Identification of Natural Metabolites in Mixture: A Pattern Recognition Strategy Based on ¹³C NMR

Because of their highly complex metabolite profile, the chemical characterization of bioactive natural extracts usually requires time-consuming multistep purification procedures to achieve the structural elucidation of pure individual metabolites. The aim of the present work was to develop a dereplication strategy for the identification of natural metabolites directly within mixtures. Exploiting the polarity range of metabolites, the principle was to rapidly fractionate a multigram quantity of a crude extract by centrifugal partition extraction (CPE). The obtained fractions of simplified chemical composition were subsequently analyzed by (13)C NMR. After automatic collection and alignment of (13)C signals across spectra, hierarchical clustering analysis (HCA) was performed for pattern recognition. As a result, strong correlations between (13)C signals of a single structure within the mixtures of the fraction series were visualized as chemical shift clusters. Each cluster was finally assigned to a molecular structure with the help of a locally built (13)C NMR chemical shift database. The proof of principle of this strategy was achieved on a simple model mixture of commercially available plant secondary metabolites and then applied to a bark extract of the African tree Anogeissus leiocarpus Guill. & Perr. (Combretaceae). Starting from 5 g of this genuine extract, the fraction series was generated by CPE in only 95 min. (13)C NMR analyses of all fractions followed by pattern recognition of (13)C chemical shifts resulted in the unambiguous identification of seven major compounds, namely, sericoside, trachelosperogenin E, ellagic acid, an epimer mixture of (+)-gallocatechin and (-)-epigallocatechin, 3,3'-di-O-methylellagic acid 4'-O-xylopyranoside, and 3,4,3'-tri-O-methylflavellagic acid 4'-O-glucopyranoside.