Comprehensive Phenolic Profiling of Cyclopia genistoides (L.) Vent. by LC-DAD-MS and -MS/MS Reveals Novel Xanthone and Benzophenone Constituents

A high-performance liquid chromatographic (HPLC) method coupled with diode-array detection (DAD) was optimized for the qualitative analysis of aqueous extracts of Cyclopia genistoides. Comprehensive insight into the phenolic profile of unfermented and fermented sample extracts was achieved with the identification of ten compounds based on comparison with authentic reference standards and the tentative identification of 30 additional compounds by means of electrospray ionization mass spectrometry (ESI-MS) and tandem MS detection. Three iriflophenone-di-O,C-hexoside isomers, three xanthone-dihydrochalcone derivatives and one dihydrochalcone are herein tentatively identified for the first time in C. genistoides. Of special interest is one iriflophenone-di-O,C-hexoside present in large amounts. New compounds (tentatively) identified for the first time in this species, and also in the genus Cyclopia, include two aromatic amino acids, one flavone, an iriflophenone-di-C-hexoside, a maclurin-di-O,C-hexoside, two tetrahydroxyxanthone-C-hexoside isomers, a tetrahydroxyxanthone-di-O,C-hexoside, two symmetric tetrahydroxyxanthone-C-hexoside dimers, nine glycosylated flavanone derivatives and five glycosylated phenolic acid derivatives. The presence of new compound subclasses in Cyclopia, namely aromatic amino acids and glycosylated phenolic acids, was demonstrated. The HPLC-DAD method was successfully validated and applied to the quantitative analysis of the paired sample extracts. In-depth analysis of the chemical composition of C. genistoides hot water extracts gave a better understanding of the chemistry of this species that will guide further research into its medicinal properties and potential uses.


HPLC-DAD Method Development
Method development was focused on optimizing the selectivity on a high-efficiency column and thus the first step entailed selecting the optimum chromatographic support. Separation provided by the 3 µm Gemini-NX column, currently employed in the analysis of other Cyclopia species [12,13], was compared to that obtainable on the 1.8 µm Zorbax SB-C18 column under constant gradient conditions and temperature. The 1.8 µm Zorbax column was included in this study due to the known benefits of sub-2 µm phases for speeding up reversed phase (RP) LC analyses [14]. The potential gain in efficiency provided by a core-shell column (2.6 µm Kinetex C18) was also investigated. The best separation was ultimately achieved on this core-shell column, which is in line with a study on the RP-LC separation of proanthocyanidins showing that the Kinetex column kinetically outperforms solid supports over efficiencies in the practical range of ca. 25,000-250,000 plates [15].
Based exclusively on selectivity considerations, the aqueous (aq.) phase was selected as 1% aq. formic acid and the organic component as a 1:1 mixture of methanol and acetonitrile (obtained by on-line mixing). When assessed individually, the organic modifiers were both characterized by insufficient chromatographic resolution. For example, the use of 100% methanol led to perfect co-elution of the major constituent, mangiferin, and its C-4 regioisomer isomangiferin, irrespective of gradient profile and temperature conditions. As opposed to 100% methanol, which is highly viscous, the mixture also served to lower the operating pressure, which is an important consideration for the Agilent 1200 instrument (P max = 400 bar).
After selecting the optimal solid support and mobile phase components, the gradient profile and column temperature were optimized. Selectivity effects as a function of changes in mobile phase composition and temperature are often complementary [16] and therefore these parameters were optimized simultaneously. An isocratic hold period at initial conditions, followed by a flat increase in solvent strength, were required to attain sufficient retention of the highly polar, early eluting compounds, whilst also improving band spacing and resolution of the later eluting compounds. This effect was more pronounced at lower temperatures. At the selected temperature of 30 °C, the optimized gradient profile comprised of a 5 min isocratic hold period at 5% organic modifier, followed by an increase to 25% organic modifier over 40 min. The second gradient step entailed an increase to 50% organic modifier over 10 min. The total chromatographic run time, including column re-equilibration, was 65 min.
The critical effect of mobile phase composition on the separation of honeybush phenolic compounds was demonstrated during method transfer from the Agilent 1200 instrument, where a quaternary pump provides low-pressure mixing of methanol and acetonitrile, to the Waters UPLC instrument equipped with a binary pump (high-pressure mixing). To achieve separation comparable to that of the Agilent instrument on the Waters instrument, methanol and acetonitrile had to be premixed using a ratio of 45% methanol to 55% acetonitrile (v/v). Under the optimized RP-LC conditions, a large number of phenolic compounds present in hot water extracts of C. genistoides were successfully separated ( Figure 1a). Compounds that were only present in the fermented sample extract and/or only detected in positive ionization mode during LC-ESI-MS analyses, are depicted in the extracted mass chromatograms ([M+H] + , Figure 1b-d). Exclusive detection of compounds in the fermented extract does not necessarily suggest their formation during fermentation, but may be related to the challenge of analyzing samples which contain very high levels of certain compounds compared to other constituents. With fermentation a decrease in the content values of the former compounds allows minor compounds to be more easily observed due to a shift in the relative peak area ratios.  Table S1].

LC-DAD-ESI-MS and -MS/MS Identification of Compounds Present in Hot Water Extracts of Unfermented and Fermented C. genistoides
Identification of compounds was performed by assigning each peak to a compound subclass based on their characteristic UV-Vis spectra, where possible [17]. Accurate mass measurement and MS/MS fragmentation patterns were then used to tentatively identify the molecular structures. Ten compounds 1-10 were identified by co-elution with the authentic reference standards, whilst 30 additional   compounds a-cc were tentatively identified by interpretation of their UV-Vis and mass spectral data compared to relevant literature reports, as discussed per compound subclass below. A further nine compounds A-I could not be identified based on the current data (Supplementary Information, Table S1). The characteristics of authentic reference standards, not present in sample extracts, but used in the identification of unidentified constituents, can be found in the Supplementary Information (Table S2). Structures for selected compounds and/or selected subclasses are shown in Figure 2.

Benzophenone Derivatives
Compound 3 was identified as iriflophenone-3-C-glucoside ( Figure 2; Table 1), based on comparative data for the isolated reference compound. Six additional benzophenone derivatives were tentatively identified in aqueous extracts of unfermented and fermented C. genistoides (Table 1). The neutral loss of a hexoside moiety is characteristic of a flavonoid O-glycoside, whilst the cross-ring cleavage of a hexoside moiety points toward a C-glycoside [17,18]. In the lower molecular weight region of the MS/MS spectrum in negative ionization mode, fragment ions characteristic of iriflophenone-3-C-glucoside were observed [8]. Compounds b, d and f were thus broadly assigned as iriflophenone-di-O,C-hexoside isomers, with the O-hexoside moiety either linked to a hydroxyl group on the aglycone or to a hydroxyl group of the C-bound hexoside residue. The data presented herein are in agreement with those of an iriflophenone-di-O,C-hexoside tentatively identified in C. subternata [12].
Compounds y and aa eluted at t R of 37.41 min and 44.24 min, respectively, and were assigned the elemental composition C 19 H 18 O 11 , indicating that they are possible isomers of (iso)mangiferin. The . This fragmentation pattern is characteristic of C-glycosyl xanthones [22]. In positive ionization mode, the fragmentation pattern of compounds y and aa showed a stronger correlation with mangiferin than with isomangiferin, as the base peak ion was detected at m/z 273 as opposed to m/z 303. This would suggest that the position of glycosylation is at C-2 on the dibenzo-γ-pyrone skeleton. However, based on the available data, it was not possible to confirm the position of glycosylation, nor to ascertain the nature of the hexoside moiety or the hydroxylation pattern of the dibenzo-γ-pyrone skeleton. Compounds y and aa were thus broadly assigned as tetrahydroxyxanthone-C-hexoside isomers.  Table S2), analyzed under the same experimental conditions. On the other hand, compound l eluted before neomangiferin, which suggests an isomer. Compound l was thus tentatively identified as a tetrahydroxyxanthone-di-O,C-hexoside. This is the first report of the presence of tetrahydroxyxanthone-C-hexoside isomers besides mangiferin and isomangiferin, as well as a tetrahydroxyxanthone-di-O,C-hexoside, in Cyclopia spp. extracts. Based on the established anti-diabetic activity of mangiferin and neomangiferin [23][24][25], the aforementioned compounds could possibly contribute to the anti-diabetic potential of C. genistoides aqueous extracts.      200  220  240  260  280  300  320  340  360  380  400  420  440  460  480  500  520  540  560  580  600  620  640  660  680  700  720  740  760  780  800  820  840  860  880 m/z  200  220  240  260  280  300  320  340  360  380  400  420  440  460  480  500  520  540  560  580  600  620  640  660  680  700  720  740  760  780  800  820  840  860  880  This fragmentation pattern, characterized by cross-ring cleavage of the saccharide residues and subsequent loss of water molecules, corresponds to the simultaneous fragmentation of two C-hexosyl groups. It was therefore postulated that each of the tetrahydroxyxanthone monomers contains a single C-linked hexosyl group. This was confirmed by the presence of a fragment ion at m/z 419 (30%-40% intensity; [M−H−422] − ) in the negative ionization mode MS/MS spectra, which corresponds to the neutral loss of tetrahydroxyxanthone-C-hexosyl (-C 19  Compounds g and k were therefore tentatively assigned as tetrahydroxyxanthone-C-hexoside dimers. Such a symmetric homodimer of mangiferin, termed mangiferoxanthone A, was recently isolated from mango tree stem bark and exhibited moderate anti-viral activity [26]. Other xanthone dimers with a MW of 842 have thus far only been identified in Swertia punicea Hemsl. (Gentianaceae), namely swertiabisxanthone diglucopyranoside and 3-glucosylpuniceaside A [22]. Based on the mass difference of 450 amu with regards to (iso)mangiferin, it is herewith postulated that compound n could possibly be a C-C linked tetrahydroxyxanthone-C-hexoside (e.g., (iso)mangiferin) and pentahydroxydihydrochalcone-C-hexoside (e.g., aspalathin). Aspalathin (3-hydroxyphloretin-3'-C-glucoside) has not been detected in Cyclopia spp. to date, including the C. genistoides extracts of the present study. However, the related compounds, 3-hydroxyphloretin-3',5'-di-C-hexoside (compound x) and phloretin-3',5'-di-C-glucoside (compound z), were shown to be present ( Figure 2; Table 3). This hypothesis is further supported by the presence of fragment ions typical of aspalathin in the MS/MS spectrum of compound n in positive ionization mode, i.e., m/z 151, m/z 139 and m/z 123 ( Supplementary Information, Table S2). Most notable is the common fragment ion at m/z 123 [(C 7 H 7 O 2 ) + ], which represents the base peak ion in the MS/MS spectrum of aspalathin in positive ion mode. Based on the supporting evidence, compound n was tentatively identified as an aspalathin derivative of (iso)mangiferin.    Table S2). These data led us to propose that compound r is a nothofagin derivative of (iso)mangiferin.
Compound  Table 2). In the positive ionization mode, the precursor ion at m/z 841 ([M+H] + ) furthermore presented product ions common to both compounds n and r, thereby suggesting another dihydrochalcone derivative of (iso)mangiferin. The pseudomolecular ion of (iso)mangiferin detected at m/z 421 ([M−H] − ) represents the neutral loss of 418 amu from the deprotonated dimeric unit, which may correspond to a dihydroxydihydrochalcone-C-hexoside monomeric unit with a molecular formula of C 21 H 24 O 9 (420 Da). These data point towards schoepfin A [27], which merely differs from nothofagin by the absence of a hydroxyl group at C-6' on the A-ring. Compound cc was therefore tentatively identified as a schoepfin A derivative of (iso)mangiferin.
These dihydrochalcone derivatives of (iso)mangiferin are herein reported for the first time in C. genistoides extracts. The linkage of xanthones to other phenolic compounds such as flavone-C-glycosides (e.g., swertifrancheside [28]) has been reported previously, although the occurrence of bisxanthones/tetrahydroxyxanthone dimers in higher plants and fungi are more common (e.g., [22,26]).

Flavanones
Flavanones usually occur as O-glycosyl derivatives, with the sugar moiety preferentially bound to the aglycone hydroxyl group at C-7 or C-3 [29]. Three known flavanone-7-O-disaccharides were identified in the sample extracts by co-elution with the authentic reference standards, namely eriodictyol-7-O-rutinoside (eriocitrin, 7), naringenin-7-O-rutinoside (narirutin, 8) and hesperetin-7-O-rutinoside (hesperidin, 9) ( Figure 2; Table 4). In addition to these compounds, nine additional flavanones were also tentatively identified ( Table 4). The UV-Vis spectra of compounds p, q 1,2 , s, t, u, v, w and bb showed maximum absorption at ca. 280 nm with an undefined shoulder at ca. 330 nm corresponding to a flavanone structure. In negative ionization mode, the presence of the fragment ion at m/z 287 in the MS/MS spectra of compounds p, q 1,2 and s indicated eriodictyol as the aglycone, while the fragment ion at m/z 271 indicated naringenin as the aglycone for compounds t, u, v, w and bb.   [17,18,30], and show more pronounced fragmentation than their rutinose analogues [31]. In accordance with the systematic nomenclature described by Domon and Costello [32], the interglycosidic linkage for compound bb was characterized as (1→2) based on the relatively high intensities of the Y 1 − (m/z 433 at 15% intensity) and showed an intensity of 60% and 100% in the MS/MS spectra of compounds r and s, respectively, and the corresponding monosaccharide ion at m/z 145 was also present at 90%-100% relative abundance in both spectra. The loss of 120 amu observed with ions at m/z 445 (20% intensity) and m/z 299 (20% intensity) could have resulted from the RDA reaction of the flavanone, although this has previously only been observed for the neohesperidosides [30]. The loss of 148 amu could not be explained and therefore compounds t and u were only classified as naringenin derivatives.
Four eriodictyol derivatives (compounds p, q 1,2 and s) were tentatively identified in the sample extracts based on their UV-Vis spectra and base peak ions detected at m/z 287 in negative ionization mode. Compounds p and q 1 161 (85-100). Loss of 136 amu from the deprotonated molecules could signify retrocyclization of the C-ring in the eriodictyol aglycone, whilst the neutral loss of 162 amu corresponds to an O-hexosyl moiety. The corresponding monosaccharide ion was also detected at m/z 161. The difference in mass between the fragment ion at m/z 419 and the molecular ion of the eriodictyol aglycone at m/z 287 (15%-30% relative abundance) furthermore indicates the presence of an O-pentosyl moiety (132 amu). The position of these glycan substituents of different mass could not be determined, but preferential cleavage of the hexosyl-aglycone bond suggests that the hexosyl is bound at a position on the flavanone ring that is more susceptible to acid hydrolysis. Compounds p and q 1 were thus tentatively identified as eriodictyol-O-hexose-O-pentose isomers. Following a similar approach and based on the corresponding mass spectral data, compounds q 2 (t R = 22.90 min) and s (t R = 24.38 min) were identified as eriodictyol-O-hexose-O-deoxyhexose isomers.
The presence of these flavanone derivatives in extracts of C. genistoides exemplifies the high degree of structural variation in natural plant extracts. Such glycosylated derivatives of eriodictyol and naringenin, other than the two most common compounds eriocitrin and narirutin, have been detected in extracts of C. genistoides (unidentified compound 7 [21]) and C. subternata (eriodictyol-di-C-hexoside, eriodictyol-O-glucoside, naringenin-di-C-hexoside and naringenin-O-dihexoside [12]). The flavanone aglycones, eriodictyol and naringenin, as well as their rutinoside derivatives, have phytoestrogenic activity, indicating the potential of these glycosylated flavanone derivatives to contribute to the phytoestrogenic potential of C. genistoides extracts (as reviewed by [33]).

Amino Acids
Compounds 1 (t R = 2.83 min) and 2 (t R = 4.89 min) were only detected in positive ionization mode and produced protonated molecular ions at m/z 182 and m/z 166, respectively (Figure 1b; Table 3). Compounds 1 and 2 were identified as the aromatic amino acids tyrosine and phenylalanine (Figure 2), respectively, based on comparison with authentic standards. To our knowledge, this is the first report of the presence of these aromatic amino acids in Cyclopia spp. The key role of phenylalanine as intermediate in the Shikimate pathway, central to biosynthesis of phenolic compounds in plants, could possibly explain the presence of these aromatic amino acids in the analyzed honeybush extracts.

Glycosylated Phenolic Acids
Phenolic acids are aromatic secondary plant metabolites possessing at least one carboxylic acid functionality. This group of organic acids contains two distinctive carbon frameworks, namely the hydroxybenzoic acid structure (C6-C1) and hydroxycinnamic acid structure (C6-C3) (Figure 2). Two glycosylated hydroxybenzoic acid derivatives (compounds e and h) and two hydroxycinnamic acid derivatives (compounds m and o) were tentatively identified in C. genistoides hot water extracts (Table 3). Under the RP-LC conditions reported in this study, and in line with literature [17], the hydroxybenzoic acids eluted first. The presence of these compounds in extracts of Cyclopia could impart additional health benefits, stemming from the potent antioxidant activity recorded for their underivatized counterparts [34]. One glycosylated phenylpropanoic acid (compound j) was also tentatively identified in the sample extracts (Table 3).
Compound e eluted at 11.14 min and exhibited maximum UV absorption at 314 nm. Its molecular formula was assigned as C 12  . This compound has been isolated from the stems of Spatholobus suberectus (family Fabaceae; traditional Chinese medicine) [37] and also from Lens culinaris Medik. (lentil cultivars) [38].
These two dihydroxybenzoic acid glycosides are herein tentatively identified for the first time in Cyclopia spp. Benzoate and 3-hydroxybenzoate are also key molecules in the biosynthetic pathway for mangiferin, isomangiferin and iriflophenone-3-C-glucoside [5], and therefore the presence of compounds e and h as hydroxylated and glycosylated derivatives of these key molecules (enhancing their water-solubility and storage in "inactive forms" [29]) is not unexpected.
In  [40,41], and thus it is very likely that compound m has the same glycosylation pattern. This is further supported by the presence of related compounds such as p-coumaric acid, phenylethanol-3-O-apiosylglucoside and benzaldehyde-4-O-apiosylglucoside in extracts of fermented C. intermedia [42,43] and unfermented C. subternata [44]. Based on the available data, however, it was not possible to establish the identity or the absolute configuration of the individual monosaccharides, nor the position of glycosylation or the type of interglycosidic linkage. The MS characteristics of compound m are in line with data reported for an unidentified compound 7 in extracts of C. subternata [12].
In a similar manner, the identity of compound o (t R = 21.52 min) with a pseudomolecular ion at m/z 473 ([M−H] − ) and fragment ions at m/z 179 and m/z 135 corresponding to a caffeic acid aglycone [39] was tentatively assigned to caffeic acid O-(pentosyl)hexoside. No information on such a compound could be found in literature. The linkage of an apiosylglucoside to the propenoic side chain (esterification), rather than the hydroxyl group on the aromatic ring of caffeic acid has, however, been reported (1-O-caffeoyl-β-D-apiofuranosyl-(1→6)-β-D-glucopyranoside [45]).
Compound j (t R = 16.77 min) was assigned a molecular formula of C 15 Figure 2). By taking the accurate mass and proposed molecular formula into account, it is hereby proposed that compound j is a phenyllactic acid-2-O-hexoside. The glucoside derivative, phenyllactic acid 2-O-β-D-glucopyranoside, has previously been isolated from Helleborus niger L. leaves [46].

HPLC-DAD Method Validation
Method validation was performed to ensure that the optimized HPLC-DAD method can produce reliable and reproducible quantitative results. The method was deemed specific for the eighteen peaks selected for quantification as their UV-Vis and MS spectra matched those of authentic reference standards, or were in accordance with literature. Linearity was assessed by performing single measurements at several analyte concentrations (µg on-column). Six to eight concentration levels were considered which conform to guidelines specifying a minimum of five levels [51]. Linearity of the calibration curves for authentic standards was excellent, with correlation coefficients larger than 0.999. The y-intercept values were also relatively low ( Table 5). The stability of the phenolic compounds in the standard calibration mixture and unfermented and fermented sample extracts was very good over the considered 24 h period (% RSD <2%; Table S3). The intra-and inter-day precision values were also excellent for most phenolic compounds (Table S3), complying with the precision criteria of % RSD <2% [51]. The stability (% RSD <3%) and precision (% RSD <6%) for maclurin-di-O,C-hexoside (a) were, however, slightly poorer, but were still deemed acceptable for such a complex sample matrix.

Quantification of Phenolic Compounds
The validated HPLC-DAD method was subsequently applied to the analysis of freeze-dried hot water extracts of unfermented and fermented C. genistoides plant materials. The unfermented and fermented plant materials originated from the same individual plant. The content values of the major phenolic compounds, expressed as g per 100 g soluble solids, are summarized in Table 6.  Hesperidin 0.374 0.268 a g maclurin equivalents/100 g soluble solids. b g iriflophenone-3-C-glucoside equivalents/100 g soluble solids. c g hesperidin equivalents/ 100 g soluble solids. d g mangiferin equivalents/100 g soluble solids. e nq = not quantified. f g eriocitrin equivalents/100 g soluble solids. g g narirutin equivalents/100 g soluble solids. h aspalathin equivalents/100 g soluble solids. i g nothofagin equivalents/100 g soluble solids.
The optimized, species-specific HPLC-DAD method was suitable for the quantification of eighteen phenolic compounds, which is a major improvement with regards to other HPLC methods previously employed in the quantitative analysis of C. genistoides extracts [3,11]. To date, quantitative data for the individual monomeric phenolic constituents of C. genistoides extracts have been mostly limited to four of the major compounds, i.e., mangiferin, isomangiferin, hesperidin and iriflophenone-3-Cglucoside [3,5,10,11].
The results in Table 6 show that the major constituents of freeze-dried hot water extracts of C. genistoides are the xanthones, mangiferin and isomangiferin, and the benzophenones, iriflophenone-di-O,C-hexoside and iriflophenone-3-C-glucoside. Collectively, these compounds comprised more than 20% of the aqueous soluble solids of the unfermented plant material, which enhances the nutraceutical potential of this species as a rich source of both xanthones and benzophenones. These content values were, however, markedly reduced with fermentation. Interestingly, the new compound iriflophenone-di-O,C-hexoside represented the second most abundant phenolic constituent in both the unfermented and fermented sample extracts and appeared relatively stable during the high-temperature fermentation process (Table 6).
Other phenolic compounds present in significant amounts in the analyzed C. genistoides sample extracts include vicenin-2, hesperidin and compound w, tentatively identified as a naringenin-Ohexose-O-deoxyhexose and quantified in terms of narirutin equivalents (Table 6).

LC-DAD-ESI-MS and -MS/MSAnalyses
LC-DAD-ESI-MS and -MS/MS analyses were conducted on an Acquity UPLC system equipped with a binary solvent manager, sample manager, column heating compartment and photodiode-array detector coupled to a Synapt G2 Q-TOF system equipped with an electrospray ionization source (Waters). For front end separation, the optimized HPLC method and a premix of methanol and acetonitrile (45:55, v/v) was used. Data were acquired in resolution mode (scanning from 150-1500 amu) and MS/MS scanning mode and processed using MassLynx v.4.1 software (Waters). The instrument was operated in positive and negative ionization modes and calibrated using a sodium formate solution. Leucine enkephalin was used for lockspray (lock mass 556.2771). The MS parameters were as follows: capillary voltage 2.5 kV, sampling cone voltage 15.0 V, source temperature 120 °C, desolvation temperature 275 °C, desolvation gas flow (N 2 ) 650 L/h, cone gas flow (N 2 ) 50 L/h. For MS/MS experiments, the trap collision energy (CE) was set to obtain sufficient fragmentation for selected precursor ions (30 or 45 V). The eluent was split 3:1 prior to introduction into the ionization chamber. The injection volume was 10 µL and UV-Vis spectra were acquired over 220-400 nm at 20 Hz.

HPLC-DAD Method Validation
Method validation was performed in terms of specificity, linearity and range, analyte stability, as well as intra-and inter-day analytical precision. The paired sample extracts of unfermented and fermented C. genistoides and a standard calibration mixture were used.
LC-DAD-ESI-MS data were used to identify compounds suitable for quantification. In both sample extracts, peak purity of the selected compounds was established by comparing their MS spectra with those of the authentic reference standards, where possible. For the additional phenolic compounds, specificity was determined by critically evaluating both the UV-Vis and MS spectra of the peaks in the sample extracts.
Calibration curves were set up for the selected standard compounds to test the linearity of the DAD response. UV-Vis spectra were recorded between 200-700 nm with selective wavelength monitoring at 288 and 320 nm. The dihydrochalcone, aspalathin, and the flavanones, eriocitrin, narirutin and hesperidin, were monitored at 288 nm, while the xanthone, mangiferin, the benzophenones, maclurin, and the flavone, vicenin-2, were monitored at 320 nm. The standard calibration mixture was diluted to obtain six different analyte concentrations which were injected at 10 µL each. The most diluted calibration mixture was also injected at 5 µL and the undiluted calibration mixture at 15 and 20 µL. On-column levels ranged between 0.0060 and 3.6422 µg. Concentration ranges were selected to cover the different quantities of the compounds present in the C. genistoides sample extracts. Linear regression, using the least squares method (Microsoft Excel 2010, Microsoft Corporation, Redmond, WA, USA), was performed to determine the slope, y-intercept and correlation coefficients (r 2 ).
The stability of the phenolic compounds, both as part of the standard calibration mixture and the C. genistoides extracts, was assessed by repeat injections over a 24 h period (n = 6). The percentage relative standard deviation (% RSD) over the time points during this period was used to evaluate the stability of the compounds.
Intra-day precision was determined by consecutive repeat injections (n = 6) of the calibration mixture and each of the extracts on the same day. To determine the inter-day precision, the same procedure was repeated over three consecutive days (n = 3). The % RSD was determined for replicate injections on each day (intra-day precision) and for mean values per day (inter-day precision) by considering the respective peak areas.

Quantification of Phenolic Compounds in Freeze-Dried Aqueous Extracts of C. genistoides
Quantitative analyses were conducted on the Agilent 1200 instrument using the optimized HPLC-DAD method. Sample extracts were prepared for HPLC analyses as described in Section 3.2 and injected in duplicate. Injection volumes of 5 and 15 µL were used for the unfermented C. genistoides extract, whilst the injection volumes were 10 and 25 μL for the fermented sample. The lower injection volumes were required to provide responses for the major constituents within the linear range.
Calibration curves were constructed as described in Section 3.6. Due to limited quantities available of the iriflophenone-3-C-glucoside, isomangiferin and nothofagin standards, these and related compounds were quantified using previously determined response factors with regards to hesperidin, mangiferin and aspalathin, respectively. In the absence of authentic reference standards, additional phenolic compounds were quantified using calibration curves of the most closely related reference standard.

Conclusions
Optimization of a species-specific HPLC-DAD method for the analysis of hot water extracts of unfermented and fermented C. genistoides provided high-resolution chromatographic separation of a large number of phenolic compounds. Characteristic profiles for these two types of extracts were described. Ten compounds were identified by co-elution with the authentic reference standards, while MS data enabled tentative identification of 30 additional compounds. A total of 31 phenolic compounds were identified for the first time in C. genistoides, including 28 identified for the first time in Cyclopia spp. The optimized HPLC-DAD method was successfully validated and applied to the quantitative analysis of the same sample extracts. The major phenolic constituents were the well-known xanthones, mangiferin and isomangiferin, the known benzophenone iriflophenone-3-C-glucoside and an iriflophenone-di-O,C-hexoside (unidentified to date). Future applications of this method will include the quantification of a large number of samples to obtain representative content values and, from a qualitative perspective, authentication of nutraceutical extracts based on their phenolic profiles. Moreover, tentative identification of these phenolic compounds by ESI-MS and tandem MS detection will prove invaluable in subsequent studies on the bio-activity of C. genistoides hot water extracts.

Supplementary Materials
Supplementary materials can be accessed at: http://www.mdpi.com/1420-3049/19/8/11760/s1. Acknowledgments F Joubert (Koksrivier, Pearly Beach, Western Cape, South Africa) is acknowledged for supplying plant material. This work is based on the research supported in part by the South African Department of Science and Technology (DST/CON 0133/2012). Other financial support to E.J. was received through a grant from the Economic Competitive Support Package for Agroprocessing to the ARC by the South African Government. NRF-DST Professional Development Programme Doctoral Scholarship to Theresa Beelders is acknowledged.

Author Contributions
Theresa Beelders performed all experimental work as part of her PhD (Food Science) study, performed data analysis and interpretation, wrote the first draft of the manuscript and prepared all tables and figures. Dalene de Beer contributed to study design, editing of tables and editing of the manuscript. Maria A. Stander contributed to data analysis and interpretation. Elizabeth Joubert, as project leader, conceptualized the study, wrote the Introduction section and contributed to editing of the manuscript. All authors have read and approved the final version of the manuscript.