Thesis (MScFoodSc)--Stellenbosch University, 2013. / ENGLISH ABSTRACT: The phenolic composition of Cyclopia species is believed to be partially responsible for the numerous health promoting properties associated with their extracts. Current quality control measures do not accommodate variation in phenolic profiles of Cyclopia species. In this study, comprehensive high performance liquid chromatography (HPLC) methods were developed for the improved characterisation of the phenolic composition of aqueous extracts of two Cyclopia species (C. subternata and C. maculata). The methods were developed to be suitable for both routine quantitative analysis on conventional HPLC instrumentation, and the construction of chromatographic fingerprints for further data analysis. The latter entailed similarity analysis and prediction of total antioxidant capacity (TAC). Using a methodical approach, two separate HPLC methods, using diode array detection (DAD), were developed and validated for the analysis of aqueous extracts prepared from unfermented (green) and fermented plant material of C. subternata and C. maculata. Separation was achieved using the same method parameters (column, temperature, mobile phases), except for differing mobile phase gradients. Hyphenation of the developed HPLC methods with mass spectrometry (MS) and tandem MS allowed the confirmation of phenolic compounds previously identified in Cyclopia, and the tentative identification of several additional compounds in Cyclopia species, which are reported here for the first time. These included apigenin-6,8-di- C-glucoside, 3-hydroxyphloretin-30,50-di-C-hexoside, eriodictyol-di-C-glucoside, iriflophenone-di-O,C-hexoside, hydroxymangiferin and hydroxyisomangiferin. Subsequently, a large number of aqueous extracts of randomly selected green C. subternata (n = 64) and C. maculata (n = 50) plant material samples were analysed. Large quantitative variations were observed on intra- and inter-species levels. Cyclopia maculata extracts contained almost six times more mangiferin than extracts from C. subternata. HPLC-DAD analysis produced duplicate fingerprints for each extract which were consequently used for further analysis. The chromatographic fingerprint of a bioactive extract of each species was included in the respective data sets. Similarity analysis was conducted between the fingerprints from the randomly selected extracts and the corresponding active extract. For each species several extracts were determined to have similar “activity” as that of the active extract (n = 15 for C. subternata and n = 45 for C. maculata). Compounds potentially responsible for the activity were tentatively identified with the aid of principal component analysis (PCA) in combination with similarity analysis. PCA was more effective in identifying small differences between fingerprints than similarity analysis based on the correlation coefficients (r) alone. Furthermore, multivariate data analysis was used to construct partial least squares (PLS) regression models for the prediction of TAC from fingerprint data of each species, and available data from two microplate TAC assays. The construction of the models was successful with reasonable errors (< 10%), and permitted the determination of compounds of interest for future research. These included compounds of known identity that had large positive contributions toward the predictions of TAC, or unknown compounds that had small UV signals, but relatively large positive contributions to the models. / AFRIKAANSE OPSOMMING: Die talle gesondheidbevorderingseienskappe van ekstrakte van Cyclopia spesies word gedeeltelik geassosieer met hul fenoliese samestelling. Huidige kwaliteitskontrolemaatreëls is nie in staat om die variasie wat in die fenoliese profiele van die spesies voorkom, te akkommodeer nie. Omvattende hoë druk vloeistof chromatografiese (HPLC) metodes is vir twee Cyclopia spesies, naamlik C. subternata en C. maculata, in hierdie studie ontwikkel vir beter karakterisering van die fenoliese samestelling van waterekstrakte van dié spesies. Die metodes moes ook geskik wees vir roetine analise van C. subternata en C. maculata ekstrakte op konvensionele HPLC instrumentasie, en vir die opstel van chromatografiese vingerafdrukke (fenoliese samestellingsprofiele) vir verdere data analise, soos gelykvormigheidsanalise en die voorspelling van die totale antioksidantkapasiteit (TAC). Twee HPLC metodes, wat van ’n ultraviolet-diode detektor (DAD) gebruik maak, is ontwikkel deur ’n sistematiese benadering te volg. Die onderskeie metodes is vir die ontleding van waterekstrakte van groen (ongefermenteerde) en gefermenteerde plantmateriaal van C. subternata en C. maculata gevalideer. Ongeag die spesie is optimale skeiding met dieselfde kolom, mobiele fase en kolom-temperatuur bereik, maar met verskillende mobiele fase gradiënte. Analise met massaspektrometrie (MS) en tandem MS het die teenwoordigheid van fenoliese verbindings, wat voorheen in Cyclopia spesies geidentifiseer is, bevestig. Verder is ook ’n aantal verbindings vir die eerste keer in Cyclopia tentatief geidentifiseer. Dit sluit apigenien-6,8-di-C-glukosied, 3- hidroksiefloretien-30,50-di-C-heksosied, eriodiktiol-di-C-glukosied, iriflofenoon-di-O,C-heksosied, hidroksiemangiferien en hidroksie-isomangiferien in. Vervolgens is ’n groot aantal ewekansig gekose waterekstrakte van beide groen C. subternata (n = 64) en C. maculata (n = 50) plantmateriaal geanaliseer, en groot kwantitatiewe variasie op intra- en inter-spesievlak waargeneem. Cyclopia maculata ekstrakte het byvoorbeeld byna ses maal die mangiferieninhoud van C. subternata ekstrakte gehad. HPLC-DAD analise van die ekstrakte het duplikaat vingerafdrukke van elke ekstrak geproduseer, wat vir verdere data analise gebruik is. Die chromatografiese vingerafdruk van ’n bioaktiewe ekstrak van elke spesie was by die onderskeie datastelle ingesluit. Gelykvormigheidsanalise is tussen vingerafdrukke van die ewekansig gekose ekstrakte en die ooreenstemmende aktiewe ekstrak uitgevoer. Vir elke spesie is ’n aantal “aktiewe” ekstrakte aangewys (n = 15 vir C. subternata en n = 45 vir C. maculata). Die verbindings wat potensieel verantwoordelik kan wees vir die aktiwiteite is met behulp van hoofkomponentontleding (PCA) in kombinasie met gelykvormigheidsanalise, tentatief aangewys. PCA was egter meer effektief om klein verskille tussen vingerafdrukke aan te dui, in vergelyking met gelykvormigheidsanalise wat slegs op die korrelasie koëffisiënt (r) gebaseer is. Meerveranderlike data analiese is gebruik om “gedeeltelike kleinste kwadrate” (PLS) regressiemodelle, vir die voorspelling van die TAC van beide spesies te bou. Die voorspelling is gebaseer op hul vingerafdruk data en TAC data van twee TAC mikroplaat metodes. Die model-konstruksie was suksesvol met aanvaarbare voorspellingsfoute (< 10%). Verbindings van belang kon ook bepaal word. Dit sluit bekende verbindings in wat groot positiewe bydraes ten opsigte van die voorspelling van TAC getoon het, asook ongeidentifiseerde verbindings wat klein UV-seine getoon het, maar relatiewe groot bydraes tot die modelle gehad het.