically fragment length and amplifiability—is critical for successful event-specific PCR. This focus on DNA integrity as a predictive marker for PCR success in heat-processed matrices represents the central novelty of our research.

Consequently, selecting a robust isolation method capable of maximizing recovery of quality DNA is essential for reliable molecular detection in thermally processed foods.”

Results of the qualitative analyses

GMO analyses were conducted on biscuits formulated with 0%, 0.1%, 0.5%, 0.9%, 2%, 50%, and 100% Roundup Ready^® soy flour (RRSF). The data were statistically evaluated via SPSS version 22.0.

All test samples were analyzed in triplicate for the soybean Lectin gene and respective GMO concentrations. Each qPCR plate included a 1% GMO-certified reference material (CRM) as a positive control. Qualitative real-time PCR assays were performed to detect both the plant-specific gene and the CaMV 35 S promoter region. The Lectin gene, serving as an endogenous control, was consistently detected across all samples but not quantified. No amplification was observed in extraction-negative controls or in no-template controls (NTCs), confirming assay specificity.

CP values were calculated using the Second Derivative Maximum (SDM) method, which identifies the point of maximum acceleration in fluorescence during exponential amplification¹. This algorithm is widely implemented in commercial qPCR platforms such as the LightCycler data for plant-specific gene, Lectin gene, and 35 S promoter region are presented in Table 7. CP values greater than 38 (CP > 38) were interpreted as negative, whereas values equal to or below 38 (CP ≤ 38) were considered positive detections.

Table 7 Crossing point (CP) data of plant gene screening, lectin gene screening, 35 S screening regions performed in real-time PCR in biscuit samples.

This study demonstrates that the sensitivity of genetically modified organism (GMO) detection in a biscuit matrix is critically influenced by both the initial GMO concentration and the intensity of the applied thermal processing. The findings shed light on important parameters that must be considered for the accurate assessment of GMO presence in processed foods using real-time PCR (qPCR).

A key finding of our work is the distinct inverse correlation observed between the CP value for the 35 S promoter region and the RRSF concentration in the sample. The systematic decrease in CP values as the RRSF amount increased from 0.1% to 100% confirms that the qPCR methodology provides a dose-dependent and quantitative measurement even within a complex food matrix like biscuits. Low CP values are a classic indicator of reaching the exponential amplification phase in early cycles due to a high initial DNA template concentration. This result confirms the method’s reliability and consistency for high GMO levels.

However, the situation becomes more complex at low GMO levels, particularly for the 2% RRSF sample. In these samples, the observed increase in CP values when the baking temperature was raised from 200 °C to 210 °C provides clear evidence of the destructive effect of thermal processing on the integrity of genetic material. It is well-established that high temperatures during food processing can cause DNA fragmentation and degradation, which directly impacts PCR amplification efficiency by reducing the number of intact target molecules^21,37. This degradation impedes the efficient amplification of the target region, leading to the signal being detected in later cycles (higher CP) or not being detected at all. This is of paramount importance as it poses a risk of false-negative results, particularly for products near the legal threshold for labeling (0.9%).

Another critical finding of our study emphasizes the necessity of manual verification in qPCR data analysis. Our negative control sample (0% RRSF) baked at 200 °C was incorrectly flagged as positive (CP = 23.91) by the automated SDM algorithm. This demonstrates that baseline fluctuations can be misinterpreted by the software as an early amplification curve, especially in samples with a low signal-to-noise ratio. Fortunately, manual inspection of the amplification curves using the Fit-Point method corrected this error and confirmed the sample was truly negative (CP = 39.01). This highlights the imperative for comprehensive data validation, which includes visual assessment of amplification curves rather than relying solely on automated CP values, as also recommended in established guidelines^38,39.

In conclusion, this study has shown that GMO analysis in processed foods requires a holistic approach that encompasses not only a precise laboratory technique but also consideration of production process parameters (thermal processing) and rigorous data analysis protocols. At low GMO levels, the degrading effect of thermal processing on DNA can reduce detection sensitivity and lead to misleading results. Therefore, it is recommended that automated qPCR data, especially for samples near threshold values, must always be supported by manual inspection to ensure the reliability of the results.

Impact of thermal processing on GMO quantification and matrix-dependent degradation dynamics

GMO analyses were conducted on biscuits formulated with 0%, 0.1%, 0.5%, 0.9%, 2%, 50%, and 100% Roundup Ready^® soy flour (RRSF).

All test samples were analyzed in triplicate for the soybean Lectin gene and respective GMO concentrations. Each qPCR plate included a 1% GMO-certified reference material (CRM) as a positive control and calibrator. The corresponding %GMO levels and CP values for Roundup Ready soy (epsp gene) and Lectin genes detected in dough and biscuit samples are summarized in Table 8. CP values greater than 38 (CP > 38) were interpreted as negative, whereas values equal to or below 38 (CP ≤ 38) were considered positive detections.

The data were statistically evaluated.

Table 8 CRM-Normalized relative EPSPS signal and crossing point (CP) values for RR soybean and lectin gene detected in dough and biscuit Samples.

Explanation of CRM-Normalized Relative EPSPS Signal (%) %RRS values were calculated using the ΔΔCt method with a 1% GMO Certified Reference Material (CRM) as the calibrator. The differing %RRS results obtained from similar CP values (e.g., Dough-0.9 and Dough-50) are a consequence of the differential DNA degradation of the target (epsps) and reference (Lectin) genes during processing. This phenomenon violates the core assumptions of the ΔΔCt method, indicating a systematic error that complicates accurate quantification in thermally processed complex food matrices.

The findings of this study demonstrate that the CRM-Normalized Relative EPSPS Signal (reported as “%GMO”) in thermally processed food matrices is critically influenced by processing parameters rather than reflecting the true initial GMO concentration. The systematic decrease in this normalized signal with increasing baking temperature—as clearly observed in Table 8, where it drops from 64.53% at 190 °C to 49.74% at 210 °C even in 100% RRSF samples—highlights a fundamental limitation of standard ΔΔCt-based quantification.

This trend is a direct consequence of differential and random DNA fragmentation induced by thermal processing, which preferentially degrades the longer epsps transgene target relative to the shorter endogenous Lectin reference. This violates the core assumption of equivalent PCR efficiency and intactness for both targets, unpredictably biasing the quantitative ratio. Consequently, the measured value does not represent absolute GMO content but instead functions as a process-induced signal, primarily indicating the relative post-processing integrity of the transgenic DNA target.

Therefore, the reported percentages should be interpreted not as accurate quantifications but as indicators of DNA degradation severity, revealing a systematic methodological bias that leads to the underestimation of true GMO levels in complex, processed matrices.

This methodological challenge is not a simple issue that can be resolved merely by switching from relative to absolute quantification. The core of the problem lies in the fact that, as also underlined by Gryson²¹, chemical reactions in complex matrices (such as Maillard reactions and protein-DNA cross-linking) affect target and reference genes at different rates, thereby invalidating the fundamental assumption of the ΔΔCt method. Indeed, the higher Cp values measured for the RR Soybean gene compared to the Lectin gene in our high-temperature processed samples provide clear evidence of this differential degradation. Our findings are in complete agreement with the view expressed by Bauer et al.⁴⁰ that DNA fragmentation accelerates beyond a certain temperature threshold and assumes a random character.

The detrimental impact of this DNA degradation on quantification in processed foods is corroborated by other studies in the literature. Baran Ekinci and Özçelik⁵ reported that in highly processed maize chips and flakes, results largely remained at “trace (TR)” levels due to severe DNA degradation, often making reliable quantification unfeasible. Similarly, Dişli and Yılmaz⁴⁰ emphasized that DNA quality seriously deteriorates in thermally processed soybean products, weakening detection reliability, including through false positives. The difficulties we encountered with automated Cp determination and the consequent necessity for manual verification using the Fit-Point method support this finding.

Another critical finding of this study is the decisive role of food matrix complexity in shaping analytical outcomes. In contrast to the findings of Ballari and Martin³⁶, who reported that thermal processing did not affect relative quantification in purified maize flour, a pronounced decrease in transgene detectability was observed in the present study using a complex biscuit matrix. This discrepancy underscores, in line with Cottenet et al.⁴¹, that quantitative PCR-based analyses become increasingly challenging in complex food systems, where not only DNA fragmentation but also interactions with matrix components may generate a “masking effect” that limits polymerase accessibility to the target sequence.

The quantitative strategy applied in this study, which relies on normalization to a species-specific endogenous reference gene, is inherently unsuitable for absolute GMO quantification in mixed food matrices. Accurate absolute quantification would require either plant-universal reference genes or calibration standards composed of ingredient mixtures with defined GMO proportions. Accordingly, the objective of the present work was not to estimate the GMO fraction of the entire biscuit matrix, but rather to evaluate, at the soybean component level, the impact of thermal processing on the relative persistence and PCR detectability of the transgenic DNA signal.

Within this methodological framework, the data presented should therefore be interpreted as indicators of relative changes in transgene signal integrity rather than as measures of absolute GMO content.

Conclusion

This study provides a comprehensive assessment of the detection and analysis of genetically modified organisms (GMOs) in both raw and thermally processed soybean-containing food products. Our findings demonstrate that thermal processing exerts a critical and multifaceted influence on the reliability of GMO analysis, posing substantial challenges for the interpretation and enforcement of strict regulatory thresholds, such as the 0.9% labeling limit, when applied to processed foods.

The results indicate that methodological factors, particularly DNA extraction performance, play a decisive role in the successful recovery of amplifiable DNA from baked matrices. Different extraction kits showed markedly different efficiencies, underlining the importance of protocol selection for processed food analysis. Beyond extraction efficiency, our data reveal a fundamental analytical challenge associated with sequence-specific DNA degradation. The cp4 epsps and 35 S promoter targets were consistently more susceptible to thermal fragmentation than the soybean-specific lectin reference gene.

This differential degradation does not reflect a failure of real-time PCR as a technique, but rather a violation of one of its key underlying assumptions—namely, that target and reference sequences are affected uniformly by processing. When this assumption is not met, quantitative outputs derived from comparative approaches (such as ΔΔCq-based analyses) no longer represent absolute GMO content, but instead reflect a process-altered, normalized signal (e.g., the CRM-Normalized Relative EPSPS Signal) that is a direct function of differential DNA integrity. Consequently, observed signal shifts in processed foods are driven primarily by sequence-specific integrity rather than by true differences in GMO proportion.

Consistent with previous studies on processed food matrices, our findings indicate that while GMO detection and relative signal comparison remain robust in raw or minimally processed materials, thermal processing compromises DNA integrity to an extent that limits the direct comparability of quantitative results with those obtained from unprocessed reference materials. Importantly, the data emphasize that the pattern and extent of DNA fragmentation, rather than total DNA concentration alone, is the principal determinant of reliable detection and interpretation.

Taken together, these results highlight the need for two parallel and complementary developments:

• (i) Advancement of analytical strategies for processed foods.

Future methodologies should explicitly account for non-uniform DNA degradation. This may include the incorporation of internal quality controls for DNA amplifiability (e.g., DNA Quality Indices), the use of shorter amplicon targets, or the validation of multi-target or multi-copy reference systems specifically designed for thermally processed matrices.

• (ii) Refinement of regulatory and interpretive frameworks.

Current labeling policies are largely derived from analytical assumptions validated in raw agricultural commodities. Our findings suggest that applying identical quantitative thresholds to highly processed foods without contextual interpretation of the underlying normalized signals may lead to methodological bias. Regulatory guidance that distinguishes between raw and processed products could improve both enforcement fairness and scientific robustness.

In conclusion, this study provides strong evidence that the quantitative interpretation of GMO-related PCR signals in thermally processed foods—specifically, the normalized relative signals used for quantification, such as the one derived from the epsps/lectin ratio—is intrinsically influenced by differential DNA degradation. Recognizing this limitation is essential for improving analytical practice, refining GMO labeling policies, supporting food control laboratories, and ensuring transparent communication to consumers. Future research should focus on developing analytical frameworks that remain reliable across diverse processing conditions and complex food matrices.