Beruflich Dokumente
Kultur Dokumente
35 (2004) 117126
a SCYNEXIS Europe Ltd., Fyfield Business and Research Park, Fyfield Road, Ongar, Essex CM5 0GS, UK
Biological Chemistry, Biomedical Sciences Division, Imperial College London, Sir Alexander Fleming Building,
Exhibition Road, South Kensington, London SW7 2AZ, UK
c Oxford Natural Products Plc., Cornbury Park, Charlbury, Oxfordshire OX7 3EH, UK
d Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine,
Keppel Street, London WC1E 7HT, UK
Received 1 October 2003; received in revised form 24 December 2003; accepted 28 December 2003
Abstract
We describe the application of 1 H NMR spectroscopy and chemometrics to the analysis of extracts of Artemisia annua.
This approach allowed the discrimination of samples from different sources, and to classify them according to anti-plasmodial
activity without prior knowledge of this activity. The use of partial least squares analysis allowed the prediction of actual values
for anti-plasmodial activities for independent samples not used in producing the models. The models were constructed using
approximately 70% of the samples, with 30% used as a validation set for which predictions were made. Models generally
explained >90% of the variance, R2 in the model, and had a predictive ability, Q2 of >0.8. This approach was also able to
correlate 1 H NMR spectra with cytotoxicity (R2 = 0.9, Q2 = 0.8).
This work demonstrates the potential of NMR spectroscopy and chemometrics for the development of predictive models of
anti-plasmodial activity.
2004 Elsevier B.V. All rights reserved.
Keywords: Artemisia annua; Malaria; Chemometrics; 1 H NMR spectroscopy; Pattern recognition; Plasmodium falciparum; Biological
activity; Prediction
Abbreviations: ToxED50 , measure of extract toxicity; FID, free induction decay; IC50 , measure of extract anti-plasmodial activity; NMR,
nuclear magnetic resonance; PCA, principal components analysis; PLS, partial least squares; PLS-DA, partial least squares discriminant
analysis
Corresponding author. Tel.: +44-1277-367036; fax: +44-1277-367099.
E-mail address: nigel.bailey@scynexis.com (N.J.C. Bailey).
1 Present address: GW Pharmaceuticals Plc., Porton Down Science Park, Salisbury, Wiltshire SP4 0JQ, United Kingdom.
2 Present address: Celltech R&D Ltd., 216 Bath Road, Slough, Berkshire SL1 4EN, United Kingdom.
0731-7085/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jpba.2003.12.024
118
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
1. Introduction
Artemisia annua L. (also known as sweet wormwood, or by its Chinese name qing hao) and the
isolated constituent artemisinin (Fig. 1) have well
documented anti-plasmodial activity [13]. Indeed,
a whole class of anti-plasmodial drugs derived from
artemisinin are now in widespread use in single and
combination therapies, particularly where resistance
to other anti-plasmodials is present [1,4].
Whilst artemisinin has been established as an
important component of A. annua with respect to
anti-plasmodial activity [2], it has also been suggested that the efficacy of the A. annua plant itself
derives from a synergistic effect, and that it is a
combination of constituents in the plant that confer
the total anti-plasmodial activity [5,6]. Several polymethoxyflavones have been found to have activity in
combination with artemisinin [7], and it has been reported that flavonoids enhance the anti-plasmodial activity of artemisinin [8]. Further, the clinical efficacy
of A. annua extracts as a treatment for malaria has
been demonstrated, with 92% of malaria patients in a
study showing disappearance of parasitaemia within
4 days [9]. As a result, the potential use of A. annua,
particularly in areas where large-scale pharmaceutical production is not possible, is clearly of interest.
However, where plant extracts themselves are to be
used, it is necessary to determine the reproducibility
and information regarding the content of such extracts
[9]. It has been shown, for example that the levels of
artemisinin and related compounds fluctuate due to
both seasonal and geographical variation [10,11].
Therefore, it is important to develop analytical
methodology that is capable of providing information relating to a whole extract with respect to antiplasmodial activity, and reproducibility. 1 H NMR
H
O
O
O
H O
O
Fig. 1. Structure of artemisinin.
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
2. Methods
2.1. Acquisition of plant samples
Nineteen A. annua accessions sourced from different locations were obtained by Oxford Natural Products Plc. (Oxford, UK). Samples in the form of dry
herb material, powder or tincture were deposited in
the herbarium in the Pharmacognosy laboratories at
Kings College London, UK. Voucher specimen numbers were as follows for sample identities 119 of A.
annua L. (Compositae): AR17 10 I1, AR17 11 I2,
AR17 12 I3, AR17 13 I4, AR17 14 I5, AR17 15 I6,
AR17 16 I7, AR17 17 I8, AR17 19 I10, AR17 20 I11,
AR17 21 I12, AR17 22 I13, AR17 23 I14, AR17 24
I15, AR17 25 I16, AR17 26 I17, AR17 27 I18, AR17
28 I19, AR17 29 I20.
2.2. Sample preparation
Samples were prepared by Advanced Phytonics (UK) using an extraction method covered
under International patent Application Numbers
PCT/GB95/00554, and International Publication
Number WO 95/26794.
For dry herb material, the samples (90135 g)
were packed into a 560 ml jacketed vessel and extracted at 25 C with a constant pumped stream of
1,1,1,2-tetrafluoroethane solvent (2664 bed vol-
119
120
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
was repeated in order to ensure all extracts were excluded once. Models were constructed to predict values for IC50 and ToxED50 .
121
Fig. 2. Representative 1 H NMR spectra for the three IC50 classes, IC50 < 0.1 g/ml (extract 7), IC50 > 0.1 g/ml, <1 g/ml (extract 9)
and IC50 > 1 g/ml (extract 19). Region 2.58.5 ppm is expanded vertically by a factor of 6 to allow observation of lower level aromatic
resonances. Resonances attributable to artemisinin are indicated with an A.
122
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
Fig. 3. PCA scores plot for A. annua plant extracts. Samples are separated according to original extract number.
4
3
2
PC2 (14%)
1
-10
-5
0
-1 0
IC50<0.1ug/ml
5
10
15
-2
-3
-4
-5
-6
PC1 (64%)
Fig. 4. PCA scores plot for A. annua plant extracts. Samples are separated into three groups, IC50 < 0.1 g/ml (), IC50 > 0.1 and
<1 g/ml () and IC50 > 1 g/ml ().
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
123
1.4
1.2
Predicted Class
1
0.8
0.6
0.4
0.2
0
-0.2 0
10
20
30
40
50
-0.4
Sample Number
Fig. 5. PLS-DA Y-predicted scatter plot for classes IC50 > 0.1 g/ml (class 0, triangles) and IC50 < 0.1 g/ml (class 1, circles). Training
set represented by closed shapes, () and (), test set represented by open shapes, () and ().
124
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
Table 1
Summary of predicted IC50 and ToxED50 values for a series of A. annua extracts
Extract
IC50 predicted
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
0.01
0.01
0.020
0.02
0.020
0.02
0.0296
0.04
0.169
0.13
0.29
0.30
0.31
0.32
0.47
8.55
24.67
4.2
3.9
0.05
0.01
0.027
0.00
0.065
0.06
0.0262
0.03
0.209
0.17
0.16
0.20
0.22
0.22
0.23
0.21
0.20
0.10
0.243
0.04
0.02
0.00
0.03
0.005
0.02
0.0003
0.02
0.008
0.01
0.04
0.03
0.04
0.02
0.04
0.03
0.03
0.01
0.002
ToxED50 value
(g/ml)
ToxED50 predicted
27
9
8
19
65
50
171
11
69
26
41
66
8
76
50
46
72
20
42
20
22
17
26
33
12
54
28
75
62
33
72
60
30
40
48
6
3
6
6
11
2
2
4
5
4
2
6
3
5
1
12
14
Relative
artemisinin levela
0.47
0.66
1
0.69
0.44
0.34
0.45
0.91
0.18
0.14
0.15
0.21
0.18
0.20
0.12
0.04
0.02
0.02
0.04
a Obtained from the 1 H NMR peak intensity for the artemisinin peak at ca. 6.1 ppm. Values expressed relative to the highest peak, that
of extract 3.
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
125
1.4
1.2
Predicted Class
1
0.8
0.6
0.4
0.2
0
-0.2
20
40
60
Sample Number
Fig. 6. PLS-DA Y-predicted scatter plot for classes ToxED50 > 65 g/ml (class 0, squares) and ToxED50 < 65 g/ml (class 1, circles).
Training set represented by closed shapes, ( ) and (), test set represented by open shapes, () and ().
new models using the ToxED50 rather than IC50 values, it is possible to predict values for ToxED50 using
the same NMR spectra. An example PLS-DA plot is
shown in Fig. 6, with the samples separated into those
with a ToxED50 value of > or <65 g/ml (one extract,
extract 7, with a ToxED50 value of 171.3 g/ml was
excluded for the same reasons as the extracts excluded
with IC50 of >1 g/ml. One further extract, number 19,
was excluded due to insufficient material being available for the ToxED50 measurement). The analysis was
repeated five times (with samples split into different
training and test sets each time), and >90% of samples
were correctly classified in all models, with >84% of
samples predicted with >99% confidence. R2 values
ranged from 0.78 to 0.94, and Q2 values from 0.69 to
0.84. Three, four or five components were used in all
models. Although the statistics reveal that the models
are not as good as the ones created for the IC50 values,
these models are still able to give a good indication as
to the likely toxicity of a particular plant extract. That
said however, the classes used are, as for the IC50 data
analysis, arbitrary classes. In order to obtain predicted
ToxED50 values for the extracts, PLS was again performed using the assay data to construct models.
Four, five, or six components were used for all PLS
models, with >78% of samples predicted with >99%
confidence. R2 values ranged from 0.89 to 0.98, and
Q2 values from 0.73 to 0.94. The model construction
process was repeated in order to exclude every extract
from the training set once (with additional samples
being removed on a random basis). The average pre-
126
N.J.C. Bailey et al. / Journal of Pharmaceutical and Biomedical Analysis 35 (2004) 117126
References
[1] V. Dhingra, K.V. Rao, M.L. Narasu, Life Sci. 66 (2000) 279
300.
[2] D. Klayman, Science 228 (1985) 10491055.
[3] R. Haynes, Curr. Opin. Infect. Dis. 14 (2001) 719726.
[4] G. Balint, Pharmacol. Therap. 90 (2001) 261265.
[5] G. Kirby, Trop. Doctor 27 (1997) 711.
[6] B. Gilbert, L.F. Alves, Curr. Med. Chem. 10 (2003) 1320.
[7] R. Bhakuni, D. Jain, R. Sharma, S. Kumar, Curr. Sci. 80
(2001) 3548.
[8] B. Elford, M. Roberts, J. Phillipson, R. Wilson, T. Roy, Soc.
Trop. Med. H. 81 (1987) 434436.
[9] M. Mueller, I. Karhagomba, H. Hirt, E. Wemakor, J.
Ethnopharmacol. 73 (2000) 487493.
[10] T. Wallaart, N. Pras, W. Quax, Planta Med. 65 (1999) 723
728.
[11] T. Wallaart, N. Pras, A. Beekman, W. Quax, Planta Med. 66
(2000) 5762.
[12] J.K. Nicholson, J. Connelly, J.C. Lindon, E. Holmes, Nat.
Rev. Drug Disc. 1 (2002) 153160.
[13] F. Bamforth, V. Dorian, H. Vallance, D. Wishart, J. Inher.
Metab. Dis. 22 (1999) 297301.