fraction I was initiated at a 0.001 AU threshold. The upslope was gradual during the following 20 min, the signal showed a maxi- mum with a near plateau between 60 and 75 min the maximal sig- nal was at 380 mV, corresponding to 0.991 AU absorbance. A sharp downslope was observed between 80 and 85 min into the run, and collection of fraction I was terminated after ca. 90 min after a total of 9.5 l had been collected. Fraction II was collected between 98 and 102 min. This signal showed a sharp upslope and maximum, but a tailing downslope. This eluate appeared yellow as opposed to the other fractions, which appeared colourless (fraction I) or nearly colourless (fraction III). Fraction III was collected starting at the valley between second and third peak at 105–115 min. When the 3 l of solvent B were applied, the baseline was back to near zero and flat, and the run was terminated at 120 min. Results from the HPLC/ELS analysis of the starting sample solution and the three collected fractions are shown in Fig. 3.
Chromatogram in Fig. 3A shows the sample solution which contained minor amounts of 2 and a second signal with similar chromatographic properties, but a slightly shorter retention time (peak 6). Fig. 3B shows the analysis of fraction I, which lacks these signals but contains 1. Considering the dilution factor of 2.4 due to elute volume, and a detection limit of less than 0.5 ppm for limonin, these results suggests the successful removal of 2 or at least lowering of its content by more than 10 fold. Fig. 3C shows the analysis of fraction II. The chromatogram shows multiple signals with retention times longer than that of 1. Upon further analysis of this fraction by LC–MS, peaks 3 and 4 were identified, based on their exhibited masses and in concert with comparison to authentic standards as nomilinic acid glucoside (peak 3, m/z 711.3) and obacunone glucoside (peak 4, m/z