Qualitative analysis of Puerariae Lobata Radix granules using partial least square discriminant analysis

Ka H Wong , Valentina Razmovski-Naumovski , Kong M. Li , George Q Li , Kelvin Chan 1 Faculty of Pharmacy, University of Sydney, NSW 2006, Australia; 2 Discipline of Pharmacology, Bosch Institute, University of Sydney, NSW 2006, Australia; 3 Centre for Complementary Medicine Research, School of Science & Health, University of Western Sydney, NSW 2560, Australia
1 1,3 2 1 1,3

Introduction
Puerariae Lobata Radix (PLR; Gegen) is a traditional Chinese medicine (TCM) for treating diabetes and cardiovascular diseases1. PLR is often substituted with a closely related species named Puerariae Thomsonii Radix (PTR; Fenge)1. Recent studies have revealed that PLR has distinct chromatographic characteristics and a significantly higher antioxidant activity than PTR1. There is lack of clinical evidence and in vitro study to support the use of PTR as a medicinal herb, and hence its safety and efficacy are questionable. The emergence of commercial granules in TCM has provided patients a more convenient way for administrating herbal extracts. However, there is a limited research comparing the quality of granules to the respective aqueous extracts prepared from raw materia medica. Puerarin is the major chemical component from Puerariae Radix. It was used as the chemical marker for the authentication of Puerariae Radix suggested by the Pharmacopoeia of Peoples Republic of China (PPRC)2.

Aim
To determine the raw materia medica used in manufacturing commercial Puerariae Radix granules.

Methods & Materials

Gegen
Materials Forty-seven dried Puerariae Radix and seventeen commercial granules were collected from herbal pharmacies in various regions of China, Australia and USA . Sample preparation The powdered samples were extracted with water under reflux for 3 hrs. The mixtures were then filtered and evaporated to obtain the water extracts. Subsequently, the water extracts and granules were extracted with methanol in a sonication bath for 30 mins to eliminate excipients such as dextran and starch powder. UPLC Column: Waters BEH C18 (1.7 m, 2.1 x 150 mm). Mobile phase: 0.3% acetic acid in water (A) and 0.3% acetic acid in acetonitrile (B). Gradient: 0-7 min, B 10-25%; 7-8 min, B 25-45%; 8-14 min, B hold on 45%; 14-15 min, B 45-10%; 15-19 min, hold on 10%. Chemometric analysis The chromatographic fingerprints were pre-processed using correlation optimised warping, baseline subtraction, mean centering and standard normal variate. The data matrix was separated into calibration (60%) and validation (40%) set using Kennard -Stone algorithm. Partial least square discriminant analysis (PLS-DA) and principal components analysis (PCA) were performed on Matlab 2011b. PLS-DA models were constructed using the entire chromatogram and puerarin.

Fenge

Granules

Results

Figure 2. PLS-DA scores plots of 39 Puerariae Radix

Figure 3. (a) Pre-processed chromatograms of Puerariae Radix; (b) Latent variable (LV) 1 PLS-DA loading plot; (c) LV2 PLS-DA loading plot; (d) five reference standards Table 3. Classification of commercial Puerariae Radix using PLS-DA

Figure 1. UPLC chromatograms of Puerariae Radix methanolic extracts (a) raw data; (b) after pre-processed with correlation optimized warping, baseline subtraction and concentration subtraction. Table 1. Statistical performance of various PLS-DA models on the calibration and validation set 2 LV(s) RMSEC RMSECV RMSEP r 0.1183 0.1452 0.0854 0.9440 Entire chro- 2 matogram 3 0.2893 0.3725 0.2049 0.6653 Puerarin Table 2. Species classification result of various PLS-DA on the validation set Entire chromato- Puerarin graphic Predicted class Predicted class True class N PLR PTR PLR PTR 9 9 8 1 PLR 6 6 1 5 PTR

Figure 4. Pre-processed chromatograms of (1) 23 aligned Puerariae Lobata Radix; (2) 20 aligned Puerariae Thomsonii Radix; (3) G3; (4) G6; (5) G9; (6) G17

Samples Entire chromatogram PLR G1 PLR G2 PTR G3 PLR G4 PLR G5 PTR G6 PLR G7 PLR G8 PTR G9 PLR G10 PLR G11 PLR G12 PLR G13 PLR G14 PLR G15 PLR G16 PTR G17

Puerarin RPL RPL PTR RPL RPL PTR RPL RPL PLR RPL RPL RPL RPL RPL RPL RPL PTR

Discussion & Conclusion

Several data pre-processing techniques were performed on the raw data matrix to reduce noise and inconsistency as well as to improve the data quality (Figure 1). PCA was applied to the pre-processed data matrix to identify possible anomalous samples since such samples could affect the quality of the model. Four outliers (PLR9, PTR5, PTR13 and PTR23) were found and removed for the subsequent analysis. The PLS-DA scores plot differentiated PLR from PTR by forming two distinct clusters (Figure 2). The statistical results (Table 1) shown that the model constructed by the entire chromatogram has a lower RMSEC, RMSECV and RMSEP as well as a greater regression coefficient, suggesting that it provides superior models stability, simplicity and performance as compared to puerarin. The performance of the established model was evaluate using the validation set. (Table 2). The model using puerarin alone misidentified two samples with a correct prediction rate of 86.67%. From the entire chromatogram PLS-DA model (Table 3), the herbal materials of four brands of commercial granules were classified as PTR and the remaining were classified as PLR. The UPLC chromatograms of these product further confirmed the PLS-DA model results (Figure 4). In contrast, the puerarin model correctly identified 3 out of 4 brands of granules, which were manufactured using PTR. These results suggested that the use of puerarin alone was insufficient in authenticating Puerariae Radix and its commercial products. The use of the entire chromatographic fingerprints should be adopted in the future studies for differentiating PLR and PTR. This study indicated that PLS-DA is a promising technique for the quality control of Pueraria Radix and other herbal material used in the manufacture of granules and other dosage forms in the pharmaceutical industry.

1.Wong, et.al. (2011). J Ethnopharmacol 134, 584-607. 2. Pharmacopoeia of Peoples Republic of China The authors would like to sincerely thank Dr Samiuela Lee and Mr Jarryd Pearson (Centre for Complementary (2010), Peoples Medical Publishing House, Vol1, pp 273-4;363-4. 3. Wong, et.al. (2013). J Pharm Bio- Medicine Research, University of Western Sydney, Australia) for their technical support in UPLC analyses. med Anal 84, 5-13.

References

Acknowledgements