A Toxicophenomic Assessment of Glyphosate Effects on Wildflowers Using Multispectral Imaging

Abstract

Pesticide spray drift is a threat to wildflowers and hence to biodiversity and pollinator habitat. Nontarget species are included in risk assessment of pesticides, yet current ecotoxicological studies rely on destructive and/or low-throughput measurements that do not capture the entire phenotype over time. They also rarely include a large variety of species. The use of novel approaches such as toxicophenomics could tackle these limitations. The present study aimed to validate multispectral indices derived from toxicophenomic measurements and assess interspecific variation in glyphosate sensitivity. The method development focused on two species of wildflowers (Centaurea cyanus and Sinapis arvensis) and was applied to four more wildflower species (Papaver rhoeas, Geranium molle, Veronica agrestis, Prunella vulgaris) and one crop species (Hordeum vulgare). Plants were exposed to eight doses of the herbicide glyphosate and monitored in an automated phenotyping platform. High-throughput multispectral measurements were used to calculate vegetation indices. Four high-throughput indices covering various plant physiological processes were chosen based on the strength and consistency of their correlation with well-established low-throughput endpoints, and they were used for a dose-response analysis and benchmark dose estimation. Out of the four indices, those related to pigments (Red-edge chlorophyll index for chlorophyll content and reflectance at 570nm for anthocyanin content) performed best, with the reflectance at 570mn yielding having the highest concordance with low-throughput estimations. Further fine-tuning is recommended for the other two high-throughput indices related to water (Normalized water index) and photosynthesis-related processes (Pigment corrected photochemical reflectance index). The high-throughput data allowed to identify large differences in sensitivity between species. Including a wide range of species in risk assessment is recommended to account for this variability. Toxicophenomics using multispectral imaging and pigment reflectance indices is promising for ecotoxicological studies as it offers high-resolution data that accurately reflect the physiological responses of plants exposed to pesticide spray drift.