Discrimination of Healthy and Botrytis cinerea-Infected Grapes Using Untargeted Metabolomic Analysis with Direct...

Botrytis cinerea infections of grapes significantly reduce yield and quality and increase phenolic compound oxidation, resulting in color loss, off-flavors, and odors in wine. In this study, metabolites were extracted from grape homogenates comprising healthy or infected grapes from different vintages, cultivars, regions, and maturity stages. Samples were randomly analyzed by direct injection into an ion trap mass spectrometer, with data collected from 50 to 2000 m/z for 1 min. Molecular feature abundances from 0.1 to 0.4 min were normalized prior to Principal Components Analysis assessment of workflow. Samples were randomly assigned to a calibration and independent test sample set, with feature reduction, a two-class model Partial Least Squares-Discriminant Analysis, cross-validation, and permutation testing performed with the calibration data set. Prediction of sample class in the independent test samples demonstrated an overall predictive error of less than 5%. Feature importance was assessed using a combined variable importance in projection and selectivity ratio plot. Annotation of important molecular features using a high-resolution LC-QTOF mass spectrometry MS/MS of selected samples enabled key metabolites palmitic, oleic, linoleic and linolenic acids, succinate, and epicatechin to be identified and associated with infection. The proposed workflow establishes sensitive high-throughput rapid MS-based methods for phytosanitary testing of grape and fruit samples.