Phytochemistry 106 (2014) 104–115

Contents lists available at ScienceDirect

Phytochemistry

journal homepage: www.elsevier .com/locate /phytochem

What the devil is in your phytomedicine? Exploring species substitutionin Harpagophytum through chemometric modeling of 1H-NMR andUHPLC-MS datasets

http://dx.doi.org/10.1016/j.phytochem.2014.06.0120031-9422/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Department of Pharmaceutical Sciences, TshwaneUniversity of Technology, Private Bag X680, Pretoria 0001, South Africa. Tel.: +27 12382 6373; fax: +27 12 382 6243.

E-mail address: viljoenam@tut.ac.za (A.M. Viljoen).

Nontobeko P. Mncwangi a, Alvaro M. Viljoen a,b,⇑, Jianping Zhao c, Ilze Vermaak a, Wei Chen a,Ikhlas Khan c,d

a Department of Pharmaceutical Sciences, Tshwane University of Technology, Private Bag X680, Pretoria 0001, South Africab Department of Natural Products and Alternative Medicine, Faculty of Pharmacy, King Abdulaziz University, Jeddah 21589, Saudi Arabiac National Center for Natural Products Research, Research Institute of Pharmaceutical Sciences, School of Pharmacy, The University of Mississippi, MS 38677, USAd Department of Pharmacognosy, School of Pharmacy, The University of Mississippi, MS 38677, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 29 March 2014Received in revised form 24 May 2014Available online 17 July 2014

Keywords:Harpagophytum procumbensHarpagophytum zeyheriPedaliaceaeDevil’s ClawChemometricsNuclear magnetic resonanceUltra-high performance liquidchromatography-mass spectrometryIridoid glycosidesHarpagoside

Harpagophytum procumbens (Pedaliaceae) and its close taxonomical ally Harpagophytum zeyheri, indige-nous to southern Africa, are being harvested for exportation to Europe where phytomedicines are devel-oped to treat inflammation-related disorders. The phytochemical variation within and between naturalpopulations of H. procumbens (n = 241) and H. zeyheri (n = 107) was explored using proton nuclear mag-netic resonance (1H-NMR) and ultra-high performance liquid chromatography coupled to mass spec-trometry (UHPLC-MS) in combination with multivariate data analysis methods. The UHPLC-MS resultsrevealed significant variation in the harpagoside content: H. procumbens (0.17–4.37%); H. zeyheri(0.00–3.07%). Only 41% of the H. procumbens samples and 17% of the H. zeyheri samples met the pharma-copoeial specification of P1.2%. Both principal component analysis (PCA) and orthogonal projections tolatent structures discriminant analysis (OPLS-DA) indicated separation based on species (UHPLC-MS dataOPLS-DA model statistics: R2X = 0.258, R2Y (cum) = 0.957 and Q2(cum) = 0.934; 1H-NMR data OPLS-DAmodel statistics: R2X = 0.830, R2Y = 0.865 (cum) and Q2(cum) = 0.829). It was concluded that two speciesare not chemically equivalent and should not be used interchangeably.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The manufacturing of any phytomedicine starts with the rawmaterial supply chain and a thorough understanding of thechemical composition and variation of the plant material is funda-mentally important to ensure a standardised, efficacious and safeconsumer product. This issue is of particular significance in thecase of Devil’s Claw where two taxonomic allies are being usedinterchangeably and one cannot assume that a taxonomic similar-ity relates to chemical homology and pharmacological equivalence.The taxonomic assessment of Harpagophytum by Ihlenfeldt andHartmann (1970) include two species and five subspecies:Harpagophytum procumbens (Burch.) DC. ex Meisn subsp. procum-bens, H. procumbens subsp. transvaalensis Ihlenf. and H. Hartm.,

Harpagophytum zeyheri Decne. subsp. zeyheri, H. zeyheri subsp.schiffii Ihlenf. and H. Hartm. and H. zeyheri subsp. sublobatum(Engl.) Ihlenf. and H. Hartm (Ihlenfeldt and Hartmann, 1970).H. procumbens and H. zeyheri are collectively known as Devil’s Claw,referring to the hook-like protrusions on the fruit (van Wyk, 2008).The British and European Pharmacopoeias make provision for theuse of either species as long as the harpagoside standard is met;however, H. procumbens is commercially preferred because it hasbeen reported to contain higher levels of the biologically activeconstituents which include harpagoside, harpagide, procumbideand acteosides (Kemper, 1999; Stewart and Cole, 2005).

H. procumbens (Pedaliaceae) is a valuable African medicinalplant growing in southern Africa (Fig. 1). The San and Khoi peopleof Southern Africa were amongst the first people to use the plantand its use was subsequently adopted by Bantu-speakers (Cole,2003). According to ethnobotanical information, secondary roottubers are used to treat rheumatism and arthritis, back-pain,gastroenterological disturbances and it is also used as a tonic(van Wyk, 2008). Topically, an ointment prepared with an animal

Fig. 1. Geographical distribution of Harpagophytum species and subspecies; (1) H. procumbens subsp. procumbens; (2) H. procumbens subsp. transvaalense; (3) H. zeyheri subsp.zeyheri; (4) H. zeyheri subsp. schiffii; (5) H. zeyheri subsp. sublobatum (adapted from Stewart and Cole, 2005).

N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115 105

fat base or petroleum is applied to treat sores, ulcers and boils (vanWyk and Gericke, 2000). Other ethnobotanical uses of H. procum-bens include the treatment of fever, diabetes, diarrhoea and blooddisease (Stewart and Cole, 2005). Devil’s Claw is of commercialimportance in Namibia and contributes to its annual income. Theraw material is exported to Europe where further product develop-ment and value adding is done. Devil’s Claw remains a widely usedand sought after phytomedicine in Europe, especially Germany(Stewart and Cole, 2005). Most of the raw material exported isobtained from the wild, and due to the unsustainable method ofharvesting there has been a notable decline in natural populationsof H. procumbens. Therefore, traders are opting for H. zeyheri due tolocal availability and ecological pressure on natural populations ofH. procumbens (Stewart and Cole, 2005). In addition, the uninten-tional substitution of H. procumbens with H. zeyheri also occursdue to morphological similarity and the absence of vegetative partsduring the preferred harvesting season. Several methods such ashigh performance thin layer chromatography (HPTLC) and highperformance liquid chromatography (HPLC) have been employedin the quality assessment of Devil’s Claw (Gunther and Schmidt,2005; Schmidt, 2005). However these methods are targeted andusually focus on a single compound, for example, the biomarker,harpagoside. A more holistic quality control approach is metabolo-mics which is considered to be the global and unbiased survey ofthe complement of small molecules (<1 kDa) in an organism(Beyoglu and Idle, 2013). The lack of knowledge about the chemicalcomposition and variation of phytomedicines limits the explora-tion of their full potential and true quality control and accountabil-ity can only be attained if standardisation exists.

Metabolomics may be performed through mass spectrometry(MS) and nuclear magnetic resonance (NMR), usually applied asparallel technologies. These techniques provide an overview ofthe metabolome as well as high-power structure elucidation(Moco et al., 2007). Plant secondary metabolites are chemicallydiverse, occurring with high variability even within the same spe-

cies. Mass spectrometry is favoured due to different instrumenta-tion characteristics such as ionisation, resolution, scan speed andfragmentation. These parameters make MS a versatile techniquewhich is applicable to many disciplines including agriculture, biol-ogy and medicine (Steinmann and Ganzera, 2011). LC–MS togetherwith multivariate data analysis methods has been shown to be apowerful tool in unraveling subtle chemistries and the classifica-tion of complex and evolving plant species (Kim et al., 2011). Unbi-ased detection is the major advantage of using proton nuclearmagnetic resonance analysis (1H-NMR) for metabolomics applica-tions. It is the only technique which produces signals directly cor-relating with the amount of analytes in the sample (Steinmann andGanzera, 2011). In 1H-NMR, the signal intensity is only dependenton the molar concentration in the solution, thus enabling the directcomparison of concentrations of all compounds present in thesample. NMR-based metabolomics have been shown to be a pow-erful tool in discriminating intricate species such as Ilex species,which could be separated into major groups using both supervisedand unsupervised methods (Kim et al., 2010). Georgiev et al. (2011)differentiated and classified Verbascum species which contain irid-oid glycosides and phenylethanoid iridoids using NMR thus show-ing that the type of iridoids present in a species can also be adifferentiating factor for the species. Wolfender et al. (2013)recently reviewed the process of identifying biomarkers fromholistic data, i.e. plant metabolomics. The holistic and comprehen-sive analysis of the plant metabolome following a systems biologyapproach is based on the acquisition of mass spectrometric (MS) ornuclear magnetic resonance (NMR) data. These ‘omics’ techniquesare most useful to identify and quantify the numerous metabolitesthrough the application of data mining that includes various untar-geted and targeted methods (Wolfender et al., 2013). These com-plementary methods have been successfully used by variousresearchers to gain a better understanding of comprehensive phy-tochemical variation and to identify discriminatory biomarkers.Safer et al. (2011) and Porzel et al. (2013) used such methods to

106 N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115

compare metabolomic fingerprints and to conclude taxonomicalrelationships between Leontopodium species and Hypericum spe-cies, respectively, and it was also used to determine secondarymetabolites in green tea (Napolitano et al., 2014). Recently, Martiet al. (2014) applied these techniques to cold-pressed lemon oilsfor which the characterisation is of great importance in the Flavourand Fragrance industry. Multivariate analysis revealed that separa-tion occurred according to geographical location as the two clus-ters of samples showed significant chemical variation betweensamples from Italy and Argentina. In addition, discriminatingmarkers (furocoumarins, flavonoids, terpenoids, fatty acids) wereidentified using the S-plot constructed after orthogonal projectionsto latent structures discriminatory analysis (OPLS-DA) (Marti et al.,2014).

In the present study, ultra-high performance liquid chromatog-raphy coupled to mass spectrometry (UHPLC-MS) and protonnuclear magnetic resonance (1H-NMR) in combination with multi-variate data analysis were used to explore the phytochemical var-iation and identify biomarkers in Harpagophytum species. Anuntargeted approach was adopted where all metabolites were ana-lysed simultaneously.

2. Results and discussion

2.1. UHPLC-MS analysis

Previous phytochemical investigations have led to the identifi-cation, isolation and characterisation of constituents from H. proc-umbens. These constituents include iridoid glycosides such asharpagoside, procumbide and harpagide; phenylpropanoid glyco-sides such as acteoside and isoacteoside; and other substancesincluding harpagoquinones, amino acids, flavonoids, phytosterolsand carbohydrates (Gruenwald, 2002). The biological activity ofDevil’s Claw has been ascribed to iridoid glycosides which are alsofound in other plant species and have well documented anti-inflammatory activity (Viljoen et al., 2012). Fig. 2A and B showstypical UHPLC-MS chromatograms of H. zeyheri and H. procumbens,

Fig. 2. A typical UHPLC-MS ES� chromatogram

respectively. Harpagide, decaffeoylverbascoside, verbascoside, p-coumaroyl-procumbide, isoverbascoside, 8-O-p-coumaroyl-harpa-gide, acetylacteoside and harpagoside were identified from theH. procumbens extract. Only harpagide, decaffeoylverbascoside,verbascoside, isoverbascoside and acetylacteoside were presentin detectable levels in the H. zeyheri extract. Literature availableon the phytochemistry of the species is based on H. procumbens,thus MS/MS was done to identify compounds present in themethanolic extract. The fragmentation patterns are listed in Table 2and the chemical structures are depicted in Fig. 3.

Although the crude extracts contain other iridoid glycosides,harpagoside is used as the biomarker for quality assessment pur-poses. The results confirmed that both species contain harpagosideand that it is present in higher amounts in H. procumbens with arange of 0.17–4.37%, whereas in H. zeyheri the range was 0.00–3.07%. A previous study analysing the harpagoside content of com-mercial products of Devil’s Claw showed that the harpagoside con-tent ranged from 0.50% to 3.00% (Fiebich et al., 2001). Thepercentage of harpagoside in secondary tubers reportedly variesseasonally by plant age, among plants in a given area, and evenamong tubers of the same plant (Kemper, 1999; von Willert andSanders, 2004). According to the standard set by the Europeanand British Pharmacopoeias, the harpagoside content is requiredto be at least 1.2% (dried drug). All of the H. procumbens samplestested contained harpagoside but only 41% met the P1.2% specifi-cation. Only 25% of the H. zeyheri samples contained detectableharpagoside and only 17% met the specification. The 17% of H.zeyheri samples which contained harpagoside and met the specifi-cation were from a single locality (Makgabeng, North West Prov-ince, South Africa). This evidence confirms that Harpagophytum isa large and variable taxon, and that the presence or absence of bio-logically active constituents is influenced by many factors includ-ing geographical location. The plant’s major chemicalconstituents are contained within the secondary tubers while ithas been determined that the flowers, stems, and ripe fruits donot contain harpagoside. Traces of harpagoside together with someunidentified iridoid compounds are present in the leaves. The sec-ondary root tubers contain approximately twice as much harpago-

of (A) H. zeyheri and (B) H. procumbens.

O

O

OH

OH

OO

OOH

OH OH

OHO

OH

O

O

OH OOH

OHO

O

O

OH

OHOH

O

O

OH

OH

OH

OH

Acetylacteoside Isoacteoside (syn. isoverbascoside, isokusaginin)

OO

O O

OHOH

OH

OH

O

OH

OH

OH

OH

OH

OHOH

O

O O OH

OHOHOH

Coumaroylharpagide Chebuloside II

O

O

OH

OH

OO

OOH

OH OH

OO

OH

O

O

O

OH

Diacetylacteoside

OHO

O

OOH

OH OH

OHO

OH

OH

OH

Decaffeoylverbascoside

O

OH

O O

OHOH

OH

OH

OH

OH

O

OH

OO O

OHOH

OH

OH

O

OH

Harpagide Harpagoside

O

O

OH

OH

OO

OOH

OH OH

OHO

OH

OH

OH

Verbascoside

Fig. 3. Chemical structures of compounds identified by LC–MS/MS from a methanolic extract of H. procumbens.

N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115 107

side compared to primary tubers (Czygan and Kruger, 1977). Inpractice, only the secondary root tubers are harvested whilst theprimary tubers remain intact to ensure the sustainability of thespecies (Stewart and Cole, 2005), however, this is not alwaysadhered to leading to harvesting of all the tubers thus resultingin unsustainability of the species.

Multivariate analysis techniques were applied to further unra-vel chemical differences between the two Devil’s Claw species.Principle component analysis (unsupervised), which provides valu-able information regarding similarities and differences in samples,revealed clear cluster formation according to species, althoughpartial overlapping was noted (Fig. 4A). Introgression within thespecies is reported by other researchers, suggesting that biosyste-matic studies are necessary to elucidate the circumscription of thespecies (van Wyk, 2008). The PCA model with four principal com-ponents had an R2X(cum) value of 0.494 and Q2(cum) of 0.183.

In order to identify putative biomarkers orthogonal projectionsto latent structures discriminant analysis (OPLS-DA) was

performed. The single PLS-factor model constructed based on thePCA model had an R2X value of 0.258, R2Y (cum) of 0.957 andQ2(cum) of 0.934 with the species separating into two majorgroups (Fig. 4B). An S-plot was constructed to determine the differ-entiating features (Fig. 4C) at the extreme ends of the S-plot andthe selected ion intensity trend plot (Fig. 5) shows the variablesresponsible for the separation into the groups observed. In group1 (H. zeyheri), the compounds with the retention time-mass/chargeratio (Rt_m/z) pairs of 5.43_517.1689 and 1.10_705.1861 was dom-inant. Group 2 (H. procumbens) was characterised by compoundswith the following retention time-mass/charge ratio pairs:5.43_517.1689; 1.10_705.1861 and 5.42_1011.3562. Compara-tively, both species contain the retention time-mass/charge ratiopairs of 5.43_517.1689 and 1.10_705.186. Higher ion-intensitiesof 5.43_517.1689 were observed in group 2 (H. procumbens)whereas higher ion-intensities of 1.10_705.1861 were observedin group 1 (H. zeyheri). The compound with the retention time-mass/charge ratio pair 5.42_1011.3562 is completely absent in

Fig. 4. (A) Principle component analysis score plot showing two clusters of H. procumbens (Blue) and H. zeyheri (Red). (B) OPLS-DA score plot based on UHPLC-MS data showingseparation of the two species into Group 1 – H. zeyheri (Red) and Group 2 – H. procumbens (Blue). (C) S-plot showing retention time-mass/charge ratio (Rt_m/z) pairs of5.43_517.1689 and 1.10_705.1861 (Group 1, H. zeyheri) and Group 2 (H. procumbens) characterised by compounds with the following retention time-mass/charge ratio pairs:5.43_517.1689; 1.10_705.1861 and 5.42_1011.3562. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

108 N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115

Fig. 5. Selected ion intensity trend plots indicating the retention time-mass/charge ratio (Rt_m/z) pairs responsible for cluster separation where group 1 represents H. zeyheriwith and group 2 represents H. procumbens.

N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115 109

H. zeyheri and therefore indicative of H. procumbens. Harpagoside(5.43_517.1689) contributes to the separation of the species basedon concentration. Quantification revealed higher concentrations inH. procumbens and a minimal presence in H. zeyheri. By applyingdiscriminatory analysis putative biomarkers could be identifiedand/or extrapolated.

Although the biosynthetic pathway of iridoid glycosides ispoorly understood (Georgiev et al., 2013), there is consensus thatiridoids are composed of a six-membered ring containing oxygenwhich is bound to a cyclopentane ring. An iridotrial is formedthrough the cyclisation of 10-oxogeranial biosynthesised fromgeraniol, to 10-hydroxygeraniol. Further oxidation, methylationand glycosylation converts iridotrial to iridoid compounds. Phenyl-ethanoid glycosides (acteoside and isoacteoside) on the otherhand, belong to a large class of plant phenols produced throughthe shikimic acid pathway. The synthesis of phenylethanoid glyco-sides involves the initial step which is the deamination of phenyl-alanine to cinnamic acid catalysed by phenylalanine ammonialyase (Korkina, 2007). The differences in the major biosyntheticpathways may be responsible for the accumulation of iridoidglycosides and phenylethanoid glycosides in H. procumbens andH. zeyheri, respectively.

2.2. 1H-NMR analysis

Nuclear magnetic resonance-based metabolomics is robust andrelatively easy to use, thus ensuring that herbal medicine can beauthenticated and the quality controlled in an efficient way(Wolfender et al., 2013). Methanol was used in the current studyto extract the constituents which could be identified by bothUHPLC-MS and 1H-NMR. Fig. 6 shows typical 1H-NMR spectra ofH. procumbens and H. zeyheri methanolic extracts. Phytochemicalstudies of the species have led to the isolation and identificationof several compounds which include iridoid glycosides and phenyl-propanoid glycosides (Mncwangi et al., 2012). Their proton NMRspectra illustrated the characteristic signals for iridoids in the d6–8 ppm region. For instance, harpagoside and harpagide showedthe characteristic signals at d 6.41 (d, J = 6.37 Hz) and d 6.38 (d,J = 6.20 Hz), respectively, ascribed to the olefinic proton on the irid-oid skeleton. Most of H. procumbens extracts contained iridoid gly-cosides as depicted by chemical shifts in the aromatic region, d 6–8 ppm (enlarged region, Fig 6A), where quantitative differences

were observed. H. zeyheri extracts were largely devoid of iridoidglycosides but they were observed in some extracts (enlargedregion, Fig 6B). All samples contained sugars as denoted by chem-ical shifts d 3.5–5.5 ppm. Principal component analysis (PCA) wasperformed for the dataset derived from the spectra of samples ofthe two species with a good statistical outcome: R2X(cum) = 0.998 and Q2 (cum) = 0.992 (Fig. 7A). In this model, someof the H. zeyheri samples were identified as strong outliers, fallingoutside the 95% confidence interval, suggesting that these arechemically different. It was determined that these outliers werecollected from a single location, Makgabeng (North West Province,South Africa). Quantitative and qualitative differences in Harpago-phytum species with regards to harpagoside content have beenreported, where harpagoside is reportedly more abundant in H.procumbens than in H. zeyheri (Georgiev et al., 2013; Stewart andCole, 2005); quantification of harpagoside in the current study pro-duced congruent findings. Compositional differences among spe-cies were further explored using supervised methods. Thearomatics (d 6–8 ppm) cause the major differences seen at specieslevel, but the type of sugars (d 3–5.5 ppm), and fatty acid esters (d0.5–2 ppm) present or absent also contribute to the intraspeciesvariation as deduced from the PCA loading analysis. A study byCzygan and Kruger (1977) showed that harpagoside can be com-pletely absent from some populations of H. zeyheri despite the factthat this species is regarded as an acceptable alternative to H. proc-umbens (Czygan and Kruger, 1977; von Willert and Sanders, 2004).When a supervised method, OPLS-DA, was used, a model with anR2X of 0.830, R2Y of 0.865 and a Q2(cum) of 0.829 was constructed.Two distinct clusters (H. procumbens – blue and H. zeyheri – red)were observed as shown in Fig. 7B. The 1H-NMR signals of com-pounds responsible for the separation of the two species weredemonstrated through the construction of an S-plot (Fig. 7C). Theextreme points on the S-plot represent the signals which makethe most significant contributions for the separation between thetwo species: the positive chemical shifts d 1.53; 7.59; 7.39 ppmare associated with H. procumbens and the negative loadings d3.47; 3.13; 2.39; 5.67 ppm correspond to H. zeyheri (Fig. 7C). Hier-archical clustering analysis also showed the presence of two majorgroups (Fig. 8A) when all the samples were included. Sub-clusterswere observed when each species were analysed separately, indi-cating phytochemical variation within each species (Fig. 8B andC). Van Wyk (2008) stated that biosystematic studies were needed

0.00.51.01.52.02.53.03.54.04.55.05.56.06.57.07.58.08.59.0

f1 (ppm)

0

50

100

150

200

250

300

350

400

6.06.16.26.36.46.56.66.76.86.97.07.17.27.37.47.57.67.77.87.98.0f1 (ppm)

0

50

100

150

200

250

300

350

400A: H. procumbens

0.00.51.01.52.02.53.03.54.04.55.05.56.06.57.07.58.08.5

f1 (ppm)

-20

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

6.06.16.26.36.46.56.66.76.86.97.07.17.27.37.47.57.67.77.87.9f1 (ppm)

-20

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

B: H. zeyheri

Fig. 6. Typical 500 MHz 1H-NMR spectra of deuterated methanol extracts of (A) H. procumbens and (B) H. zeyheri.

110 N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115

Fig. 7. (A) Principle component analysis score plot showing two clusters of H. procumbens (Blue) and H. zeyheri (Red). (B) OPLS-DA score plot of Harpagophytum spp. spectra(H. procumbens – Blue; H. zeyheri – Red), (C) S-plot showing positive loadings: chemical shifts d 1.53; 7.59; 7.39 ppm (H. procumbens – Blue) and negative loadings: d 3.47;3.13; 2.39; 5.67 ppm (H. zeyheri – Red).

N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115 111

Fig. 8. Hierarchical clustering showing, (A) All samples, Group 1–3: H. zeyheri; Group 4–7: H. procumbens; (B) H. procumbens only, which separates into four groups; (C) H.zeyheri only, which separates into four groups.

112 N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115

although the basic taxonomy of Harpagophytum was known. Thecircumscription of some of the species and infraspecific taxa arenot clear, and the full morphological and chemical diversityremains to be explored. To our knowledge, this is the first studyto show that H. procumbens and H. zeyheri have both similarities

and differences chemically, based on a large sample pool(n = 348) and using highly advanced methods. In a recent studyby Mncwangi et al. (2014) where single-point vibrational spectro-scopic and hyperspectral imaging techniques were employed, itwas concluded that H. procumbens and H. zeyheri, are not chemi-

N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115 113

cally congruent even though they may contain similar compounds(Mncwangi et al., 2014).

Applying the two techniques in a parallel manner offers comple-mentary benefits, where 1H-NMR is reproducible with rich struc-tural information and has the ability to detect a wide range ofplant metabolites; this complements the disadvantage of UHPLC-MS experiments where insufficient ionisation may prevent thedetection of some metabolites. The signal overlapping challenge in1H-NMR data is circumvented with UHPLC-MS (Holmes et al.,2006). Both 1H-NMR and UHPLC-MS datasets showed complemen-tary results whereby in the UHPLC-MS data, harpagoside, was a dis-tinguishing factor based on the ion intensity plot and with regards tothe 1H-NMR data, signals associated with iridoid glycosides whichare aromatic in nature, (d 1.53; 7.59; 7.39 ppm), were observed.

3. Conclusions

The British Pharmacopoeia definition for Devil’s Claw states:‘‘Cut and dried, secondary tuberous roots of H. procumbens DC.and/or H. zeyheri Decne’’ (BP, 2012). However, our results clearlydemonstrate a phytochemical disparity between the two specieswhich may impact on their biological properties. Furthermore, itappears that the harpagoside content specification maybe unreal-istic, especially when applied to H. zeyheri, where harpagosidewas detectable in only 25% of the samples. The UHPLC-MS and1H-NMR results illustrated the chemical variations within andbetween natural populations of Devil’s Claw. UHPLC-MS and 1H-NMR techniques were useful for identifying metabolites, and dif-ferentiating the metabolite patterns of the two Harpagophytumspecies. Quantification of the biomarker, harpagoside, showed thatharpagoside is highly variable, higher in H. procumbens species anddetected in low concentrations in some populations of H. zeyheri aspreviously reported by Czygan and Kruger (1977). Both UHPLC-MSand 1H-NMR methods produced congruent findings. The chemo-metric analysis results showed that the two species are not chem-ically equivalent, particularly, that harpagoside is not alwayspresent in H. zeyheri, suggesting that the therapeutic outcomemay be different, thus they should not be used interchangeablyuntil pharmacological equivalence has been confirmed.

Table 1Locality data and number of samples of Harpagophytum procumbens and H. zeyheri.

Species Location Number of samples (n = 348)

H. procumbens Cassel*. 19Ganyesa*h 117Moswana*h 48Molopo nature reserve*h 24Terra Firma*h 28Namibia 05

H. zeyheri Ferrolands*h 28Springbokfontein*h 21Makgabeng*h 46Victoria Falls (Zimbabwe) 12

* South Africa.h North West Province.. Northern Cape.

4. Experimental

4.1. General

4.1.1. UHPLC-MS analysisUHPLC analysis was performed on a Waters Acquity ultra-high

performance liquid chromatographic system with PDA detector(Waters, Milford, MA, USA). UHPLC separation was achieved onan Acquity UHPLC BEH C18 column (150 mm � 2.1 mm, i.d.,1.7 lm particle size, Waters) maintained at 35 �C. The mobilephase consisted of 0.1% formic acid in water (solvent A) and aceto-nitrile (solvent B) at a flow rate of 0.3 ml/min and gradient elutionwas applied as follows: 85% A: 15% B to 65% A: 35% B in 4 min,changed to 50% A: 50% B in 2 min, to 20% A: 80% B in 1 min, main-taining for 1 min and back to initial ratio in 0.5 min. The analysistime was 9 min. Samples were introduced into the mobile phasewith an injection volume of 1.0 ll (full-loop injection) for samplesand 2.0 ll for reference standards. The UHPLC system was inter-faced with a Xevo G2QTof MS (Waters, USA). The following massspectrometry operating conditions were applied: source – ESI neg-ative mode; capillary voltage – 3 kV; cone voltage 30 V; calibration– sodium formate; lock spray – leucine enkaphalin and scan massrange – 200–1500m/z.

4.1.2. 1H-NMR analysisSpectra were obtained using a 500 MHz Agilent DD2-500 NMR

spectrometer (Santa Clara, CA, USA) with Agilent Vnmrj 3.2 soft-ware. The following parameters were maintained during spectraldata acquisition: water peak saturation at 4.85 ppm; spectral width:8012 Hz; number of scans: 64; relaxation delay: 3 s; acquisitiontime: 8.179 s; pulse width (90�): 7.20 ls; Temperature: 25 �C. Thespectra were manually phased and baseline corrected using Mnovasoftware (MestReNova 8.0.0, Mestrelab Research). The 1H-NMRspectra were calibrated to the signal of TSP at the chemical shiftd = 0.00 ppm. Binning was applied to the chemical shifts d 0.00–9.00 ppm, with the width of each integral region set at 0.02 ppmusing the average sum. The 1H-NMR spectra were normalised bylargest peak (100) and total spectral area; after normalising thespectra were converted to ASCII files. Spectral regions d 4.75–5.05 ppm and d 3.27–3.37 ppm were excluded to eliminate theeffects of water suppression and the methanol residual signal,respectively.

4.2. Chemicals and reagents

Chemicals of analytical grade were used in the study: deuter-ated methanol (CD3OD, 99.8% D) was obtained from CambridgeIsotope Laboratories, Inc. and an internal standard (3-(trimethyl-silyl)-propionic 2,2,3,3-d4 acid, NaCl salt) [TSP] and harpagosidewere purchased from Sigma Aldrich, Inc (South Africa). HPLC-gradesolvents for UHPLC-MS analysis were obtained from Waters Corpo-ration, South Africa.

4.3. Plant material

Sampling took place during 2010 throughout the natural distri-bution of the taxa; samples were obtained from ten different local-ities in three countries; South Africa (North-West Province andNorthern Cape), Namibia and Zimbabwe. In total 348 samples werecollected and analysed, 241 H. procumbens and 107 H. zeyheri sam-ples (Table 1). Voucher specimens are retained at the NationalBotanical Museum, Bloemfontein (South Africa). Secondary roottubers were chopped, dried and further ground to a fine powderusing a Retsch� 400 MM ball mill at a frequency of 30 Hz for 60 sand stored in airtight vials.

4.4. UHPLC-MS solvent extraction

Ground secondary root tubers were weighed (100 mg) andextracted with 5 ml AR-Grade methanol by ultrasonication(Thermo Scientific, South Africa) for 30 min at 40 �C. The processwas repeated three times. Extracts were then combined and dried

Table 2MS/MS fragmentation selected compounds from H. procumbens.

Peak No. Rt (min) UVmax Pseudo-molecular ion [M�H]� (m/z) MS/MS fragmentation [M�H]� Compound

1 0.84 – 665 (665), 503, 485, 443, 383, 341, 179 Acetylacteoside2 0.9 199 377 (341), 215, (179), 161, 119, 89, 71 Derivative of caffeoyl glucoside3 1.11 199 399 (388), 364, 327, 309, 276, 219, 179, (61) Harpagide4 1.84 228; 283 503 503, (461), 443, 315, 161, 135, 113, 71 Decaffeoylverbascoside5 2.8 328 623 (623), 461, 161 Acteoside/isoacteoside6 3.58 219; 339 665 (665), 623, 503, 461, 161 Verbascoside7 3.97 217; 325 665 (503), 478, 345, 329, 183, 163, 145, 119 Chebuloside II/nigaichigoside8 4.06 – 509 (509), 163 Coumaroylharpagide9 4.25 234; 327 665 (665), 623, 503, 461, 162, 161 Isoverbascoside10 4.89 220; 330 707 (707), 665, 545, 503, 161, 135, 103 Diacetylacteoside11 4.98 201; 215; 278 528 (493), 383, 345, 179, 147,119, 89 Harpagoside12 5.57 274 763 (763), 729, 593, 530, 482, 416, 344, 269, 207, (96) Unidentified

Values in bold signify the fragments with high intensity.

114 N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115

in a vacuum oven (VISMARA, Labotec, Switzerland). Extracts werere-dissolved in MeOH (HPLC grade) at a concentration of 1 mg/mlwhich was diluted to a final concentration of 100 lg/ml and fil-tered through 0.2 lm syringe filters for UHPLC-MS analysis.

4.5. 1H-NMR solvent extraction

Homogenised plant powder was weighed (300 mg), 600 ll CD3-

OD containing TSP was added and then vortexed. All samples weresonicated at 25 �C for 5 min and left for 60 min, thereafter centri-fuged at 13,400 rpm for 9 min using a high speed mini-centrifugeand the supernatant transferred into appropriately labeled NMRtubes for 1H-NMR analysis.

4.6. Statistical analysis

UHPLC-MS data were collected and processed by chromato-graphic software Masslynx 4.1 (Waters, Milford, MA, USA). Thequantitative aspect of the method was validated by determiningthe linearity, precision, accuracy, limit of detection (LOD) and thelimit of quantification (LOQ). Chemometric analysis was performedusing Markerlynx 4.1 and SIMCA-P+ software (12.0, Umetrics). All1H-NMR data were exported to SIMCA-P+ software (12.0, Umet-rics) for multivariate data analysis. Principal component analysis(PCA) enables the data to be transformed to a new coordinate sys-tem such that the significant variances by any projection of thedata can be visualised based on principal components. All metabo-lites and interactions associated with the changes of metabolitelevels are accounted for, thus making this approach more favour-able compared to a targeted approach. The first principle compo-nent (PC1) represents the largest component of variance and PC2the second largest, etc. (Beyoglu and Idle, 2013). Principle compo-nent analysis is an unsupervised method which provides valuableinformation regarding similarity and difference of the samples. Thechemotaxonomic classification of the species can be visualised byboth PCA score plots and hierarchical cluster analysis (dendro-gram). Hierarchical cluster algorithm analysis groups a set ofobjects in such a manner that objects in the same group (cluster)contain more similarities to each other than those in other groups(clusters). It is a useful tool in data mining especially for patternrecognition purposes. This grouping is thus more ‘natural’ sincethe objects are not pre-assigned to classes. In order to obtain a bet-ter understanding of the differences, similarities and to possiblyfind potential biomarkers discriminating between the two species,orthogonal partial least squares discriminant analysis (OPLS-DA)was performed. This is a supervised method where the sampleinformation is assigned to the species classes, and the method is

useful for identifying the significant variables that are responsiblefor the differentiation of the two species.

Acknowledgements

We are thankful to Dr Z. Zietsman (National Botanical Museum,Bloemfontein) for assistance with the sourcing of raw materialsand the National Research Foundation (Thuthuka Programme) forfinancial support.

References

Beyoglu, D., Idle, J.R., 2013. Metabolomics and its potential in drug development.Biochem. Pharmacol. 85, 12–20.

BP, 2012. British Pharmacopoeia Herbal Drugs, Herbal Drug Preparations andHerbal Medicinal Products. British Pharmacopoeia Commission, London.<http://bp2012.infostar.com.cn/Bp2012.aspx?tab=browse&a=display&n=98&id=5089> (Retrieved).

Cole, D., 2003. The Impact of Certification on the Sustainable Use of Devil’s Claw(Harpagophytum procumbens) in Namibia. Non-Wood Forest ProductsProgramme. <http://www.fao.org/forestry/foris/pdf/NWFP/Namibia>(Retrieved).

Czygan, F.C., Kruger, A., 1977. Mitteilung: Zur Verteilung des Iridoid-GlycosidsHarpagosid in den einzelnen Organen von Harpagophytum procumbens DC undHarpagophytum zeyheri Decne. Planta Med. 31, 305–307.

Fiebich, B.L., Heinrich, M., Hiller, K.O., Kammerer, N., 2001. Inhibition of TNF-alphasynthesis in LPS-stimulated primary human monocytes by Harpagophytumextract SteiHap 69. Phytomedicine 8, 28–30.

Georgiev, M., Ali, K., Alipieva, K., Verpoorte, R., Choi, Y.H., 2011. Metabolicdifferentiations and classification of Verbascum species by NMR-basedmetabolomics. Phytochemistry 72, 2045–2051.

Georgiev, M.I., Ivanovska, N., Alipieva, K., Dimitrova, P., Verpoorte, R., 2013.Harpagoside: from Kalahari Desert to pharmacy shelf. Phytochemistry 92, 8–15.

Gruenwald, J., 2002. Expanding the Market for Devil’s Claw in Europe. Paperpresented at the Namibian National Devil’s Claw Conference (Retrieved).

Gunther, M., Schmidt, P.C., 2005. Comparison between HPLC and HPTLC-densitometry for the determination of hapagoside from Harpagophytumprocumbens CO2-extracts. J. Pharm. Biomed. Anal. 37, 817–821.

Holmes, E., Tang, H.R., Wang, Y.L., Seger, C., 2006. The assessment of plantmetabolite profiles by NMR-based technologies. Planta Med. 72, 771–785.

Ihlenfeldt, H.-D., Hartmann, H., 1970. Die Gattung Harpagophytum (Burch) DC. exMeissn. Mitteilungen aus dem Institut für Allgemeine Botanik in Hamburg 13, 15–69.

Kemper, K.J., 1999. Devil’s Claw (Harpagophytum procumbens) (Online). Availablefrom: <http://www.mcp.edu/herbal/default.htm>.Kemper (Retrieved).

Kim, H.K., Khan, S., Wilson, E.G., Kricun, S.D.P., Meissner, A., Goraler, S., Deelder,A.M., Choi, Y.H., Verpoorte, R., 2010. Metabolomic classification of SouthAmerican Ilex species by NMR-based metabolomics. Phytochemistry 71, 773–774.

Kim, H.K., Choi, Y.H., Verpoorte, R., 2011. NMR-based plant metabolomics: where dowe stand, where do we go? Trends Biotechnol. 29, 267–275.

Korkina, L.G., 2007. Phenylpropanoids as naturally occurring antioxidants: Fromplant defense to human health. Cell. Mol. Biol. 53, 15–25.

Marti, G., Boccard, J., Mehl, F., Debrus, B., Marcourt, L., Merle, P., Delort, E., Baroux, L.,Sommer, H., Rudaz, S., Wolfender, J.-L., 2014. Comprehensive profiling andmarker identification in non-volatile citrus oil residues by mass spectrometryand nuclear magnetic resonance. Food Chem. 150, 235–245.

Mncwangi, N., Chen, W., Vermaak, I., Viljoen, A.M., Gericke, N., 2012. Devil’s Claw –A review of the ethnobotany, phytochemistry and biological activity ofHarpagophytum procumbens. J. Ethnopharmacol. 143, 755–771.

N.P. Mncwangi et al. / Phytochemistry 106 (2014) 104–115 115

Mncwangi, N., Vermaak, I., Viljoen, A.M., 2014. Mid-infrared spectroscopy and shortwave infrared hyperspectral imaging – A novel approach in the qualitativeassessment of Harpagophytum procumbens and H. zeyheri (Devil’s Claw).Phytochem. Lett. 7, 143–149.

Moco, S., Bino, R.J., De Vos, R.C.H., Vervoort, J., 2007. Metabolomics technologies andmetabolite identification. Trends Anal. Chem. 26, 855–866.

Napolitano, J.G., Gödecke, T., Lankin, D.C., Jaki, B.U., McAlpine, J.B., Chen, S.N., Pauli,G.F., 2014. Orthogonal analytical methods for botanical standardization:determination of green tea catechins by qNMR and LC–MS/MS. J. Pharm.Biomed. Anal. 93, 59–67.

Porzel, A., Farag, M.A., Mülbradt, J., Wessjohann, L.A., 2013. Metabolite profiling andfingerprinting of Hypericum species: a comparison of MS and NMRmetabolomics. Metabolomics, 1–15.

Safer, S., Cicek, S.S., Pieri, V., Schwaiger, S., Schneider, P., Wissemann, V., Stuppner,H., 2011. Metabolomic profiling of Leontopodium species (Astaraceae) by meansof 1H NMR and HPLC-ESI-MS. Phytochemistry 72, 1379–1389.

Schmidt, A.H., 2005. Fast HPLC for quality control of Harpagophytum procumbens byusing a monolithic silica column: method transfer from conventional particle-based silica column. J. Chromatogr. A1073, 377–381.

Steinmann, D., Ganzera, M., 2011. Recent advances on HPLC/MS in medicinal plantanalysis. J. Pharm. Biomed. Anal. 55, 744–757.

Stewart, K.M., Cole, D., 2005. The commercial harvest of Devil’s Claw(Harpagophytum spp.) in southern Africa: the devil’s in the details. J.Ethnopharmacol. 100, 225–236.

Van Wyk, B.-E., 2008. Abroad review of commercially important southern Africanmedicinal plants. J. Ethnopharmacol. 119, 342–355.

Van Wyk, B.-E., Gericke, N., 2000. People’s Plants: A Guide to Useful Plants ofSouthern Africa. Briza Publications, Pretoria, South Africa.

Viljoen, A., Mncwangi, N., Vermaak, I., 2012. Anti-inflammatory iridoids of botanicalorigin. Curr. Med. Chem. 19, 2104–2127.

Von Willert, D.J., Sanders, J., 2004. Devil’s Claw: conservation through cultivation.Results of worldwide ecological studies. In: Proceedings of the SecondSymposium of the A.S.W. Schimper Foundation.

Wolfender, J.-L., Rudaz, S., Choi, Y.H., Kinm, H.K., 2013. Plant metabolomics: fromholistic data to relevant biomarkers. Curr. Med. Chem. 20, 1056–1090.