Metabolites Fingerprinting Variations and Chemotaxonomy of Related South African Hypoxis Species

Abstract

Hypoxis hemerocallidea (Hypoxidaece) is thoroughly researched and well documented for its plethora of anecdotal and scientifically backed pharmacological potentials. Its anecdotal uses and pharmacological activities are attributed to its extract’s inherent bioactive compounds like hypoxoside, rooperol, and β-sitosterol. This study aimed at conducting a targeted and holistic phytochemical profiling of variations in Hypoxis hemerocallidea (H. hemerocallidea) and related species. The chemotaxonomic classifications of H. hemerocallidea and seven other related species were also carried out to avert the possibility of over harvesting H. hemerocallidea and the encouragement of species inter-change. The plant extracts were analysed with reverse phase ultra-pure liquid chromatography quadrupole time-of-flight mass spectrometry and gas chromatography, as well as high-performance thin-layer chromatography. The generated chromatographic data were made compatible for chemometric computation using Principal Component Analysis (PCA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) models. The results obtained unveil orcinol glycoside, curculigoside C, hypoxoside, dehydroxyhypoxoside, bisdehydroxy hypoxoside, hemerocalloside, galpinoside, cholchicoside, geraniol glycoside, β-sitosterol, oleic acid, and 2-hydroxyethyl linoleate as target phytochemicals that define the profiles of the Hypoxis species. In addition, three distinct chemotypes defined by hemerocalloside, galpinoside, and colchicoside, respectively, were observed, as well as holistic variations in all secondary metabolites. Due to similarities in the phytochemical constituents of selected species, species inter-change seems imminent if further research confirms the findings of this study.

1. Introduction

Plants contain complex mixtures of secondary metabolites with diverse chemical properties and variations in their concentrations [1]. Geographic locations, genetic traits, harvesting time, processing methods, and other biotic factors largely influence the chemical variation in plant metabolites [2]. These factors necessitate the classification of plants into different chemotypes based on their phytochemical content using a variety of methods. It can be assumed that plants with different chemistries will also have different efficacies as medicines, since the individual or synergistic actions of molecules are responsible for the biological activities of these plants [3]. The complexity of plant metabolites makes it unfeasible to use a single technique to classify these plants [4]. Quality assurance and standardisation protocols aimed at providing consumers with safe, high-quality products and enhancing profit for the manufacturer are critical. One of the methods of standardising plant extracts is the use of chemical fingerprinting, since this provides a unique pattern representing the presence of a particular chemical composition that defines the plant [5]. Chemical fingerprints can be linked to the efficacy of plant extracts towards a specific biological activity, thereby taking synergism into account [6]. The quantitative analysis of many indigenous Hypoxis species to determine several polar metabolites and β-sitosterol indicated similar phytochemical profiles for H. obtusa and H. hemerocallidea, on the one hand, and for H. galpinii and H. rigidula var. rigidula, on the other. The quantitative data corroborate morphological similarities described for these Hypoxis species [7]. Since many samples were available in this study, multivariate analysis of the chromatographic data was used to provide a holistic view of both intra- and interspecies phytochemical variations within the indigenous members of the genus.

Metabolomics has become a useful and powerful tool for the analysis of plants and related natural products [8]. The Plant Kingdom alone is estimated to produce 100,000–200,000 metabolites [9]. This technique is comprehensive for the simultaneous and systematic evaluation of metabolite levels in the metabolome and their periodic changes [10]. Metabolomics is a valuable tool for comparing variations in metabolites in biological systems and has found application in human, plant, and microbial systems [11]. A metabolome is a complete set of small-molecule metabolites which are either intermediates or the products of metabolism. Primary metabolites, such as carbohydrates, fatty acids, and amino acids, are associated with plant growth and reproduction. However, secondary metabolites including terpenes, alkaloids, polyphenols, and flavonoids are considered to be an expression of the individuality of the plant species [12]. Secondary metabolites serve a variety of functions in plants, including protection from predators, disease pathogens, and the attraction of insect pollinators [13]. Metabolomics has been successfully used for the standardisation of herbal materials and related products. Data from several analytical methods, including Nuclear Magnetic Resonance (NMR) spectroscopy, Liquid Chromatography (LC) and Gas chromatography (GC), are suitable for the metabolomic assessment of phytomedicines and related raw materials [14,15].

The principle underlining metabolomics is either a succinct targeted approach or a random non-targeted or holistic perspective [16]. The targeted approach involves the identification and determination of plant or other natural extracts that belong to a specific class of compounds or have desired properties, using known standard compounds. Such compounds, if present in the extracts, can be used as biomarkers in the standardisation of traditional medicines [14]. On the contrary, the untargeted approach measures the “whole” metabolite regardless of the pharmacological value of the constituents. In the latter approach, the identification of all the metabolites is not mandatory prior to data processing. In both metabolomic approaches, many samples are investigated, and the data generated often contain multiple variables [17]. This type of data is difficult to analyse with classical statistical techniques [18], because, in Multivariate data (MVD, the number of samples analysed are usually smaller than the number of variables generated [19]. To circumvent this challenge with classical statistical analysis, chemometric computation of MVD is almost indispensable. Chemometrics are suitable for summarising, visualising, and interpreting complex data in a robust and statistically appropriate way [17]. The unsupervised principal component analysis (PCA) is applied to identify trends, variables, clusters, and outliers within a dataset [20]. Without the introduction of bias to the data, PCA allows the researcher to project and extract systematic variation in MVD to reveal clustering and allow for the recognition of metabolite variation [17]. This type of modelling does not rely on class information for the discrimination of samples. In contrast, Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) is a supervised technique that is able to account for differences between classes (each species in a genus could be seen as a class). In addition, the marker compounds responsible for these differences are revealed by the model. Once an OPLS-DA model has been constructed, it can be used as a prediction model to allocate unknown samples to a specific class [19].

Although many reports detail the phytochemistry of a particular Hypoxis species [20], only a few comparative studies have been performed [21,22]. Considering the wide range of distribution of the species in South Africa [23], variations in the chemical profiles of specimens harvested from different regions can be expected because of differences in genetic traits, climate, soil, and precipitation. These differences could drastically impact the medicinal efficacy of that specimen for a particular purpose. Boukes et al. [21] determined the levels of β-sitosterol and other sterols and hypoxoside in corms of H. hemerocallidea, H. stellipilis and H. sobolifera var. sobolifera from the eastern and southern Cape regions of South Africa. Hypoxis sobolifera var. sobolifera and H. hemerocallidea were found to contain the most β-sitosterol and hypoxoside, respectively. Since the plants, regrown from the purchased and collected corms, were exposed to the same conditions for six months prior to analysis, and the analysed samples representing each individual species originated from pooled corms, any phytochemical variations between specimens that may have been present were voided. A later study by [24] using extracts from the same three species indicated differences in their cytotoxicity towards various cell lines, suggesting phytochemical differences. The need for fundamental studies that contribute to knowledge of the biosystematics of the genus was mentioned by Van Wyk as far back as 1997 [25]. Singh [26] has already made a valuable contribution towards this knowledge by revising the taxonomy of the southern African species. Phytochemical variation studies are vital to achieve successful and sustainable commercialisation, since initial plant selection, quality control, and proof of efficacy are intrinsically linked to the chemical composition of the medicinal product [27].

The objective of the manuscript is reporting on ten identified metabolites and their variations in eight Hypoxis species, as well as classifying the eight Hypoxis species into three distinct chemotypes.

2. Materials and Methods

2.1. Plants Collection and Comminution

The location, altitude of each site where Hypoxis corms were collected by a licenced collector, and their voucher specimen numbers post-identification are displayed in

Table 1. The location and altitude of each site where Hypoxis corms were collected.

Hypoxis Species	Province	Town	Locality	Voucher Specimen No.	No. of Specimen	Altitude (mASL)
H. argentea	Northern Cape	Kimberley	L3	HA 007-HA 009	3	1555
H. colchicifolia	Eastern Cape	Port Edward	L22	HC 070-HC 072	3	415
H. colchicifolia	KwaZulu-Natal	Midmar	L24	HC 076-HC 078	3	1036
H. galpinii	Mpumalanga	Wakkerstroom	L8	HG 024-HG 026	3	1791
H. galpinii	KwaZulu-Natal	Groenvlei	L9	HG 027-HG 029	3	1914
H. galpinii	KwaZulu-Natal	Wagendrift	L17	HG 054-HG 057	4	1497
H. galpinii	KwaZulu-Natal	Ukahlamba	L18	HG 058-HG 060	3	1662
H. galpinii	KwaZulu-Natal	Impendle	L26	HG 082-HG 084	3	1599
H. galpinii	KwaZulu-Natal	Ukahlamba	L29	HG 091-HG 093	3	1714
H. hemerocallidea	Northern Cape	Boshof	L2	HH 004-HH 006	3	1555
H. hemerocallidea	KwaZulu-Natal	Mount Frere	L14	HH 042-HH 044	3	1100
H. hemerocallidea	KwaZulu-Natal	Kokstad	L20	HH 064-HH 066	3	1340
H. hemerocallidea	Eastern Cape	Port Edward	L23	HH 073-HH 075	3	415
H. hemerocallidea	Gauteng	Walter Sisulu Garden	L31	HH 097-HH099	3	1550
H. hemerocallidea	Gauteng	Tshwane Market	L32	HH 100-HH 102	3	1280
H. hemerocallidea	Gauteng	Marabastad Market	L33	HH 103-HH 105	3	1280
H. multiceps	Mpumalanga	Breyten	L7	HM 021-HM 023	3	1784
H. obtusa	Mpumalanga	Breyten	L6	HO 018-HO 020	3	1785
H. obtusa	KwaZulu-Natal	Chelmsford	L11	HO 033-HO 035	3	1284
H. obtusa	Gauteng	Randfontein	L13	HO 039-HO 041	3	1709
H. obtusa	KwaZulu-Natal	Mount Frere	L15	HO 045-HO 050	6	1102
H. obtusa	KwaZulu-Natal	Kokstad	L19	HO 061-HO 063	3	1451
H. obtusa	KwaZulu-Natal	Ukahlamba	L28	HO 088-HO 090	3	1637
H. rigidula var. polissisima	KwaZulu-Natal	Dundee	L12	HRP 036-HRP 037	3	1284
H. rigidula var. polissisima	KwaZulu-Natal	Wagendrift	L16	HRP 051-HRP 053	3	1345
H. rigidula var. rigidula	Kwazulu-Natal	Boshof	L1	HRP 001-HRP003	3	1556
H. rigidula var. rigidula	Northern Cape	Boshof	L4	HRR 010-HRR 012	3	1562
H. rigidula var. rigidula	Mpumalanga	Breyten	L5	HRR 013-HRR 017	5	1780
H. rigidula var. rigidula	KwaZulu-Natal	Groenvlei	L10	HRR030-RHH 032	3	1915
H. rigidula var. rigidula	Eastern Cape	Port Edward	L21	HRR 67-HRR 069	3	418
H. rigidula var. rigidula	KwaZulu-Natal	Impendle	L25	HRR 79-HRR 081	3	1598
H. rigidula var. rigidula	KwaZulu-Natal	Ukahlamba	L27	HRR 085-HRR 087	3	1455
H. rigidula var. rigidula	Free State	Sterkfontein	L30	HRR 94-HRR 096	3	1694

. Each fresh corm was separated from the Hypoxis plant and the tap roots removed from it. After washing, the corms were chopped into small pieces and oven dried at 30 °C for 36 h prior to extraction. The resulting dried plant material was then pulverised using a Retsch1MM 400 ball milling frequency of 30.0 Hz for 120 s to yield fine brown powders. These powders were sieved using a 500 mm mesh size (Endcotts Filters Ltd., London, UK) to ensure uniform particle size.

2.2. Plant Extraction

Approximately 5 g mass of one specimen of each of the eight species (H. hemerocallidea, H. colchicifolia, H. obtusa, H. rigidula var. rigidula, H. rigidula var. polissisima, H. galpinii, H. argentea, and H. multiceps) was accurately weighed in duplicate into 50.00 mL Erlenmeyer flasks. A 10.0 mL aliquot of Analytical Reagent (AR) grade methanol (MeOH) was added, and the mixture was sonicated at 45 °C for 30 min. After filtering through a filter paper (No 4, Whatman Ltd., Maidstone, UK), the filtrate was kept aside, and the residue was returned to the flask. This process was repeated twice, whereafter the filtrates were combined, and the solvent was evaporated until dryness. The dried extracts were weighed and stored at 25 °C for analysis. Thereafter, the same procedure was followed to prepare chloroform (CHCl₃) extracts of a second set of the weighed powdered plant material. The MeOH extracts were used for the determination of the polar analytes by reverse phase ultra-pure liquid chromatography quadrupole time-of-flight photodiode array detector/mass spectrometry (RP-UPLC-QTOF-PDA/MS) and high-performance thin layer chromatography (HPTLC), while the chloroform extracts were used for the determination of non-polar extracts GC-MS.

2.3. RP-UPLC-QTOF-MS Analysis

The RP-UPLC-QTO-MS method and conditions used in this study have previously been reported [28].

2.4. GC-MS Analysis

An Agilent 7890A gas chromatograph fitted with a mass selective detector (5975C GC/MSD) was used for the analysis of the CHCl₃ extracts of the wild South African Hypoxis. The GC system was equipped with an autosampler (Model 7693) and a split–splitless injector. Samples were introduced via split injection with a split ratio of 1:15, using an injector temperature of 300 °C. Separation was achieved on a DB-5MS-fused silica capillary column (30 m × 0.32 mm id., 0.25 µm film thickness; J & W Scientific, Folsom, CA, USA). The thermal conditions for the oven were 80 °C for 2 min, 10 °C min⁻¹ to 300 °C (held for 14 min), while the helium carrier gas flow was maintained at 1 mL min⁻¹. A transfer temperature of 300 °C, a scan range of m/z 50–600, a sampling rate of 1.1 scans⁻¹, and an ionisation potential of 70 eV were applied. A detector temperature of 300 °C was maintained over the total run time of 30 min.

2.5. HPTLC Analysis

The HTLC analysis reported by K. Bassey et al. [28] was used in this study.

2.6. Chemometric Data Analysis

Chemometric analysis of the data was performed using Simca-P 13.0 software (Umetrics AB, Malmo, Sweden). The H. argentea and H. multiceps samples were excluded from the chemometric analysis due to the too small sample size (n = 6) that would generate zero components, a statistically insignificant parameter. Removing small samples in chemometrics analysis could lead to bias in overfitting a model, imprecise estimates, reduced statistical power, and limited generalizability. Such biases were prevented through the construction of a five-component PCA model and using the first two components for the data analysis.

After alignment of the RP-UHPLC-Q-TOF-MS and GC-MS chromatographic data, PCA models were constructed and used to analyse the data obtained for the individual Hypoxis specimens. Data were pareto scaled, but no data pre-processing was performed. The presence of outliers was investigated, and the effects of excluding these on model statistics were ascertained. The PCA model was constructed to indicate trends (similarities and/or differences) within the dataset. Thereafter, the samples were assigned to one of six separate classes (Classes 1–6) based on the Hypoxis species under investigation. Each separate class was tentatively assumed to contain a unique set of phytochemicals. An OPLS-DA model was constructed and used for analysis of the class-assigned dataset. This model is the most suitable for the discrimination of data that show similarities from an unsupervised PCA plot [29]. This is because samples that are unrelated (orthogonal) to those that are most similar are excluded by the OPLS-DA model. External validation of all the models was performed by randomly removing a subset of the data (test set), constructing a new model with the remaining observations (training set), and subsequently using the new model to predict the class to which each member of the test sets belongs [30]. The accuracy of prediction was then determined.

3. Results

3.1. Chemical Fingerprinting of Wild Hypoxis by HPTLC

The mobile phase used was suitable for the resolution of all the phytocomponents in the Hypoxis extracts investigated. Even compounds with very close Rf-values such as colchicoside and hypoxoside (Rf 0.28 and 0.29) and galpinoside and curculigoside C (Rf 0.48 and 0.49) were resolved by the eluent (Figure 1).

After the analysis of the extracts against the standard compounds (Figure 2) using the same mobile phase, spots that were common to all the extracts were identified as hypoxoside, dehydroxyhypoxoside, bis-dehydroxyhypoxoside, and geraniol glycoside.

The hydrophobic nature of β-sitosterol, the other active metabolite in Hypoxis raw materials, motivated our investigation of the Hypoxis chloroform extracts. Even though this extract is not popular commercially due to the carcinogenicity of the solvent, ref. [21] determined β-sitosterol, stigmasterol, demosterols, and stigmastenolin in CHCl₃ extracts of H. hemerocallidea, H. sobolifera, and H. stellipilis. In the current study, using optimised HPTLC conditions, an attempt was made to overcome the obstacle of co-migration of the sterols, which was encountered when using manual thin layer chromatography (TLC) [21]. Using the semi-automated HPTLC system, the Hypoxis CHCl₃ extracts were applied to glass plates, which were developed in CHCl₃-EtOAc-HCOOH (5-4-1 v/v/v), as reported by Boukes et al. [21]. The Rf value (0.95) for β-sitosterol was found to be higher than that reported by them. Most of the compounds migrated with the solvent front. It was decided that the eluent was not appropriate for good resolution of the CHCl₃ extract of Hypoxis. So, this study applied chloroform as the HPTLC mobile phase and a better result, displayed in Figure 3 using the standard whose structures are detailed in Figure 4, was the outcome.

3.2. Chemical Fingerprinting of Wild Hypoxis by RP-UPLC-QTOF-MS Using a Targeted Approach

Contrary to previous reports that Hypoxis MeOH extracts contain only hypoxoside, dehydroxyhypoxoside, and bis-dehydroxyhypoxoside polar compounds [31], we anticipated that with our RP-UPLC-QTOF-MS, other previously unresolved and undetected secondary metabolites, would be determined from our investigation. Analysis of the MeOH extracts of 105 Hypoxis samples in negative ion mode by RP-UPLC-QTOF-MS yielded well resolved peaks. As was observed by HPTLC, some peaks were common to all the species, while other peaks were species specific. By comparing the pseudo-molecular ion mass to charge (m/z) ratios, mass fragmentation patterns, and UV absorbance and retention times (Rt) of the standard compounds (

Table 2. Data obtained by RP-UPLC-QTOF-MS, reflecting the chemical fingerprints of methanol extracts of Hypoxis corms.

Peaks	Retention Time (min)	Mass to Charge Ratio [M − 1]− (m/z)	UV (λmax) (nm)	Hypoxis Species Detected	Mass Fragmentation Pattern [M − H]−	Fragment Identification	Compound
A	1.64	299	279	H. colchicifolia	299, 163, 137	[M − H]−, [ M − C7H5O3]−, [M − C6H11O5]−	orcinal glycoside
B	2.48	517	210	H. colchicifolia	517, 481, 330, 197, 162	[M − Cl2]−, [M − H]−, [M − C8H9O3]−, [M − C13H17O7]−, [M − C16H15O7]−	curculigoside C [32]
C	3.50	519	255	H. hemerocallidea and H. obtusa	519, 461, 315, 299, 221, 205 163	[M − H]−, [M − C2H3O2]−, [M − C8H13O6]−, [M − C8H13O7]−, [M − C17H15O5] −, [M − C17H15O6]−, [M − C17H15O6 − C2H3O − 2H]−	hemerocalloside
D	3.88	663	278	Predominantly H. colchicifolia	663, 627, 465, 315, 217	[M − H]−, [M − C2H2O2 − 5H]−, [M − C9H13O6 − H2O]−, [M − C9H13O6 − C8H15O2]−, [M − C22H23O10]−.	colchicoside
E	4.27	623	256, (299)	H. galpinii and H. rigidula var. rigidula	623, 461, 325, 299, 163, 160	[M − H]−, (M − C9H7O3]−, [M − C17H15O5]−, [C15H17O8]−, [M − C23H25O10]−, [M − C17H15O8 − C9H7O3]−	galpinoside
F	4.58	605	257	All	605, 443, 281	[M − H]−, [M − C6H11O5]−, [M − 2C6H11O5]−,	hypoxoside [33]
G	4.89	589	258	Most	589, 427, 256, 163	[M − H]−, [M − C6H11O5] −, [M− 2C6H11O5]−, [M − C23H23O8	dehydroxyhypoxoside [31]
H	5.20	609	259	Most	609, 411, 248, 163	[M − Cl]−, [M − C6H11O5]−, [M − 2C6H11O5]−, [M − C23H23O7]−	bis-dehydroxyhypoxoside [31]
I	7.60	447	200	All	483, 447, 295, 163, 155	[M − Cl2], [M − 1]−, [M − C10H19O], [M − C10H19O− C11H19O9–C5H9O4]−, [M − C5H9O4]−	geraniol glycoside [34]

), as well as the literature references to the Hypoxis extracts, nine typical peaks in the Hypoxis MeOH extracts were annotated (Figure 5). These peaks are herein reported as the chemical signature of MeOH extracts of the South African Hypoxis species.

3.3. Chemical Fingerprinting of Wild Hypoxis by GC-MS Using a Targeted Approach

Analysis of CHCl₃ extracts (10 mg mL⁻¹) of the wild Hypoxis by GC-MS yielded a satisfactory resolution and sensitivity for the separation of the different phytocomponents from the Hypoxis corms. By comparing the pseudo-molecular ion mass to charge (m/z) ratios, ion fragmentation patterns, and retention times (Rt) to the literature data [21], the known compounds (campesterol, and β-sitosterol) were identified. (Figure 6)

Table 3. Relative retention times, fragmentation ions, and mass to charge ratios used for the identification of the phytoconstituents in the CHCl₃ extracts of H. hemerocallidea, H. obtusa, H. rigidula var. rigidula, H. rigidula var. polissisima, H. galpinii, H. colchicifolia, and H. multiceps.

Peaks	Retention Time (min)	Mass to Charge Ratio [M]+ (m/z)	Mass Fragmentation Pattern	Compound
A	11.29	224	224, 111, 97, 69, 57	2-hexadecanol
B	13.33	248	248, 242, 225, 185, 143, 129, 102	isopropyl-propyl 12-methyltetradecanoate
C	14.68	288	199, 185, 171, 166, 143, 129, 115	tetradecanoic acid
D	15.32	242	242, 213, 199, 157, 143, 129, 97	pentadecanoic acid
E	16.50	256	239, 227, 213, 199, 185, 171, 157	n-hexadecanoic (palmitic) acid
F	17.11	312	283, 270, 253, 241, 227, 205, 199	propyl-14-methylhexadecanoate
G	17.34	270	241, 227, 213, 171, 143, 129, 115	heptadecanoic acid
H	17.71	282	264, 222, 180, 137, 125, 111, 97	cis-13-octadecanoic oleic acid
I	18.16	264	246, 235, 222, 207, 193, 180, 165	oleic acid
J	18.33	284	284, 255, 241, 227, 199, 185, 171	octadecanoic acid
K	18.72	310	259, 241, 121, 199, 173, 157, 147	cis-13-eicosenoic acid
L	20.21	268	341, 313, 272, 241, 199, 147, 129	hexanedioc acid-bis(2-ethylhexyl) ester
M	21.20	341	380, 351, 323, 295, 267, 239, 211	heptacosane
N	22.64	365	341, 322, 281, 264, 250, 207, 183	oleic acid-3-(octadecyloxy) propyl ester
O	24.05	490	462, 418, 392, 348, 320, 292, 267	17-pentatriacontene
P	24.79	380	337, 309, 281, 253, 225, 197, 169	heptacosane
Q	25.45	400	367, 253, 213, 185, 159, 145, 121	ethylisoallocholate
R	25.65	484	424, 365, 330, 304, 287, 244, 227	7,8-epoxylanostan-11-ol
S	26.65	450	421, 393, 365, 337, 309, 281, 253	dotriacontane
T	26.95	400	382, 367, 340, 315, 289, 273, 255	campesterol
U	27.77	364	337, 295, 281, 251, 225, 197, 183	octadecane-3-ethyl-5-(2-ethylbutyl)
V	28.02	414	396, 354, 329, 303, 255, 231, 213	β-sitosterol
W	28.49	490	462, 418, 391, 348, 320, 292, 267	17-pentatriacontene
X	29.54	410	395, 368, 296, 241, 229, 187, 174	stigmasta-3,5-dien-7-one
Y	29.91	484	424, 365, 330, 304, 287, 244, 227	8-epoxylanostan-11-ol
Z	31.73	490	460, 432, 390, 362, 334, 306, 285	pentatricontanol
AA	32.95	428	413, 400, 328, 285, 269, 227, 213	citrost-7-en-3-ol

Table 3. Relative retention times, fragmentation ions, and mass to charge ratios used for the identification of the phytoconstituents in the CHCl₃ extracts of H. hemerocallidea, H. obtusa, H. rigidula var. rigidula, H. rigidula var. polissisima, H. galpinii, H. colchicifolia, and H. multiceps.

Peaks	Retention Time (min)	Mass to Charge Ratio [M]+ (m/z)	Mass Fragmentation Pattern	Compound
A	11.29	224	224, 111, 97, 69, 57	2-hexadecanol
B	13.33	248	248, 242, 225, 185, 143, 129, 102	isopropyl-propyl 12-methyltetradecanoate
C	14.68	288	199, 185, 171, 166, 143, 129, 115	tetradecanoic acid
D	15.32	242	242, 213, 199, 157, 143, 129, 97	pentadecanoic acid
E	16.50	256	239, 227, 213, 199, 185, 171, 157	n-hexadecanoic (palmitic) acid
F	17.11	312	283, 270, 253, 241, 227, 205, 199	propyl-14-methylhexadecanoate
G	17.34	270	241, 227, 213, 171, 143, 129, 115	heptadecanoic acid
H	17.71	282	264, 222, 180, 137, 125, 111, 97	cis-13-octadecanoic oleic acid
I	18.16	264	246, 235, 222, 207, 193, 180, 165	oleic acid
J	18.33	284	284, 255, 241, 227, 199, 185, 171	octadecanoic acid
K	18.72	310	259, 241, 121, 199, 173, 157, 147	cis-13-eicosenoic acid
L	20.21	268	341, 313, 272, 241, 199, 147, 129	hexanedioc acid-bis(2-ethylhexyl) ester
M	21.20	341	380, 351, 323, 295, 267, 239, 211	heptacosane
N	22.64	365	341, 322, 281, 264, 250, 207, 183	oleic acid-3-(octadecyloxy) propyl ester
O	24.05	490	462, 418, 392, 348, 320, 292, 267	17-pentatriacontene
P	24.79	380	337, 309, 281, 253, 225, 197, 169	heptacosane
Q	25.45	400	367, 253, 213, 185, 159, 145, 121	ethylisoallocholate
R	25.65	484	424, 365, 330, 304, 287, 244, 227	7,8-epoxylanostan-11-ol
S	26.65	450	421, 393, 365, 337, 309, 281, 253	dotriacontane
T	26.95	400	382, 367, 340, 315, 289, 273, 255	campesterol
U	27.77	364	337, 295, 281, 251, 225, 197, 183	octadecane-3-ethyl-5-(2-ethylbutyl)
V	28.02	414	396, 354, 329, 303, 255, 231, 213	β-sitosterol
W	28.49	490	462, 418, 391, 348, 320, 292, 267	17-pentatriacontene
X	29.54	410	395, 368, 296, 241, 229, 187, 174	stigmasta-3,5-dien-7-one
Y	29.91	484	424, 365, 330, 304, 287, 244, 227	8-epoxylanostan-11-ol
Z	31.73	490	460, 432, 390, 362, 334, 306, 285	pentatricontanol
AA	32.95	428	413, 400, 328, 285, 269, 227, 213	citrost-7-en-3-ol

3.4. Investigation of Phytochemical Variation by RP-UPLC-QTOF-MS Using a Targeted Approach

Eight Hypoxis samples were initially studied using UPLC-QTOF-MS to obtain an overview of the qualitative metabolite variations. The UPLC-QTOF-MS profiles of the eight Hypoxis species investigated were used to summarise the variation in the phytoconstituents within the samples. Hypoxoside (4) and geraniol glycoside (7) were present in all the Hypoxis species (

Table 4. An overview of metabolite variations in eight Hypoxis species.

Hypoxis Species	Secondary Metabolites Identified in the MeOH Extract of Hypoxis Samples
	1	2	3	4	5	6	7
H. hemerocallidea	√	X	X	√	X	√	√
H. obtusa	√	X	X	√	√	√	√
H. rigidula var. rigidula	X	X	√	√	√	X	√
H. rigidula var. polissisima	X	X	√	√	√	√	√
H. galpinii	X	√	X	√	√	X	√
H. colchicifolia	X	√	√	√	√	√	√
H. multiceps	X	X	X	√	√	X	√
H. argentea	X	X	√	√	√	√	√

√: presence and X: absence of secondary metabolite. 1: hemerocalloside, 2: colchicoside, 3: galpinoside, 4: hypoxoside, 5: dehydroxyhypoxoside, 6: bis-dehydroxyhypoxoside, and 7: geraniol glycoside.

).

3.5. Qualitative Variation in the Methanol and Dichloromethane Extracts of the Hypoxis Samples

The qualitative variation study of the MeOH extracts of the 105 Hypoxis species encompassing 33 populations in South Africa was conducted using the area under the curve (AUC) of the identified peaks obtained from UPLC-QTOF-MS. The relative amounts of the marker compounds, namely, hypoxoside, dehydroxyhypoxoside, bis-dehydroxyhypoxoside, colchicoside, hemerocalloside, galpinoside, and orcinal glycoside, could be visualised using a column plot (Figure 7).

Similarly, the qualitative variations in β-sitosterol, campesterol, and some of the fatty acids, expressed as area under the curve (AUC), were investigated in the non-polar extracts of Hypoxis samples, and Figure 8 depicts the results.

3.6. Chemotype Identification and Variations as Determined by RP-UPLC-QTOF-MS and Metabolomic Analysis

After converting the RP-UPLC-QTOF-MS data to Excel and exporting to Simca-14^®, chemometric analysis was performed. Two of the species (H. argentea and H. multiceps) were excluded from the dataset since the number of samples (three corms each) was too few to provide reliable statistics. A five-component PCA model of the unsupervised UPLC-QTOF-MS data grouped the samples based on their chemical composition into three clusters (A, B, and C; Figure 9), which are not well defined.

The scores plot of the OPLS-DA model revealed clear differences between some species (Figure 10), although others were clustered together. This plot allowed for the identification of distinct groupings that can be viewed as having distinct chemical profiles and therefore as chemotypes. The OPLS-DA model was able to separate the H. galpinii, H. rigidula var. rigidula (except one), and H. rigidula var. polissisima (Figure 10) samples (Cluster A) from those of H. hemerocallidea and H. obtusa (Cluster B) in the first component. The H. colchicifolia samples were grouped in Cluster C, which was separated from most of the Cluster A samples by the second component. The colchicifolia had a higher intensity or concentration of dehydroxyhypoxoside (peak G) and bis-dehydroxyhypoxoside (peak H) than all the other Hypoxis species. This factor could be responsible for the colchicifolia samples being placed outside the 95% confidence level of the analysis to underscore a possible difference in the species chemistry and morphology compared with all the other species analysed.

H. rigidula var. rigidula and H. rigidula var. polissisima samples from location L25 and L27 and two samples from L16 were grouped into chemotype B and not in chemotype A like the rest. The reasons for the slight variation in the chemotype of these samples were thought to be related to the Global Positioning System. This is because these samples were harvested from Impendle (−29.5996955, 29.8670563), Ukahlamba (−29.3767, 29.5377), and Wagendrift (−23.467, 29.650), all in the mountainous Drakensberg’s Area. This locality is characterised by an identical longitude of twenty-nine degrees and a similar altitude. These factors can influence the chemistry of these samples away from the other H. rigidula var. rigidula and H. rigidula var. polissisima samples that were placed in chemotype A.

Further confirmation of the three defined chemovars was performed via a hierarchical analysis (HCA), which was performed with no predefined classes of the dataset. The HCA of the non-predefined data results agreed with the predefined OPLS-DA analysis, with all the observations grouped into the three distinctly defined chemotypes A, B, and C (Figure 11), with each cluster consisting of species with similar (A, B) or unique (C) phytochemicals.

Considering the premise that clustering was based on similarities or differences in the structures of the compounds in the different Hypoxis species per chemotype, an OPLS-DA loading scores plot (Figure 12) was constructed using the same data. This plot was able to single out the compound that has the largest influence on the separation of clusters (chemotypes). This compound or chemical marker compounds could therefore be used to define each chemotype. From the loadings plot, the three clusters were annotated with a mass number 623 (for Cluster A), 519 (for Cluster B), and 663 (for Cluster C) by the model as compounds responsible for the separation of each chemotype. The assigned mass numbers matched the molecular weight of galpinoside (m/z 623), hemerocalloside (m/z 519), and colchicoside (m/z 663) at Rt of 4.26 min, 3.49 min, and 3.91 min, respectively.

A thorough scrutiny of each chemotype revealed that chemotype A predominantly consists of H. galpinii, H. rigidula var. rigidula, and H. rigidula var. polissisima; chemotype B encompasses H. hemerocallidea and H. obtusa, while chemotype C encompasses only H. colchicifolia. Comparing the RP-UPLC-QTOF-MS fingerprints of Hypoxis species within each chemotype (

Table 5. Classification of South African wild Hypoxis into three chemotypes from OPLS-DA analysis. Interchangeably used species (H. hemerocallidea and H. obtusa) were placed in the same chemotype.

Hypoxis Species	Province	Town	Locality Id	Voucher Specimen No	Chemotype
H. hemerocallidea	Gauteng	Westcliff	L23	HH 073 HH 074 HH 075	B B B
	Gauteng	Walter Sisulu	L31	HH 097 HH 098 HH 099	B B B
	Gauteng	Tshwane market	L32	HH 100 HH 101 HH 102	B B B
	Gauteng	Marabastad Market	L33	HH 103 HH 104 HH 105	B B B
	Eastern Cape	Mount Frere	L14	HH 042 HH 043 HH 044	B B B
	KwaZulu-Natal	Kokstad	L20	HH 064 HH 065 HH 066	B B B
	Northern Cape	Boshof	L2	HH 004 HH 005 HH 006	B B B
H. obtusa	Gauteng	Randfontein	L13	HO 039 HO 040 HO 041	B B B
	KwaZulu-Natal	Chelmsford	L11	HO 033 HO 034 HO 035	B B B
	Eastern Cape	Mount Frere	L15	HO 045 HO 046 HO 047 HO 048 HO 049 HO 050	B B B B B B
	KwaZulu-Natal	Kokstad	L19	HO 061 HO 062 HO 063	B B B
	KwaZulu-Natal	Ukahlamba	L28	HO 088 HO 089 HO 090	B B B
	Mpumalanga	Breyten	L6	HO 018 HO 019 HO 020	B B B
H. rigidula var. rigidula	Eastern Cape	Port Edward	L21	HRR 067 HRR 068 HRR 069	A A A
	Free State	Sterkfontein	L30	HRR 094 HRR 095 HRR 096	A A A
	KwaZulu-Natal	Groenvlei	L10	HRR 030 HRR 031 HRR 032	A A A
	KwaZulu-Natal	Impendle	L25	HRR 079 HRR 080 HRR 081	B B B
	KwaZulu-Natal	Ukahlamba	L27	HRR 85 HRR 86 HRR 87	B B B
	Mpumalanga	Breyten	L5	HRR 013 HRR 014 HRR 015 HRR016 HRR 017	A A A A A
	Northern Cape	Boshof	L4	HRR 010 HRR011 HRR 012	A A A
H. rigidula var. polissisima	KwaZulu-Natal	Boshof	L1	HRP 001 HRP 002 HRP 003	A A A
	KwaZulu-Natal	Dundee	L12	HRP 036 HRP 037 HRP 038	A A A
	KwaZulu-Natal	Wagendrift	L16	HRP 051 HRP 052 HRP 053	B A B
H. galpinii	KwaZulu-Natal	Groenvlei	L9	HG 027 HG 028 HG 029	A A A
	KwaZulu-Natal	Wagendrift	L17	HG 054 HG 055 HG 056 HG 057	A A A A
	KwaZulu-Natal	Ukahlamba	L18	HG 058 HG 059 HG 060	A A A
	KwaZulu-Natal	Impendle	L26	HG 082 HG 083 HG 084	A A A
	KwaZulu-Natal	Ukahlamba	L29	HG 091 HG 092 HG 093	A A A
	Mpumalanga	Wakkerstroom	L8	HG 024 HG 025 HG 026	A A A
H. colchicifolia	Eastern Cape	Port Edward	L22	HC 070 HC 071 HC 072	C C C
	Eastern Cape	Port Elizabeth	L24	HC 076 HC 077 HC 078	C C C

) clearly indicated close similarities between the chemical profiles. The chemical similarities correspond to morphological similarities between the species [6] within each of the three chemotypes. Table 5 shows the classification of the 105 Hypoxis samples into the three chemotypes.

4. Discussions

The HPTLC profiles of the methanol extract could be grouped into H. hemerocallidea-type, H. galpinii-type, and H. colchicifolia-type, with common bands of 0.29, 0.35, 0.39, and 0.74 at Rf values and species-specific uncommon bands. Bands at a Rf below 0.2 and above 0.85 could not be identified due to a lack of standards. The chloroform extract, on the other hand, mainly indicated the presence of β-sitosterol [35], with trace amounts of campesterol and stigmasterol [21], oleic acid, and 2-hydroxyethyl linoleate.

Chemical profiling of the Hypoxis methanol extract using RP-UPLC-ESI-Q-TOF-MS, due to its high sensitivity, throughput capacity, and high resolution [36], revealed the common peaks as hypoxoside, dehydroxyhypoxoside, bis-dehydroxyhypoxoside, and geraniol glycoside with retention times of 4.59, 4.90, 5.21, and 7.60 min, respectively. On the other hand, the species-specific peaks were identified as orcinal glycoside, galpinoside, hemerocalloside, colchicoside, and curculigoside C with corresponding retention times of 1.64, 3.47, 3.49, 3.91, and 4.96 min. The unidentified peaks by HPTLC resolved at 2.60 (m/z 621), 4.03 (m/z 643), 5.67 (m/z 611), and 6.25 min (m/z 773). Using the mass fragmentation patterns of these peaks, a thorough search of LC-MS databases, including the National Institute of Health (NIH), Pubchem, Scifinder, and Chemspider, provided no matching information for possible compounds. All the unidentified peaks had a large molecular weight of m/z = 621, 611, 773, and 643. However, three of the peaks had very minor intensities and one (m/z 643) had a moderate one. None of the peaks were nominated by the PCA and OPLS-DA chemometric modals as the marker compound responsible for the three Hypoxis clusters. Hence, none of the four compounds could influenced the holistic classification of the Hypoxis chemotype.

From the GC-MS analysis, twenty-four volatile compounds, mostly C-18 fatty acids (saturated and unsaturated) and their derivatives, were identified by using the NIST 8 spectral library. Several of these additional 24 metabolites were not identified in previous studies on Hypoxis plants.

Palmitic acid and oleic acid were dominant over campesterol and β-sitosterol. Sterolins were not detected in either the MeOH or CHCl₃ extracts of South African Hypoxis, which is contrary to previous reports that claimed that 100 g of the enriched aqueous extract of Hypoxis contained 9 mg [37] and less than 6 mg [38] of sterolins.

Hypoxoside is the principle active component in the Hypoxis genus. Therefore, its availability in all the species investigated is of immense clinical importance as the presence or absence of the compound is likely to affect the efficacy of the products formulated from Hypoxis. Due to the clinical importance of hypoxoside, H. hemerocallidea was identified as containing the highest levels of hypoxoside (60–65%). In contrast, H. colchicifolia contained 3% of hypoxoside and the highest 70% of colchicoside.

The other recorded clinically important β-Sitosterol was detected in all of the samples, with the exception of the H. hemerocallidea sample from the Gauteng province, H. rigidula var. rigidula from the KwaZulu-Natal Province, and H. obtusa from the Mpumalanga Province. It was of significant interest that H. hemerocallidea also afforded the largest amount of β-sitosterol (53% of the total), while that of H. rigidula var. polissisima was found to contain only 8.0% of this compound. The relative amounts of campesterol, stigmasterol, heptadecanoic acid, hexadecanoic acid, oleic acid, and other fatty acids esters varied between 16 and 31%.

Principal component analysis models and OPLS-DA [39] were used to reveal clustering of all the Hypoxis samples arising from their chemical similarities or differences thereof without any interference [18]. Such clustering of the Hypoxis species was expected, because morphological analysis revealed such patterns with the Hypoxis samples analysed [7]. Whereas H. galpinii, H rigidula var. rigidula, and H. rigidula var. polissisima formed Cluster A, H. hemerocallidea and H. obtusa formed Cluster B, and H. colchicifolia constituted Cluster C. The clustering was not based on geographic origin, because each cluster contains samples from all the geographic localities. Although bias can easily be introduced into OPLS-DA models because of class assignment, the clustering evident on the PCA plot indicated the presence of clear chemical composition differences between the samples, thereby ruling out bias.

Hence, the South African Hypoxis species were classified into three chemotypes, namely, chemotype, chemotype B, and chemotype C. The assigned m/z ratio at the specified retention time of the marker compounds was interpolated to the RP-UPLC-QTOF-MS chromatograms of the Hypoxis samples to further confirm the information of the marker compounds that defined each Hypoxis chemotype. Hence chemotypes A, B, and C were defined by galpinoside, hemerocalloside and colchicoside, respectively. This study, like similar studies, has underpinned the importance of using metabolomics in the identification of plant secondary metabolites or markers that define plant chemotypes or mitigate oxidative stresses, cancer, and diabetes [40,41,42].

5. Conclusions

The following conclusions could be made after completing this study: the six Hypoxis species investigated can be grouped into three chemotypes. Chemotype A, consisting of H. galpinii, H. rigidula var. rigidula, and H. rigidula var. polissisima; chemotype B, comprising H. hemerocallidea and H. obtuse; and chemotype C, made up of H. colchicifolia. Chemotype A is characterised by galpinoside as a marker compound and chemotype B by hemerocalloside, while colchicoside serves as the biomarker for chemotype C. All studies on the biological activities of Hypoxis raw material and products have hitherto associated the activity with the presence of rooperol, the aglycone formed from hypoxoside in the gastrointestinal tract, and β-sterol or sterolins. However, it is necessary that the chemotype marker compounds (galpinoside, hemerocalloside, and colchicoside) be fully explored for their biological activities. The presence of a variety of bioactive compounds could explain the plethora of uses of Hypoxis for traditional medicine and as a health supplement. For the purpose of the standardisation of Hypoxis raw material, this study recommends that H. obtusa and H. hemerocallidea, classified as chemotype B, be cultivated. This will ensure the presence of hypoxoside and other metabolites in optimum yields. The chemical similarities between H. hemerocallidea and H. obtusa (chemotype B) and between H. galpinii and H. rigidula var. rigidula (chemotype A) mirror the morphological similarities and justifies their potential interchangeable use of H. hemerocallidea and H. obtusa.