Ultra-High Performance Liquid Chromatography for the analysis of steviol glycosides

When analyzing natural products, such as steviol glycosides, one must always be aware of the fact that the analytical method used might not reveal every compound that is actually present in the sample. For instance, a chromatogram representing the separation of steviol glycosides on two 20 cm C18 columns in series (dp= 5.0 μm) indicates the presence of 12 compounds; four of them not being identified yet. Some compounds may be "invisible" to the instrument, because they are hidden under larger peaks, or because they elute very late and the band broadening renders them undetectable. The van Deemter equation predicts a dramatic increase in column performance(and hence: resolution) by using much smaller particle diameters to pack the column. However, these columns will generate a substantially larger back pressure in the system, so a pump is needed that can deliver this pressure. Hence the name of the technique: Ultra-High Performance (or: Pressure) Liquid Chromatography (UHPLC). Moreover, a high inlet pressure allows the use of longer column lengths,further increasing the resolution of the separation. An automatic column coupling device, installed on an Agilent Infinity UHPLC system with a maximum inlet pressure of 1200 bar, was used to couple several Acquity BEH C18 columns (particle size: 1.7 μm) in series. This set-up was used to evaluate the separation of a steviol glycosides sample on different column lengths, ranging between 5 cm and 25 cm. Various gradients of aqueous formic acid (0.1%) - acetonitrile were first run on the shortest column length (5 cm), allowing selection of the best gradient conditions in the shortest possible time. The gradient conditions resulting in the best separation were subsequently repeated on a 10, 15, and 20 cm support. A total of 26 peaks could be separated on the longest column length (20 cm), obtained by coupling two 2.1 x 100 mm columns. None of the extra peaks could be identified at this moment, but it is hoped to identify them using LCMS.