Influence of direct and sequential extraction methodology on metabolic profiling

A systematic comparison was made of the detected metabolite profiles for two plant materials (black beans and soybeans) and a dietary supplement (black cohosh) extracted using sequential (hexane, ethyl acetate, and 50% aqueous methanol) and direct extraction with three solvent systems (80% aqueous methanol, methanol/chloroform/water (2.5:1:1, v/v/v) and water). Extracts were analyzed by LC–MS (without derivatization) and GC-FID (with BSTFA/TMCS derivatizations). For sequential extraction, HPLC-UV and BSTFA/TMCS-derivatized GC-FID detection were more responsive to the polar molecules with a rough distribution of 10%, 10%, and 80% of the total signals in hexane, ethyl acetate, and 50% aqueous methanol, respectively. With HPLC–MS detection, the distribution of signals was more balanced, roughly 40%, 30%, and 30% for the same extracts (hexane, ethyl acetate, and 50% aqueous methanol). For direct extraction, HPLC-UV and BSTFA/TMCS-derivatized 4GC-FID provided signals between 60% and 150% of the total sequential extracted signals. The overlap of signals for the 3 sequential extracts ranged from 1% to 3%. The overlap of the signals for direct extraction with the total for sequential extraction ranged from 15% to 98%. With HPLC–MS detection, signals varied from 30% to 40% of the total signals for sequential extraction. Multivariate analysis showed that the components for the sequential and direct extracts were statistically different. However, each extract, sequential or direct, allowed discrimination between the 3 plant materials.