Multi-Component Analysis of Ilex Kudingcha C. J. Tseng by a Single Marker Quantification Method and Chemometric Discrimination of HPLC Fingerprints

The quantitative analysis of multiple components with a single marker (QAMS) method was firstly established for simultaneous determination of 18 active components in Ilex kudingcha C. J. Tseng by HPLC. Using rutin, isochlorogenic acid A and kudinoside A as internal refererence substances (IRS), compatibility results showed that the relative correction factors (RCFs) of all compounds showed good reproducibility under different chromatographic conditions. On the basis of previous studies, the accuracy of the QAMS method was systematically evaluated by investigating the influences of curve intercept, analytes and IRS concentration. The results showed that the concentration (especially at low level) of analytes and curve intercept were the major influencing parameters for the LRG-QAMS method (LRG = linear regression), whereas the influence of IRS concentration seemed more apparent in terms of the AVG-QAMS method (AVG = average). The two approaches were complementary with each other. In addition, hierarchical clustering analysis (HCA), principal components analysis (PCA) and similarity analysis (SA) were performed to differentiate and classify the samples based on the contents of 18 marker compounds. The results of the different chemometric analyses were completely consistent with each other, and could be supported by the quantification results.


Introduction
Ilex kudingcha C. J. Tseng (Kudingcha in Chinese), the dried leaves of the genus Ilex in the family Aquifoliaceae, is widely distributed in the Hainan, Hubei, Guangdong, Guangxi and Hunan provinces of China. It has been used as an ethnomedicine for the treatment of diseases relative to dyslipidosis for millennia [1]. Nowadays it has become a hotspot in traditional Chinese medicine (TCM) research due to its significant effects on cardiovascular-related diseases and metabolic syndrome [2,3]. Ilex kudingcha C. J. Tseng is commonly applied as a strategy in the clinic for adjuvant therapy of diabetes, hypertension, obesity, and hyperlipidemia mainly based on its antioxidant and immunoregulatory properties [4], and has also been developed into various health care products to improve patient compliance.
Pharmacologically active ingredients in Ilex kudingcha C. J. Tseng have been elucidated to be phenolic acids, triterpenoids and flavonoids, which have activities against inflammatory processes,

Validation of the Traditional External Standard Method
The method specificity was assessed by comparing the consistency of the retention time and UV spectrum of each analyte between a sample and corresponding reference standard. Figure 1 shows a typical separation of a standard mixture (A)-(C) and Ilex kudingcha C. J. Tseng extracts (A)'-(C)' under the optimized chromatographic conditions. All the markers were not only well resolved from background peaks but a good resolution of adjacent peaks within the analysis time was also attained. Figure 1(A-C) and (A'C') illustrate that the eighteen peaks in the chromatogram of Ilex kudingcha C. J. Tseng could all be identified by the corresponding standards. The peak purity detection function of photo-diode array (PDA) detector was used, which confirmed the acceptable purity of the eighteen analytes' peaks in the samples.

Calibration Curves and Evaluation of RCFs Versus Concentration
A total of 20 µL of different concentration levels (n = 6) of mixed standards solution were injected into HPLC systems for the construction of calibration curves. The standard curves for the 18 components showed good linearity with correlation coefficients higher than 0.9990 within the test ranges (Table 1). Meanwhile, RCFs and relative retention time (RT R ) were calculated based on the linearity data and summarized. The relative standard deviation (RSD) values of RCFs and RT R were less than 4.77% and 0.26%, respectively, which suggested that RCFs and RT R obtained on the same instrument at different concentration level were highly repeatable.

Precision, Repeatability, Stability and Accuracy
The precision was examined from five consecutive injections of the sample placed in the autosampler (10 • C). The repeatability was tested by injecting six independently prepared samples, which were prepared according to the method outlined in the "Preparation of solutions" section. The stability was tested with a sample solution that was stored at room temperature (25 ± 5 • C) and analyzed at 0, 2, 4, 6, 8 and 12 h. The accuracy was determined by a recovery test performed by spiking all the reference standards at middle concentration level (n = 6) into a sample (0.1 g) of Ilex kudingcha C. J. Tseng powder, followed by extraction and analysis as described in "Preparation of solutions" section. The recovery for each analyte was calculated based on the equation of recovery (%) = 100 × (amount found − original amount)/amount spiked. The results are displayed in Table 2. , Waters Symmetry C 18 (4.6 mm × 250 mm, 5 µm) and Phenomenex Synergi Hydro-RP C 18 (4.6 mm × 250 mm, 5 µm) columns were used to study the effects of different columns. The experiments on different instruments were carried out in different labs, using LC-20A, Agilent 1260 and Waters Alliance E2695 instruments. The results (Table 3) showed that the RCFs and RT R from different columns had good reproducibility with RSD < 4.86% and RSD < 4.94%, respectively, while HPLC instruments had a slight influence on RCFs and RT R , with RSD < 5.17% and RSD < 4.92%, respectively. The Shimadzu LC-20A system with a Phenomenex synergi C 18 column was used to investigate the effects of different column temperatures (25, 28, 30, 32 and 35 • C) and flow rates (0.8, 0.9 and 1.0 mL·min −1 ) on RCFs. The results (Table 4) proved that baseline separation of all marker components would be realized when the column temperature was lower than 30 • C and flow rate less than 1.0 mL·min −1 . RCFs and RT R kept constant within the temperature range of 25 • C~30 • C, with RSD of 1.14%~6.24% and 0.54%~4.59%, respectively. Three flow rates were examined, and good repeatability of RCF and RT R for the target compounds was obtained with RSDs ranging from 1.45%~5.31% and 0.09%~3.40% respectively.

Sample Analysis
The content of 18 compounds was calculated according to the calibration curves, and those that were scattered in the vicinity of the lowest concentration point on the standard curve, were determined with a one point external standard method. The results (Table 5) illustrated that there were remarkable differences among the content of the 18 analytes in different Ilex kudingcha C. J. Tseng samples, which could be attributed to the variations of genetics, plant origins, environmental factors, drying process, storage conditions and so on. The respective summation of individual component content grouped by ingredient categories was performed, and the results ( Figure 2) showed that the differences of total content for each type components in Ilex kudingcha C. J. Tseng samples were less than five times for most herbal samples (except sample No. 1).  Additionally, we found that the content summation of C2, C6 and C7 accounted for so large proportion of the total phenolic acids (>75%) that the three compounds could be used as index components for quality assessment of Ilex kudingcha C. J. Tseng, and the same conclusion could be drawn for K1, K2 and K3 with more than 80% proportion of the total saponins.

Evaluation the Accuracy and Investigation of Influencing Parameters on QAMS Method
To systematically investigate the accuracy of LRG-QAMS and AVG-QAMS method, the two QAMS methods were compared with an external standard method (ES) [15], respectively, by standard method difference (SMD) calculated according to the following equation: SMD% = 100 × (C ES − C QAMS )/C ES (7), where C ES and C QAMS represent the concentration of an analyte assayed by the ES method and QAMS method, respectively. As shown in Figure 3 and Tables 6-8, a total of 675 SMD datapoints calculated from fifteen quantitative markers in 15 batches of Ilex kudingcha C. J. Tseng samples were obtained. The results showed that the two QAMS methods could obtain accordant values compared with the ES method for the assay of the different Ilex kudingcha C. J. Tseng samples, but different from previous studies [5,16], the reported deduction that LRG-QAMS method had higher accuracy in comparison with AVG-QAMS method did not seem universal judging from Figure 3.    The influences of two parameters on the accuracy of LRG-QAMS and AVG-QAMS method were investigated, including the content of each quantitative component in the plant materials and the content of IRS. SMDs (total of 675 data) were divided into two groups according to the criteria described as follows: (1) contents of IRS at high, middle and low level; (2) contents of analytes at high, middle and low level.
Scatter diagrams (Figures 4 and 5) were used to investigate the mathematical relationship between these potential influence parameters and SMDs. As shown in the scatter diagrams, the results implied that the accumulation of the component in herbal material rather than that of IRS was the main correlation factor associated with the accuracy of the LRG-QAMS method.  In the case of the AVG-QAMS method, on the contrary, the IRS content seemed a little more predominant impact factor on the accuracy, and SMDs of most components exhibited a decreasing trend with increasing IRS accumulations. However, the influence of the analytes' content on SMDs did not show a consistent and regularly tendency in the AVG-QAMS method.

Analysis of the Samples
In the HCA analysis, 'Euclidean distance' was selected as measurement, and the method of 'Within-group Linkage' was applied. As shown in Figure 6A, the fifteen tested populations were classified into two main clusters (I and II) according to their contents. No.1 sample from Hainan with characteristics of significant content differences in each type of components was clearly distinguished from other origin herbal samples. Regarding the similarity calculation, Euclidean distance, correlation coefficient and cosine were adopted as measurements to reflect the similarity of samples (summarized in Table 9). Compared with the Euclidean distance, the measurements of cosine and correlation which ignore the quantitative discrepancy of variables mainly provide qualitative information for TCMs authentication. No. 1 sample, having significant differences in each category component content, gained relatively high score value (>0.94) for cosine and correlation analysis, which proved it was collected from the same genus, and coincident with the results of pharmacognosy identification. On the other hand, the measurement of Euclidean distance also focuses on reflecting the quantitative discrepancy of characteristic variates. The relatively lower score value (approximately 0.90) in the No. 1 sample for Euclidean distance analysis implied that there was a significant discrepancy in the content of marker compounds between No. 1 and other herbal samples. In addition, the determination results of 15 samples were further analyzed and classified by PCA. The two principle components (PC1 and PC2) with more than 60.9% of the whole variances were extracted for analysis. As shown in Figure 6B, the scatter plot was noticeable that No. 1 sample was clearly discriminated from the group, and scattered on the boundary of the circle. The results of HCA, PCA and similarity analysis were exactly consistent with each other.

Discussion
According to the literature, sonication with methanol was a preferred method. The extraction duration and times were tested. Finally ultrasonic extraction with 100 fold excess of methanol for 30 min was chosen because all the analytes could not only be efficiently extracted, but also well resolved from background peaks.
In order to generate the RCFs of analytes at different wavelengths, it is essential to choose three suitable IRS. The selected IRS should be an index composition meeting four requirements as follows: (1) abundant in sample; (2) stable; (3) easily accessible; (4) having a maximum UV absorption. Meanwhile, a good separation under the chromatographic conditions should be achieved.
The validated traditional external standard method (ES) and QAMS method were employed to assay 15 batches of representative Ilex kudingcha C. J. Tseng samples from different locations, which were based on the principle of the linear relationship between a detector response and the levels of components within certain concentration ranges.
As it shown in the results of evaluation the accuracy of QAMS, we found that the SMDs calculated from LRG-QAMS method for those components, the content of which in Ilex kudingcha C. J. Tseng samples scattered near the lowest concentration point of the standard curve, suffered relatively more significant deviation from ES results compared with the AVG-QAMS method. Such deviation at low concentration level would be dramatically reduced with the decreasing of intercept values. Only if the intercept value was small enough to be ignored (relative to the product of slope and concentration), the content assayed by LRG-QAMS method would be much closer to ES method calculation. Due to the significant content variability in different source herbal materials, the standard curve usually has to be constructed with a broad liner range. Thus, SMDs values calculated from LRG-QAMS method for the components, content of which is out of or near the lowest concentration level of the standard curve, are remarkable in most cases due to the influence of standard curve intercept. Therefore, we recommended that LRG-QAMS method could be used as the substitute of AVG-QAMS method with higher accuracy for quantitative determination only under the condition of zero-tending impact of intercept. Otherwise, the AVG-QAMS method was more suitable and general.
Interestingly, as for the LRG-QAMS method, the content of the IRS had minimal influence on the SMD values, while a clear numerical reduction trend in the low content region of scatter diagrams was mostly observed when the content of the analytes increased, which coincided with the conclusion mentioned in Wang's and our previous research [16,17]. We also noticed that the larger the intercept, the more obvious this trend was, even within a relatively wide concentration range. Our previous study has proved that there is no significant influence on the SMDs by changing different component as IRS.

Reagents and Chemicals
Acetonitrile (HPLC grade) was obtained from Merck KGaA (Darmstadt, Germany), and distilled water was purchased from Huaren Yibao Drinks Co.

Plant Materials
Fifteen commercial herbal samples of Ilex kudingcha C. J. Tseng (identified with serial numbers 1-15, Table 10) were purchased from Guangzhou drug stores or markets in 2015, and were authenticated by Professor Jin-song Zhou from Guangzhou University of Traditional Chinese Medicine. The voucher specimens were deposited at the Herbarium at Guangzhou University of Traditional Chinese Medicine. The air-dried samples were smashed into powder (40 mesh) and stored in a desiccator.

Instrument and Chromatographic Conditions
Analyses were primarily performed using Waters Alliance E2695 (Waters Corp., Milford, MA, USA) liquid chromatographic system before Shimadzu is indicated comprised of a quaternary solvent delivery system, an online degasser, an auto-sampler and a photodiode array detector (PDA) (Waters Corp.). The separation was performed on Phenomenex Synergi Hydro-RP C 18 columns (4.6 mm × 250 mm, 5 µm). The mobile phase was composed of 0.05% aqueous phosphoric acid (A, pH 3.0), acetonitrile (B) and methanol (C) using a gradient elution of 6% B at 0-5 min (A-B)

Preparation of Solutions
Approximately 0.2 g of finely ground sample powder (40 mesh) was accurately weighted into a brown stoppered flask and immersed with 20 mL methanol (A.R) at room temperature (25 ± 3 • C) for 30 min, then extracted using ultrasonication in an ice bath for 30 min. After cooling to room temperature, the lost solvent in the extraction solution was replenished with methanol and mixed well.

Calculation of the Relative Correction Factors (RCFs) and Relative Retention Time (RT R )
RCFs of the components co-existing in a TCM were calculated using the standard substance of each analyte by taking a typical active one with characteristic of sufficient abundance in chromatograms and easy availability as the internal reference substance (IRS) according to formula (1) as reported previously. In most QAMS related studies, the final RCF of an analyte is calculated using the average of several RCFs from the IRS and the analyte measured under multiple concentration levels, which would be referred to as average method (AVG-QAMS) in the following passages [12]. In the present study, a novel RCF calculating method, namely linear regression method (LRG-QAMS) was applied based on Equation (3), which used the linear relationship between C x (i.e., concentration of the analyte) and (A x × C i )/A i (i.e., peak areas of internal referring substance-A i and the analyte-A x ; concentration of the internal referring substance-C i ) to calculated RCFs by linear regression. The relative retention time was calculated as the ratio of retention time of the analyte versus to that of IRS: where A x and C x represent the peak area and concentration of the analyte, respectively. A i and C i are the peak area and concentration of the IRS, accordingly.

HCA, PCA and SA Analysis of Samples
The HCA analysis was realized using the SPSS 19.0 software (IBM, Armonk, NY, USA). The between-group linkage method was applied, and euclidean distance was selected as a measurement. Dendrograms resulting from the 18 marker components content were derived from the HPLC profiles of the tested samples.
The PCA analysis was done by the SIMCA-P 12.0 software (Umetrics, Sweden). In this study, content of the detective eighteen markers in 15 batches of samples composed a data matrix with 15 rows and 18 columns, which were used for the PCA analysis after normalization. The first two principal components were extracted, and the scatter plot was obtained by plotting scores of PC1 versus PC2. Euclidean distance, correlation coefficient and cosine were adopted to calculate the similarity of samples as expressed by the following formulae (4)-(6), respectively: where X ik and X rk represent the kth variate of the ith sample and mean vector in common pattern, respectively; X i and X r are the mean values of them.

Conclusions
In our study, a validated and sensitive QAMS method with high precision, stability, and repeatability was firstly developed for simultaneous quantification of eighteen analytes in Ilex kudingcha C. J. Tseng, as well as to study the influence parameters on the accuracy of the LRG-QAMS and AVG-QAMS method, respectively. Some interesting findings different from previous studies attributed to the diverse results of varied ingredients were illustrated and discussed in our paper. The AVG-QAMS and LRG-QAMS methods could be used as a substitute for the external standard method when standard substances are lacking, and had their respective particulars. Additionally, in order to ensure the high accuracy of the QAMS method, the choice of AVG or LRG was introduced. Moreover, the content of the eighteen analytes from HPLC profiles was applied for HCA, PCA and similarity analysis to identify particular herbs from related species and evaluate the similarities and differences among Ilex kudingcha C. J. Tseng samples. This high throughput method has a promising potential to play a very substantial and enhanced role in the development of standards for the quality control of Ilex kudingcha C. J. Tseng.