Multicomponent Analysis of Liuwei Dihuang Pills by a Single Marker Quantification Method and Chemometric Discrimination of Fingerprints

An effective and comprehensive quality evaluation method for Liuwei Dihuang pills (LDP) was established by the simultaneous determination of 8 active components in LDP by the quantitative analysis of multicomponents by single marker (QAMS) method and high-performance liquid chromatography (HPLC) fingerprint combined with chemometrics. These 8 active components were determined by QAMS and the external standard method (ESM), and the quantitative results of the two methods were compared to validate the accuracy and feasibility of the QAMS method. 8 active components showed good linear relationships within their ranges, whose average recoveries were 99.7∼102.3%. No significant difference was found (P > 0.05) in the quantitative results determined by QAMS and ESM. Furthermore, the fingerprint of LDP was also established, with 11 common peaks identified, and the similarity of the fingerprints of 21 batches of LDP was greater than 0.95. The 21 batches of LDP were basically divided into 3 groups by hierarchical cluster analysis (HCA) and principal component analysis (PCA), and 3 differential markers were screened out by orthogonal partial least squares discriminant analysis (OPLS-DA). The established QAMS method is accurate, economical, fast, and convenient and can simultaneously determine the content of 8 active components in LDP. HPLC fingerprint combined with chemometric analysis more comprehensively evaluated the quality consistency of different batches of LDP and analyzed the markers that cause quality differences between batches. It can provide a scientific basis and reference of quality consistency evaluation for the manufacturers and drug regulatory departments of the preparation.


Introduction
Liuwei Dihuang pills (LDP) is a classic prescription for nourishing yin and tonifying kidney.It is composed of 6 herbs: Rehmannia glutinosa, Cornus ofcinalis (manufactured), Cortex Moutan, Rhizoma Dioscoreae, Poria cocos, and Alisma orientalis [1].LDP has thousands of years of clinical application history in China.It is used to treat kidney yin defciency, dizziness, tinnitus, soreness and weakness of waist and knees, bone steaming and hot fashes, essence, and qi night sweats.In addition, it has anti-infammatory, hypoglycemic, antioxidant, enhancing immune function, delaying aging, anticancer, and other efects [2].Recently, LDP is more and more extensively used and has played a good role in the prevention, treatment, and adjuvant treatment of tumors [3].Compared with western medicine, LDP combined with western medicine can better reduce systolic blood pressure and diastolic blood pressure and has advantages in clinical efcacy and antihypertensive efect [4].
Te quality evaluation indicators of LDP in the "Chinese Pharmacopoeia" (2020 edition) are 3 active ingredients: morroniside, loganin, and paeonol [1], which are too few to refect its quality comprehensively and accurately.Some studies have also used the external standard method (ESM) for the simultaneous determination of multiple active ingredients of LDP by high-performance liquid chromatography (HPLC) [5][6][7][8][9], but the ESM requires a large number of standards and is costly.Te application of ESM and internal standard methods will be limited when the reference is insufcient or unstable, especially in multicomponent quantitative and quality control [10].
Te TCM fngerprint can refect the results of a multicomponent and multitarget synergistic treatment of TCM based on the overall understanding of its chemical compositions.Te internal quality of the compound can be refected to a certain extent by identifying the fngerprint, so as to control the overall quality of TCM compound preparations [33,34].Chemical pattern recognition is considered to be a more objective and efective method for identifying the specifc consistency and stability of TCM products and evaluating the determination of single or multiple markers.Chemometric analysis methods such as similarity analysis, hierarchical clustering analysis (HCA), and principal component analysis (PCA) are widely used in chemical classifcation and chromatographic profle analysis [35].Currently, there is no literature report on the quality consistency of LDP and the screening of quality diferential markers between various batches by fngerprint combined with chemometrics.
In this research, to establish an efcient, economical, and practical quality evaluation method for LDP, 11 common peaks were identifed in the established fngerprint, and chemometrics approaches were used to extract the covered information and knowledge from chemical systems.Te method is accurate, reliable, simple, and robust.Combined with QAMS, it can comprehensively and scientifcally evaluate the quality of LDP and provide a scientifc basis and reference for the quality consistency control of LDP for manufacturers and drug regulatory authorities.

Software Methods. "Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese
Medicine" (version 2012A, Chinese Pharmacopoeia Commission, Beijing, China) was used to establish the fngerprint and analyze the similarity by importing the chromatogram of 21 batches of LDP."Excel" was used to conduct the radar plot analysis."SPSS 23.0" data analysis software was used to cluster samples."SIMCA 14.1" software was used to identify the principal component.LDP was crushed by a multifunction pulverizer for 2 min.1.65 g of fne powder of LDP was accurately weighed and placed in a 50 mL measuring fask, and then about 40 mL of 70% methanol was added.Te mixture was sonicated for 1 h, cooled, diluted with 70% methanol to scale, and fltered with a 0.22 μm microporous membrane, and the subsequent fltrate was taken as the sample solution.

Validation of Analytical Method.
Te mixed standard solution and the sample solution were injected with 20 μL, respectively, and analyzed under the 200-400 nm full scanning wavelength to test the system suitability and specifcity.Te mixed standard solution was injected with 20 μL to obtain the peak areas, and the linear regression was carried out with the concentration as the abscissa (X) and the peak area as the ordinate (Y).Te mixed standard solution was injected 6 times continuously to evaluate the precision.Te same batch of LDP (S3) was taken to prepare 6 sample solutions in parallel and determined to investigate the repeatability of the established method.Te stability was tested by injecting the same sample solution (S3), respectively, after being placed at room temperature for 0, 4, 8, 12, 18, 24, and 36 h.Te known contents of LDP (S3) powder were precisely weighed and placed in a 50 mL volumetric fask (n � 9) and then divided into three groups on average.Each group was precisely added with about 80%, 100%, and 120% of the standard solution equivalent to the content of each component in the sample.Te peak area was recorded, and the recovery of each component was calculated.

Establishment of HPLC Fingerprint and Chemometrics
Analysis.Te 21 batches of LDP were prepared according to the preparation method of the sample solution.Te samples were determined, and the chromatograms were recorded.After the unifed integration of all chromatograms, the obtained chromatograms were imported into the "Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (2012 Edition)" software in AIA format to establish the HPLC fngerprint.Te similarity of fngerprints of 21 batches of LDP was also calculated by the abovementioned software.Taking the peak area of 11 common peaks of 21 batch samples as variables, using SPSS 23.0 data analysis software, the intergroup connection method and Euclidean distance were selected for systematic HCA.Te peak areas of common peaks in 21 batches of samples were imported into SIMCA 14.1 software for PCA, and the data were ftted by Ctr's scaling method to obtain the corresponding score map.In order to fnd the characteristic components that cause quality diferences between various batches, SIMCA.14.1 was used to conduct orthogonal partial least squares discriminant analysis (OPLS-DA).

Methodology Verifcation.
Tere were chromatographic peaks at the corresponding positions in the chromatograms of the sample solution and the mixed standard solution, and the spectral characteristics of the chromatographic peaks were consistent with those of the standards.Te peak purity of each target component was detected, and the results met the requirements (the purity angle was less than the purity threshold).Te separation degree of each peak is better than 1.5, and the number of theoretical plates is not less than 3000.As shown in Table 1, each r of the 8 components within their linear range was not less than 0.9991, indicating that the method had a satisfactory linear relationship.Te RSDs of the 8 components were no more than 0.33%, 0.95%, and 0.92%, demonstrating the instrument had good precision, the method had satisfactory repeatability, and the sample solution was stable within 36 h at room temperature.Te average recoveries of the 8 ingredients were in the range of 99.1%∼102.2%(Table 2), manifesting the presented method was accurate.Te abovementioned experimental results indicated that the HPLC method was accurate and reliable for the quantitative analysis of LDP.

Optimization of Chromatographic Condition.
Te mobile phase composition of this method was mainly referred to the "Chinese Pharmacopoeia" 2020 edition, using an acetonitrile-0.3%phosphoric acid solution for gradient elution.Considering the acidity of the 0.3% phosphoric acid is not conducive to the protection of the chromatographic column, 0.1% and 0.2% phosphoric acid aqueous solutions were also investigated.Te results showed that 0.1% phosphoric acid reduced the separation degree by reducing the number of column plates with a peak deformation diference.When 0.2% phosphoric acid was used, the performance of each chromatographic peak was similar to 0.3% phosphoric acid, and the acidity of the mobile phase was weakened, which was benefcial to prolong the life of the column.Consequently, an acetonitrile-0.2%phosphoric acid solution was selected for gradient elution.Te chromatographic peaks of each component were scanned at full 4 Journal of Analytical Methods in Chemistry wavelength using a diode array detector.Gallic acid had the maximum absorption at 216 nm and 271 nm, 5hydroxymethylfurfural had the maximum absorption at 230 nm and 285 nm, morroniside had the maximum absorption at 241 nm, paeoniforin had the maximum absorption at 246 nm, loganin had the maximum absorption at 237 nm, paeoniforin had the maximum absorption at 232 nm and 274 nm, cornuside had the maximum absorption at 218 nm and 273 nm, and paeonol had the maximum absorption at 211 nm, 228 nm, 274 nm, and 312 nm.Due to the diferent contents of each component in the sample, taking into account the response value of each component, 240 nm was selected as the detection wavelength.Under the chromatographic conditions, the chromatogram showed a smooth baseline and an efcient separation degree of the target components which meet the system suitability requirements.

Calculation of Relative Correction Factor (RCF).
Te dominant advantage of QAMS is that only one standard is needed in the daily inspection work to quantitatively determine the multicomponent as long as the method was established.It greatly saves the number of standards and has the characteristics of simplicity, easy operation, and low cost.
Te selection of internal standards should be in accordance with the principles of stability, low price, availability, low toxicity, and efectiveness.After comparison, among the components to be measured, paeonol had high content, stable property, and optimal chromatographic peak shape and are nontoxic, and the standard was the cheapest and easy to obtain, and hence, paeonol was selected as the internal standard.Te relative correction factors calculated under various instruments and columns with paeonol as the internal standard were satisfactory, manifesting that the established QAMS method had great durability and stability.
Te relative correction factors of gallic acid, 5-HMF, morroniside, sweroside, loganin, paeoniforin, and cornuside were calculated with paeonol as the internal standard.Te results are shown in Table 3.
Te formula is as follows: where A s , C s , A i , and C i represent the peak area of the internal standard, the concentration of the internal standard, the peak area of component i, and the concentration of component i.According to formula (1), the concentration of a component i(C i ) can be calculated by using the following formula:    4; the RSD values of all RCFs under diferent experimental conditions were less than 5%, indicating that the QAMS had good durability and stability.Diferent instruments and chromatographic columns had no signifcant infuence on the relative correction factors of each component.

Te Location of the Analytes' Chromatographic Peaks.
With the paeonol chromatographic peak as the reference, the relative retention times (RRT) of gallic acid, 5hydroxymethylfurfural, morroniside, sweroside, loganin, paeoniforin, and cornuside were calculated, respectively.Te reproducibility of the RRT of each component under diferent HPLC conditions and diferent chromatographic columns was investigated.As shown in Table 5, the RRT fuctuation of each component was small.In this study, the relative retention time was used as the positioning index of the target chromatographic peak of QAMS, and the position of the target peak was further accurately determined by combining the UV absorption characteristics of each chromatographic peak in the sample solution.

Sample Determination.
Te concentrations of the 8 efective components in 21 batches of LDP determined by the ESM and QAMS methods are listed in Table 6.Te content of 8 active components in 21 batches of LDP was diferent.Paeonol had the highest content, followed by 5-HMF, morroniside, and loganin.Paeoniforin, sweroside, and cornuside had relatively low content, and paeoniforin was not detected in some batches.Among the 21 batches of samples, gallic acid had the largest content diference, followed by paeoniforin, cornuside, 5-HMF, loganin, and paeonol.It shows that there are some diferences in the content of each active component between diferent batches, which may be related to the source of medicinal materials and the extraction and production processes of medicinal materials.For manufacturers, the source of herbs should be strictly controlled, and the extraction and production processes should be standardized to ensure the consistency of drug quality.6. Te confdence interval was 95%, and the P values were all greater than 0.05, indicating that there was no signifcant diference in the content measured by the two methods, indicating that QAMS can be well used for the simultaneous determination of eight active components.Te QAMS is a reliable and convenient method for the determination of multicomponent content, especially in the absence of reference substances.Tis strategy reduced the experimental cost and assay time, and the QAMS method was applied accurately to the quantitative analysis of LDP.

Establishment and Similarity Evaluation of HPLC
Fingerprint.Te chromatogram of the S1 sample with a steady baseline, good peak shape, and separation was selected as the reference.Te multipoint correction method was used to match the whole peak of the chromatogram with the time window width of 0.2, the control fngerprint (R) was generated by the median method, the HPLC fngerprint of 21 batches of samples was generated, and 11 common peaks were identifed (Figure 3).Among the 11 common peaks, peaks 1, 2, 6, 7, 8, 10, and 11 were identifed as gallic acid, 5hydroxymethylfurfural, morroniside, sweroside, loganin, cornuside, and paeonol.Using the chromatogram of the S1 sample as the reference chromatogram, the RSDs of the retention time of each common peak were less than 3.19%, and the RSDs of each common peak area were distinct.Te similarity results of 21 batches of samples were more than 0.95, indicating that the consistency of LDP quality was good and the chemical composition of LDP was similar, but the content of common peak components varied greatly among batches.Consequently, chemical pattern recognition should be carried out to refect the intrinsic quality of LDP more objectively.

Hierarchical Cluster Analysis
. HCA, which is to say, "objects are clustered by class," is a practical method in multivariate statistics [36].It mainly achieves the purpose of classifcation by using the principle that the same kind of  samples are similar to each other, and the distance of similar samples in multidimensional space is smaller, while the distance of diferent samples in multidimensional space is larger.HCA was performed on SPSS 23.0 data analysis software by selecting the intergroup connection method and Euclidean distance with the peak area of 11 common peaks in 21 batches of LDP as variables.As shown in the Figure 4(a), the Euclidean distance "10" was selected as the judgment basis, 21 batches of samples were clustered into 3 categories, S1-S6 were clustered into one category, S16, S21, and S17 were clustered into the same category, and the rest of the batches were clustered into another category.Adjacent batches of LDP are relatively close in production time, and the harvesting season, sources of medicinal materials, and processing methods of the prescription medicinal materials may be more similar, so they are more likely to be grouped together.Since it is impossible to guarantee the production conditions are exactly the same in every production, it is reasonable to have some quality diferences between samples from batch to batch.In addition, diferences in the quality of the samples may be infuenced by the source of the herbs, the production process, and the conditions of transport and storage.So the production of Chinese patent medicines should be considered comprehensively, taking into account various variable factors, so that the production quality of each batch of medicinal materials is consistent.

Principal Component Analysis.
PCA is an unsupervised pattern recognition method.On the basis of the dimension reduction idea, it can transform the multiple indicators into several independent synthetic indicators which contain most of the information in the original ones.After the transformation, the data matrix is simplifed, the dimension is reduced, and a few principal components linearly combined by the original variables are found, which is convenient for the multivariate statistical method of extracting chemical information [37].Te results of the scatter plot of principal component scores (Figure 4(b)) showed that the contribution rates of principal component 1 and principal component 2 are 53.9% and 19.3%, respectively, and the cumulative contribution rate reaches 73.2%.Terefore, the two principal components can comprehensively refect the total compositional content of the peak.Te 21 batches of samples can be divided into three groups: samples S16 and S17 were in one group, S1-S6 belonged to another group, and the rest of the samples were in another group.Except for S21, the results of PCA are generally consistent with the results of HCA.Te results indicate that samples produced at similar times have high similarity.Te scatter diagram of the principal component loading (Figure 4(c)) revealed the proportion of each chromatographic peak in the principal component, where the abscissa represents the loading of each substance on the principal component 1 and the ordinate represents the loading of each substance on the principal component 2. Te farther the peak is from the Y-axis, the greater the contribution to the principal component 1, and peak 2 (5-HMF), peak 6 (morroniside), peak 7 (sweroside), and peak 10 (cornuside) had an important contribution to the principal component 1 when the absolute value of the X-axis distance was limited to 3.5.Te farther the peak is from the X-axis, the greater the contribution to principal component 2, and peaks 3 and 4 had an important contribution to the principal component 2 when the absolute value of the Y-axis was bounded by 3.5.5-HMF, morroniside, sweroside, and cornuside were all bioactive components in LDP and evaluated in QAMS.

Orthogonal Partial Least Squares Discriminant
Analysis.OPLS-DA is a supervised pattern recognition method for multivariate statistics of massive data to explore the diferences of indicators between groups [38].As shown in Figures 4(d in the prescription of LDP, Rehmannia glutinosa is the principal drug and Cornus ofcinalis is the adjuvant drug.Te research results just showed that the quality control of the principal drug and adjuvant drug should be more important. 3.9.Radar Plot Analysis.Radar plot is an important descriptive tool for multivariate data.Generally speaking, radar plots are a circular graphic approach that projects a series of spokes or rays from the center point, each ray represents diferent variable labels [39].Te determination results of 21 Te results of OPLS-DA showed that gallic acid, 5-HMF, and morroniside were the main markers afecting the quality of LDP.From Figure 5(d), it can also be found that gallic acid, 5-HMF, and morroniside were the 3 components with the greatest variation in content among the 8 active components.It was suggested that the components with large content diferences may be the main quality markers that afect the quality diferences of diferent batches of LPD.

Conclusions
Te established QAMS method used paeonol as the internal standard, and the RCF calculation method was used to determine the content of 8 active components in 21 batches of LDP, and the methodology was verifed.Te RCF values of diferent chromatographic columns and instruments were stable.Te results showed that there was no signifcant diference between QAMS and ESM.Te QAMS method is economical and fast and can evaluate the quality of LDP more comprehensively and scientifcally than the quality control index of only 3 components in the existing standard.Te analytical method proposed in this paper was established for the quality control of LDP for the frst time.Te HPLC fngerprint of LDP was also established with 11 common peaks were identifed and the similarities of fngerprints were greater than 0.95, demonstrating the quality of LDP is consistent between diferent batches.Te 21 batches of LDP were basically divided into 3 groups by HCA and PCA, and OPLS-DA screened out 3 diferential markers in various batches.HPLC fngerprint combined with chemometrics information more comprehensively evaluated the quality consistency of diferent batches of LDP and analyzed the markers that caused quality diferences between batches.Consequently, it can provide a scientifc basis and reference for the quality, consistency, and evaluation of LDP for the manufacturers and drug regulatory authorities of the preparation.
Te 21 batches of LDP were prepared as a sample solution and injected for determination.Te contents of 8 components

Table 2 :
Results of recovery test.

Table 3 :
Relative correction factors of various components.

Table 6 ,
there is no signifcant diference between the content values measured by the ESM and QAMS methods.Te results of ESM and QAMS were tested by SPSS 17 statistical software.Te results are shown in Table

Table 5 :
Relative retention times determined by diferent instruments and columns.

Table 4 :
Relative correction factors determined by diferent instruments and columns.

Table 6 :
Contents of 8 components in LDP by ESM * and QAMS * .