Fingerprint Analysis of Desmodium Triquetrum L. Based on Ultra Performance Liquid Chromatography with Photodiode Array Detector Combined with Chemometrics Methods

Abstract

A fingerprinting approach was developed by means of ultra high-performance liquid chromatography with photodiode array detector for the quality control of Desmodium triquetrum L., an herbal medicine widely used for clinical purposes. Ten batches of raw material samples of D. triquetrum were collected from different regions of China. All UPLC analyses were carried out on a Waters ACQUITY UPLC BEH shield RP18 column (2.1 × 50 mm, 1.7 µm particle size) at 60°C, with a gradient mobile phase composed of 0.1% aqueous formic acid and acetonitrile at a flow rate of 0.45 mL/min. The method validation results demonstrated the developed method possessing desirable reproducibility, efficiency, and allowing fingerprint analysis in one chromatographic run within 13 min. The quality assessment was achieved by using chemometrics methods including similarity analysis, hierarchical clustering analysis and principal component analysis. The developed method can be used for further quality control of D. triquetrum.

Introduction

Desmodium triquetrum L. is widely distributed in sub-tropical and Pacific regions of the world and was recorded in the 1977 edition of “China Pharmacopoeia”. Traditionally, its leaves have been used for the treatment of diabetes, obesity, flu fever, sore throat, nephritis, cholestatic hepatitis, enteritis, bacillary dysentery, pregnant vomiting and prostatic hyperplasia worldwide (1–3). With increased application, it becomes more important to control and evaluate its quality.

As a promising tool in qualitative and quantitative analysis of complex natural products, fingerprint technology provides a comprehensive picture of all active biological and pharmaceutical components in the plant, and it has been accepted by World Health Organization (WHO) (4), the State Food and Drug Administration (SFDA) of china (2000) (5) and other authorities (6, 7). Traditionally, high-performance liquid chromatography (HPLC) has played an important role in fingerprint technology. However, HPLC cannot meet the requirements of high-throughput analysis because of the low efficiency and long analysis time, generally more than 1 h (8–10). Recently, ultra performance liquid chromatography (UPLC) as a newly developed and efficient technique has been used in the research as a chromatographic fingerprint (11, 12). Compared with the traditional HPLC method, it has several unparalleled advantages. It can enhance the effectiveness of liquid chromatography separation, not only to improve the resolution, but also to check the sensitivity; moreover, the speed of analysis is greatly improved. All of the advantages of UPLC make it a powerful tool for the phytochemical analysis (13–18).

Due to the limited studies on the quality control of D. triquetrum, an investigation of the fingerprint analysis method was performed in this paper. UPLC was applied for discriminating D. triquetrum from different cultivated regions, and chemometrics methods including similarity analysis (SA), hierarchical clustering analysis (HCA) and principal component analysis (PCA) were used to further confirm the performance of experimental results.

Experimental

Plant material

Ten batches of raw material samples of D. triquetrum were collected from Anhui, Fujian, Jiangsu, Yunnan, Hunan, Guangxi and Guangdong provinces of China at different time period. The various sources of samples were shown in Table I. The samples were identified based on morphological characteristics by the director Jia-jian Zhang from Zhejiang Academy of traditional Chinese medicine.

Table I.

Raw Samples of D. triquetrum Investigated in This Work

No.	Source	Collection date
S1	Anhui, China	May 2014
S2	Anhui, China	July 2014
S3	Fujian, China	July 2014
S4	Jiangsu, China	May 2014
S5	Guangdong, China	June 2014
S6	Hunan, China	June 2014
S7	Yunnan, China	July 2014
S8	Guangxi, China	June 2014
S9	Guangxi, China	May 2014
S10	Guangdong, China	July 2014

Chemicals and reagents

Standard nicotiflorin (purity ≥98.0%) was purchased from Tianjin Yifang Technology Co., Ltd (Tianjin, China); rutin (purity ≥98.0%) was acquired from National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China); kaempferol (purity ≥98.0%) was obtained from Aladdin Chemistry Co., Ltd (Shanghai, China). HPLC grade acetonitrile and methanol were purchased from Merck (Darmstadt, Germany). All other analytical grade chemicals used in this paper were purchased from Yongda Chemical Reagent Company (Tianjin, China). Ultrapure water with resistivity above 18 MΩ·cm was obtained by Barnstead TII super Pure Water System (MA, USA).

Apparatus and chromatographic conditions

Chromatographic analysis was performed by a Waters Acquity UPLC™ system (Waters, Milford, MA, USA) equipped with binary solvent delivery pump, an auto sampler and photodiode array detector (PAD) and connected to Empower 2 Pro software (Waters). Chromatographic separation was performed on a Waters Acquity UPLC BEH shield RP18 (2.1 × 50 mm i.d., 1.7 µm). The mobile phase was composed of 0.1% aqueous formic acid (v/v, solvent A) and acetonitrile (solvent B). The eluting conditions were optimized as follows: 0–4.5 min, 5–15% B; 4.5–6 min, 15% B; 6–9 min, 15–25% B; 9–11 min, 25–50% B; 11–13 min, 50% B; the total flow rate was 0.45 mL/min and the column temperature was maintained at 60°C. The monitoring wavelength was set at 340 nm, and the online ultraviolet absorption spectra were recorded in the range of 190–400 nm. The injection volume was 1.0 µL.

Sample preparation

All the samples were milled into powder and oven-dried at 50°C until they reached a constant weight. About 0.5 g powder of dried D. triquetrum was accurately weighed into a 10 mL conical flask, with adding precisely 8 mL methanol. Then, it was subjected to a 20 min ultrasonic extraction at 50°C. The extracted solution was prepared by the method of weight relief by which the weight lost in the extraction procedure was compensated. The sample solution was subsequently filtered through a 0.22 µm PTFE membrane (Welch, Shanghai, China) and injected into the UPLC system for analysis.

Standard solution preparation

Accurately weighed three compounds rutin, nicotiflorin and kaempferol were dissolved in acetonitrile to prepare stock solutions. The concentrations of stock solutions were 512 µg/mL for rutin, 262 µg/mL for nicotiflorin and 260 µg/mL for kaempferol. The stock solutions were serially diluted, mixed and used for preparation of standard solutions. The solutions were then diluted stepwise with acetonitrile to give seven different concentrations in the range of 1.02–102 µg/mL for rutin, 1.05–105 µg/mL for nicotiflorin and 0.208–20.8 µg/mL for kaempferol. All the solutions were filtered through 0.22 µm filter and stored at 4°C before analysis.

Validation of UPLC method

The calibration curve was created by running three mixed standards of different concentrations in triplicate. The correlation coefficient was determined using a linear regression model. The limit of detection (LOD) and limit of quantification (LOQ) for three compounds were estimated at signal-to-noise ratios (S/N) of 3 and 10, respectively, by injecting a series of dilute solutions with known concentration.

The repeatability, reproducibility and stability experiments were performed on sample S1 to evaluate the method. The repeatability was evaluated by six repeated runs of the same sample solution within a single day. The reproducibility was measured by running six replicates of the same sample solution prepared independently in a single day. Meanwhile, solution stability was determined by running five times of the same sample in the same day (every 2 h). The relative standard deviations (RSD) of relative retention time (RRT) and relative peak area (RPA) of 18 typical components of each test were calculated.

Data analysis

Data analysis was accomplished through three steps. SA was performed by the Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (Version 2004A), which was recommended by SFDA (19, 20). HCA and PCA were applied to demonstrate the variability of the chromatographic fingerprinting analysis in 10 batches of D. triquetrum samples collected from different localities by using SPSS statistics software (SPSS for Windows 19.0, IBM, USA) and the Unscrambler X 10.0 software from Camo AS (Trondheim, Norway), respectively.

Results

Optimization of extraction conditions

To achieve efficient extraction of active components, various factors such as extraction solvent [methanol, ethanol, 50% aqueous methanol, 50% aqueous ethanol, ethyl acetate, 50% ethyl acetate-methanol (v/v)], temperature (30, 40, 50 and 60°C) and extraction time (10, 20, 30 and 60 min) were investigated. The influence of each factor was studied by single factor experiment. The results suggested that ultrasonic-assisted with methanol at 50°C for 20 min was effective and convenient for extraction.

Optimization of UPLC condition

To achieve accurate and valid chromatographic conditions, different UPLC parameters were investigated and optimized, including various columns (Acquity UPLC BEH shield RP18 2.1 mm × 50 mm i.d., 1.7 µm; UPLC HSS T3 2.1 × 50 mm i.d., 1.8 µm; UPLC BEH Phenyl 2.1 × 50 mm i.d., 1.7 µm), mobile phases (methanol–water and acetonitrile–water with different modifiers, including formic acid, acetic acid and phosphoric acid), column temperature (30, 35, 40, 50, 55 and 60°C) and the flow rate (0.15, 0.20, 0.25, 0.30, 0.35, 0.40 and 0.45 mL/min). Additionally, a wavelength ranging from 190 to 400 nm was scanned by PAD and the wavelength of 340 nm was chosen, where most compounds could be detected and had adequate absorption. These optimized conditions were further developed for resolution, baseline and analysis time.

Validation of the method

Linearity and detection limit

The linearity for three analytes was established by plotting the peak area (y) versus concentration (x) of each analyte which was expressed by the equation given in Table II. The linearity of the calibration curves had been verified by correlation study and the correlation coefficients were all better than 0.9999 within test ranges. The LODs and LOQs for the three analytes were <0.21 and 0.52 µg/mL, respectively.

Table II.

Linearity, LODs and LOQs for Three Analytes

Peak no.	Analyte	Regression equation (n = 3)	R2	Linear range (μg/mL)	LOD (μg/mL)	LOQ (μg/mL)
9	Rutin	y = (3346 ± 12)x − 632.7 ± 6.4	1.0000	1.02–102	0.20	0.51
13	Nicotiflorin	y = (3796 ± 4)x + 335.2 ± 7.1	0.9999	1.05–105	0.21	0.52
18	Kaempferol	y = (5732 ± 19)x − 456.6 ± 5.1	1.0000	0.208–20.8	0.06	0.21

Repeatability, reproducibility and stability

The repeatability, reproducibility and stability tests were performed on sample S1. Peaks that existed in all 10 samples with reasonable heights and good resolution were assigned as “characteristic peaks” for the identification of the plant. There were 18 characteristic peaks (1–18) within 13 min, shown in Figure 1. Peaks 9, 13 and 18 were identified as rutin, nicotiflorin and kaempferol, respectively, by using the standards and UPLC-TOF/MS (Waters AcquityUPLC-BrukermicrOTOF-QII). Peak 13 (nicotiflorin) was signed as reference peak to calculate RRT and RPA. The RSDs of RRT and RPA of each common peak were calculated for the estimation of repeatability, repeatability and stability. The RSDs of RRT and RPA of characteristic peaks in the repeatability test were below 0.21 and 2.81%, whereas in the reproducibility test, the RSDs of RRT and RPA were not exceeding 0.17 and 3.25%, respectively. And in the stability test, the RSDs of RRT and RPA were <0.32 and 2.83%, respectively. These results were summarized in Table III.

Figure 1:

The UPLC chromatograms of mixed standards (A) and sample extract (B). Peak 9, rutin; Peak 13, nicotiflorin; Peak 18, kaempferol.

Table III.

The Validation Data of the UPLC–PAD Fingerprint Method

Peak no.	Repeatability (RSD, %, n = 6)		Reproducibility (RSD, %, n = 6)		Stability-RSD (RSD, %, n = 6)
	RRT	RPA	RRT	RPA	RRT	RPA
1	0.20	1.17	0.17	2.52	0.32	2.83
2	0.21	2.59	0.14	3.25	0.26	1.81
3	0.17	1.19	0.09	1.49	0.13	1.62
4	0.21	2.48	0.06	1.89	0.11	1.56
5	0.19	0.97	0.08	1.06	0.06	0.31
6	0.17	1.04	0.10	0.84	0.07	0.69
7	0.14	0.52	0.07	0.79	0.05	0.60
8	0.11	0.38	0.06	0.93	0.03	0.98
9	0.12	2.81	0.05	3.22	0.04	1.72
10	0.13	1.42	0.06	1.69	0.04	0.49
11	0.13	0.82	0.05	1.57	0.04	0.67
12	0.15	1.33	0.05	1.61	0.05	1.24
13	0.15	0.57	0.05	1.67	0.05	0.77
14	0.13	0.51	0.04	1.02	0.04	0.69
15	0.11	1.04	0.03	0.86	0.04	0.48
16	0.08	0.75	0.03	0.25	0.03	0.25
17	0.06	1.49	0.04	2.98	0.01	1.91
18	0.02	0.69	0.01	0.79	0.01	0.62

UPLC fingerprint of D. triquetrum and SA

In order to evaluate the similarities and differences in these samples, a mean chromatogram was established to calculate as a representative object, and the similarity was evaluated by the calculation on the correlation coefficient of original data. The closer the similarity values approached 1, the more similar the two chromatograms were. Chromatograms of 10 batches of D. triquetrum samples were shown in Figure 2. The correlative coefficient (r) between each chromatogram of D. triquetrum sample and the simulative mean chromatogram was 0.979, 0.997, 0.987, 0754, 0.990, 0.959, 0.939, 0.944, 0.983 and 0.981, respectively. These results (Table IV) indicated that the internal qualities of the samples from four areas harvested in different times were different. The sample from Anhui province with harvesting time at July 2014 (S2 in Table I) had the biggest correlation coefficient (0.997) among these samples, while the sample from Jiangsu province at May 2014 (S4) had the adverse result.

Figure 2.

UPLC–PAD fingerprints of samples from various sources.

Table IV.

The SA Results of D. triquetrum Samples

	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10	Sa
S1	1
S2	0.983	1
S3	0.969	0.991	1
S4	0.698	0.767	0.788	1
S5	0.976	0.988	0.977	0.747	1
S6	0.943	0.957	0.921	0.726	0.948	1
S7	0.858	0.927	0.924	0.737	0.908	0.892	1
S8	0.876	0.934	0.950	0.741	0.919	0.846	0.974	1
S9	0.992	0.982	0.975	0.694	0.976	0.932	0.877	0.900	1
S10	0.973	0.977	0.945	0.716	0.974	0.988	0.898	0.878	0.964	1
Sa	0.979	0.997	0.987	0.754	0.990	0.959	0.939	0.944	0.983	0.981	1

^aS represents the mean chromatography.

Hierarchical clustering analysis

In order to evaluate the resemblance and differences in these samples, a hierarchical agglomerative clustering analysis of ten batches of D. triquetrum was performed based on RPA of the 18 characteristic peaks, in which nicotiflorin (Peak 13) was assigned as the reference peak. Between-groups linkage method was used as the amalgamation rule and the squared Euclidean distance was used as the metric. The HCA results were presented as a dendrogram (Figure 3), providing clearer visualization of data in a high-dimensional matrix (10 objects × 18 variables). Depending on the distance, it was clear that the samples were divided into three clusters obviously. Cluster Ι was formed by the sample S2, S3, S7 and S8. Cluster ΙΙΙ consisted of S4, and Cluster ΙΙ consisted of the remaining samples. Cluster ΙΙΙ of sample S4 from Jiangsu province had the longest distance to other clusters, showing the obviously different chemical fingerprints and internal quality of S4 from other samples.

Figure 3.

HCA results for the chemical fingerprints of 10 batches of D. triquetrum samples.

Principal component analysis

In order to evaluate the quality variation and differentiate the sources of D. triquetrum samples, PCA, a method for feature extraction and dimensionality reduction, was carried out. PCA was processed by using RPA of 18 characteristic peaks from 10 chromatograms as input data for calculation and all data were mean-centered. On the basis of eigenvalues >1, the result was illustrated in Figure 4, PC1 and PC2 (first and second principal components) were chosen to provide the highest variation of data objects (51 and 31% of the variation) for convenient visualization and differentiation. As shown in the score plot of PC1 (x) and PC2 (y), PCA displayed the results that 10 batches of D. triquetrum samples were clustered into three domains, similar to HCA. S2, S3, S7 and S8 were in Domain Ι, S1, S5, S6, S9 and S10 were in Domain ΙΙ, while S4 was in Domain ΙΙΙ. Moreover, the results of the correlation loading plot of the PCA (Figure 5) indicated that Peaks 2–4, 6, 7, 9–14, 17 and 18 might have more influence on the discrimination of the samples from different localities than other compounds. Therein, Peaks 9 (rutin), 13 (nicotiflorin) and 18 (kaempferol) could be identified accurately.

Figure 4.

PCA score plot of D. triquetrum samples.

Figure 5.

The correlation loadings plot from PCA for the 18 characteristic peaks.

Discussion and conclusion

This work developed for the first time a UPLC method for the fast, simultaneous determination of rutin, nicotiflorin, kaempferol and the UPLC-based fingerprint profile in D. triquetrum. Our method validation results demonstrated the developed method possessing desirable repeatability (<2.81% RSD), reproducibility (<3.25% RSD), stability (<2.83% RSD), which indicated that the method of UPLC fingerprint analysis was valid and satisfactory.

For the fingerprint analysis of D. triquetrum, chemometrics techniques, such as SA, HCA and PCA, were successfully applied to comprehensive chemical analysis of D. triquetrum samples from different regions to explain the difference. The SA results were in accordance with the actual patterns and peak shapes of the chromatograms. Moreover, the results of the HCA were in agreement with the results obtained by the PCA, both of them clustered the samples into three classes, showing the resemblance and differences among the 10 samples. S4 from Jiangsu province was a little differentiated from other samples, which was probably due to some conceivable factors, such as the different geographical origins, cultivated soils and harvesting times. The results from three chemometrics methods above were basically consistent with each other and could provide more references for the quality evaluation of D. triquetrum samples.

In short, this work developed for the first time a UPLC method for the fast, simultaneous determination of rutin, nicotiflorin, kaempferol and the UPLC-based fingerprint profile in D. triquetrum. Furthermore, the analytical method described here was fast, simple and sensitive with good accuracy and precision. The advantages of UPLC with shorter analysis times, higher sensitivity, reduced quantity of solvent consumption as well as simpler manageability facilitated the quality control of D. triquetrum which also provided an important reference to establish the quality control method for other related traditional Chinese medicine herbs.