UMR LISAH, Université Montpellier, INRAE, IRD, Institut Agro, 34060, Montpellier, France.
Univ. Bordeaux, INRAE, UMR1332, BFP, 33882, Villenave d'Ornon, France; Bordeaux Metabolome, MetaboHUB, PHENOME-EMPHASIS, 33140, Villenave d'Ornon, France.
Chemosphere. 2023 Oct;337:139302. doi: 10.1016/j.chemosphere.2023.139302. Epub 2023 Jun 27.
Sorption regulates the dispersion of pesticides from cropped areas to surrounding water bodies as well as their persistence. Assessing the risk of water contamination and evaluating the efficiency of mitigation measures, requires fine-resolution sorption data and a good knowledge of its drivers. This study aimed to assess the potential of a new approach combining chemometric and soil metabolomics to estimate the adsorption and desorption coefficients of a range of pesticides. It also aims to identify and characterise key components of soil organic matter (SOM) driving the sorption of these pesticides. We constituted a dataset of 43 soils from Tunisia, France and Guadeloupe (West Indies), covering extensive ranges of texture, organic carbon and pH. We performed untargeted soil metabolomics by liquid chromatography coupled with high-resolution mass spectrometry (UPLC-HRMS). We measured the adsorption and desorption coefficients of three pesticides namely glyphosate, 2,4-D and difenoconazole for these soils. We developed Partial Least Square Regression (PLSR) models for the prediction of the sorption coefficients from the RT-m/z matrix and conducted further ANOVA analyses to identify, annotate and characterise the most significant constituents of SOM in the PLSR models. The curated metabolomics matrix yielded 1213 metabolic markers. The prediction performance of the PLSR models was generally high for the adsorption coefficients Kd (0.3 < R < 0.8) and for the desorption coefficients Kf (0.6 < R < 0.8) but low for n (0.03 < R < 0.3). The most significant features in the predictive models were annotated with a confidence level of 2 or 3. The molecular descriptors of these putative compounds suggest that the pool of SOM compounds driving glyphosate sorption is reduced compared to 2,4-D and difenoconazole, and these compounds are generally more polar. This approach can provide estimates of the adsorption and desorption coefficients of pesticides, including polar pesticide, for contrasted pedoclimates.
吸附作用调节了从种植区到周围水体的农药分散以及它们的持久性。评估水污染风险和评估缓解措施的效率，需要精细分辨率的吸附数据和对其驱动因素的良好了解。本研究旨在评估一种新方法的潜力，该方法结合化学计量学和土壤代谢组学来估计一系列农药的吸附和解吸系数。它还旨在确定和表征驱动这些农药吸附的土壤有机质（SOM）的关键成分。我们组成了一个由来自突尼斯、法国和瓜德罗普（西印度群岛）的 43 种土壤组成的数据集，涵盖了广泛的质地、有机碳和 pH 范围。我们通过液相色谱与高分辨率质谱（UPLC-HRMS）进行了非靶向土壤代谢组学研究。我们为这些土壤测量了三种农药（草甘膦、2,4-D 和吡虫啉）的吸附和解吸系数。我们从 RT-m/z 矩阵开发了偏最小二乘回归（PLSR）模型，用于预测吸附系数，并进行了进一步的方差分析，以确定、注释和表征 PLSR 模型中 SOM 的最显著成分。经过校对的代谢组学矩阵产生了 1213 个代谢标志物。PLSR 模型对吸附系数 Kd（0.3<R<0.8）和解吸系数 Kf（0.6<R<0.8）的预测性能通常较高，但对 n（0.03<R<0.3）的预测性能较低。预测模型中最显著的特征用置信水平为 2 或 3 进行注释。这些假定化合物的分子描述符表明，与 2,4-D 和吡虫啉相比，驱动草甘膦吸附的 SOM 化合物库减少，并且这些化合物通常更具极性。这种方法可以为包括极性农药在内的不同农区的农药吸附和解吸系数提供估计值。
