Phytochemical Analysis Using UPLC-MS/MS Combined with Network Pharmacology Methods to Explore the Biomarkers for the Quality Control of Lingguizhugan Decoction

As a classic TCM prescription, LGZG has been widely used in clinical prevention and treatment of heart failure, nonalcoholic fatty liver, and hyperlipidemia. However, there are few studies on chemical components in recent years, and the basis of quality evaluation is not sufficient. This study was to find the active ingredients of the Lingguizhugan decoction using UPLC-MS/MS and network pharmacology. By comparing the retention time and MS dates of the reference and self-building database, the cleavage rules of chemical composition whose mass errors are less than 1 ppm(FL less than 3 ppm) are analyzed. On this basis, a network pharmacology method was used to find biomarkers for quantitative analysis. The results show that 149 compounds were preliminaries identified or inferred, including 63 flavonoids, 30 triterpenes, 22 phenylpropanoids, 13 organic acids, 6 lactones, 5 alkaloids, 4 anthraquinones, and 6 other compounds. According to the network pharmacology results, 20 chemical constituents were selected as the biomarkers, which were determined simultaneously for the first time, including poricoic acid A, poricoic acid B, glycyrrhizic acid, glycyrrhetinic acid, liquiritin, isoliquiritin, liquiritigenin, isoliquiritin apioside, cinnamic acid, caffeic acid, neochlorogenic acid, chlorogenic acid, cryptochlorogenic acid, isochlorogenic acid A, B, and C, atractylenolide I, II, and III, and coumarin. The methodological results show that the linearity, stability, precision, repeatability, and recovery of the method are satisfactory. Therefore, a comprehensive quality assessment system for LGZG was established on the basis of a systematic study of chemical substances and network pharmacology, which provided an important reference for the foundation of pharmacological action and its mechanics.


Introduction
Lingguizhugan decoction (LGZG), coming from Zhang Zhongjing's Synopsis of Preions of the Golden Chamber, has already been listed as the first batch of " e Catalogue of ancient classical prescription" by the National Administration of Traditional Chinese Medicine (TCM) in 2018.
LGZG is prepared by four different herbal medicines, namely Fuling (the dry sclerotia of Poria cocos (Schw.), FL), Guizhi (the dry twigs of Cinnamomum cassia Presl, GZ), myocardial infarction [5], nonalcoholic fatty liver [6], obesity [7], hyperlipidemia [8], gastrointestinal function regulation [9], and many other diseases [10,11]. However, combined with the present stage of research, most of the studies on LGZG have been restricted to the main research index of licorice-related components, while the related components in other herbal medicines are rarely involved, or the overall quality evaluation indexes are not comprehensive, or the range of mass error is excessive, which results in a limited accuracy [12,13]. Considering that licorice is only used as a guide medicine in the prescription, the comprehensiveness and objectivity of the method cannot be guaranteed.
As we all know, compound prescriptions of TCM are related to many different herbal medicines, which contain a variety of chemical components and more complex interactions. Generally speaking, after oral administration, the chemical components of the drugs can produce specific synergism or antagonism through blood circulation in multiple organs and targets of the whole body. Its prototype components or metabolites play a corresponding therapeutic role through specific pathways, such as participating in the metabolic regulation, or regulating the homeostasis of microorganisms, and finally causing the overall level of metabolites to change to the normal level. erefore, it is particularly necessary to establish a comprehensive and systematic research for the chemical constituents of LGZG. us, our study established a quality evaluation system for LGZG using both UPLC-MS/MS and network pharmacology method, which has important guiding significance to help us clarify the pharmacodynamic material basis and its mechanism.
Methanol of the HPLC grade was obtained from Fisher Scientific (Fair Lawn, NJ, USA). Acetonitrile of the HPLC grade, together with formic acid of the LC-MS grade, was purchased from Merck (Darmstadt, Germany). Purified water was purchased from Watsons Corporation (Guangzhou, China). e vacuum freeze-dryer was purchased from Beijing Boyikang Experimental Instrument (Beijing, China). All other chemicals and reagents used were of HPLC grade.

Sample Preparation.
Following the ancient prescription, herbal medicines, including FL 55.2 g, GZ 41.4 g, BZ 41.4 g, and GC 27.6 g, were taken together with 1200 mL purified water. Until the extracted solution remained 600 mL, it was filtered through a 200-mesh sieve and freezedried. For details of the origin of 18 batches of sample medicinal materials, see Table 1 e sample of LGZG powder(1.0 g) was accurately weighed, dissolved in 70% methanol-water (25 mL), and subjected to ultrasonic extraction for 15 min. e solution was weighed again and after the loss in weight was made up with methanol, the obtained solution was then subjected to centrifugation at 4000 rpm for 10 min, and the supernatant was filtered through a 0.22 µm nylon membrane (Millipore, USA) prior to analysis. All solutions are stored at 4°C before use. B. e flow rate and the injection volume were set at 0.3 mL/min and 2 μL, respectively. e mass analysis was implemented by a Triple TOFTM 6600 + (AB SCIEX, Foster City, CA, USA), which is equipped with an electrospray ionization (ESI) source. e MS analysis was performed in both positive ionization mode and negative ionization mode by full scan mode. e optimized parameters were as follows: ion spray voltage (ISV), 5.5 kV (ESI+) or −4.5 kV (ESI-); ion source temperature (TEM), 550°C; ion source gas 1 (GS1), 55 psi; ion source gas 2 (GS2), 55 psi; curtain gas (CUR), 35 psi, scan range of TOF-MS, m/z 100-1500 Da; and scan range of product ion, m/z 100-1500 Da. IDA-MS/MS conditions were as follows: accumulation time was 0.05 s; high sensitivity mode was set, excluding isotopes were within 4 Da; declustering potential (DP) was 80V (ESI+) or −80V (ESI-); collision energy (CE) was 10 eV (MS mode), and 40 and 80 eV (MS2 mode); and collision energy spread (CES) was 20 eV. Moreover, in order to ensure accurate mass measurements during the MS experiments, the instrument performed mass accuracy calibration through the CDS system before each experiment. During the experiment, the mass accuracy was calibrated for every three samples. e data acquisition and processing will be analyzed by SCIEX OS-Q 2.0 Software (AB SCIEX, Foster City, CA, USA) and Peak View ® 2.2 Software (AB SCIEX, Foster City, CA, USA).

UPLC-QQQ-MS/MS Analysis. UPLC-QQQ-MS/MS
analysis was performed on the LC-30A UPLC system(Shimadzu, Kyoto, Japan), including DGU-30A3 type online vacuum degasser, LC-30AD-type binary pump, SIL-30AC-type automatic sampler, and CTO-30A-type column incubator. e experimental conditions were as follows: Shimpack GIST C18(100 mm × 2.1 mm, 2 μm) chromatographic column; column temperature, 35°C; flow rate, 0.3 mL/min; and injection volume, 2 μL. e mobile phase was selected to be water containing 0.1% formic acid (A) and acetonitrile (B) with the following gradient elution programme: 0-17 min, 5-17% B; 17-31 min, 17-75% A; 31-32 min, 75% A; and 32.1-35 min, 5% A. In terms of mass spectrometry, we chose QTRAP 4500 (AB SCIEX, Foster City, CA, USA) coupled with an electrospray ionization (ESI) source. e analytical method adopts the MRM mode, with the positive and negative ion switching detection method. e ion spray voltage (ISV) was 5.5 kV (ESI+) or −4.5 kV (ESI-); ion source temperature, (TEM) 550°C; ion source gas 1 (GS1), 55 psi; ion source gas 2 (GS2), 55 psi; and curtain gas (CUR), 35 psi. e mixed reference substance solution was diluted step by step to form a mixed reference substance solution with six concentration gradients. According to the condition of item 2.3.2, the extracted ion current chromatogram of each reference substance was obtained. With the concentration of each reference substance (X) and peak area (Y), the standard Evidence-Based Complementary and Alternative Medicine 3 curves were drawn and linear regressions were carried out. e mixed reference solution was diluted step by step and determined until the quantification limit (LOQ) and detection limit (LOD) were determined when the signal-tonoise ratio (S/N) was 10 and 3, respectively.

Precision, Repeatability, Stability, and Recovery.
In order to investigate the precision, the mixed reference solution and sample solution were injected continuously for 6 times, the peak areas were recorded, and the RSD values were calculated. In the repetitive investigation, 6 samples of the same batch were taken and prepared in parallel. rough the sequential injection analysis of the samples, the peak area and the RSD value were calculated. According to the stability test, the solution of the same batch of samples was injected and analyzed at 0, 4,8,12,18,24,36, and 48h, respectively. e recovery rates were investigated by adding high, medium, and low levels of reference solution (150, 100, and 50% of the known amount, respectively) to the same batch of samples, with 3 parallel samples at each level. Recovery (%) � (detected value -original value)/added amount × 100%.

Strategy.
rough searching and sorting out the previous literature, the chemical composition information database of four kinds of TCM in LGZG was established, including compound name, molecular formula, exact molecular weight, structural formula, and other essential information. For compounds with reference substances, it was confirmed by comparing the chromatographic retention time and mass spectrometry data. For unknown compounds, the OS software was used to compare the self-built composition database, TCM MS/MS database (supplied by AB SCIEX) and online Chemspider database. e screening conditions were that the quality deviation was less than 1 ppm (FL less than 3 ppm). e fragmentation pattern of representative components was analyzed by the MS/MS spectrogram of the compound.
According to the results of chemical composition research, a network pharmacology research was carried out to screen biomarkers, which were taken as the references to carry out the quality evaluation research. On this basis, the study established a quantitative study with high accuracy and good stability. Combined with the method of statistical analysis, the differences between different areas of herbal medicines were analyzed, which will provide a theoretical basis for the quality evaluation of TCM.

Identification of the Chemical Components in LGZG by UPLC-ESI-Q-TOF-MS.
By introducing the UPLC-ESI-Q-TOF-MS analysis method based on DBS-IDA data acquisition technology, the spectrum signals were 45,205 and 43,188 measured in positive mode and negative ion detection mode, respectively. Combined with reference substance comparison, literature research, and MS/MS information, a total of 149 chemical constituents in LGZG were identified, including 63 of flavonoids, 30 of triterpenes, 22 of phenylpropanoids, 13 of organic acids, 6 of lactones, 5 of alkaloids, 4 of anthraquinones, and 6 of other compounds. e typical total ion chromatograms of positive and negative ion modes are shown in Figure 1. All the compounds and related information are shown in Tab.S1.

Identification of Flavonoids.
As the most abundant compounds in nature, flavonoids are not only the main chemical components of many kinds of TCM, but also the most important therapeutic active components. In this study, a total of 63 flavonoids were identified or preliminarily inferred, mainly from GZ and GC. ese components can be further divided into flavonoid aglycones, O-glycosyl flavonoids, and C-glycosyl flavonoids.
(1) Flavonoid Aglycones. e main cracking mode of flavonoids is the Retro-Diels-Alde (RDA) fragmentation mechanism, which can also eliminate small neutral molecular fragments, such as CH 3  Meanwhile, the molecule can remove the isopentenyl group (C 5 H 9 ) from the benzene ring to form the m/z 268.0377 fragment ion (m/z -69), which can continue to remove -CO 2 to form m/z 224.0480 fragment ion. Finally, on the basis of 368 fragments, m/z 135.0091 and m/z 117.0346 fragments are produced through the RDA fragmentation mechanism. e fragmentation pathway of Eurycarpin A is shown in Figure 2.
As a branch of flavonoids, isoflavanes do not contain CO in the C 3 ring, which cannot form a conjugated system. is leads to its unstable structure and easy fracture at the C 3 ring. We explain the fragmentation mechanism of isoflavane through compounds 7,4′-dihydroxy-3′-methoxyisoflavan (C 16 e component exhibited a parent ion at m/z 271.1000, which yielded daughter ions at m/z 256.0719 and 241.0512 by losing a CH 3 and CH 2 O group, respectively. Besides, components were cleaved by RDA reaction to form m/z 121.0296 and 149.0600. e most important thing is that the C-C and C-O bonds at C 3 can be broken, so that the molecule is divided into two new fragment ions m/z 109.0305 and 135.0447. e fragmentation pathway of 7,4′-dihydroxy-3′methoxyisoflavan was shown in Figure 3.
According to the rules mentioned above, a total of 46 flavonoids were identified or inferred, including  For O-flavonoid glycosides, the glycosyl connection position is related to the parent nuclear structure of flavonoids. In terms of flavonoids, dihydroflavonoids and isoflavones, most of them form monosaccharides on 7-OH. While flavonols and dihydroflavonol glycosides form monosaccharides on 3-OH, 7-OH, 3′-OH and 4′-OH, or disaccharides on 3-OH and 7-OH, 3-OH and 4′-OH or 7-OH and 4′-OH.
(3) C-Glycosyl Flavonoids. In C-glycosides flavonoids, the glycosyl groups are mostly linked at the position of C 6 or C 8 , or at both C 6 and C 8 . Because the C-C bond of flavonoid C-glycosides is unstable, the ring-opening reaction of glycosyl groups mainly occurs in the process of cleavage. e continuous neutral loss of H 2 O(m/z 18.0106 Da) is dominant in the positive ion mode, and the glycosyl neutral loss is mainly in the negative ion mode, such as hexose neutral loss (C 4  e specific cracking process is shown in Figure 5. On the basis of the above rules, 17 flavonoid glycosides were identified,

Identification of Triterpenoids.
e basic skeleton of triterpenoids is connected by 6 isoprene structural units. Considering the skeleton with multiple six-membered rings in the structure is extremely stable and not easy to break bonds, higher collision energy is needed in the process of mass spectrometry pyrolysis. Triterpenes and their glycosides are easy to remove sugar groups (such as glucose) and ring-linked substituents, including OH (lost in the form of H 2 O, m/z 18.0106 Da), COOH(lost in the form of CO 2 , m/z 43.9898 Da), and other substituent groups [18].
For our study, most of triterpenoids in the prescription mainly come from FL and GC. With reference to tumulosic acid (C 31 H 50 O 4 , [M-H] -, P130), the difference between parent ion and daughter ion is m/z 62.0004. at is to say, the compound loses CO 2 and H 2 O at one time (parent iondaughter ion: m/z 485.3639-423.3196). e following is to compare the differences of fragmentation mechanism between the positive and negative ion mode with pachymic acid (C 33 H 52 O 5 , P146). In the positive ion mode, when the collision energy is set to 40, the parent ion peak is not displayed, and the daughter ion is m/z 511.3770, 451.3562, 355.2625, and 295.2422. Among them, m/z 511.3770 is formed by removing H 2 O on the basis of the parent ion, and the subsequent daughter ions lose CH 3 COOH, branchedchain C 7 H 12 , and C 2 H 4 O 2 . However, in negative ion mode, when the collision energy is set to 40, basically only 527 single peak can be displayed (there is one peak at 465, and the intensity is only 2.37%). Increasing the collision energy to 80, the parent ion is at m/z 527.3737 and the daughter ion peaks are at m/z 467.3529, 465.3368, 405.3179, and 293.1899 in sequence. On the basis of the parent ion, on the one hand, it can lose CO 2 and H 2 O to form the m/z 465.3368 ion. On the other hand, it can also lose the CH 3 COOH to form the m/z 467.3529 ion, which can continue to complete the loss of another group of ions, forming a 405.3179 peak. Finally, the branched-chain C 8 H 16 was removed again to form the 293.1899 peak. e detailed cracking process is shown in Figure 6. According to the rules discussed above, a total of 30 compounds are inferred (labeled as P82, 86, 88, 89, 90, 91, 92,  102, 105, 106, 107, 108, 111, 121, 128, 129, 130, 133, 134, 136,  137, 138, 139, 142, 143, 144, 145, 146, 148, 149).      ) produced an m/z peak at 162.0326 upon the elimination of CH 3 , and this was followed by the loss of COOH groups in succession to give peaks at m/z 117.0339 (Figure 7(a)). e basic nuclear structure of coumarin is benzopyranone (cis-2-hydroxycinnamic acid lactone). According to the different substituents and connection modes, coumarin can be divided into simple coumarin, furocoumarins, pyranocoumarins, and other coumarins. Coumarins generally have many CO, CO 2 , OH, H 2 O, CH 3, and OCH 3 connected to aromatic rings, so that a series of neutral ion peaks with continuous loss often appear. In addition, coumarins often have common functional groups such as isopentenyl, acetoxy, and 5-carbon unsaturated acyloxy groups, which are also the main characteristics of coumarin compounds [21,22] Lignans are natural compounds of derivatives formed by the oxidative polymerization of phenylpropanoids, usually in the form of dimers, as well as a few number of trimers and tetramers. Most of them exist in free form in plant wood and resin, and also combine with glycosyl groups to form glycosides. As the phenylpropanoids polymers, the typical characteristic ion fragments of lignans are M/2 or M/3 peak, whose response intensity is relatively high. e other cleavage rules are consistent with phenylpropanols [23]. In the mass spectrogram of secoisolariciresinol (C 20 (Figure 8). According to the rules discussed above, a total of 13 compounds are inferred (labeled as P2, 3,9,10,11,12,14,15,19,25,26,36,79 [26][27][28][29][30]. Given the large differences in the amounts of the various components, they are hereby grouped into other categories.

Biomarkers Screening and Validation by Network Pharmacology and Components Absorbed into Blood.
A total of 284 potential targets related to the 149 identified compounds were obtained from BATMAN-TCM (http://bionet.ncpsb. org.cn/batman-tcm/) databases(with the score cutoff > 20). rough the keyword "Hyperlipidemias," a total of 1,691 disease targets were obtained from DrugBank (https://go. drugbank.com/), OMIM (https://omim.org/), GeneCards (https://www. genecards.org/), and DisGeNET (https:// www.disgenet.org/) databases. rough protein-protein interaction analysis, 71 targets with higher association were obtained as shown in Figure 9(    shown in Table 2. ese reserved proteins were further imported into the ClueGo plugin of Cytoscape 3.7.2 software for KEGG enrichment analysis, and relatively important disease pathways were found as shown in Figure 9(b). Finally, the "component-target-function" network was visualized in Figure 9(c). ere are a total of 20 compounds (including poricoic acid A, poricoic acid B, glycyrrhizic acid, glycyrrhetinic acid, liquiritin, isoliquiritin, liquiritigenin, isoliquiritin apioside, cinnamic acid, caffeic acid, neochlorogenic acid, chlorogenic acid, cryptochlorogenic acid, isochlorogenic acid A, isochlorogenic acid B, isochlorogenic acid C, atractylenolide I, atractylenolide II, atractylenolide III, and coumarin), 71 proteins, and 20 pathways involved. Of the 20 pathways, the score of PPAR pathway was the highest, which plays an important role in lipid metabolism. On this basis, we verify it by the method of molecular docking. e results show that the free energies of docking binding of 20 compounds are all greater than -6.471 kcal/ mol. e typical components of the four TCM we selected are displayed in Figure 10. Besides, all the 20 compounds were detected in the rat serum by using UPLC-QQQ-MS/ MS, which further verified that these compounds were proper biomarkers.

Quantification of the Major Constituents in LGZG by UPLC-QQQ-MS/MS.
On the basis of the qualitative study and network pharmacology, the characteristic parameters of 20 main chemical components were investigated by the MRM analysis method, including parent ion, daughter ion, declustering potential, and collision energy. In order to satisfy the simultaneous quantification of multiple components, taking into account the response of different chemical components in MRM mode and the content differences in LGZG, the DP and CE of some chemical components were dynamically adjusted, as detailed in Table 3 [31][32][33][34][35][36][37]. Multiple reaction monitoring chromatograms of the control and sample solutions are shown in Figure 11.

Method Validation
(1) Calibration Curve, LOD, and LOQ. According to the concentration (X) and peak area (Y) of each reference substance, the calibration curves were drawn and linear regressions were carried out, which showed good linearities(R 2 > 0.9937). In addition, the LOD and LOQ values were determined, and the obtained results are presented in Table 4.
(2) Precision, Repeatability, Stability, and Recovery. e precision, repeatability, stability, and recovery of the quantitative method were investigated. e RSD of the precision of the reference substance and the sample did not exceed 4.29% and 5.02%, respectively. e repetitive result is less than 6.47%. e stability test results show that the LGZG sample solution is stable within 48 hours, and the RSD is less than 6.01%. e average recovery is between 96.22% and 104.19%. e results show that the established quantitative method is accurate, reliable, and reproducible and can be used to evaluate the quality of LGZG (Table 5).

Conclusion
Based on UPLC-MS/MS and network pharmacology, this article established a research method of chemical composition research-network pharmacology predictive analysisquality evaluation system. First of all, the 149 components of LGZG were preliminarily identified or inferred by UPLC-Q/ TOF-MS, including flavonoids, triterpenoids, phenylpropanoids, organic acids, quinones, and other types of components. en, a total of 20 biomarkers were screened by network pharmacological screening based on the presumed chemical composition. Finally, the quantitative determination method of biomarkers in LGZG was established by UPLC-QQQ-MS/MS, which covered all 4 herbal medicines in LGZG. Combined with statistical analysis, the quality evaluation system of LGZG was established. Following the results of this study, its pharmacological effects point to lipid metabolism-related targets and pathways. erefore, the experimental group intends to carry out further studies on LGZG metabolomics and lipidomics to provide further data support for its pharmacodynamics and pharmacological mechanism.
Data Availability e figure and table data used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Baolin Li and Shuaishuai Fan contributed equally to this work.