M. Maulidiania , Bassem Y. Sheikhb , Ahmed Mediania , Leong Sze Weia , Intan Safinar Ismaila , Faridah Abasa , Nordin H. Lajisa,b, * a Laboratory of Natural Products, Institute of Bioscience, Universiti Putra Malaysia, 43400 UPM, Serdang, Selangor, Malaysia bAl-Moalim BinLaden Chair for Scientific Miracles of Prophetic Medicine, Scientific Chairs Unit, Taibah University, P.O. Box 30001, Madinah al Munawarah 41311, Saudi Arabia A R T I C L E I N F O Article history: Received 15 April 2015 Received in revised form 28 June 2015 Accepted 6 July 2015 Available online xxx Keywords: Nigella sativa Metabolomics Nitric oxide inhibition Radical scavenging a-Glucosidase inhibition A B S T R A C T Authenticity and quality coherence are the major elements in ensuring the consistency of the expected beneficial outcomes from the use of traditional or herbal remedies. Metabolomics offers the possibility of addressing these issues. Principal Component Analysis (PCA) was applied to select the best solvent system for sample extraction. Partial Least Square (PLS) regression analysis was found useful in evaluating the relationship between Nigella sativa seeds from four different origins on the basis of their metabolite profiles. In this study, different bioactivities were displayed by different samples with the Qasemi and Syrian samples exhibited high a-glucosidase inhibitory activity, which was correlated to the high fatty acid contents based on the PLS model. The Ethiopian sample exhibited high DPPH radical scavenging and nitric oxide (NO) inhibition activities, which may be related to the presence of high levels of thymoquinone and thymol. The method was also successfully used to classify “new test” samples into their proper groups. ã 2015 Published by Elsevier B.V. on behalf of Phytochemical Society of Europe. 1. Introduction In recent years, there has been a growing interest in the use of herbal medicines for treatment of illnesses and health care purposes, in both the developed and developing countries. This owes much to their steady therapeutic effects, perceived safety and efficacy, and the escalating costs of modern medicines. Currently, herbal medicinal products are widely marketed around the world, and increasingly so in Western nations (Jordan et al., 2010). The sales of traditional medicine have significantly increased during the last decade with the growth of global demand for medicinal plant-based raw materials was estimated at 10–15% per annum. The annual industrial output for China, as listed on the herbal database Chinese Materia Medica was at US$ 47.84 billion in 2010, up 29.5% from the previous year (WHO, 2012). Despite the perceived safety and efficacy of herbal medicinal products, few cases of adverse effects from their use have been reported, which could originate from contamination of products with toxic metals, adulteration with pharmacologically active synthetic compounds,misidentificationor substitutionoftheherbal ingredients, or improperly processed or prepared products (Ankli et al., 2008; Ernst, 2004). Concerned by such issues, the public now demands for the improvement in the over-all quality of herbal medicines through strict standardization and proper quality control of both the raw materials and finished products. This is reflected by the increased discussion on safety assessment of herbs, such that several initiatives have beenconducted (Cordell, 2011). Furthermore, in 2009 a Consortium on Good Practice in Traditional Chinese Medicine Research was established and funded by the European Union to address some fundamental issues including the develop- ment of guidelines as well as identification of priorities, challenges and opportunities (Uzuner et al., 2012). Owing to the potential variation in metabolite content and chemical changes during plantation, post-harvest and processing, it is imperative that many medicinal plants properly standardized, especially for those that are widely used and grown in different regions and conditions. Nigella sativa, also commonly known as “Black cumin” is a plant belonging to Ranunculaceae family, which grows in countries bordering the Mediterranean Sea. In the Middle Eastern countries, apart from being applied for medicinal purposes the seeds are often used as seasoning for vegetables, legumes and different types * Corresponding author at: Laboratory of Natural Products, Institute of Bioscience, Universiti Putra Malaysia, 43400 UPM, Serdang, Selangor, Malaysia and Al-Moalim BinLaden Chair for Scientific Miracles of Prophetic Medicine, Scientific Chairs Unit, Taibah University, P.O. Box 30001, Madinah al Munawarah 41311, Saudi Arabia. E-mail address: nordinlajis@gmail.com (N.H. Lajis). http://dx.doi.org/10.1016/j.phytol.2015.07.012 1874-3900/ã 2015 Published by Elsevier B.V. on behalf of Phytochemical Society of Europe. Phytochemistry Letters 13 (2015) 308–318 Contents lists available at ScienceDirect Phytochemistry Letters journal homepage: www.elsevier.com/locate/phytol of baked products (Atta, 2003). Numerous studies on the non- volatile extracts and essential oils of the seeds revealed the presence of a wide spectrum of chemical constituents such as monoterpenes, glycolipids, sterols (Cheikh-Rouhou et al., 2008) and fatty acids (Nergiz and Otles, 1993) have been reported. Valuable pharmacological properties of the seed extracts including anti-oxidant, anti-cancer, anti-diabetic and anti-inflammatory activities have also been summarized (Gali-Muhtasib et al., 2006; Ramadan and Mörsel, 2003; Ramadan, 2007; Zribi et al., 2014). The presence of thymoquinone and its derivatives such as thymol and dihydrothymoquinone in the essential oils of the seeds has been associated with protective effects against induced toxicity (Mohamadin et al., 2010), in addition to other pharmaco- logical activities (Harzallah et al., 2011). Besides these, other monoterpenes such as p-cymene, a-thujene, b-pinene, terpenol, carvacrol and terpinene constitute the major components of the essential oils. It has been shown that the biological properties of the essential oils are directly associated with the chemical composition, which varies depending on the origin (Bourgou et al., 2010). Several of the pharmacologically interesting substances have also been isolated from Nigella seeds including isoquinoline and indazole alkaloids (Atta, 2003; Atta-ur-Rahman et al., 1995), saponins (Mehta et al., 2009), cycloartenols (Mehta et al., 2008), and flavonoid glycosides (Merfort et al., 1997). The complex nature of herbal preparations has made standardi- zation a difficult task. In most cases, therapeutic activities and sometimes toxic effects of herbal medicines are derived from synergistic actions of multiple secondary metabolites in a single or among multiple herbs used in the formulation (Georgiev et al., 2011; Kim et al., 2010a; Razmovski-Naumovski et al., 2010; Zahmanov et al., 2015). Metabolomics approach offers a new hope towards standardization and efficacy validation of the herbal and comple- mentary medicines, while at the same time strengthens its public acceptance (Buriani et al., 2012;Yamazaki et al., 2003).Itis a toolthat allows qualitative and quantitative determination of all the metabolites of an organism based on the spectral and physicochem- ical data generated. Statistical tools such as principal component analysis (PCA), partial least square (PLS) and hierarchical cluster analysis (HCA) are used to recognize the pattern generated from comparison of the data belonging to different sample groups, to identify correlation between variables and to observe chemotype relationship between varieties or species, respectively. Notwithstanding the numerous reports on of N. sativa seeds, comprehensive studies toward correlating the metabolite con- stituents with geographic origin, genetics, agricultural aspects and potential biological implications have been lacking. Recently, the variation of yields and composition of N. sativa seed oil from different geographical origins has been reported (Farag et al., 2014). In the current study we intended to utilize metabolomics approach to differentiate samples of N. sativa seeds originating from four different countries (Saudi Arabia, Yemen, Syria and Ethiopia) based on their NMR fingerprint, and to use the data sets for the conformation of the origin of unknown samples. Proton nuclear magnetic resonance ( 1 H NMR) spectrometric data was used as the basis for metabolomics analyses in this study owing to its simplicity and rapid data analysis. Furthermore, we also intended to establish the relationship between the constituents and bioactivities exhibited by three assays including the DPPH A B C 8.00 7.50 7.00 6.50 7.00 6.90 6.80 6.70 6.60 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 8.00 7.50 7.00 6.50 7.10 7.00 6.90 6.80 6.70 6.60 6.50 1 1 2 3 4 3 3 1 2 2 1 1 2 8 7 8,7 5 4 5 3 2 1 2 1 3 3 6 3 4 3 9 Fig.1. The representative 1 H NMR spectra of N. sativa seeds extracts of the Ethiopian origin using (A) CDCl3, (B) CD3OD and (C) CD3OD + D2O as solvents (signals not assigned). Identified signals: 1, thymol; 2, thymoquinone; 3, fatty acids; 4, p-cymene; 5, thymol-b-glucopyranoside; 6, sucrose; 7, quercetin; 8, kaempferol; and 9, cycloartenol. M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 309 radical scavenging (DPPH-RS), nitric oxide inhibition (NOI), and a-glucosidase inhibitory (AGI) activities. 2. Results and discussion 2.1. Assignment of compounds from the 1 H NMR spectra Three different solvent systems including chloroform, metha- nol and aqueous methanol (1:1) were tested for the extraction of each sample and the optimum sample discrimination was compared using multivariate analysis. The representative spectra of the Ethiopian samples extracted using three different solvent systems are shown in Fig. 1. Visual inspection of the 1 H NMR spectra of the extracts showed the presence of phenolic, flavonoid, carbohydrate, terpenoid, and fatty acid compounds. In the chloroform extract, no signal for sugar or carbohydrate (chemical shift region of 3–4 ppm) was observed. However, sugars and carbohydrate signals were observed in the methanol and became more prominent in the spectra of the aqueous methanol extracts. New signals at 7.08 and 6.62 ppm, which may be attributed to thymols and thymoquinone were observed. The signal attributable to olefinic signal at 5.95 ppm was also present in the spectrum of methanol extracts along with the high-field signals at 0.65 and 0.39 ppm, which may be assigned to the cyclopropane methylene of cycloartenol type saponins. The methyl signals at 0.80–0.90 ppm and olefinic signals at 5.20–5.50 ppm could be attributed to triterpenoid aglycones as well as unsaturated fatty acids. In addition, aromatic signals belonging to flavonoid aglycone (quercetin and kaempferol) were also detected in this highly polar and protic solvent. The spectra of the aqueous methanol extracts exhibited even stronger signals of these metabolites. These observations are consistent with the fact that the triterpenoid saponins and flavonoid glycoside are more soluble in aqueous methanol. The assignment of compounds were supported by comparison with the literature data [thymol and thymol-b-glucopyranoside (Ahmed and Jakupovic, 1990), thymo- quinone (de Sousa et al., 2011), fatty acids (Knothe and Kenar, 2004), p-cymene (Schwarz and Ernst, 1996), cycloartenol (Mehta et al., 2009), sucrose, quercetin, and kaempferol (Maulidiani et al., 2012)]. Identification of thymol, thymoquinone, fatty acids, p- cymene, thymol-b-glucopyranoside, and sucrose was also con- firmed by 2D NMR data (HMBC and J resolved). The 1 H NMR assignments of the identified compounds in each sample of different origin are shown in Table 1. 2.2. Selection of the solvent system for extraction based on PCA. The 1 H NMR spectra of twelve samples representing three from each origin, were taken for preliminary evaluation and selection of solvent system for extraction. Principal component analysis (PCA) was conducted on the 1 H NMR data (Fig. 2) to assess their suitability based on the inclusion of most representative metab- olites, resulting discriminant performance, simplicity of procedure and cost. Thus, three sets of PCA model representing each of the solvent system used were developed using the entire spectral range of dH 0.52–10.00 ppm, and the results were analyzed. By comparison, the methanol extracts (Fig. 2B) provided the best degree of discrimination where its first two principal components (PCs) cumulatively accounted for 88.5% of the total variance in the original data (PC1 = 69.8% and PC2 = 18.7%). Lower performance was shown by chloroform extracts (Fig. 2A) with 70.8% (PC1 = 52.4% and PC2 = 18.4%) of total variance. However, the PCA could only generate one principal component for the aqueous methanol (not shown here). When the overlapping sugar spectral region at 3.12–3.96 ppm was excluded, class separation of the methanol extracts (Fig. 2D) was further improved with 91.9% (PC1 = 75.5% and PC2 = 16.4%), followed by the aqueous methanol extracts (Fig. 2C) with 59.3% (PC1 = 43.6% and PC2 = 15.7%) of total of variance. The best separation and higher values of total variance indicate that the input spectra has provided more information, which is an important factor in the multivariate analysis. On this basis, the extraction using methanol and exclusion of the sugar signal region (at 3.12–3.96 ppm) was decided as the best condition and selected method for further model development (PLS and PLSDA) in next section. 2.3. PLS model for bioactivity description The binned 1 H NMR chemical shifts were used as the input (X or independent variables), while three assay results including the percent of inhibition values for AGI and NOI, and the IC50 values for DPPH-RS activities were the output (Y or dependent variables). The total number of observation data sets used for this model was twelve comprising three samples each from the four different origins (Syria, Yemen, Ethiopia, and Qasem). The biological activity results for all three assays (AGI, NOI and DPPH-RS activities) of the methanol extracts, which are presented as mean SD, are summarized in the Supplementary. The goodness of fit, prediction of Y, and permutation test were used to validate the PLS model. Autofit of PLS in the SIMCA resulted in two components with the goodness of fit for Y variables at R2 Y = 0.97 and the goodness of prediction of the model with respect Table 1 The 1 H NMR spectral data and relative quantification of identified constituents of N. sativa seeds (in CD3OD extract). No Compound Chemical shift 1 Thymol 7.08 (2H, dd, J = 7.0, 2.0Hz) 6.61 (1H, d, J = 2.0Hz) 3.18 (1H, m) 2.28 (3H, s) 1.22 (6H, d, J = 7.0Hz) 2 Thymoquinone 6.62 (1H, d, 1.5Hz) 6.53 (1H, d, 1.0Hz) 3.02 (1H, hept, J = 6.5, 1.0Hz) 2.00 (3H, d, 1.5Hz) 1.14 (6H, d, 7.0Hz) 3 Fatty acid 5.35 (d, J = 5.0Hz) 2.34 (t, J = 7.5Hz) 2.04 (m) 1.33 (m) 1.28 (m) 0.90 (s) 4 p-Cymene 2.85 (1H, sept, J = 7.0Hz) 2.29 (3H, s) 1.27 (6H, d, 7.0Hz) 5 Thymol-b-glucopyranoside 7.09 (d, J = 8.5Hz) 6.94 (dd, J = 8.0, 2.0Hz) 6.69 (dd, J = 8.0, 2.0Hz) 5.11 (d, J = 8.0Hz) 4.38 (dd, J = 3.5, 12.0Hz) 4.18 (dd, J = 6.5, 12.0Hz) 6 Sucrose 5.39 (d, J = 3.5Hz) 4.10 (d, J = 8.0Hz) 7 Quercetin 7.71 (1H, d, J = 1.5Hz) 7.63 (1H, dd, J = 8.5, 2.0Hz) 6.37 (1H, d, J = 1.5Hz) 8 Kaempferol 8.07 (2H, d, J = 8.5Hz) 6.80 (2H, d, J = 8.5Hz) 6.39 (1H, d, J = 2.0Hz) 6.19 (1H, d, J = 2.0Hz) 9 Cycloartenol 5.95 (d, J = 6.5Hz) 5.38 (d, J = 6.5Hz) 1.68 (3H, s) 0.85 (3H, s) 0.80 (3H, s) 0.65 (1H, dd, J = 4.5, 8.0Hz) 0.39 (1H, dd, J = 4.5, 8.0Hz) 310 M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 to Y variables at Q2 Y = 0.95. The PLS model for the three assays indicated that it fits reasonably well based on the values of the resulting regression coefficient at 0.97, 0.97 and 0.94 for the AGI, NOI and DPPH-RS activities, respectively, as shown in Fig. 3. The model was further validated by permutation test. The results from permutation test on the second component of PLS for AGI (Supplementary, Fig. 1) clearly showed that the Y-axis intercepts of R2 and Q2 were at 0.09 and 0.35, respectively. Similar results were also obtained for the permutation test on NOI of the DPPH-RS activities, where the Y-axis intercepts of R2 and Q2 were at 0.22 and 0.3 for the former and 0.15 and 0.4 for the latter (not show here). The intercepts of R2 < 0.3 and Q2 < 0.05 on Y-axis, and the R2-line being far from being horizontal suggested that the PLS model does not over-fit (Eriksson et al., 2006). 2.4. Relationship between metabolites, bioactivity and the origin based on PLS model The relationship between chemical constituents and bioactivity were evaluated using the variable importance in the projection (VIP) analysis and the regression coefficient of the PLS. The VIP and its ranking, which are listed in Table 2 indicate how strong the contribution of each identified constituent to the model. The VIP values of almost all of the signals representing major constituents (e.g. fatty acid, thymoquinone, and thymol) were larger than 0.8, indicating their high degree of contribution (Eriksson et al., 2006). These metabolites were ranked among the highest five, which also suggested of their most influential contribution. Other detected constituents such as p-cymene, thymol-b-glucopyranoside, cyclo- artenol, as well as derivatives of quercetin and kaempferol contributed less effects. The regression coefficient of the PLS indicates the positive and negative effect of X variables to the output or Y variables. A larger effect is shown by the high value, whereas a lowest effect is given by the near zero value. As shown in Table 2, the major constituents such as thymoquinone and thymol contributed positive effect to all the three bioactivities, while fatty acid gave positive contribution to AGI, but negative to NOI and DPPH-RS activities. The results seemed to agree with the previous results, which found that N. sativa inhibits intestinal glucose absorption (Meddah et al., 2009). Moreover, thymoquinone was also reported to exhibit hepatic glucose reducing property via a-glucosidase inhibition (Fararh et al., 2002), reduction of the nitric oxide production (El-Mahmoudy et al., 2002) and antioxidant activities (Badary et al., 2003). Oleic acid and linoleic acid were also reported to have a strong anti-a-glucosidase activity (Su et al., 2013). Other detected constituents such as p-cymene, thymol- b-glucopyranoside, cycloartenol, and the derivatives of quercetin and kaempferol showed identical influences, which were positive effects for the NOI and DPPH-RS activities but negative to the AGI (Table 2). The presence of sucrose did not influence the activities of all assays. Fig. 2. The PCA scores plots of N. sativa seeds extracts from four different origins. (A) chloroform, (B) methanol, and (C) and (D) for methanol–water and methanol as extractant, respectively, with sugar spectral region (3.12–3.96 ppm) excluded. Keys to figure: Et – Ethiopian (orange diamonds); Qa – Qasemi (green triangles); Sy – Syrian (red circles); Ye – Yemeni (blue boxes). M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 311 The relationship between the origin and bioactive constituents is described in the PLS scores and loadings plots of the first two components, which accounted for 85.7% (t[1] = 69.9% and t [2] = 15.8%) of the total variance, as presented in Fig. 4. The scores plot (Fig. 4A) showed that three main classes were developed distinctly corresponding to four different origins. The Ethiopian samples were clustered in the first (top-right) quadrant, while the Qasemi and Syrian samples were in the second (top-left) quadrant. The cluster for the Yemeni samples was located at the middle of the bottom half. Signals of the constituents that contributed to the characteristics of the samples are shown in the loadings plot at the locations, which correspond to the respective clusters of the scores plot. The distances of the signals from the origin (0,0 point) of the loadings plot indicate the extent of their contribution to the characteristics of the respective clusters. Inspection of the loadings plot (Fig. 4B) indicated that thymoquinone and thymol contributed most in the Ethiopian sample, which also led to its higher NOI and DPPH-RS activities. Relatively higher concentration of fatty acid was indicated in the Qasemi and Syrian samples, and thus their AGI were also found to be higher (Supplementary Table 1). The Yemeni samples were found to exhibit the least activity in all the three assays (Supplementary Table 1). These findings were in agreement with the relative quantification and analysis of variance (ANOVA) of the major identified compounds, thymol and thymoquinone, which were very significant (p < 0.001) in the Ethiopian as compared to the samples from other origins. The presence of fatty acids was very significant in the Qasemi and Syrian (p < 0.001) as compared to the Yemeni and Ethiopian samples (Supplementary Table 2). 2.5. Hierarchical cluster analysis (HCA) To assess the distance between classes of samples represented in the scores plot ofthe PLS (Fig. 4A), a hierarchical cluster analysis was performed. The HCA of the samples in the scores plot was calculated using Ward’s minimum variance, and the results are presented as dendrogram tree, which are sorted based on size and heat map as shown in Fig. 5. Heat map represents the averaged binned 1 H NMR integrals of the varying metabolites from minimum (dark green) to maximum (dark red) values. The developed heat map showed that the Ethopian sampleswere characterized by thehighconcentrations of thymol and thymoquinone but low in fatty acids, which agreed well with the results from the regression coefficient and loadings plot of the PLS model. In contrast, the Syrian and Qasemi samples contained high level of fatty acids but low concentrations of thymol and thymoquinone. The heat map also clearly showed that the biological replicates from the same origin are grouped closely together, indicating an excellent reproducibility in both the sample extraction and 1 H NMR measurement. Moreover, the HCA dendo- gram resulted in three clustering groups of the Qasemi + Syrian, the Yemeni and the Ethiopian. The Syrian and Qasemi samples were metabolicallyintimatelyrelated, and bothofthesewerequite closely related to the Yemeni sample. On the other hand, the Ethiopian sample was more distantly related to the other three. These analyses were in agreement with the separation of the former from the latter groups by t[1] of the PLS scores plot (Fig. 4C). The close metabolic relationship of the Yemeni, Syrian and Qasemi samples may be attributedtothe similarity in the intrinsic (such geneticmakeup) or extrinsic (such as climate and agronomics) factors. Variation of the chemical profiles and biological activities of the essential oils of N. sativa seeds from different sources has also been reported (Bourgou et al., 2010). 2.6. PLSDA for discrimination of sample from different origins. A total of fifty-four samples representing thirteen each from the Syrian, Yemeni and Qasemi (S. Arabia) origins, and 15 from the Ethiopian were used to develop the supervised (PLSDA) model in the discrimination of the samples, using CD3OD as the solvent for extraction. Forty samples (representing ten from each origin) were used as the training, while three samples each from the Syrian, Yemeni and Qasemi, and five from Ethiopian were randomly selected for the testing sets. The accuracy of the PLS-DA model was evaluated by means of a “probability of class membership”, which are presented in the classification list in Table 3 for both the training and testing data sets as well. Each sample was classified in the form of “Predicted Y value for dummy variables”, which indicates the probability of each sample/member to belong to the specified group. Based on the guidelines outlined in the SIMCA+ 13 user Fig. 3. The PLS models for the observed and predicted activities. (A) a-Glucosidase inhibition, (B) nitric oxide inhibition and (C) DPPH radical scavenging (1/IC50) activities. 312 M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 guide, the members with predicted values of below 0.35 are not considered to belong to the class, while those with predicted values of between 0.35 and 0.65 are considered borderline, and those with predicted values of above 0.65 are members that belong to the class. Using these guidelines, the training and testing results showed that all data sets were classified in their respective class correctly, indicating a 100% of overall accuracy. The robustness of this technique was further tested on “new samples” purchased from the same retailer at different time (of nine months gap), which were labeled as the Yemeni and Qasemi (S. Arabia) origins. Each origin was represented by five samples. All of the samples labeled as the Yemeni origin were found to fit the class correctly, while the Qasemi (S. Arabia) labeled samples were also classified as belonging to the Yemeni group, but with the lower scores as compared to the former. Comparison of the Y values between the classes found that the scores for the samples in Yemeni group were consistently higher than those in the other groups. The classification results suggested that the new Qasemi test sample was very closely related to the Yemeni group albeit not exactly the same. The reliability of PLSDA for the classification of testing data sets and that after the introduction of new samples (data are shown in Table 3) was assessed using contingency matrix and the Cohen’s Kappa analysis as depicted in Table 4. The lowest class accuracy was obviously shown by the Qasemi (Saudi) sample at 37.5%, which gave 62.5% of omission error (error of exclusion from producer). The low percentage of this sample may suggest the possible discrepancy of the sample, or due to genetic and agronomic factors resulting in metabolic compatibility with the Yemeni sample. Despite these, the percentage of the overall agreement was = 100% * ( PXii)/N = 100% * (3 + 8 + 5 + 3)/24 = 79.17%. The Cohen’s kappa was subsequently used to measure the reliability of the class analysis of PLSDA by subtracting out the agreement due to chance from the overall agreement between the PLSDA and the reference classes (Allouche et al., 2006; Carletta, 1996). Thus, the Cohen’s Kappa coefficient,that was calculated by k = [(PXii) ( PXi+X+i)/N]/ [N ( PXi+X+i)/N] = [19 6.75]/[24 6.74] was equal to 0.71 (see Table 4). According to criteria of Landis and Koch (1977), the k value of 0.6–0.8 is considered reliable for classification. On the basis of these, it may be concluded that the class analysis of PLSDA for the discrimination of the sample origin of N. sativa was good, and may be used as a standardization tool. In conclusions, the 1 H NMR based metabolomics utilizing the PCA, PLS and PLSDA models was successfully applied for the selection of most appropriate solvent for extraction of N. sativa seeds originating from four different countries, for the establish- ment of the sample origin and bioactivity relationship, and for the discrimination of samples and determination of their origin, respectively. The best separation of clusters was achieved by methanol extraction in combination with the exclusion of the sugar region in the NMR spectra prior to the PCA. The developed PLS model could describe all the three bioactivities reasonably well without the tendency of over-fitting. Almost all of the signals representing the major constituents (fatty acid, thymoquinone, and thymol) were the most influential constituents in sample discrimination based on the PLS model. The Ethiopian samples were found to be relatively richer in thymoquinone and thymol, which led to the higher NOI and DPPH-RS activities. The Qasemi and Syrin samples were richer in fatty acids, and exhibited higher AGI activity. The HCA showed that the Syrian, Yemeni and Qasemi samples were closely related metabolically, while the Ethiopian was a distance away. The PLSDA could classify the training data sets into their origins correctly, and verify the quality status of new samples. The overall accuracy of testing data set reached 79.2% with the kappa coefficient of 0.71, which indicated a reliable PLSDA model and its suitability for use as a standardization tool. The study has also revealed the possibility of herbal products from different sources to exhibit different biological activity. For the practical use, application the proposed method needs to be extended to products from more sources in order to expand the database. Table 2 The VIP rank, values and coefficient regression of PLS model for a-glucosidase inhibitory (AGI), nitric oxide inhibition (NOI), and DPPH radical scavenging (DPPH-RS) activities. Var ID dH Metabolite VIP rank VIP [2] Regression coefficient [2] AGI NOI oxide DPPH-RS 1/IC50 1.28 Fatty acid 1 6.66 7.75 3.62 0.49 1.32 Fatty acid 2 5.69 2.68 5.92 1.96 1.12 Thymoquinone 3 3.98 0.37 5.28 2.07 1.20 Thymol 4 3.28 0.26 4.33 1.69 0.88 Fatty acid 8 2.43 2.24 1.79 0.43 7.08 Thymol 9 2.24 0.34 3.02 1.20 2.04 Fatty acid 10 2.18 0.26 2.68 1.00 2.28 Thymol 13 1.80 1.35 2.77 1.22 2.32 Fatty acid 16 1.66 3.28 0.79 0.68 5.36 Sucrose 20 1.47 1.23 2.28 1.02 5.32 Fatty acid 21 1.46 1.03 2.23 0.98 1.24 Cymene 23 1.37 0.59 1.46 0.49 3.00 Thymoquinone 26 1.22 0.05 1.60 0.62 2.84 Cymene 34 1.08 0.18 1.31 0.48 3.20 Thymol 48 0.87 1.72 0.42 0.36 6.64 Thymoquinone 62 0.75 0.10 1.01 0.40 4.08 Sucrose 64 0.74 1.32 0.10 0.19 4.12 Thymol-b-glucopyranoside 68 0.72 0.41 0.70 0.22 6.56 Thymoquinone 79 0.64 0.11 0.87 0.35 0.64 Cycloartenol 85 0.54 0.03 0.68 0.26 6.92 Thymol 89 0.52 0.04 0.68 0.27 5.12 Thymol-b-glucopyranoside 92 0.48 0.09 0.58 0.21 6.36 Quercetin deriv. 122 0.26 0.02 0.32 0.12 6.40 Kaempferol deriv. 128 0.18 0.04 0.22 0.08 8.08 Kaempferol deriv. 137 0.14 0.12 0.11 0.03 6.80 Kaempferol deriv. 139 0.12 0.05 0.14 0.05 M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 313 3. Experimental 3.1. Chemicals Deuterated chloroform (CDCl3), deuterated methanol-d4 (CH3OH-d4), non deuterated KH2PO4, sodium deuterium oxide (NaOD), tetramethylsilane (TMS), trimethylsilylpropionic acid-d4 sodium salt (TSP) and deuterium oxide (D2O) were supplied by Merck (Darmstadt, Germany). a-Glucosidase (from Saccharomyces cerevisiae), 4-nitrophenyl-a-D-glucopyranoside (pNPG), acarbose, 1,1-diphenyl-2-picrylhydrazyl (DPPH), lipopolysaccharide (LPS), phosphate buffered saline (PBS) and recombinant murine IFN-g were purchased from Sigma Chemical Co. (St. Louis, USA). The cell culture media, Dulbecco’s Modified Eagle’s Medium (DMEM) containing HEPES and L-glutamine (with phenol red and without), as well as penicillin-streptomycin antibiotic solution, foetal bovine serum (FBS), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenylte- trazolium bromide (MTT) and triple Express enzyme were purchased from Gibco/BRL Life Technologies Inc. (Eggenstein, Germany). 3.2. Sample materials Four samples of N. sativa seeds originating from four different origins including Ethiopia, Syria, Yemen and Qasem (S. Arabia) were purchased from the Uthmani Herbal Medicinal Store in Madinah al-Munawarah in June 2012. The samples were kept at 80 C and freeze dried prior to extraction. Another batch of test samples was purchased from the same retailer in February, 2013 and processed following the same protocol. Representative specimens of each sample were deposited at the Herbarium of the Institute of Bioscience, UPM for reference. Fig. 4. The PLS (A) scores and (B) loadings plots. Keys to figure: Et – Ethiopian (orange diamonds); Qa – Qasemi (green triangles); Sy – Syrian (red circles); Ye – Yemeni (blue boxes). 314 M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 3.3. Sample preparation and extraction Exactly 50 mg of each sample was ground using mortar and pestle, and 1 mL of NMR grade deuterated solvent was immediately added. The amount of seeds used was found appropriate on the basis of the detectability of metabolites in the extracts and the cost of deuterated solvents used. The suspension was then transferred into 1.5 mL microcentrifuge eppendorf tubes and ultra-sonicated at room temperature for 30 min. Following this, the suspension was centrifuged at 13,000 rpm for 5 min. The supernatant was pipetted and transferred into 5 mL NMR tube to be ready for analysis. Twelve samples representing three from each origin were extracted, each batch using deuterated chloroform, deuterated methanol and deuterated aqueous methanol (1:1), for the preliminary selection of the best solvent for extraction. Adopting the method established by Kim et al. (2010b), the pH of KH2PO4 buffer in D2O for the aqueous methanol extraction was adjusted to 6.0 using 1.0 M of NaOD. For the development of model, further fifty-four extracts representing thirteen each from the Syrian, Yemeni and Qasemi (S. Arabia), and fifteen from Ethiopian origins were prepared using methanol as the solvent. 3.4. 1 H NMR measurement The 1 H NMR spectra were acquired at 25 C on a Varian Unity INOVA 500 MHz spectrometer (Varian Inc., CA) operating at 499.89 MHz. All spectra were manually phased and baseline corrected. For each sample, 64 scans were recorded using an acquisition time of 220 s, a pulse width of 3.75ms, and a relaxation delay of 1.0 s. The spectral width was adjusted to the range between 1.00 and 20.00 ppm. For all CDCl3 and CD3OD extracts, tetramethylsilane (TMS) was used as an internal standard (concentration of TMS 0.2% in CDCl3 or CD3OD) for chemical shift reference and intensity scaling of all NMR signals. 3-Trimethylsi- lylpropanoic acid (TSP) instead of TMS was used for CD3OD + D2O extracts (at the same concentration) for the same purposes. In the case of all CD3OD and CD3OD + D2O extracts, pre-saturation method was applied after acquiring the 1 H NMR in order to suppress the residual water signal with low power selective irradiation. Additionally, the two dimensional J-resolved and heteronuclear multiple-bond correlation (HMBC) experiments were also carried out to confirm the identity of the constituents. 3.5. a-Glucosidase inhibition assay The a-glucosidase inhibitory assay was conducted according to the method described previously (Deutschländer et al., 2009), with slight modifications. The enzyme (0.03 U/mL) was dissolved in 30 mM Na-phosphate buffer (pH 6.5), and 4-nitrophenyl b-D- glucuronide (p-NPG) in the same buffer was used as the enzyme and substrate solutions, respectively. A stock solution of five milligrams of the extracts in 1 mL of DMSO was prepared, and Fig. 5. Dendogram and heat map of HCA using Ward’s minimum variance method and Euclidean distance. The dendogram is shown on the left side of the heat map [Et – Ethiopian (red), Ye – Yemeni (light blue), Sy – Syrian (blue), and Qa – Qasemi (green)]. The concentration of each metabolite varies from minimum (dark green) to maximum (dark red). These values presented along the the chemical shifts column are the averaged integrals of the binned signals. Major contributing metabolites:thymol dH 7.08, 6.60, 1.20; thymoquinone dH 6.62, 6.56, 2.00, 3.00, 1.12; p-cymene dH 2.84, 2.28; fatty acids dH 1.28, 1.32, cycloartenol dH 1.64, 0.88. M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 315 250mg/mL of quercetin was dissolved in 1 mL buffer to be used as a positive control. Subsequently, the stock solution was further diluted with 30 mM phosphate buffer (pH 6.5) to achieve the final concentrations of ranging from 2.5 to 0.08 mg/mL. The test sample (10mL) in phosphate buffer (or phosphate buffer only, for blank control) with concentrations of 200,100, 50, 25,12.5, 6.25, 3.12mg/ mL and 15mL of the enzyme were diluted with 100mL of a 30 mM phosphate buffer (pH 6.5) and pre-incubated in 96-well plates at room temperature for 10 min. The reaction was initiated by adding 75mL of p-NPG (50 mM concentration in 50 mM phosphate buffer, pH 6.5). The mixture was incubated for additional 15 min under the same conditions, followed by the addition of 50mL of 2 M glycine (pH 10) to stop the reaction. The enzymatic activity was evaluated spectrophotometrically by recording the absorbance at 405 nm, on a 96 microplate spectrophotometer (SPECTRAmax PLUS, USA). The inhibitory activity of the sample was calculated by the following formula: % Inhibition of sample 1⁄4 An Bs An 100 where, An is the absorbance of the control and Bs is the absorbance of the sample. 3.6. Nitric oxide (NO) inhibition assay The RAW 264.7 murine macrophage cells, which were obtained from American Type Culture Collection (ATCC, Rockville, MD) were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) containing phenol red, HEPES and L-glutamine, were supplemented with 10% foetal bovine serum (FBS) and 1% antibiotic solution (Gibco/BRL), and cultured in the plastic culture flasks under 5% CO2 atmosphere at 37 C. After 3–4 days of culture, the cells were collected from the flask by using TrypLETM Express enzyme, followed by centrifuga- tion at 1000 rpm for 10 min at 4 C. The medium was then discarded and the cells were suspended with fresh DMEM (without phenol red) containing HEPES, L-glutamine and other similar supplements. The cells were then counted and the viability was determined by implementing the standard trypan blue cell counting method. The cell concentration was adjusted to 1 106 cells/mL in the same medium. The cells were then cultured in the 96 well plates, in the previously mentioned media (50mL) with the addition of triggering agents comprising 200 U/mL of recombinant murine IFN-g and 10mg/mL lipopolysaccharide. 50mL of the serially diluted extracts (10.0, 5.0, 2.50, 1.25, 0.63, 0.31, and 0.156 mg/mL) containing 0.8% DMSO were loaded into the wells to yield a final concentration of DMSO at 0.4% per well. The analyses were carried out in six replicates and the cells were incubated in a humidified incubator for 17 h at 37 C in a 5% CO2 atmosphere. For the control, the cells were cultured in media containing triggering agents and 0.4% DMSO (Abas et al., 2006). The stable nitric oxide conversion product, nitrite (NO2 ) was measured using Griess reagent. After 17 h of incubation, 50mL aliquots were taken from the supernatant of the cultured cells and incubated with an equal volume of Griess reagent (1% sulfanilamide, 0.1% N-L-naphthyle- thylenediamine dihydrochloride, 2.5% H3PO4) at room tempera- ture for 15 min. The absorbance was taken at 550 nm using Spectramax Plus (Molecular Devices) UV/Vis microplate reader. The percentage of inhibition was then calculated for all samples. Fresh culture medium was taken as a blank in every experiment (Abas et al., 2006). The cytotoxicity assay was performed to eliminate false positive results due to the cytotoxic effects. In this experiment, the pale yellow 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bro- mide (MTT) substrate is converted into dark blue insoluble formazan product after being reduced by living cells. The reduction only takes place in active mitochondria and thus, MTT will not be Table 3 Classification of membership list for the discriminant analysis models. Predicted Y value for the dummya Origin Syria Yemen Ethiopia Qasem Training Syria 0.99 0.03 0.02 0.05 Syria 1.02 0.08 0.00 0.09 Syria 0.77 0.15 0.02 0.09 Syria 1.00 0.04 0.01 0.06 Syria 1.13 0.15 0.03 0.02 Syria 0.76 0.05 0.05 0.15 Syria 0.67 0.25 0.04 0.13 Syria 1.19 0.16 0.00 0.03 Syria 0.79 0.08 0.01 0.12 Syria 0.95 0.11 0.00 0.06 Yemen 0.02 0.97 0.09 0.04 Yemen 0.17 1.01 0.02 0.18 Yemen 0.09 0.93 0.01 0.14 Yemen 0.19 0.88 0.04 0.02 Yemen 0.19 0.88 0.04 0.03 Yemen 0.10 1.15 0.02 0.02 Yemen 0.10 0.92 0.02 0.04 Yemen 0.27 0.81 0.07 0.16 Yemen 0.09 0.96 0.03 0.09 Yemen 0.12 1.13 0.04 0.03 Ethiopia 0.09 0.10 0.92 0.11 Ethiopia 0.02 0.05 1.03 0.00 Ethiopia 0.08 0.07 0.97 0.04 Ethiopia 0.05 0.09 0.93 0.07 Ethiopia 0.14 0.06 1.06 0.14 Ethiopia 0.06 0.06 1.07 0.06 Ethiopia 0.18 0.00 0.95 0.12 Ethiopia 0.04 0.02 1.04 0.06 Ethiopia 0.16 0.06 0.96 0.14 Ethiopia 0.05 0.06 1.01 0.00 Qasem 0.03 0.00 0.01 0.98 Qasem 0.06 0.03 0.01 0.96 Qasem 0.04 0.05 0.04 1.03 Qasem 0.04 0.04 0.02 1.03 Qasem 0.07 0.04 0.01 1.10 Qasem 0.06 0.01 0.01 0.93 Qasem 0.08 0.10 0.00 0.82 Qasem 0.20 0.02 0.03 0.79 Qasem 0.11 0.13 0.02 1.01 Qasem 0.02 0.03 0.01 1.02 Testing Syria 0.90 0.13 0.04 0.06 Syria 0.94 0.08 0.02 0.00 Syria 1.10 0.07 0.01 0.02 Yemen 0.14 1.22 0.04 0.04 Yemen 0.33 0.89 0.08 0.13 Yemen 0.41 0.98 0.07 0.45 Ethiopia 0.06 0.08 0.96 0.02 Ethiopia 0.35 0.05 0.82 0.13 Ethiopia 0.38 0.06 1.02 0.30 Ethiopia 0.17 0.18 0.96 0.05 Ethiopia 0.21 0.07 1.08 0.22 Qasem 0.08 0.07 0.03 1.12 Qasem 0.27 0.16 0.03 0.54 Qasem 0.01 0.04 0.01 1.02 Testing with new sample Yemen_u 0.29 1.20 0.08 0.17 Yemen_u 0.51 1.09 0.02 0.43 Yemen_u 0.17 0.66 0.03 0.20 Yemen_u 0.06 0.87 0.08 0.27 Yemen_u 0.15 0.89 0.06 0.32 Qasem_u 0.05 0.55 0.10 0.40 Qasem_u 0.15 0.43 0.06 0.36 Qasem_u 0.31 0.39 0.06 0.25 Qasem_u 0.01 0.58 0.13 0.27 Qasem_u 0.17 0.63 0.07 0.46 a The valuesdisplayedinthe classificationlistfor thediscriminant analysismodels are the prediction values of training and testing data sets. Members with predicted values below 0.35 are not belonging to the class, predicted values between 0.35 and 0.65 are borderline (values in the Table are shown in bold italic letters), and predicted values above 0.65 are belonging to the class (values in the Table are shown in bold letters). 316 M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 reduced by dead cells. After the removal of the culture medium, 100mL of complete DMEM was loaded into the wells. Following this, 20mL of 5 mg/mL solution of MTT in phosphate buffered saline (PBS) 7.2 was added. The cells were incubated at 37 C in 5% CO2 atmosphere for 4 h. The medium was then removed and the formazan salt was dissolved in DMSO. The absorbance was read at 570 nm. The absorbance of formazan salt in control (untreated cells) was taken as 100% viability, and the percentages of dead cells were calculated comparatively to the control group (Abas et al., 2006). 3.7. DPPH radical scavenging activity assay The DPPH radical scavenging activity assay of N. sativa seeds was performed in triplicates, in the 96-well microplate following the method described previously (Abas et al., 2006). A stock solution at 4 mg/mL concentration of the sample was prepared. The test samples were prepared in 7 dilutions starting from 2000 down to 31.25mg/mL concentrations (in MeOH) with the final volume of each well reached 100mL. A set of wells containing medium only were designated as a blank for background correction. Finally, 5mL of DPPH solution (2.5 mg/mL of DPPH in MeOH) was added into each well and incubated in the dark for 30 min. The absorbances were then read at 517 nm with a microplate reader (SPECTRAmax PLUS). The percentage of inhibi- tion was calculated as [Ab As/Ab] 100 where Ab is the absor- bance of the reagent blank and As is the absorbance of the samples. The results were expressed as IC50 (the half maximal inhibitory concentration). 3.8. Data analysis The 1 H NMR spectra were automatically converted to ASCII file using Chenomx software (v. 5.1, Alberta, Canada). Spectral intensities were scaled to TMS (or TSP) and binned into regions of 0.04 ppm width for the spectral region of dH 0.52–10.00, giving a total of 233 integrated regions per NMR spectrum. The averaged binned integral of the 1 H NMR data were then subjected to multivariate data analysis (PCA, PLS, HCA, and PLSDA), performed using SIMCA-P+ version 13.0 (Umetrics AB, Umeå, Sweden). PCA was applied on the 1 H NMR spectra of the extracts of different solvent systems, and selection of the best system was made based on the class separation of samples and the value of the total variance for the first two principal components (PCs). Subsequent- ly, the PLS model was developed to obtain correlations between the phytochemical profile of samples and their biological activities from the three assays. The HCA was used to observe the metabolic proximities between samples and their potential relationship. Heat map was generated from the HCA using MetaboAnalyst 2.5, a freely available metabolomics data analytical tool (http://www. metaboanalyst.ca), which allows a closer investigation of the metabolite variations of the samples (Xia and Wishart, 2010). Finally, PLS-DA was performed for the discrimination of samples originating from different geographical regions. Pareto scaling was applied in all the analyses. Conflict of interest The authors have declared that there are no competing interests exist in this project. Acknowledgment The authors (NHL and BYS) wish to thank the Al-Moalim BinLaden Chair for Scientific Miracles of Prophetic Medicine, Scientific Chairs Unit, Taibah University, P.O. Box 30001, Madinah al Munawarah 41311, Saudi Arabia for the financial support for this project. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. phytol.2015.07.012. References Abas, F., Lajis, N.H., Israf, D., Khozirah, S., Umi Kalsom, Y., 2006. Antioxidant and nitric oxide inhibition activities of selected Malay traditional vegetables. Food Chem. 95, 566–573. Ahmed, A.A., Jakupovic, J., 1990. Sesqui- and monoterpenes from Jasonza Montana. Phytochemistry 29, 3658–3661. Allouche, O., Tsoar, A., Kadmon, R., 2006. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol 43, 1223–1232. Ankli, A., Reich, E., Steiner, M., 2008. Rapid high-performance thin-layer chromatographic method for detection of 5% adulteration of black cohosh with Cimicifuga foetida,C. heracleifolia, C. dahurica, or C. americana. J. AOAC Int. 91, 1257–1264. Atta, M.B., 2003. Some characteristics of nigella (Nigella sativa L.) seed cultivated in Egypt and its lipid profile. Food Chem. 82, 63–68. Atta-ur-Rahman, Malik, S., Hasan, S., Choudhary, M.I., Chao-Zhou Ni, C., Clardy, J., 1995. Nigellidine – a new indazole alkaloid from the seeds of Nigella sativa. Tetrahedron Lett. 36, 1993–1996. Badary, O.A., Taha, R.A., Gamal El-Din, A.M., Abdel-Wahab, M.H., 2003. Thymoquinone is a potent superoxide anion scavenger. Drug Chem. Toxicol. 26, 87–98. Bourgou, S., Pichette, A., Marzouk, B., Legault, J., 2010. Bioactivities of black cumin essential oil and its main terpenes from Tunisia. S. Afr. J. Bot. 76, 210–216. Table 4 The contingency table and the Cohen’s Kappa analysis for reliability of PLSDA from testing data sets and introduction of new samples. PLSDA class Producer accuracy Origin Syria Yemen Ethiopia Qaesim Total Percent correct Omission error (%) Reference class Syria 3 0 0 0 3 100 0 Yemen 0 8 0 0 8 100 0 Ethiopia 0 0 5 0 5 100 0 Qaseim 0 5 0 3 8 37.50 62.50 Total 3 13 5 3 24 Users accuracy Percent correct 100 61.54 100 100 79.17 Commission error (%) 0 38.46 0 0 Agreement 3 8 5 3 19 By chance 0.38 4.33 1.04 1 6.75 Cohen’s Kappa, k 0.71 M. Maulidiani et al. / Phytochemistry Letters 13 (2015) 308–318 317 Buriani, A., Garcia-Bermejod, M.L., Bosisio, E., Xu, Q., Li, H., Dong, X., Simmonds, M.S. J., Carrara, M., Tejedor, N., Lucio-Cazana, J., Hylands, P.J., 2012. Omic techniques in systems biology approaches to traditional Chinese medicine research: present and future. J. Ethnopharmacol. 140, 535–544. Carletta, J., 1996. Assessing agreement on classification tasks: the kappa statistic. Comput. Ling. 22, 249–254. Cheikh-Rouhou, S., Besbes, S., Lognay, G., Blecker, C., Deroanne, C., Attia, H., 2008. Sterol composition of black cumin (Nigella sativa L.) and Aleppo pine (Pinus halepensis Mill.) seed oils. J. Food Compos. Anal. 21, 162–168. Cordell, G.A., 2011. Phytochemistry and traditional medicine – a revolution in process. Phytochem. Lett. 4, 391–398. de Sousa, D.P., Nóbrega, F.F.F., Santos, C.C.M.P., Benedito, R.B., Vieira, Y.W., Uliana, M. P., Brocksom, T.J., de Almeida, R.N., 2011. Anticonvulsant activity of thymoquinone and its structural analogues. Rev. Bras. Farmacogn. 21, 427–431. Deutschländer, M., Van de Venter, M., Roux, S., Louw, J., Lall, N., 2009. Hypoglycaemic activity of four plant extracts traditionally used in South Africa for diabetes. J. Ethnopharmacol. 124, 619–624. El-Mahmoudy, A., Matsuyama, H., Borgan, M., Shimizu, Y., El-Sayed, M., Minamoto, N., Takewaki, T., 2002. Thymoquinone suppresses expression of inducible nitric oxide synthase in rat macrophages. Int. Immunopharmacol. 2, 1603–1611. Eriksson, L., Johansson, E., Kettaneh-Wold, N., Trygg, J., Wikstrom, C., Wold, S., 2006. Multi- and Megavariate Data Analysis Part I: Basic Principles and Applications. Umetrics, Inc., Umea. Ernst, E., 2004. Risks of herbal medicinal products. Pharmacoepidemiol. Drug Saf. 13, 767–771. Farag, M.A., Gad, H.A., Heiss, A.G., Wessjohann, L.A., 2014. Metabolomics driven analysis of six Nigella species seeds via UPLC-qTOF-MS and GC–MS coupled to chemometrics. Food Chem. 151, 333–342. Fararh, K., Atoji, Y., Shimizu, Y., Takewaki, T., 2002. Isulinotropic properties of Nigella sativa oil in Streptozotocin plus Nicotinamide diabetic hamster. Rech. Vet. Sci. 73, 279–282. Gali-Muhtasib, H., El-Najjar, N., Schneider-Stock, R., 2006. The medicinal potential of black seed (Nigella sativa) and its components. In: Khan, M.T.H., Ather, A. (Eds.), Lead Molecules from Natural Products. Elsevier, 133-153. Georgiev, M.I., Ali, K., Alipieva, K., Verpoorte, R., Choi, Y.H., 2011. Metabolic differentiations and classification of Verbascum species by NMR-based metabolomics. Phytochemistry 72, 2045–2051. Harzallah, H.J., Kouidhi, B., Flamini, G., Bakhrouf, A., Mahjoub, T., 2011. Chemical composition, antimicrobial potential against cariogenic bacteria and cytotoxic activity of Tunisian Nigella sativa essential oil and thymoquinone. Food Chem. 129, 1469–1474. Jordan, S.A., Cunningham, D.G., Marles, R.J., 2010. Assessment of herbal medicinal products: challenges, and opportunities to increase the knowledge base for safety assessment. Toxicol. Appl. Pharmacol. 243, 198–216. Kim, H.K., Khan, Saifullah, Wilson, S., Kricun, E.G., Meissner, S.D.P., Goraler, A., Deelder, S., Choi, A.M., Verpoorte, Y.H., 2010a. Metabolic classification of South American Ilex species by NMR-based metabolomics. Phytochemistry 71, 773– 784. Kim, H.K., Choi, Y.H., Verpoorte, R., 2010b. NMR-based metabolomic analysis of plants. Nat. Protoc. 5, 536–549. Knothe, G., Kenar, J.A., 2004. Determination of the fatty acid profile by 1 H-NMR spectroscopy. Eur. J. Lipid. Sci. Technol. 106, 88–96. Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics 159–174. Maulidiani, H., Khatib, A., Shaari, K., Abas, F., Shitan, M., Kneer, R., 2012. Discrimination of three pegaga (Centella) varieties and determination of growth-lighting effects on metabolites content based on the chemometry of 1 H nuclear magnetic resonance spectroscopy. J. Agric. Food Chem. 60, 410–417. Meddah, B., Ducroc, R., Faouzi, M.E.A., Eto, B., Mahraoui, L., Benhaddou-Andaloussi, A., Martineau, L.C., Cherrah, Y., Haddad, P.S., 2009. Nigella sativa inhibits intestinal glucose absorption and improves glucose tolerance in rats. J. Ethnopharmacol. 121, 419–424. Mehta, B.K., Mehta, P., Gupta, M., 2009. A new naturally acetylated triterpene saponin from Nigella sativa. Carbohydr. Res. 344, 149–151. Mehta, B.K., Sharma, U., Agrawal, S., Pandit, V., Joshi, N., Gupta, M., 2008. Isolation and characterization of new compounds from seeds of Nigella sativa. Med. Chem. Res. 17, 462–473. Merfort, I., Wray, V., Barakat, H.H., Hussein, S.A.M., Nawwar, M.A.M., Willuhn, G., 1997. Flavonol triglycosides from seeds of Nigella Sativa. Phytochemistry 46, 359–363. Mohamadin, A.M., Sheikh, B.Y., Abd El-Aal, A.A., Elberry, A.A., Al-Abbasi, F.A., 2010. Protective effects of Nigella sativa oil on propoxur-induced toxicity and oxidative stress in rat brain regions. Pest. Biochem. Physiol. 98, 128–134. Nergiz, C., Otles, S., 1993. Chemical composition of Nigella sativa L. seeds. Food Chem. 48, 259–261. Ramadan, M.F., 2007. Nutritional value, functional properties and nutraceutical applications of black cumin (Nigella sativa L.): an overview. Int. J. Food Sci. Technol. 42, 1208–1218. Ramadan, M.F., Mörsel, J., 2003. Analysis of glycolipids from black cumin (Nigella sativa L.), coriander (Coriandrum sativum L.) and niger (Guizotia abyssinica Cass.) oil seeds. Food Chem. 80, 197–204. Razmovski-Naumovski, V., Tongkao-on, W., Kimble, B., Qiao, V.L., Beilun, L., Li, K.M., Roufogalis, B., Depo, Y., Meicun, Y., Li, G.Q., 2010. Multiple chromatographic and chemometric methods for quality standardisation of Chinese herbal medicines. Mode. Tradit. Chin. Med. Mater. Med. 12, 99–106. Schwarz, K., Ernst, H., 1996. Evaluation of antioxidative constituents from thyme. J. Sci. Food Agric. 70, 217–223. Su, C., Hsu, C., Ng, L., 2013. Inhibitory potential of fatty acids on key enzymes related to type 2 diabetes. Biofactors 39, 415–421. Uzuner, H., Bauer, R., Fan, T., Guo, D., Dias, A., El-Nezami, H., Efferth, T., Williamson, E. M., Heinrich, M., Robinson, N., Hylands, P.J., Hendry, B.M., Cheng, Y., Xu, Q., 2012. Traditional Chinese medicine research in the post-genomic era: good practice, priorities, challenges and opportunities. J. Ethnopharmacol. 140, 458–468. WHO, 2012. Traditional Medicine: Regional Progress in Traditional Medicine 2010– 2011. World Health Organization Western Pacific Region, Hong Kong. Xia, J., Wishart, D.S., 2010. MSEA: a web-based tool to identify biologically meaningful patterns in quantitative metabolomic data. Nucleic Acids Res. 38, W71–W77. Yamazaki, M., Nakajima, J., Yamanashi, M., Sugiyama, M., Makita, Y., Springob, K., Awazuhara, M., Saito, K., 2003. Metabolomics and differential gene expression in anthocyanin chemo-varietal forms of Perilla frutescens. Phytochemistry 62, 987–995. Zahmanov, G., Alipieva, K., Simova, S., Georgiev, M.I., 2015. Metabolic differentiations of dwarf elder by NMR-based metabolomics. Phytochem. Lett. 11, 404–409. Zribi, I., Omezzine, F., Haouala, R., 2014. Variation in phytochemical constituents and allelopathic potential of Nigella sativa with developmental stages. S. Afr. J. Bot. 94, 255–262.