Untargeted ¹H NMR metabolomics to distinguish milk from different feeding systems for haymilk authentication

This study investigated the distinction between haymilk and milk from silage feeding using proton nuclear magnetic resonance (¹H NMR) spectroscopy, given that the use of silage in cow diet is not allowed in haymilk production as well as in several PDO cheese regulations. A total of 245 raw milk samples from three feeding systems - haymilk, with no fermented feed in the ration (n = 49), maize silage milk with maize silage as the only fermented feed (n = 98), and grass silage milk with grass silage as the only fermented feed (n = 98) - collected between 2019 and 2021 in South Tyrol (Northern Italy) were analyzed using untargeted ¹H NMR of both aqueous and lipid fractions. PCA showed clustering of milk samples by season and sampling period in the aqueous ¹H NMR spectra, especially for haymilk, whereas lipid spectra showed no such clustering, indicating stable fatty acid profiles across seasons and years of collection. Untargeted ¹H NMR metabolomics combined with multivariate supervised classification based on milk lipid profile showed a good discriminant capacity between haymilk and milk from silage feeding, achieving sensitivity and specificity for haymilk samples equal to 87% and 94%, respectively. Furthermore, the proposed modelling approach enabled discrimination between milk from maize silage and grass silage feedings. Conjugated linoleic acid and cyclopropane fatty acids were identified as potential biomarkers to discriminate among the three milk groups. Overall, these results demonstrate that ¹H NMR lipid metabolomics represents a robust approach for the authentication of haymilk for the presence of silage in the ration.