Qualitative and quantitative gas chromatography-mass spectrometry analysis of the principal toxic constituents of some South African plants in human biological fluid

Abstract

In South Africa there are regular cases of human and animal poisoning where plants used in ethnic medicine are suspected to be the source of the poison. This project aims to develop methods for identifying and quantifying the principal toxins of four such plants in human urine. In this way a steppingstone for further research and development on this subject is produced.

This project focuses on toxins from four plants that fall into two broad classes of compounds: cardiac glycosides (Acokanthera oppositifolia and Urginea sanguinea) and alkaloids (Boophane disticha and Gloriosa superba). The principal toxic compounds within the respective plants are: acovenoside A, scillaren A, buphanidrine and colchicine. In this study, buphanidrine was isolated from a dichloromethane-methanol plant extract of B. disticha, through the utilisation of multiple chromatographic techniques, and its structure confirmed by nuclear magnetic resonance spectroscopy and high-resolution mass spectrometry. Samples of known analyte concentrations were prepared by spiking blank urine samples with specific amounts of analyte stock solution. Extraction of colchicine and cardiac glycosides from urine was accomplished by solid-phase extraction. The extracted cardiac glycosides were hydrolysed in hydrochloric acid and the reaction products extracted via liquid-liquid extraction. Following extraction, the products of the hydrolysis reaction were silylated for gas chromatography-mass spectrometry analysis.

Extracts from urine samples containing colchicine were prepared without a hydrolysis process. Urine samples spiked with buphanidrine were prepared by buffering to alkaline pH, and subsequently conducting a liquid-liquid extraction. The organic phase extracts for both alkaloids were concentrated by evaporation and reconstitution in a small volume of an organic solvent. Both alkaloids were analysed without derivatisation following up-concentration. Response models were generated by duplicate analyses of three batches of each analyte in a de-ionised aqueous solution and urine, respectively. The applicability of the linear response models to the measured data was evaluated for each respective analyte. The relative standard deviation was evaluated to establish the variance in the linear models. The response models, developed for colchicine, acovenoside A and scillaren A, were problematic and further investigated to determine the contributing factors. In the case of colchicine and acovenoside A, the utilisation of log-linear response models was less problematic in describing the trends in the data.

In analysing the trends in the processed data, the limits of detection and quantification were calculated statistically for buphanidrine, acovenoside A and colchicine. These were then compared to limits observed in the qualitative analysis of the analytes based on ion ratios. For each analyte five characteristic ion ratios were selected, considering the molecular ion signal area, relative to that of the base peak ion for each analyte. Applying multivariate Gaussian statistical analysis techniques, the correlation of ion ratios and their dependence upon analyte concentration was established. This proved insightful regarding the mechanism of analyte fragmentation.

Suggested further work, following this project, should look at factors concerning the optimisation of sample preparation methods, especially the purification via solid-phase extraction. The project provided an opportunity for the investigation of trends in ion ratios, to determine the fundamental limits of identification.