Chemical Comparison of Two Drying Methods of Mountain Cultivated Ginseng by UPLC-QTOF-MS/MS and Multivariate Statistical Analysis

In traditional Chinese medicine practice, drying method is an essential factor to influence the components of Chinese medicinal herbs. In this study, an ultra-performance liquid chromatography quadrupole time-of-flight tandem mass spectrometry (UPLC-QTOF-MS/MS)-based approach was used to compare the content of chemical compounds of mountain cultivated ginseng that had been natural air dried (LX-P) and vacuum freeze-dried (LX-L). Multivariate statistical analysis such as principal component analysis (PCA) and supervised orthogonal partial least squared discrimination analysis (OPLS-DA) were used to select the influential components of different samples. There were 41 ginsenosides unambiguously identified and tentatively assigned in both LX-L and LX-P. The results showed that the characteristic components in LX-P were ginsenoside Rb1, ginsenoside Rc, ginsenoside Rg6, dendrolasin, and ginsenoside Rb2. The characteristic components in LX-L were malonyl-ginsenoside Re, malonyl-ginsenoside Rb1, malonyl-ginsenoside Rc, malonyl-ginsenoside Rb1 isomer, malonyl-ginsenoside Rb2, malonyl-ginsenoside Rb3, malonyl-ginsenoside Rd isomer, gypenoside XVII, and notoginsenoside Fe. This is the first time that the differences between LX-L and LX-P have been observed systematically at the chemistry level. It was indicated that vacuum freeze-drying method can improve the content of malonyl-ginsensides in mountain cultivated ginseng.


Introduction
The root and rhizome of ginseng, Panax ginseng C.A. Meyer (Araliaceae), has been widely used as a traditional Chinese medicine and a functional food to prevent various diseases in the Orient [1]. Numerous research has shown that Panax ginseng possesses many pharmacological properties relating to the central nervous system [2], cardiovascular system [3], and aging process [4], which exhibits antioxidant [5], anticancer [6], and immunomodulatory effects [7]. The active components of ginseng are attributed to polysaccharides, ginsenosides, and volatile oil.
Mountain cultivated ginseng (MCG), which is grown in forests and mountains, can be considered to mimic mountain wild ginseng (MWG) [8]. Normally, MCG is harvested at the age of 10-20 years

UPLC-MS Analysis
As shown in previous articles, the ACQUITY BEH C 18 column has frequently been used to analyze ginsenosides from various ginseng products. Figure 1 shows the Based Peak Intensity (BPI) chromatograms obtained from the analysis of LX-L and LX-P in positive ion mode. The resultant peaks indicate that the components were complex in both MCG samples. There were 41 ginsenosides identified in LX-L and LX-P, including protopanaxatriol, panoxadiol, and their derivates. Among these ginsenosides, eight compounds were assigned by comparing them to standard ginsenosides, and 33 ginsenosides were identified by comparing their retention times and mass spectra with the reference compounds. The ginsenosides were further confirmed through ion fragmentation patterns. As illustrated in Table 1 peaks indicate that the components were complex in both MCG samples. There were 41 ginsenosides identified in LX-L and LX-P, including protopanaxatriol, panoxadiol, and their derivates. Among these ginsenosides, eight compounds were assigned by comparing them to standard ginsenosides, and 33 ginsenosides were identified by comparing their retention times and mass spectra with the reference compounds. The ginsenosides were further confirmed through ion fragmentation patterns. As illustrated in Table 1

PCA Analysis
Due to the similar components contained in each sample, the differences between LX-L and LX-P were hard to identify only from the BPI chromatograms (shown in Figure 1). In this case, MVA was commonly applied to process the data, and we can clearly see the difference between LX-L and LX-P from the PCA score plot.
A two-component PCA score plot of UPLC-QTOF-MS data was utilized to depict general variation of components among the mountain cultivated ginseng samples ( Figure 2). The PCA scores plot in Figure 2 can be readily divided into two big clusters. The LX-L and LX-P samples were clearly separated by the principal component 1 (PC1). Figure 3 shows the hierarchical cluster analysis (HCA) dendrogram of mountain cultivated ginseng samples. It appears that the components of them are indeed differential.

PCA Analysis
Due to the similar components contained in each sample, the differences between LX-L and LX-P were hard to identify only from the BPI chromatograms (shown in Figure 1). In this case, MVA was commonly applied to process the data, and we can clearly see the difference between LX-L and LX-P from the PCA score plot.
A two-component PCA score plot of UPLC-QTOF-MS data was utilized to depict general variation of components among the mountain cultivated ginseng samples ( Figure 2). The PCA scores plot in Figure 2 can be readily divided into two big clusters. The LX-L and LX-P samples were clearly separated by the principal component 1 (PC1). Figure 3 shows the hierarchical cluster analysis (HCA) dendrogram of mountain cultivated ginseng samples. It appears that the components of them are indeed differential.

Marker Ions Analysis
It is evident from Figure 2 that the samples were clearly clustered into two groups: one is LX-L, the other is LX-P, confirming that the components of LX-L and LX-P were indeed different in level and occurrence.
To explore the potential chemical markers that contributed most to the differences between LX-L and LX-P, UPLC-QTOF-MS/MS data were processed by supervised OPLS-DA. In the S-plot ( Figure  4), each point of an exact mass retention time (EMRT) pair could be the potential markers. The X-axis and the Y-axis show the variable contributions and sample correlations, respectively. Therefore, the further away a data point is from the 0 value, the more it contributes to sample variance and the better its correlation from injection to injection. As shown in the S-plot in Figure 4 ion (tR 6.28 min, m/z 917.5517) in the top right corner of the "S" were ions from LX-L that contributed most to the difference between LX-L and LX-P. These ions could be used as potential chemical markers to distinguish LX-L from LX-P.
Moreover, we can further confirm these spectral variables using the ion intensity plot ( Figure 5

Marker Ions Analysis
It is evident from Figure 2 that the samples were clearly clustered into two groups: one is LX-L, the other is LX-P, confirming that the components of LX-L and LX-P were indeed different in level and occurrence.
To explore the potential chemical markers that contributed most to the differences between LX-L and LX-P, UPLC-QTOF-MS/MS data were processed by supervised OPLS-DA. In the S-plot (Figure 4), each point of an exact mass retention time (EMRT) pair could be the potential markers. The X-axis and the Y-axis show the variable contributions and sample correlations, respectively. Therefore, the further away a data point is from the 0 value, the more it contributes to sample variance and the better its correlation from injection to injection. As shown in the S-plot in Figure 4, the first five ions, 1 ion (t R 4.56 min, m/z 1109.6176), 2 ion (t R 4.76 min, m/z 1079.6058), 3 ion (t R 4.98 min, m/z 1079.6069), 4 ion (t R 5.56 min, m/z 947.5623) and 5 ion (t R 10.00 min, m/z 219.1749) at the lower left of the "S" were the ions from LX-P that contributed most to the differences between LX-L and LX-P. Analogously, the first nine ions, 6 ion (t R 2.98 min, m/z 1033.5633), 7 ion (t R 4.65 min, m/z 1195.6194), 8 ion (t R 4.85 min, m/z 1165.6094) 9 ion (t R 4.91 min, m/z 1195.6187), 10 ion (t R 5.10 min, m/z 1165.6088), 11 ion (t R 5.37 min, m/z 1165.6067), 12 ion (t R 5.93 min, m/z 1033.5693), 13 ion (t R 5.95 min, m/z 947.5623) and 14 ion (t R 6.28 min, m/z 917.5517) in the top right corner of the "S" were ions from LX-L that contributed most to the difference between LX-L and LX-P. These ions could be used as potential chemical markers to distinguish LX-L from LX-P.
Moreover, we can further confirm these spectral variables using the ion intensity plot ( Figure 5) which was generated by Marker Lynx software. It was the convenient instrument to aid the profiling of marker ions. The marker t R 10.00 min, m/z 219.1748 ( Figure 5A) was from the LX-P sample and the marker ion t R 4.85 min, m/z 1165.6094 ( Figure 5B) was from the LX-L sample. The representative ion intensity plot illustrated the abundance of marker ions t R 10.00 min, m/z 219.1748 and t R 4.85 min, m/z 1165.6094 over 19 MCG samples. The ions fulfilled the criteria of marker ions because they were found to have significant difference in the content levels of the samples.

Maker Ions Assignment
Once having obtained the potential markers, element composition calculation was performed for the target markers. The molecular formula of the markers can be easily obtained by calculating their accurate masses. The next step was to search against a database and use the retention times as correlation references to identify the markers. Finally, the structure of the markers was confirmed by the fragments which appeared in the high capillary electrophoresis (CE) scan. The results are in Table 2.  , and ion 14 (t R 6.28 min, m/z 917.5517) in the LX-L samples were affirmed to be malonyl-ginsenoside Re, malonyl-ginsenoside Rb1, malonyl-ginsenoside Rc, malonyl-ginsenoside Rb1 isomer, malonyl-ginsenoside Rb2, malonyl-ginsenoside Rb3, malonyl-ginsenoside Rd isomer, gypenoside XVII, and notoginsenoside Fe, respectively.
After assigning the maker ions, we could easily find that mountain cultivated ginseng processed in different drying methods have a significant different in their chemical components. Malonyl-ginsenosides, which are naturally present in ginseng, were abundant in LX-L. However, LX-P contained a large number of major ginsenosides which were derived from malonyl-ginsenosides by natural air drying. This study indicated that the drying method is an essential factor to controlling the quality of mountain cultivated ginseng, and the vacuum freeze-drying method was found to improve the content of malonyl-ginsensides in mountain cultivated ginseng.

Ginseng Samples and Sample Processing
There were 19 MCG samples which were cultivated for 15 years before being collected from Ji'an city of the Jilin province of China. All these samples were fresh ginseng, which were then processed by natural air drying or by vacuum freeze-drying, respectively. All of these processed samples were identified by Professor Xiangri Li (School of Chinese Materia Medica, Beijing University of Chinese Medicine) and deposited in the specimen cabinet of traditional Chinese medicine of Beijing University of Chinese Medicine.

Sample Preparation
The dried roots were powdered to a homogeneous size, and sieved through a No. 65 mesh. The amount of 0.4 g of ginseng powder was accurately weighed and then placed in a triangular flask with 50 mL methanol, filled with a plug, weighed, and ultrasonic-extracted for 30 min. After cooling to room temperature, the loss of weight was replenished with methanol and then the sample was filtrated. Precision draw subsequent filtrate 25 mL and concentrated it into residue, which was then dissolved in methanol in a 10-mL volumetric flask. Finally, the extraction solution was injected into the UPLC system after being filtered through a 0.22-µm filter membrane.

Reagents
Fisher Optima grade acetonitrile, methanol, and isopropanol were purchased from Thermo Fisher Co. (Waltham, MA, USA). Formic acid and leucine enkephaline were purchased from Sigma Aldrich (St. Louis, MI, USA). Ultra-pure water was obtained in our laboratory via a Milli-Q water purification system (Millipore Corporation, Bedford, MA, USA). Ginsenoside Rg1, Re, Rb1, Rf, Rb2 and Rb3 standards were purchased from the National Institute for the Pharmaceutical and Biological Products (Beijing, China). Ginsenoside Rc, Rg2 standards were obtained from the Beijing Xiantong era Pharmaceutical Co. Ltd. (Beijing, China). The standards were dissolved in methanol and stored at 4 • C until analysis.

Mass Spectrometry Conditions
MS detection was performed on a quadrupole orthogonal acceleration time-of-flight tandem mass spectrometer (Waters Synapt MS System). The data acquisition mode was MS E and the ion polarity was set to positive mode (ESI + ). The optimized condition was desolvation gas at 480.0 L/h at a temperature of 350 • C, cone gas at 50 L/h and source temperature at 120 • C, capillary and cone voltage at 3.0 kV and 20 V, respectively. The lock mass compound used was leucine enkephaline. The low-energy scan collision energy was set at 5 eV in order to collect information on the intact precursor ions, and high-energy scan energy was set at 20 eV-30 eV to obtain the fragment ions. The UPLC-MS data acquisition was controlled by Mass Lynx 4.1 Mass Spectrometry Software (Waters Corporation).

Data Processing Procedure
For post-acquisition data processing, the MVA such as PCA and OPLS-DA were performed by Marker Lynx XS, which is an application manager for Mass Lynx software. The structural elucidation was performed by the Mass Fragment tool provided by Mass Lynx.

The PCA Scores Plot of Samples
From the chromatographic trace, we actually acquired three-dimensional (3-D) data which represented retention time, m/z, and intensity. It was necessary to convert each data point into a 2-D matrix, i.e., an exact mass retention time (EMRT) pair. After the EMRT 2-D matrix was obtained, the MVA interface was launched with all EMRT information automatically imported so that the extended statistics module PCA could be executed.

The Scatter Plot (S-Plot) from OPLS-DA Analysis
The loading plot (S-plot) of every group pair was processed by OPLS-DA analysis. In the S-plot, the leading contributing EMRT pairs could be captured selectively so that a list of top contributing markers from each sample group was generated and saved as a text file.

The Elemental Composition Calculation for the Targeting Markers
The matched elemental composition of markers was obtained by calculating the exact mass. Then, we searched against an existing database to acquire the chemical structure. Once the identity of a marker was tentatively identified, its fragment ions could be easily obtained by going back to the raw data file to investigate the high capillary electrophoresis (CE) scan of the samples. The fragment ions which we obtained through the Mass Fragment tool of Mass Lynx was used for elucidating the structure.

Conclusions
The multivariate statistical analysis (MAV) and UPLC-QTOF-MS/MS were combined to analyze mountain cultivated ginseng subjected to either natural air drying or vacuum freeze-drying. The combination of the high-resolution UPLC separation and high-resolution MS detection along with the multivariate statistical analysis (MAV) details of the samples proved able to identify and select the important marker ions in both samples, even at low concentration levels. As a result, this is the first time that the differences between LX-P and LX-L have been observed systematically at the level of their chemical components.