Chemotaxonomic Study of Medicinal Taxus Species with Fingerprint and Multivariate Analysis

• Abstract
• Abbreviations
• Introduction
• Materials and Methods
• Reagents and chemicals
• Plant materials
• Apparatus and analytical conditions
• Fingerprinting method validation
• Data extraction and processing
• Supporting information
• Results and Discussion
• Acknowledgements
• References

Abstract

Species delimitation in Taxus has been controversial and it is very difficult to distinguish yew materials by their morphological characters. In this paper, a valid HPLC fingerprinting method coupled with multivariate analysis was used to define a framework for Taxus species identification and classification. Fingerprint-based similarity was employed for a chemotaxonomic study by hierarchical clustering analysis (HCA) and principal component analysis (PCA). Based on the PCA loadings, twelve chemical constituents were selected as chemotaxonomic markers which can be used to establish a more practical classification. Finally, eight studied species could be divided into six well-supported groups and most samples can be assigned to the correct species. Additionally, twelve markers were tentatively identified by LC/MS.

Abbreviations

10-DAB: 10-deacetylbaccatin III

10-DAP: 10-deacetylpaclitaxel

10-DAXP: 10-deacetyl-7-xylosylpaclitaxel

9-DHB: 9-dihydro-13-acetylbaccatin III

HCA: hierarchical clustering analysis

ITS: internal transcribed spacer

PCA: principal component analysis

Introduction

Yew trees, species of Taxus L. (Taxaceae), as sources of the anticancer agent paclitaxel have attracted considerable attention and are under intense investigation [1], [2]. Nowadays, paclitaxel is mainly produced by semi-synthesis from several precursors such as 10-deacetylbaccatin III (10-DAB) [3]. Paclitaxel and its precursors belong to a group of typical secondary metabolites named the taxane diterpenoids or taxoids which can be isolated from renewable resources such as yew needles and cell culture, but the distribution and level are very variable with species [4], [5], [6]. Correct Taxus species identification is very important for Good Agricultural Practice (GAP) and Good Manufacturing Practice (GMP) [7]. Unfortunately, Taxus is a problematic genus. Ten Taxus species are generally recognized (see Supporting Information, Table 1S), but the taxonomy has been difficult and controversial, as most yews look very similar and there are only a few reliable morphological characteristics for species diagnosis. Recently, some articles reported that Taxus species can be carefully separated by using several molecular markers [8], [9], but they are inapplicable for yew extracts. Furthermore, the gene variations cannot represent the variations at the metabolite level which are closely related to the manufacturing process of taxanes. Novel classifications based on metabolic analysis are thus highly desirable.

During the past twenty years, some substantial quantitative and qualitative variations among different Taxus species have been discovered using modern analytical techniques [4], [6], [10], [11], [12]. However, to date there is no practical chemical classification that can be applied to yew species identification. It must be noted that systemic analysis of yew constituents is a big challenge, due to the numerous constituents from different classes, and varying metabolite levels caused by many non-genetic factors such as developmental stage, climate, elevation, slope exposure, and others. In this case, a holistic approach of fingerprint analysis, profile similarity-based clustering, and choice of taxonomic markers capable of capturing the greatest chemical variations were proposed for Taxus classification in this study. As a rapid, cost-efficient and popular analysis method, HPLC fingerprinting has been regarded as the first choice for medicinal plant identification and quality control [13], [14], [15]. Herein, a valid HPLC fingerprinting method that allows analysis of samples from different Taxus species is developed and applied to yew classification.

The major goals of this study were to investigate the chemical diversity among various Taxus species and to define a framework for the classification of these important medicinal species. Thirty samples representing eight Taxus species were collected and analyzed, and the fingerprint-based data were extracted and processed by hierarchical clustering analysis (HCA) and principal component analysis (PCA). Specifically, we aimed to construct a practical classification for yew samples based on several typical chemical markers by a combination of fingerprint and multivariate analyses.

Materials and Methods

Reagents and chemicals

Millipore water, HPLC grade acetonitrile and methanol were used throughout; other reagents were of analytical grade. All taxane standards were purchased from Shanghai Jinhe Bio-technology Co. Ltd, 10-deacetylbaccatin III (10-DAB, ≥ 98 %), 10-deacetylpaclitaxel (10-DAP, ≥ 98 %) and paclitaxel (≥ 98 %) were used throughout; other standards (≥ 95 %) used in this study included 7-epi-10-decetylbaccatin III (epi-DAB), 9-dihydro-13-acetylbaccatin III (9-DHB), taxinine M, 7-epi-10-decetylcephalomannine, 7-epi-10-decetylpaclitaxel and 10-deacetyl-7-xylosylpaclitaxel (10-DAXP).

Plant materials

One commercial clipping extract (MA1) and 29 needle specimens from eight Taxus species were collected ([Table 1]). For each species, two or more individuals were sampled to check intraspecific variation. All specimens were deposited in our lab and carefully authenticated by Dr. Da-Cheng Hao, based on morphological characters and molecular markers including 18S rDNA and internal transcribed spacer (ITS) by comparison with literature data. Sample extraction and preparation are described in the Supporting Information.

Apparatus and analytical conditions

A Shimadzu HPLC system equipped with a quaternary pump, a photodiode array detector (DAD) and an autosampler was used for HPLC fingerprinting; an ODS column (4.6 mm × 200 mm × 5 μm, Kromasil) was used as solid phase, the mobile phase consisted of water (A) and CH₃CN (B), and the gradient elution had the following profile: 0 - 30 min, 75 - 55 % A; 30 - 50 min, 55 % A. The flow rate was 1.0 mL/min and the column temperature was room temperature; the injection volume was kept at 10 μL; scan wavelength was set from 190 to 370 nm and the detection wavelength was set at 227 nm.

Mass detection was performed on an LC-MS instrument equipped with a Waters SQ detector via an electrospray complex ionization (ESCi) interface. The LC effluent was introduced into the ESCi source with a split ratio of 1 : 10. Detection was carried out in both negative and positive ion modes from m/z = 100 to 1000; the cone voltage was set at 40 V in order to obtain some important fragment ions. Other MS detection conditions were as follows: capillary voltage, 3.5 kV; source temperature, 120 °C; desolvation temperature, 350 °C; and desolvation gas flow was 800 L/h.

Fingerprinting method validation

The precision was determined by six consecutive injections of sample 1 and relative standard deviations (RSDs) of retention times and peak areas of all selected peaks were below 0.2 % and 1.1 %, respectively. The reproducibility was performed with six sample solutions prepared from one batch of yew needles, the RSDs of retention times and the peak area percentages were both less than 2.9 %. In addition, the intra-day reproducibility was performed on the basis of analyzing 12 repeated analyses at four concentration points, the RSDs of retention times and peak area percentages were 0.5 % and 3.2 %, respectively. The inter-day reproducibility was also evaluated and the results showed acceptable RSD values (<5 %). The limit of detection (LOD) and limit of quantification (LQD) were determined through analysis of three taxane standard solutions (10-DAB, 10-DAP and paclitaxel), the LOD were 4.0, 4.0 and 8.0 ng, respectively, and LQD ranged from 15 to 30 ng. The stability test indicated that the sample solutions were stable during three days at room temperature. The recovery, evaluated by addition of known amounts of three standard solutions (10-DAB, 10-DAP and paclitaxel) to sample 1, provided satisfying results. These results indicate that the fingerprinting method is valid and applicable.

Data extraction and processing

The area percentages (the area ratio of each selected peak to the total chromatographic peaks between 10 - 50 min) of 33 characteristic peaks were extracted as raw data while the total peak area was considered as 100. A 33 × 30 matrix was thus obtained, and then logarithm transformed before multivariate analysis. Hierarchical clustering analysis (HCA) and principal component analysis (PCA) were performed using SPSS software (SPSS for Windows 11.0, SPSS Inc.). A cluster method called average linkage between groups was applied and Pearson correlation was selected as measurement. The PCA plot was drawn by MATLAB 7.0.

Supporting information

The information on characteristic peaks and the principal component loadings are available as Supporting Information.

Results and Discussion

Traditional chemotaxonomic and chemosystematic studies are frequently used to infer relationships among plant taxa, by using the average concentration of several pre-selected compounds. However, they could not be used to examine the variations within species and may result in wrong conclusions in cases where the intraspecific variation is large [16], [17], [18]. In contrast, profile-based classification can investigate variations within and among species by comparison of fingerprints [19]. Moreover, the fingerprint similarity-based taxonomy, which relies on the ratio of selected constituents, can improve the misclassifications caused by large quantitative differences.

All samples were analyzed using the above mentioned fingerprinting method. The chromatographic region from 10 to 50 min was selected for further studies; due to the peaks eluting within the first 10 minutes representing highly polar components, their signals were affected by solvent partition during sample preparation. Thirty-three peaks within the studied region (10 - 50 min) were selected as characteristic peaks since they had relatively large areas. Most of the selected peaks have the characteristic absorption of taxanes (227 nm) (see supporting Information, Table 2S) [20]. The intraspecific variations were examined using a series of T. mairei samples (1 - 9) collected from trees grown under different developmental and environmental conditions. Consistent profiles were observed and no obvious qualitative difference was noticed. In contrast, fingerprints of different species varied significantly except for T. mairei and T. chinensis ([Fig. 1]). Only three common peaks were present and designated as 10-DAB, 10-DAP, and paclitaxel (retention times were 12.9, 35.4 and 46.6 min, respectively). But the level of these three peaks varied considerably both within and among species. It should also be noted that quantitative differences of most taxoids within species were significant and they may vary up to 400 % [4], [21], [our unpublished data]. It is evident that chemotaxonomic studies based on the average or absolute concentration of several predefined taxanes would not yield reliable results. Furthermore, the predefined compounds cannot represent the overall chemical information, and some predefined compounds may not be ideal chemotaxonomic markers. Thus, the fingerprint-based data were used in this present work. The area percentages of all selected peaks were extracted as raw data because there was no one constituent serving as reference compound, which should be abundant with a comparatively stable concentration in each of the studied species.

Two popular clustering algorithms, hierarchical clustering analysis (HCA) and principal component analysis (PCA), were used to investigate the similarity of fingerprints [22], [23], [24]. [Fig. 2] illustrates the HCA result. It is apparent that intra-species samples are more similar than inter-species samples. T. yunnanensis (12 - 14) and T. wallichiana (15 - 17) clearly fell into separate clusters while T. mairei (1 - 9) and T. chinensis (10 and 11) were grouped into one cluster due to their similar chemotypes, indicating a close relationship between T. mairei and T. chinensis. T. × media and its parental species (T. cuspidata and T. baccata) form two clusters: three T. × media samples (28 - 30) were grouped into one cluster with the subgroup of T. cuspidata (23 - 25), but the other two T. × media samples (26 and 27) were found to be closer to the subgroup of T. baccata (18 and 19). The random distribution of T. × media indicated that it had at least two chemotypes, which can contribute to the repeated hybridization among T. × media and its parental species [9]. In addition, T. canadensis samples (20 - 22) formed another cluster which was located near the cluster of T. baccata. HCA can divide all studied species into two large groups, one consisting of four South Asian species and displaying three well separated clusters, and the other including T. canadensis, T. × media and its parental species. However, the divergences among several species were not satisfying, which may be attributed to some indistinctive constituents which can negatively affect the HCA results. It provided the motivation to find some valuable taxonomic markers for yew classification.

In order to find some characteristic constituents which can serve as ideal taxonomic markers, PCA was conducted. The plot based on the first three principal components (3PCs, 66 % of variance explained) is presented in [Fig. 3]. In contrast to HCA, PCA can divide all samples into five groups; but the first three PCs were not enough to distinguish samples of T. wallichiana and T. yunnanensis. The first six PCs (85.1 % of variance explained) were thus used for data reduction. After principal component analysis, the factor loadings of each peak can delineate the relevance of this peak to species classification [25], [26], [27]. Twelve peaks corresponding to the first three loadings of six PCs (Supporting Information, Table 2S) were selected and used to form a reduced data set. The reduced matrix was then implemented in HCA and better results were displayed ([Fig. 4]). Six well-supported clusters were obtained, while samples of T. × media and T. baccata form one big cluster which is near to T. cuspidata and T. canadensis. It is obvious that the divergences among clusters were three times more than within species. The large divergences ensured the species discriminative power of this method. These twelve characteristic peaks ([Table 2]) were thus recognized as potential taxonomic markers.

Finally, LC/MS with in-source CID was used to assign these chemical markers based on their molecular and fragment ions ([Table 2]). Seven markers including 10-DAB, epi-DAB, 9-DHB, 10-DAXP, taxinine M, 7-epi-10-deacetylcephalomannine and 7-epi-10-deacetylpaclitaxel were identified by comparison of their retention times, UV and mass spectra with authentic standards. Two chemical markers were tentatively identified as phloroglucinol glucoside and taxine B (10.9 and 12.4 min, respectively) based on their mass data ([Fig. 5]); but the other three markers could not be determined without standard compounds, due to the coexistence of isomers in Taxus needles. It is clear that most markers are prominent or typical constituents of a given or several species, e. g., taxinine M (30.8 min) and 10-DAXP (32.3 min) are the two most remarkable peaks of T. mairei and T. chinensis; 9-DHB (27.8 min) is unique to T. canadensis; taxine B (12.4 min) and 10-DAB (12.9 min) are two predominant constituents of T. baccata. Additionally, the peak at 45.2 min is a typical component of T. × media and T. baccata, and this marker is helpful to separate T. cuspidata from T. × media and T. baccata.

During the past 40 years, interspecific relationships within Taxus have been implied in various taxonomic approaches including morphological, geographic classifications and molecular identification. Although chemical profiles are often considered to be inadequate for phylogenetic inference, they can provide powerful evidence for species relationship inferences [28]. A controversial point in taxonomy is the uncertain relationships among three endemic species of south China, namely T. chinensis, T. yunnanensis and T. mairei [2]. This study is able to make clear delimitations among these three endemic species. Moreover, fingerprint-based chemotaxonomy also provides similar results with the ITS-based molecular classification [8], e. g., T. mairei can be considered as a variety or subspecies of T. chinensis; a close relationship exists between T. × media and its parental species; T. canadensis is distant to other species. These results indicate that the proposed chemotaxonomy is useful for Taxus species delimitation and species relatedness study.

Acknowledgements

We thank Dr. Stephane Bailleul of the Montreal Botanical Garden, Canada, for providing T. canadensis sample (voucher 1960 - 2000). The authors are also grateful to Professor Qiu Minghua of the Kunming Institute of Botany, China, Mr. Liu Yonghua of the Chayu forestry bureau, Tibet, China, Mr. Zhang Xianzhong, Hubei, China, and other mentioned plantations ([Table 1]) for collecting material of this uncommon plant. This project was supported by the 973 program (2007CB707802) of the Ministry of Science and Technology of China, by the Fund of the Dalian Science & Technology Bureau (2007E11SF0052) and by a DICP Innovation Fund of the Chinese Academy of Sciences.

References

• 1 Baloglu E, Kingston D GI. The taxane diterpenoids.  J Nat Prod. 1999;  62 1448-72
  CrossrefPubMedSearch in Google Scholar
• 2 Appendino G. The phytochemistry of the yew tree.  Nat Prod Rep. 1995;  121 349-60
  PubMedSearch in Google Scholar
• 3 Denis J N, Greene A E, Guénard D, Guéritte-Vorgrlrin F, Mangatal L, Potier P. A highly efficient, practical approach to natural taxol.  J Am Chem Soc. 1988;  110 5917-9
  CrossrefPubMedSearch in Google Scholar
• 4 van Rozendaal E, Lelyveld G, van Beek T. Screening of the needles of different yew species and cultivars for paclitaxel and related taxoids.  Phytochemistry. 2000;  53 383-9
  CrossrefPubMedSearch in Google Scholar
• 5 Navia-Osorio A, Garden H, Cusidó R M, Palazón J, Alfermann A W, Piñol M T. Production of paclitaxel and baccatin III in a 20-L airlift bioreactor by a cell suspension of Taxus wallichiana.  Planta Med. 2002;  68 336-40
  Thieme ConnectPubMedSearch in Google Scholar
• 6 Poupat C, Hook I, Guéritte F, Ahond A, Guénard D, Adeline M T. et al . Neutral and basic taxoid contents in the needles of Taxus species.  Planta Med. 2000;  66 580-4
  Thieme ConnectPubMedSearch in Google Scholar
• 7 Khan I A. Issues related to botanicals.  Life Sci. 2006;  78 2033-8
  CrossrefPubMedSearch in Google Scholar
• 8 Li J, Davis C C, Tredici P D, Donoghue M J. Phylogeny and biogeography of Taxus (Taxaceae) inferred from sequences of the internal transcribed spacer region of nuclear ribosomal DNA.  Harvard Papers Bot. 2001;  6 267-74
  PubMedSearch in Google Scholar
• 9 Collins D, Mill R R, Moller M. Species separation of Taxus baccata, T. canadensis, and T. cuspidate (Taxaceae) and origins of their reputed hybrids inferred from RAPD and cpDNA data.  Am J Bot. 2003;  90 175-82
  CrossrefPubMedSearch in Google Scholar
• 10 Madhusudanan K P, Chattopadhyay S K, Tripathi V, Sashidhara K V, Kumar S. MS/MS profiling of taxoids from the needles of Taxus wallichiana.  Phytochem Anal. 2002;  13 18-30
  CrossrefPubMedSearch in Google Scholar
• 11 Kerns E H, Volk K J, Whitney J L, Rourick R A, Lee M S. Chemical identification of botanical components using liquid chromatography & mass spectrometry.  Drug Inf J. 1998;  32 471-85
  PubMedSearch in Google Scholar
• 12 Parmar V S, Jha A, Bisht K S, Taneja P, Singh S K, Kumar A. et al . Constituents of the yew trees.  Phytochemistry. 1999;  50 1267-304
  CrossrefPubMedSearch in Google Scholar
• 13 Lu G H, Chan K, Liang Y Z, Leung K, Chan C L, Jiang Z H. et al . Development of high-performance liquid chromatographic fingerprints for distinguishing Chinese Angelica from related Umbelliferae herbs.  J Chromatogr A. 2005;  1073 383-92
  CrossrefPubMedSearch in Google Scholar
• 14 Xie P, Chen S, Liang Y, Wang X, Tian R, Upton R. Chromatographic fingerprint analysis - a rational approach for quality assessment of traditional Chinese herbal medicine.  J Chromatogr A. 2006;  1112 171-80
  CrossrefPubMedSearch in Google Scholar
• 15 Zou P, Hong Y, Koh H L. Chemical fingerprinting of Isatis indigotica root by RP-HPLC and hierarchical clustering analysis.  J Pharm Biomed Anal. 2005;  38 514-20
  CrossrefPubMedSearch in Google Scholar
• 16 Becerra J X. Synchronous coadaptation in an ancient case of herbivory.  Proc Natl Acad Sci USA. 2003;  100 12 804-7
  PubMedSearch in Google Scholar
• 17 Grayer R J, Chase M W, Simmonds M SJ. A comparison between chemical and molecular characters for the determination of phylogenetic relationships among plant families: an appreciation of Hegnauer's ”Chemotaxonomie der Pflanzen”.  Biochem Syst Ecol. 1999;  27 369-93
  CrossrefPubMedSearch in Google Scholar
• 18 Wink M. Evolution of secondary metabolites from an ecological and molecular phylogenetic perspective.  Phytochemistry. 2003;  64 3-19
  CrossrefPubMedSearch in Google Scholar
• 19 Vieira R F, Grayer R J, Paton A J. Chemical profiling of Ocimum americanum using external flavonoids.  Phytochemistry. 2003;  63 555-67
  CrossrefPubMedSearch in Google Scholar
• 20 Theodoridis G, Laskaris G, de Jong C F, Hofte A JP, Verpoorte R. Determination of paclitaxel and related diterpenoids in plant extracts by high-performance liquid chromatography with UV detection in high-performance liquid chromatography-mass spectrometry.  J Chromatogr A. 1998;  802 297-305
  CrossrefPubMedSearch in Google Scholar
• 21 Mukherjee S, Ghosh B, Jha T B, Jha S. Variation in content of taxol and related taxanes in Eastern Himalayan populations of Taxus wallichiana.  Planta Med. 2002;  68 757-9
  Thieme ConnectPubMedSearch in Google Scholar
• 22 Massart B, Guo Q, Questier F, Massart D L, Boucon C, deJong S. Data structures and data transformations for clustering chemical data.  Trends Anal Chem. 2001;  20 35-41
  PubMedSearch in Google Scholar
• 23 Aruga R. Multivariate classification of constrained data: problems and alternatives.  Anal Chim Acta. 2004;  527 45-51
  CrossrefPubMedSearch in Google Scholar
• 24 Berman E, Kulp K, Knize M, Wu L, Nelson E, Nelson D. et al . Distinguishing monosaccharide stereo- and structural isomers with TOF-SIMS and multivariate statistical analysis.  Anal Chem. 2006;  78 6497-503
  CrossrefPubMedSearch in Google Scholar
• 25 Bailey N, Sampson J, Hylands P J, Nicholson J K, Holmes E. Multi-component metabolic classification of commercial feverfew preparation via high-field ¹H-NMR spectroscopy and chemometrics.  Planta Med. 2002;  68 734-8
  Thieme ConnectPubMedSearch in Google Scholar
• 26 Smith R M, Burford M D. Comparision of flavanoids in feverfew varieties and related species by principal components analysis.  Chemometrics Intelligent Lab Syst. 1993;  18 285-91
  PubMedSearch in Google Scholar
• 27 Máximo P, Lourenço A, Tei A, Wink M. Chemotaxonomy of Portuguese Ulex: Quinolizidine alkaloids as taxonomical markers.  Phytochemistry. 2006;  67 1943-9
  CrossrefPubMedSearch in Google Scholar
• 28 Nyman T, Julkunen-Tiitto R. Chemical variation within and among six northern willow species.  Phytochemistry. 2005;  66 2836-43
  CrossrefPubMedSearch in Google Scholar

Prof. Ling Yang

Laboratory of Pharmaceutical Resource Discovery

Dalian Institute of Chemical Physics

Chinese Academy of Sciences

457 Zhongshan Road

Dalian 116023

People’s Republic of China

Phone: +86-411-8437-9317

Fax: +86-411-8467-6961

Email: yling@dicp.ac.cn

References

• 1 Baloglu E, Kingston D GI. The taxane diterpenoids.  J Nat Prod. 1999;  62 1448-72
  CrossrefPubMedSearch in Google Scholar
• 2 Appendino G. The phytochemistry of the yew tree.  Nat Prod Rep. 1995;  121 349-60
  PubMedSearch in Google Scholar
• 3 Denis J N, Greene A E, Guénard D, Guéritte-Vorgrlrin F, Mangatal L, Potier P. A highly efficient, practical approach to natural taxol.  J Am Chem Soc. 1988;  110 5917-9
  CrossrefPubMedSearch in Google Scholar
• 4 van Rozendaal E, Lelyveld G, van Beek T. Screening of the needles of different yew species and cultivars for paclitaxel and related taxoids.  Phytochemistry. 2000;  53 383-9
  CrossrefPubMedSearch in Google Scholar
• 5 Navia-Osorio A, Garden H, Cusidó R M, Palazón J, Alfermann A W, Piñol M T. Production of paclitaxel and baccatin III in a 20-L airlift bioreactor by a cell suspension of Taxus wallichiana.  Planta Med. 2002;  68 336-40
  Thieme ConnectPubMedSearch in Google Scholar
• 6 Poupat C, Hook I, Guéritte F, Ahond A, Guénard D, Adeline M T. et al . Neutral and basic taxoid contents in the needles of Taxus species.  Planta Med. 2000;  66 580-4
  Thieme ConnectPubMedSearch in Google Scholar
• 7 Khan I A. Issues related to botanicals.  Life Sci. 2006;  78 2033-8
  CrossrefPubMedSearch in Google Scholar
• 8 Li J, Davis C C, Tredici P D, Donoghue M J. Phylogeny and biogeography of Taxus (Taxaceae) inferred from sequences of the internal transcribed spacer region of nuclear ribosomal DNA.  Harvard Papers Bot. 2001;  6 267-74
  PubMedSearch in Google Scholar
• 9 Collins D, Mill R R, Moller M. Species separation of Taxus baccata, T. canadensis, and T. cuspidate (Taxaceae) and origins of their reputed hybrids inferred from RAPD and cpDNA data.  Am J Bot. 2003;  90 175-82
  CrossrefPubMedSearch in Google Scholar
• 10 Madhusudanan K P, Chattopadhyay S K, Tripathi V, Sashidhara K V, Kumar S. MS/MS profiling of taxoids from the needles of Taxus wallichiana.  Phytochem Anal. 2002;  13 18-30
  CrossrefPubMedSearch in Google Scholar
• 11 Kerns E H, Volk K J, Whitney J L, Rourick R A, Lee M S. Chemical identification of botanical components using liquid chromatography & mass spectrometry.  Drug Inf J. 1998;  32 471-85
  PubMedSearch in Google Scholar
• 12 Parmar V S, Jha A, Bisht K S, Taneja P, Singh S K, Kumar A. et al . Constituents of the yew trees.  Phytochemistry. 1999;  50 1267-304
  CrossrefPubMedSearch in Google Scholar
• 13 Lu G H, Chan K, Liang Y Z, Leung K, Chan C L, Jiang Z H. et al . Development of high-performance liquid chromatographic fingerprints for distinguishing Chinese Angelica from related Umbelliferae herbs.  J Chromatogr A. 2005;  1073 383-92
  CrossrefPubMedSearch in Google Scholar
• 14 Xie P, Chen S, Liang Y, Wang X, Tian R, Upton R. Chromatographic fingerprint analysis - a rational approach for quality assessment of traditional Chinese herbal medicine.  J Chromatogr A. 2006;  1112 171-80
  CrossrefPubMedSearch in Google Scholar
• 15 Zou P, Hong Y, Koh H L. Chemical fingerprinting of Isatis indigotica root by RP-HPLC and hierarchical clustering analysis.  J Pharm Biomed Anal. 2005;  38 514-20
  CrossrefPubMedSearch in Google Scholar
• 16 Becerra J X. Synchronous coadaptation in an ancient case of herbivory.  Proc Natl Acad Sci USA. 2003;  100 12 804-7
  PubMedSearch in Google Scholar
• 17 Grayer R J, Chase M W, Simmonds M SJ. A comparison between chemical and molecular characters for the determination of phylogenetic relationships among plant families: an appreciation of Hegnauer's ”Chemotaxonomie der Pflanzen”.  Biochem Syst Ecol. 1999;  27 369-93
  CrossrefPubMedSearch in Google Scholar
• 18 Wink M. Evolution of secondary metabolites from an ecological and molecular phylogenetic perspective.  Phytochemistry. 2003;  64 3-19
  CrossrefPubMedSearch in Google Scholar
• 19 Vieira R F, Grayer R J, Paton A J. Chemical profiling of Ocimum americanum using external flavonoids.  Phytochemistry. 2003;  63 555-67
  CrossrefPubMedSearch in Google Scholar
• 20 Theodoridis G, Laskaris G, de Jong C F, Hofte A JP, Verpoorte R. Determination of paclitaxel and related diterpenoids in plant extracts by high-performance liquid chromatography with UV detection in high-performance liquid chromatography-mass spectrometry.  J Chromatogr A. 1998;  802 297-305
  CrossrefPubMedSearch in Google Scholar
• 21 Mukherjee S, Ghosh B, Jha T B, Jha S. Variation in content of taxol and related taxanes in Eastern Himalayan populations of Taxus wallichiana.  Planta Med. 2002;  68 757-9
  Thieme ConnectPubMedSearch in Google Scholar
• 22 Massart B, Guo Q, Questier F, Massart D L, Boucon C, deJong S. Data structures and data transformations for clustering chemical data.  Trends Anal Chem. 2001;  20 35-41
  PubMedSearch in Google Scholar
• 23 Aruga R. Multivariate classification of constrained data: problems and alternatives.  Anal Chim Acta. 2004;  527 45-51
  CrossrefPubMedSearch in Google Scholar
• 24 Berman E, Kulp K, Knize M, Wu L, Nelson E, Nelson D. et al . Distinguishing monosaccharide stereo- and structural isomers with TOF-SIMS and multivariate statistical analysis.  Anal Chem. 2006;  78 6497-503
  CrossrefPubMedSearch in Google Scholar
• 25 Bailey N, Sampson J, Hylands P J, Nicholson J K, Holmes E. Multi-component metabolic classification of commercial feverfew preparation via high-field ¹H-NMR spectroscopy and chemometrics.  Planta Med. 2002;  68 734-8
  Thieme ConnectPubMedSearch in Google Scholar
• 26 Smith R M, Burford M D. Comparision of flavanoids in feverfew varieties and related species by principal components analysis.  Chemometrics Intelligent Lab Syst. 1993;  18 285-91
  PubMedSearch in Google Scholar
• 27 Máximo P, Lourenço A, Tei A, Wink M. Chemotaxonomy of Portuguese Ulex: Quinolizidine alkaloids as taxonomical markers.  Phytochemistry. 2006;  67 1943-9
  CrossrefPubMedSearch in Google Scholar
• 28 Nyman T, Julkunen-Tiitto R. Chemical variation within and among six northern willow species.  Phytochemistry. 2005;  66 2836-43
  CrossrefPubMedSearch in Google Scholar

Prof. Ling Yang

Laboratory of Pharmaceutical Resource Discovery

Dalian Institute of Chemical Physics

Chinese Academy of Sciences

457 Zhongshan Road

Dalian 116023

People’s Republic of China

Phone: +86-411-8437-9317

Fax: +86-411-8467-6961

Email: yling@dicp.ac.cn

• Supplementary Material