Relationship between the UPLC Fingerprints of Citrus reticulata “Chachi” Leaves and Their Antioxidant Activities

Citrus reticulata “Chachi” (CRC) leaves contain abundant flavonoids, indicating that they possess good nutritional/pharmacological research and development potential. This study aims to explore chemical antioxidant quality markers based on the spectrum-effect relationship and quality control strategy of CRC leaves. The ultrahigh performance liquid chromatography (UPLC) system was used to establish chromatographic fingerprints of Citrus reticulata “Chachi” leaves. Simultaneously, they were evaluated by using similarity analysis (SA), hierarchical cluster analysis (HCA), and principal component analysis (PCA). Afterwards, the DPPH assay was adopted to study the antioxidant effects. The spectrum-effect relationship between UPLC fingerprints and DPPH radical-scavenging activities was studied with grey relational analysis (GRA). Analysis results indicated that there were twenty-one common peaks of fourteen batches of CRC leaves which were from different regions of Guangdong province, and their similarities ranged from 0.648 to 0.997. HCA results showed that fourteen batches of samples of CRC leaves could be divided into six classes at Euclidean distance of 5. The results from GRA showed that tangeretin and hesperidin were the main flavonoids responsible for the antioxidant activity in CRC leaves. In conclusion, this research established a chromatographic analysis method suitable for CRC leaves and demonstrated that chromatographic fingerprints analysis combined with the antioxidant activity could be used to evaluate the material basis of CRC leaves and may provide a reference to establish a quality standard.


Introduction
Citrus reticulata "Chachi" (CRC) is a citrus plant of Rutaceae family and can be used as medicine and food. At present, the application of CRC is mainly in the peel which is the source of citri reticulatae pericarpium (CRP) in Guangdong. CRP is used to treat respiratory and digestive system disorders including dyspepsia,acid refux, constipation, and diarrhea, as well as the symptoms of other gastrointestinal diseases [1]. In order to protect and promote fruit, the leaves of CRC are regularly pruned every year, resulting in a large number of CRC leaves getting abandoned, which is a potential waste of resources. Only a small proportion of these leaves are used as foodstufs, e.g., for tea making and spices. Preliminary studies by our research group found that the leaves of CRC also contained abundant favonoids [2], indicating that they possess good nutritional/pharmacological research and development potential.
In order to expand the market application for CRC leaves, it is necessary to establish the quality control strategy and investigate the pharmacological efcacy. Terefore, this study aims to establish ultraperformance liquid chromatography (UPLC) fngerprints of fourteen batches of CRC leaves and to determine the contents of hesperidin, nobiletin, tangeretin, and 5-demethylnobiletin in fourteen batches of samples. Simultaneously, they are evaluated using similarity analysis (SA), hierarchical cluster analysis (HCA), and principal component analysis (PCA). Afterwards, the DPPH assay was adopted to study the antioxidant efects. Te spectrum-efect relationship between UPLC fngerprints and DPPH radical-scavenging activities were studied with grey relational analysis (GRA). Te correlation coefcient between the common peaks and the chemical antioxidant activity can be used to explore the quality markers of CRC leaves.

Materials and Reagents.
Fourteen batches of CRC leaves were collected from diferent areas of Guangdong province in China for analysis, and the detailed information is listed in Table 1. Te leaves were dried (60°C), reduced to coarse homogeneous powders, and stored in sealed containers at ambient temperature until required. Samples were authenticated by Prof. Kang Chen (College of Chinese Medicine, Guangzhou University of Chinese Medicine, Guangzhou, China).

Sample no
Origins S1 Cixi county, Guangdong province S2 Chakeng county, Guangdong province S3 Zhuwan county, Guangdong province S4 Nanhuan county, Guangdong province S5 Gujing county, Guangdong province S6 Lianhe county, Guangdong province S7 Gujing county, Guangdong province S8 Shuangshui county, Guangdong province S9 Tianbian county, Guangdong province S10 Dazhe county, Guangdong province S11 Yamen county, Guangdong province S12 Chakeng county, Guangdong province S13 Xinhui city, Guangdong province S14 Lianhe county, Guangdong province 2 International Journal of Analytical Chemistry simulated chromatogram, representative of fourteen fngerprints, was generated automatically by the software using the median method. Similarities between the chromatographic data from each of the diferent batches of samples and the reference chromatogram were then calculated.
2.5. HCA. Te areas of common UPLC peaks were defned as characteristics to classify fourteen samples using SPSS Statistics for Windows, Version 26.0 software (IBM Corp., Armonk, NY, USA) [5,6]. Te between-groups linkage method and square Euclidean distance were used to measure the closeness of areas of common peaks among diferent samples.
2.6. PCA. PCA was applied to observe the distribution of samples in multivariate dimensional space and explore the relations among the independent variables. Tus, the common peaks of fourteen samples were analyzed by PCA using SPSS Statistics for Windows, Version 26.0 software (IBM Corp., Armonk, NY, USA). A principal component (PC) loadings matrix was constructed using the 21 peaks common to fourteen samples. Te loadings values of each PC were then divided by the arithmetic square root of the eigenvalues of each autonomous component to give the linear model equations. After substituting the normalized peak areas of the 21 common peaks into the linear model, a PC score was obtained for each batch of samples.

DPPH Radical-Scavenging Assay.
A working solution of DPPH (25 mg in 250 mL ethanol) was prepared and stored in the dark. L-Ascorbic acid was chosen as positive control (20 mg in 10 mL 50% ethanol), and the above solution was diluted with 50% ethanol into diferent concentrations of test solution for the later experiment. Sample extracts were serially diluted by 6.25, 12.5, 25, 50, 100, 200, and 400 times. Te DPPH working solution (100 μL) was mixed with the diluted sample extract (100 μL) in a 96-well plate. After incubating for 30 minutes in the dark, the absorbance of the reaction solutions was measured at 517 nm. Te negative control solution without antioxidants was prepared and determined as the same manner. Te free radical-scavenging capacity was calculated as follows: free radical-scavenging capacity (%) � A0 − A1/A0 × 100%, where A 0 is the absorbance of negative control solution without antioxidants, and A 1 is the absorbance of sample solution which is a mixture of free radical working solution and sample solutions. Te results were expressed as the half maximal inhibitory concentration (IC50; mg/mL) [7,8].

GRA.
Grey relation analysis is an analysis method that includes multifactor statistical analysis. It uses grey relation grade to describe the degree of correlation between the data. IC 50 was defned as the reference sequence, and the areas of common peaks from the UPLC fngerprint were defned as the comparison sequences. Standardizing the peak area and IC50 values of each sample by using GRA, the absolute diference sequence, the correlation coefcient, and the correlation degree were obtained. Finally, the potential antioxidant active components were screened according to the correlation degree.

UPLC Fingerprints
To obtain stable and reproducible UPLC fngerprints, the retention times and peak areas for all characteristic peaks were measured relative to a reference peak (hesperin). Te regression equations of hesperin, tangeretin, nobiletin, and 5-demethylnobiletin (Table 2) all showed good linearity (R 2 > 0.999). Te RSDs for the precision, repeatability, stability, and recovery of the reference favonoids did not exceed 3% (Table 3). Te correlation coefcients between the fngerprints obtained by repeated injection of the same sample extract solution and the common pattern fngerprints obtained were >0.95.

Similarity Analysis.
Te overlaid chromatograms for samples from each batch of CRC are shown in Figure 1, and 3D chromatogram showing UV absorbance spectra at 300 nm is shown in Figure 2. Tere were 21 major peaks in UPLC chromatograms of all samples. Among these, four peaks were identifed as hesperidin (peak 17), nobiletin (peak 19), tangeretin (peak 20), and 5-demethylnobiletin (peak 21; Figure 3). To compensate for any drift in retention time and change in peak area response, peak 17 was assigned as the reference to calculate the relative retention times (RRT) and relative peak areas (RPA) of each characteristic peak (Tables 4 and 5).
Te results of the similarity analysis using SESCFTCM are listed in Table 6. Te similarity data indicated that fourteen batches of samples were all diferent but within a moderate and acceptable range. When the similarity threshold was set to 0.9, samples 8 and 10 were dissimilar.

Content of Flavonoids.
Amounts of each reference favonoid in samples from each batch of CRC (Table 7) were identifed and quantifed from their UV spectra ( Figure 4) and the slope and intercept of the linear regression calibration equations, respectively. Amounts of hesperidin were signifcantly greater than the other favonoids; tangeretin showed the lowest concentrations in samples.

Antioxidant Activity.
Te antioxidant activities of fourteen batches of samples were determined using the DPPH assay. Te results showed that the DPPH clearance rate was not linear with the concentration. Terefore, the Probit method in SPSS 26.0 software was used for regression analysis to ft the corresponding equation Probit (p) � Intercept + Bx, and then the IC50 values (the concentration of the sample solution when Probit � 0.50) was obtained. Te IC50 values (Table 8; Figure 5) showed that the clearance rates for each sample followed the order as follows: S14 > S7 > S1 > S13 > S5 > S2 > S6 > S12 > S3 > S9 > S4 > S10 > S8 > S11.
International Journal of Analytical Chemistry

HCA. Te interval of intergroup connection and square
Euclidean distance were used to establish a dendrogram of HCA of fourteen samples, which is shown in Figure 6. Te HCA results show that the samples were mainly clustered into four classes at Euclidean distance of 10 and six classes at Euclidean distance of 5. According to Table 1, the results of the HCA are not signifcantly correlated with their geographical locations. For example, S5 and S7, S12 and S2 are from same location, but they are grouped in diferent clusters. Since the bases where we collected CRC leaves are relatively limited, HCA results did not show obvious regional characteristics. At present, the main breeding methods of Citrus reticulata "Chachi" are ring-branch and grafting, and the favor of fruit cultivated in diferent ways is also diferent. Moreover, the suitability of soil physical and chemical properties is crucial to the growth and development of plants and the accumulation of secondary metabolites [9]. In summary, other factors such as grafting or not, rootstock varieties for grafting and soil environment may afect quality of CRC leaves. Terefore, it is necessary to expand the scope of sample collection and comprehensively consider the factors 160  155  150  145  140  135  130  125  120  115  110  105  100  95  90  85  80  75  70  65  60  55  50  45  40  35  30  25  20  15     Sample S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S1 1  International Journal of Analytical Chemistry such as soil factors, types of rootstock, and so on in the further study of CRC leaves quality classifcation and identifcation.
3.5. PCA. Te data of the 21 common peaks from fourteen samples was subjected to PCA (SPSS 26.0). For extraction of standard eigenvalue of more than 1, the cumulative contribution rate of the frst three components was 92.94%, indicating that the three principal components (PC) could represent most of the fngerprint information about fourteen samples. As shown in Tables 9 and 10, the eigenvalue in PC1 reached 14.565 with the variance contribution rate of 69.358%, and the peaks with higher loading were peak 3, peak 7, peak 14, and peak 15, indicating that these four peaks mainly refected the information of PC1. Te eigenvalue in PC2 was 2.585 with the variance contribution rate of   International Journal of Analytical Chemistry 12.308%, and the peaks with higher loading were peak 12, peak 20 (tangeretin), and peak 21 (5-demethylnobiletin), indicating that these four peaks mainly refected the information of PC2. Te eigenvalue in PC3 was 2.367 with the variance contribution rate of 11.269%, and the peaks with higher loading were peak 12, peak 13, and peak 19 (nobiletin), indicating that these four peaks mainly refected the information of PC3. To further visualize the results, the data were imported into SIMCA software version 15.0 (Umetrics, Sweden) to obtain two-dimensional analysis plot ( Figure 7). As shown in Figure 7, the samples were mainly clustered into six categories: S10, S6, S4, S13, S3; S7, S12, S14, S11; S8; S2; S9; and S1, S5. Te results of PCA were consistent with HCA. Among the samples, S8 distributed over the circle had poor quality, which was considered as an abnormal value, and this result was consistent with "similarity analysis." 3.6. GRA 3.6.1. Fingerprint-Efcacy Relationship. Te relationship between the 21 common peaks and the DPPH radicalscavenging activity was established using the GRA model.
Te results showed that the contribution of each chromatogram peak (X) to the DPPH radical-scavenging activity was in turn (by number): (Table 11). Te results of correlation analysis between the 21 common peaks and the antioxidant activity showed that peak 6, peak 14, peak 7, peak 13, peak 16, peak 20 (tangeretin), and peak 17 (hesperidin) were the major components related to the antioxidant efcacy of the samples, and the correlation coefcients of 21 common peaks were more than 0.6, indicating that the antioxidant activity of the CRC leaves was the result of multiple ingredients synergy.

Dose-Efect Relationship.
Te relationship between the measured amounts of hesperin, tangeretin, nobiletin, and 5demethylnobiletin in each sample and DPPH radicalscavenging activity was also determined using the GRA model. Te results showed that the contents of four favonoids were not proportional to the antioxidant activity. Te average contents of peak 20 (tangeretin) and peak 21 (5demethylnobiletin) were low, but they had high correlation S4 S11 S10 S14 S13 S8 S9 S12 Figure 6: HCA dendrogram of fourteen samples.   S4 S11 S10 S14 S13  coefcient with the antioxidant activity. On the contrary, the content of peak 17 (hesperidin) was the highest compared with other three components, but it had the lowest correlation coefcient with the antioxidant activity (Table 12).
According to the reports in the literature, polymethoxylated favones are a class of highly methoxylated favonoids peculiar to citrus plants [10], which have antioxidant [11] cardiovascular and cerebrovascular disease prevention and anti-infammatory efects. In this study, 5demethylnobiletin and tangeretin had high correlation coefcient with the antioxidant activity. Terefore, the content of 5-demethylnobiletin and tangeretin in the CRC leaves can be used as the indexes for establishing the quality control strategy.

Conclusions
In conclusion, this research established a chromatographic analysis method suitable for CRC leaves and obtained good chromatographic separation. Te results of GRA demonstrated that chromatographic fngerprints analysis combines with antioxidant activity could be used to evaluate material basis of CRC leaves. 5-demethylnobiletin and tangeretin in the CRC leaves can be used as the indexes for establishing the quality control strategy.

Data Availability
Te data used to support the fndings of this study are included within the article.

Conflicts of Interest
Te authors declare that they have no conficts of interest.