Abstract
Legumes are known to exert potential health benefits with evidence stemming from epidemiological studies. Furthermore, a significant body of evidence demonstrates that isoflavone metabolites are good markers of soy intake, while research is lacking on specific biomarkers from other leguminous sources eg. peas. In this context, a randomised cross-over acute intervention study was conducted on healthy volunteers (n=11) with the objective to detect urinary biomarkers of peas intake using untargeted metabolomics. Participants consumed peas and couscous (control food) in a random order and urine samples were collected postprandially and analysed by liquid chromatography-mass spectrometry (UPLC-QTOF-MS). The data obtained was reduced to 819 features in positive mode and 1004 features in negative mode applying filtering procedures, such as data alignment and sample occurrence frequency. Following data pre-processing, multivariate statistical analysis was applied to further reduce the data dimensionality. Robust models were obtained for comparison of fasting against the postprandial time points. For example, fasting vs 4h and fasting vs 6h revealed robust PLS-DA models [(R2X=0.41, Q2=0.4) and (R2X=0.517, Q2=0.495) (-ve mode), respectively]. Subsequently, the models were validated by permutation testing. The variable importance for the projection (VIP) list was created from the PLS-DA plot and variables with scores ≥1.5 were considered discriminant between the two time points. Time series plots of these features revealed at least 26 features (14 in -ve and 12 in +ve) showing a time response following peas consumption. Overall, 8 features displayed a differential time course as compared to the control food. Further analysis will focus on identification of these discriminating features of interest by metabolite annotation using databases, such as Metlin, HMDB and confirmation by LC-MS/MS which should result in identification of specific pea markers.
Keywords
metabolomics, biomarkers, legumes