Simultaneous Quantification of Flavonol Glycosides, Terpene Lactones, Biflavones, Proanthocyanidins, and Ginkgolic Acids in Ginkgo biloba Leaves from Fruit Cultivars by Ultrahigh-Performance Liquid Chromatography Coupled with Triple Quadrupole Mass Spectrometry

On the basis of liquid chromatography coupled with triple quadrupole mass spectrometry working in multiple reaction monitoring mode, an analytical method has been established to simultaneously determine flavonol glycosides, terpene lactones, biflavones, proanthocyanidins, and ginkgolic acids in Ginkgo biloba leaves. Chromatographic separation was carried out on an Acquity BEH C18 column (100 mm × 2.1 mm, 1.7 μm) with gradient elution of acetonitrile and 0.10% formic acid (v/v) at a flow rate of 0.4 mL/min, and column temperature 30°C. The developed method was validated in terms of linearity, accuracy, precision, stability, and sensitivity. The optimized method was successfully applied to analyze twenty-two G. biloba leaf samples of fruit cultivars collected from different places in China. Furthermore, hierarchical clustering analysis (HCA) was performed to evaluate and classify the samples according to the contents of the twenty-four chemical constituents. All of the results demonstrated that the developed method was useful for the overall evaluation of the quality of G. biloba leaves, and this study was also helpful for the comprehensive utilization and development of G. biloba resources.


Introduction
e dried leaves of Ginkgo biloba L. have been used as herbal remedies in China, and now their extracts are one of the most widely used herbal products and/or dietary supplements in the world. e products are popular for their alleged tonic effect and possible curative and restorative properties [1][2][3]. Important constituents in G. biloba leaves include �avonol glycosides, terpene lactones, bi�avones, proanthocyanidins, and ginkgolic acids, of which the �avonol glycosides, terpene lactones, and proanthocyanidins are considered to be the main components for their bene�cial effects. ese compound classes have received by far the most attention [1,[4][5][6]. Flavonol glycosides are widely known for their antioxidant and free radical scavenging activity [7]. Ginkgolides are potent and selective platelet-activating-factor antagonists, and recently there has been considerable interest concerning the antagonistic effect of ginkgolides on the glycine receptor [8][9][10]. Proanthocyanidins have substantial antioxidant activity and may modulate several reactions involved in cancer processes [11]. In addition, most Gymnospermae plants were characterized mainly by the occurrence of bi�avonoids [12], while ginkgolic acids were considered as potentially hazardous constituents in G. biloba leaves because they possessed possibly mutagenic [13] and carcinogenic activity [14].
In the past decade, many scienti�c analytical methods have been studied on the chemical constituents of G. biloba leaves, mainly the �avonoids and terpene lactones. HPLC-ELSD has remained the most widely used method for the analysis of terpene lactones [15][16][17]. To date, various analytical methods have been used to analyze �avonol glycosides that can be classi�ed as two types: the direct assays and the assays aer hydrolysis. e general procedure for the quantitative analysis of �avonol glycosides with an acidic hydrolysis is followed by HPLC determination of the resulting aglycones, based on calibration curves of aglycone standards [18][19][20]. However, the measurement solely of the �avonol aglycones does not preclude the possibility of adulteration of G. biloba leaves with pure �avonol aglycones, pure �avonol glycosides, or other plant material containing �avonols or �avonol glycosides. So, it is necessary to analyze intact glycosides for an accurate evaluation of the quality of G. biloba leaves. Of course, �avonol glycosides can be analyzed by HPLC with DAD or MS [21,22]. For saving the time and cost by reducing multiple sample preparations and mobile phases, HPLC-DAD-ELSD was used for simultaneous determination of intact �avonol glycosides and terpene lactones [23]. However, its low sensitivity and uncertainty in peak identi�cation limited its use in analysis. Recent success in the use of liquid chromatography coupled with triple quadrupole mass (UPLC/MS/MS) for characterizing and quantifying a wide variety of compounds in complex samples [24][25][26] suggested that UPLC/MS/MS might be a good technique for the simultaneous determination of intact glycosides and terpene lactones. In addition, it is possible that other important constituents of G. biloba leaves, such as bi�avones, proanthocyanidins, and ginkgolic acids could also analyzed simultaneously.
e Ginkgo resources are quite rich in China. Most researches on G. biloba leaves focus on the leaves collected during July and August from trees 4-7 years old [27][28][29][30][31]; a large number of leaves from fruit cultivars (the tree of above 10 years) are ignored and become obsolete aer fruit harvest season (November). Herein, we developed methods for a systematic identi�cation of the main active components in different cultivation sources of G. biloba leaves collected in November, which will provide a scienti�c basis for the comprehensive utilization and development of G. biloba resources.
In this paper, by using UPLC/MS/MS, an efficient and sensitive method was developed and validated for simultaneous identi�cation and quanti�cation of twenty-four compounds including �avonol glycosides, terpene lactones, bi�avones, proanthocyanidins, and ginkgolic acids in G. biloba leaves from fruit cultivars. e proposed method was successfully applied to analyze twenty-two samples collected from different places in China. Furthermore, hierarchical clustering analysis (HCA) was performed to evaluate and classify the samples according to the contents of the twentyfour chemical constituents.

Preparation of Standard Solutions.
A mixed standard stock solution containing the twenty-four analytes was prepared in 70% aqueous methanol and then diluted with 70% aqueous methanol to appropriate concentrations for establishing calibration curves. Solutions containing different concentrations of the twenty-four analytes were injected in triplicate, and the calibration curves were plotted by the peak area ratio versus the concentrations of each analyte. e standard solutions were �ltered through a 0.22 m membrane prior to injection.

Sample Preparation.
Aer being dried at 55 ∘ C for seven days, the leaves were pulverized to homogenous powders (40 mesh). e dried powder (0.5 g) was weighed accurately into a 100 mL conical �ash with a stopper, and 40 mL of 70% aqueous methanol was added. Aer accurate weighting, ultrasonication (40 kHz) was performed at room temperature for 60 min, and then the same solvent was added to compensate for the weight lost during the extraction. e solution was adequately mixed and followed by centrifugation at 13000 rpm for 10 min. For the quanti�cation of compounds 20-24, the supernatants were diluted 10 or 100 times with 70% aqueous methanol, while for the others, the supernatants did not require dilution. All the solutions were stored at 4 ∘ C and �ltered through a 0.22 m membrane �lter before injection into the UPLC system for analysis. gas �ow rate set to 1,000 L/h at a temperature of 550 ∘ C, the cone gas �ow rate set at 50 L/h, and the source temperature at 150 ∘ C. e capillary voltage was set to 3,000 V. e analyte detection was performed by using multiple reaction monitoring (MRM). e cone voltage was set depending upon each speci�c MRM for each compound. Data was collected in MRM mode by screening parent and daughter ions simultaneously. Dwell time was automatically set by the soware. All ESI and MS parameters were optimized individually for each target compound and are listed in Table  2.

Hierarchical Cluster Analysis (HCA) of Samples.
In order to evaluate the variation of G. biloba leaves, HCA was carried out using SPSS 16.0 soware. Ward's method was applied, and square Euclidean distance was selected as a measurement. e dendrogram resulted from the 24 investigated compounds' contents derived from �PLC-MS/MS pro�les of the tested samples.

Method Validation.
A series of analyses such as the linearity, stability, precision, and limits of detection (LOD) and quanti�cation (LOQ) were conducted to validate the performance of the method. e standard solution containing 24 markers was prepared and diluted with 70% aqueous methanol to appropriate concentrations for the construction of calibration curves. Calibration curves were developed by plotting the peak areas versus the corresponding concentrations of each analyte. e precision of the method was evaluated by analyzing the 24 standard compounds. e relative standard deviation (RSD) of peak area was used to evaluate the precision of the developed method. LOD and LOQ for each analyte were determined at the signal-to-noise ratio ( ) of about 3 and 10, respectively. e recovery test was used to evaluate the accuracy of this method. e test was performed by adding the corresponding marker compounds at low (80% of the known amounts), medium (same as the known amounts), and high (120% of the known amounts) to the G. biloba leaves sample, which had previously been analyzed. e mixture was extracted and analyzed using the aforementioned method in triplicate.  were optimized. e results suggested that ultrasonic extraction was better than re�ux extraction, and it had the advantage of greater convenience. Using the sonication method, various solvents including water and 30, 50, 70, and 100% methanol were screened. e best solvent was 70% aqueous methanol, which resulted in fewer interfering peaks and provided the highest values for the contents of the 24 markers. In addition, the extraction time was optimized to 60 min. Each sample extracted with 40 mL was adequate. Finally, suitable extraction conditions were optimized as follows. Each sample was extracted by sonication with 40 mL of 70% aqueous methanol for 60 min. All of the samples were extracted at room temperature.

Optimization of the Chromatographic Conditions.
e chromatographic conditions such as column, mobile phase, and gradient elution were optimized in the preliminary test to achieve chromatograms with short analysis time and peak shape without excessive peak tailing. Two analytical columns, Acquity HSS T3 column (100 mm × 2.1 mm, 1.8 m) and Acquity BEH C 18 column (100 mm × 2.1 mm, 1.7 m), were compared, showing that with the latter one obtains chromatograms with better resolution than the former with the same mobile phase. Although all compounds could be separated satisfactorily with the Acquity HSS T3, the chromatographic peak is asymmetric. erefore, the Acquity BEH C 18 column was selected as the analytical column. Furthermore, various mixtures of water and methanol were used as the mobile phase but no satisfactory separation was achieved. However, when the mobile phase composed of methanol-water was replaced by acetonitrile-water, the separation resolution was greatly improved. Addition of formic acid to the mobile phase was found to enhance the resolution of target compounds at a �ow rate of 0.4 mL�min. e separation was better when column temperature was kept at 30 ∘ C rather than 25 and 35 ∘ C.   To select a proper transition for the MS/MS detection of the analyte, all the compounds were examined separately in direct infusion mode and at least two precursor/product ion pairs for each analyte were presented in this study. And then, according to the quantitative results, the highest sensitive and speci�c ion pairs were selected for the MRM determination. Once the most appropriate precursor/product ion pairs had been determined, the values of cone voltage and collision energy were optimized using the IntelliStart soware. All the MRM transitions and parameters applied in the study are listed in Table 2. Representative chromatograms for 24 analytes are shown in Figure 2. e results indicated that the MRM mode has more advantages in the quanti�cation of low content analysis and separation of overlapped constituents in complicated mixtures.

Method Validation
3.3.1. Calibration Curves, LOD, and LOQ. e linearity of the method was obtained by plotting the peak areas versus the corresponding concentrations of each analyte. Each calibration curve was constructed with at least six appropriate concentrations in duplicate. All calibration curves showed good linearity ( 2 > 0.9903) within the test ranges. LOD and LOQ for each analyte were determined at the signal-tonoise ratio ( ) of about 3 and 10, respectively. e overall LODs and LOQs were in the range of 0.04-3.50 ng/mL and 0.10-11.20 ng/mL, respectively. e results are reported in Table 3.

Precision, Repeatability, and
Stability. e precision of the method was evaluated by analyzing the 24 standard compounds. e intra-and interday precisions were examined for the mixed standards six times on the same day and the experiments were repeated on three consecutive days, respectively. e relative standard deviation (RSD) was taken as a measure of precision. e RSD values of intra-and interday variations of the 24 analytes were in the range of 1.91%-6.28% and 2.30%-6.30%, respectively. e results are shown in Table 4. e reproducibility was assessed by analyzing six independently prepared samples using the same method and one of the sample solutions was stored at 20 ∘ C and analyzed at 0, 2, 4, 8, 12, and 24 h, respectively, to evaluate the solution's stability. e repeatability and stability present as RSD [32,33] were in the range from 3.12% to 6.19% and from 0.76% to 5.68%. e results are shown in Table 4.

3.�. ��en�i��a�ion an� ��an�i��a�ion.
Identi�cation of the target peaks was performed by comparing their UPLC retention times and mass/charge ratios ( / ) with those of the standards. In order to further con�rm the structures of the constituents, standards and samples were analyzed by UPLC-MS/MS. �uanti�cation was performed using linear calibration plots of peak areas and concentration.

Sample Analysis.
To evaluate the effectiveness of the method, it was applied to the analysis of G. biloba leaves collected from the major production areas of G. biloba in China (Jiangsu and Guangxi provinces). Due to the fact that the contents of compounds 20-24 are equal to or greater than two orders of magnitude of the other target compounds in most samples, their quanti�cation was performed by diluting the sample solutions as described in the experimental section before analysis. All the contents are summarized in Table 5. e results showed that the twenty-four constituents were present in most samples. is outcome proved the necessity of the simultaneous determination of these components in order to improve the quality control of this traditional Chinese medicine. Among these compounds, bi�avones, terpene lactones, and ginkgolic acids were dominant components, which were higher than the detected �avonol glycosides and proanthocyanidins. Remarkable differences were found among the contents of the 24 target compounds in G. biloba leaves of different cultivation regions. Components 22 and 24 were dominant in most samples with contents that ranged 2.28-4.92‰and 0.87-3.00‰, respectively. e contents of total �avonoids, terpene lactones, bi�avones, proanthocyanidins, and ginkgolic acids in the analysed samples varied from 670.0 to 3524.7 g/g, 362.8 to 3668.3 g/g, 4072.1 to 8641.3 g/g, 16.9 to 525.1 g/g, and 1441.5 to 3889.7 g/g, which should result in differences in their therapeutic potential. erefore, these active components should also be quanti�ed so as to re�ect the therapeutic effect.  Ward's method, which is efficient for the analysis of variance between clusters, was applied, and the square Euclidean distance was selected as a measure of similarity. e dendrogram ( Figure 3) shows that the twenty-two tested samples of G. biloba leaves were divided into two main clusters (I and II) according to their contents. Samples 1, 4, 5, 6, 20, and 21 were in cluster II and the other samples were in cluster I. e samples with similar chemical pro�les were classi�ed into one group. It was noticeable that the distribution of the 24 target compounds in G. biloba leaves exhibited regional disparity. In the dendrogram, the northern and southern parts of China were described as cluster I and II, which indicates that the pro�les of �avonol glycosides, terpene lactones, bi�avones, proanthocyanidins, and ginkgolic acids are similar when they are cultured under the same conditions. e schematic diagram of regional disparity revealed by HCA on the Chinese map is shown in Figure 4.

Conclusions
In this study, a reliable, simple, and sensitive method was developed for the identi�cation and quanti�cation of 24 compounds in G. biloba leaves by using UPLC/MS/MS. To our knowledge, it is the �rst time that the UPLC-MS/MS method had been employed to analyze simultaneously twenty-four main bioactive components in G. biloba leaves. In this study, the method was applied to investigate G. biloba leaves of fruit cultivars of different national regions. e results demonstrate that the developed method is useful for overall evaluation of the quality of G. biloba leaves, and this study should be helpful for the comprehensive utilization and development of G. biloba resources.