High-throughput screening tools for a reinforced chemical safety surveillance of food – SENTINEL

Sample Pooling – A proof-of-concept study of this approach for chemical surveillance of food was conducted using the case of nDL-PCBs in meat, where their levels are regulated, as a model. The evaluation relied on two parallel approaches: 1- An experimental approach to assess the practical feasibility of analyzing pools of 2 to 200 samples without degrading analytical performance; 2- A numerical approach combining database concatenation, modeling of the distribution of nDL-PCB levels, and simulation of sample pooling from 1 to n samples to determine the optimal sample pool size for achieving the best compromise between sensitivity, specificity, and analytical cost.

Aptasensors – An integrated approach combining aptamer selection, biosensor development, and optimization of a rapid PCB extraction protocol in meat matrices was developed. After in silico analysis of aptamers described in the literature, structures II and III were modeled and then subjected to docking simulations to identify the aptamers with the best recognition potential. These were then evaluated using various complementary approaches (FRET fluorescence, thermophoresis, electrophoresis). In parallel, a rapid extraction method for PCBs from meat, inspired by the QuEChERS method, was developed. The different steps were adapted to optimize the extraction yield while ensuring good compatibility with future biosensors: reduced solvent use, limited interfering compounds, and optimized cleaning phases.

Exposure markers sensors – The approach combined the discovery and identification of omics markers of PCB exposure in chickens with the design of sensors dedicated to the high-throughput, low-cost detection of these markers. The discovery of the markers was based on analyses of the volatile metabolome or volatolome (volatile organic compounds), the proteome (proteins), and the non-volatile metabolome performed on chicken tissues (liver, ceca, uropygial oil, blood plasma) obtained from in vivo case/control experiments (experimental rearing of 96 chickens divided into 1 control group and 6 "case" groups contaminated with different PCB cocktails).

Benefit-risk analysis of the SENTINEL tools - In order to improve the transfer of tools to industry and public authorities and to anticipate their benefits and risks (Objective 2), the approach is structured in 2 steps: 1) the conditions for implementing the SENTINEL tools were defined through scenarios of probable evolution of the meat sector; 2) A benefit-risk analysis of a selection of these scenarios was carried out, taking into account economic, regulatory, social and health aspects to help future decisions to strengthen health surveillance.

Sample pooling – The dual approach validated sample pooling for monitoring PCBs in meat: the cost of analysis is reduced by a factor of 19 when samples are pooled in groups of 25, while maintaining the same performance as analyzing samples individually. A collaboration with ANSES (French Agency for Food, Environmental and Occupational Health & Safety) confirmed the relevance of this approach for strengthening heavy metal monitoring in meat. A collaboration with a Greek infant formula manufacturer in the frame of H2020-RIA SAFFI, 2020-2024, coordinated by E. Engel extended the proof of concept to industrial self-monitoring of mycotoxins (ochratoxin A, zearalenone) in cereal products, with analysis costs reduced by a factor of 7 for pools of 10 samples.

Aptasensors – Modeling of structures II and III and docking simulation enabled the selection of sequences with the best recognition potential for the different PCB congeners tested. The PCB/aptamer pairs identified through in silico analyses were then evaluated (postdoc C Aymard, P3). FRET measurements confirmed the predicted interactions, validating the aptamers' ability to recognize targeted conspecifics and the probable location of interaction sites. Three aptamer sensors were developed using different labeling and immobilization strategies. Of these, only two sequences generated measurable intensity variations in the presence of PCBs, but these were too weak and insufficiently reproducible to establish a calibration curve or reliable quantification in a complex environment.

Exposure marker sensors – Based on 2D NMR and SPME-GC-MS profiling, metabolomic and volatolomic analyses (C Rémy's PhD thesis) and several Master internships (P1, P2) revealed markers of chicken exposure to PCBs (VOCs, lipids) in the liver, caecum, and uropygial oil. However, the analysis of these samples using the "electronic nose" approach, based on headspace samples taken with or without agitation, unfortunately did not detect a sufficiently strong signal to distinguish exposed animals from control animals. This latter result highlights the lack of sensitivity of the current electronic nose device for detecting traces of VOCs in complex biological samples.

Benefit-Risk Analysis – Six future scenarios for the sector were proposed. A benefit-risk analysis was conducted for three scenarios: firstly, through semi-qualitative analysis using argumentation systems for the "business-as-usual" scenario (R. Chaib's thesis, pp. 8, 11); secondly, through quantitative analysis of risk exposure for the "regional magnet" and "two-faced sector" scenarios (H. Hachem's postdoctoral research, pp. 9, 1). The risk assessment showed a "health" benefit from implementing sample pooling for the most exposed populations.

Sample pooling – The proof-of-concept work on sample pooling carried out in SENTINEL has led to developments within the framework of the European H2020 SAFFI project, a Research and Innovation Action (RIA) also coordinated by Erwan Engel (P1). This work focuses on exploring sample pooling for the chemical surveillance of infant products: 1- surveillance of dioxins using biosensors (collaboration with BDS (NL)); 2- surveillance of pesticides (collaboration with ANSES (FR)); 3- surveillance of tropane alkaloids (collaboration with IRTA (SP)).

Aptasensors – SENTINEL highlighted the limitations of currently available aptamers for PCB recognition and identified several avenues for improvement in future developments. A priority is to explore the use of binary dilution media, which could improve the interaction between the target and the biosensor. Furthermore, a more realistic approach to PCB detection would be the use of mimetic bioreceptors, such as molecularly imprinted polymers (MIPs), which are compatible with nonpolar solvents and potentially better suited to the hydrophobic properties of PCBs. These MIPs could then be tested on nanostructured surfaces (carbon nanotubes, gold nanoparticles) to improve electrochemical transduction and enhance the overall sensitivity of the sensor.

Exposure marker sensors – Regarding the search for volatile metabolites capable of signaling animal exposure to PCBs, a volatolomics investigation is underway on other biological matrices (particularly lung and kidney tissue) with high informational potential, collected from the chickens produced by the in vivo case-control experiment, but which were not initially targeted in SENTINEL. Data processing will also be further developed during the final year of C. Remy's PhD (2023-2026) to explore the potential of multi-omics and multi-matrix biological processing. To improve the electronic nose's performance, particularly in terms of sensitivity and selectivity, it could be coupled with a pre-concentration system or new sensitive materials could be designed (nanomaterials, selective peptides, etc.). A collaboration between P1 and P5 is underway on this topic.

Knowledge Engineering – A current perspective, not yet conceivable at the start of the project, is the personalized use of large language models (LLMs) to automate the analysis of interviews and documents used in the SENTINEL analysis. This automation would require significant development and validation work to ensure a protocol that guarantees an equivalent level of quality.

The first highlights of the project are being published with an article submitted on scenario building, several conferences at national (2) and international (3) congresses, popularization actions (30).

The European Parliament has called on Member States to strengthen their food safety control mechanisms by 2019. In July 2018, the French government has launched a platform to strengthen the coverage and the efficiency of the surveillance. To reach this objective, it is necessary for both food safety authorities and industry to get high-throughput and cost effective screening methods to monitor priority hazards throughout the food chain. Over the past few years, there has been substantial progress in microbiological safety and fast and cheap PCR-based methods are now currently used by regulatory authorities for inspection and by manufacturers for self-monitoring. On the chemical safety side, technical and societal transition is lagging behind. The current French system is reliant on : 1/ Surveillance and inspection plans which are used to detect non-conformities i.e. above the maximum level (ML) for a given contaminant /food couple and 2/ National total diet studies which are implemented every 5 years to assess the risk related to the global chronic dietary exposure of consumers to sub-ML doses of contaminants. Due to the very low ML of most priority contaminants, both approaches revolves mainly around very sensitive reference methods which are often expensive and low-throughput, thus limiting frequency and scope of surveillance by food safety authorities and dissuading routine preventive monitoring by the industry.

Starting with the surveillance of PCBs in meat as a model issue, SENTINEL's primary objective is to develop a panel of three complementary high-throughput, sensitive, cost-effective screening tools 1/ for strengthening the detection of non-conformities, but also 2/ for monitoring these contaminants at relevant sub-ML levels. When non-conformities (>ML) are detected, confirmatory analyses by approved laboratories will be requested prior to corrective action whereas detection above targeted sub-ML levels will lead to preventive actions carried on the food chain. With the final aim to better control consumer dietary exposure to these contaminants, the implementation of these novel tools should thus boost the screening of positive samples by food safety authorities (top-down approach) and permit sample self-monitoring by the agri-food industry (bottom-up approach). Recent advances in analytical sciences enable to meet this challenge in SENTINEL via three options: i) coupling highly-sensitive mass spectrometry-based methods with novel sample pooling strategies ii) coupling up-to-date contaminant-targeted biosensors with quick, efficient and cheap extraction methods and iii) designing sensors targeted on the detection of markers of animal exposure to contaminants discovered by omic approaches.

The second objective is to determine several practical and plausible implementation scenarios for the new tools, and to anticipate the main costs and benefits of their meat sector implantation. The project will experiment an original two-stage methodology, in order to improve the technological transfer from research to industry and to food safety authorities. First, the conditions of SENTINEL tool implementation will be defined on the basis of the most probable evolutions of the French meat supply chain. Second, the cost-benefit analysis (economic, regulatory, social, public health impacts) of these implementation scenarios will be performed to support future decisions for strengthening the surveillance of food chain chemical safety.

SENTINEL is a multidisciplinary collaborative research project that will prompt developments in the fields of residue chemistry, biosensors, e-nose, omics, chemometrics, bioinformatics, social and consumer sciences, risk assessment and knowledge engineering. SENTINEL involves 11 partners from 4 scientific institutes (INRA, INRIA, IRSTEA, CNRS), 2 educational and research institutions (Perpignan University, ONIRIS) and 1 technical institute (IFIP).