A Strategy for Screening the Lipid-Lowering Components in Alismatis Rhizoma Decoction Based on Spectrum-Effect Analysis

Alismatis Rhizoma decoction (ARD), comprised of Alisma plantago-aquatica subsp. orientale (Sam.) Sam and Atractylodes macrocephala Koidz. at a ratio of 5 : 2, is a classic traditional Chinese medicine (TCM) formula with successful clinical hypolipidemic effect. This paper aimed to explore the major bioactive compounds and potential mechanism of ARD in the treatment of hyperlipidemia on the basis of spectrum-eﬀect analysis and molecular docking. Nine ARD samples with varying ratios of the constituent herbs were prepared and analyzed by UPLC-Q-TOF/MS to obtain the chemical spectra. Then, the lipid-lowering ability of the nine samples was tested in an oleic acid-induced lipid accumulation model in human hepatoma cells (HepG2). Grey relational analysis and partial least squares regression analysis were then performed to determine the correlation between the chemical spectrums and lipid-lowering efficacies of ARD. The potential mechanisms of the effective compounds were investigated by docking with the farnesoid X receptor (FXR) protein. The results indicated that alisol B 23-acetate, alisol C 23-acetate, and alisol B appeared to be the core eﬀective components on hyperlipidemia in ARD. Molecular docking further demonstrated that all three compounds could bind to FXR and were potential FXR agonists for the treatment of hyperlipidemia. This study elucidated the eﬀective components and potential molecular mechanism of action of ARD for treating hyperlipidemia from a perspective of different compatibility, providing a new and feasible reference for the research of TCM formulas such as ARD.


Introduction
Alismatis Rhizoma decoction (ARD), comprised of the rhizomes of Alisma orientale (Sam.) Juzep. (AR) and the rhizomes of Atractylodes macrocephala Koidz. (AMR) at a ratio of 5 : 2, was first described by Zhongjing Zhang in "Jingui Yaolue" [1]. Although the composition of ARD is simple, its therapeutic effect is remarkable. Recent studies show that ARD can reduce the triglyceride and total cholesterol levels in serum and liver and significantly increase the concentrations of serum HDL-cholesterol, possessing successful clinical capability of treating hyperlipidemia [2,3]. Even though the hypolipidemic activity of ARD is notable, little information is available about the active compounds in ARD, obstructing the in-depth study of the molecular mechanism of ARD for the treatment of hyperlipidemia. So far, triterpenoids, sesquiterpenoids, flavonoids, and polysaccharides have been identified in the two individual herbs AR and AMR that exhibit hypolipidemic, antioxidative, and anti-inflammatory effects [4][5][6][7]. However, the core hypolipidemic components of ARD remain to be clarified.
As a highly complex system containing multiple components, rational strategies are required to be devised to identify pharmacologically active compounds in TCM. For instance, the spectrum-effect study is increasingly being used to discern the specific biological active compounds in TCM by analyzing the correlation between the spectral profiles and biological effects [8]. Reasonable chemometric methods, such as canonical correlation analysis, grey relational analysis, and partial least squares regression analysis, are frequently utilized for the spectrum-effect research [9]. Also, multiple data analysis methods are recommended to be simultaneously applied into the process of exploring the active ingredients to obtain better accuracy. Furthermore, for the reported literature related to spectrum-effect study, most of them focused on single herb medicine, and herbal formula as the research object only accounted for a small part due to its complicated composition. So, developing appropriate and feasible approaches from a holistic perspective to carry out the spectrum-effect study of TCM prescription possesses important practical significance.
ere is considerable interest in developing novel selective FXR ligands to treat diseases related to disorders of lipid metabolism. It has been reported that ARD could activate the LXRα/ABCA1 pathway and maintain homeostatic balance of ABCA1, which played a role in promoting lipid metabolism [15]. Moreover, the chemical components in ARD such as alisol M 23-acetate, alisol A, and alisol B 23acetate were reported to act as the LXR agonist [16,17]. Furthermore, the HepG 2 cell line can express FXR, and the cells are often used to investigate the agonists or inhibitors of FXR [18,19]. e HepG 2 cell line also has the advantages of easy culture and transfection, and most features of the hepatocyte cell line are preserved in HepG 2 cells.
In the present study, we started from changing the compatibility of AR and AMR in ARD to comprehensively implement the spectrum-effect study. e chemical spectra and lipid-lowering efficacies of ARD extracts in HepG 2 cells with different compatibilities were correlated to reveal the core lipid-lowering compounds, and their potential interactions with FXR were then determined by molecular docking in order to verify the accuracy of spectrum-effect analysis and probe the mechanism of action of ARD on hyperlipidemia.

Preparation of Samples.
Nine different ARD samples were prepared by combining AR and AMR in the ratio (w/w) of 3 : 1, 5 : 2, 2 : 1, 3 : 2, 1 : 1, 2 : 3, 1 : 2, 2 : 5, and 1 : 3. e samples were crushed and refluxed thrice in 95% ethanol for 2 h each. e extracts were filtered, combined, and concentrated to obtain a 10% ethanol extract. To prepare working solutions for chromatography, 50 mg of each ARD crude extract was diluted in 10 mL 80% methanol to yield a solution of concentration 5 mg/mL. e standard solution was prepared by dissolving accurately weighted portions of the five reference compounds in MeOH.

2.3.2.
e Establishment of the Spectrum. Under the abovementioned UPLC-MS conditions, the bask peak ion (BPI) chromatograms of the nine ARD samples were obtained as the chemical spectrum in the spectrum-effect study. en, nine chemical chromatograms were automatically matched by the professional software named Similarity Evaluation System for Chromatographic Spectrum of Traditional Chinese Medicine (2004A version, Committee of Chinese Pharmacopeia) to generate the common peaks of the different ARD samples [20].

Qualitative Analysis of ARD.
e UPLC-MS-based qualitative analysis of ARD was performed in our previous study. e chromatographic and mass spectrometric conditions were the same as the conditions above. e results could be confirmed in the published paper in detail [21].

Lipid-Lowering Effect Investigations
2.4.1. Cell Viability Assay [21]. HepG 2 cells from the laboratory of Chinese Academy of Medical Sciences and Peking Union Medical College (Beijing, China) were maintained in complete DMEM (10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin) at 37°C under 5% CO 2 . e cells were seeded in a 96-well plate at the density of 1 × 10 4 cells/ well and cultured for 24 h with 10, 20, 40, 80, 160, 320, and 640 μg/mL ARD. After washing thrice with PBS, fresh medium with 100 μL 10% CCK-8 solution was dispensed per well, and the cells were incubated for another hour [22]. e absorbance at 450 nm was measured. e experiment was repeated at least thrice, with six replicates per dose.

Inhibition of Lipid Accumulation in HepG 2 Cells of ARD Samples.
e HepG 2 cells were seeded in a 96-well cell plate at the density of 1 × 10 4 cells per well and cultured for 24 h. After replacing with fresh medium supplemented with 0.2 mM oleic acid and the suitable ARD dosage, the cells were incubated for 24 h [20,21,23]. After washing thrice with PBS, the cells were stained with 0.3 mM Nile red for 15 min in the dark and rinsed with PBS to remove the excess dye [24,25]. e fluorescence emission at 572 nm was read using a microplate reader (Tecan, Switzerland).

Statistical Analysis.
All the data were analyzed by oneway analysis of variance (ANOVA) using SPSS 23.0 (SPSS, USA). Results were expressed as mean ± standard deviation (S.D.). A value of P < 0.05 was considered to be a significant difference.

Grey Relational Analysis (GRA)
. GRA, originating from grey system theory, is used to calculate the correlation between multiple variables on the basis of the geometric similarities [26]. Owing to its superiority in analyzing the complicated relationships of multiple variables, GRA is particularly suitable to analyze the mixture consisting of known and unknown information such as TCM which contains a myriad of chemicals [9]. For these reasons, the online software SPSSAU was utilized to synthetically evaluate the correlation between the common peaks from the chromatographic spectrums and lipid-lowering action to seek out the active compounds of ARD. In this analysis, grey relational coefficients were used to assess the degree of correlation. A high grey relational coefficient implies a high correlation between the peaks and the efficiency. us, peaks with high grey relational coefficients were selected [27,28].

Partial Least Squares Regression (PLSR) Analysis.
PLSR is another data analysis technique which is frequently applied to determine the correlation between a set of dependent variables and a larger set of independent variables [29]. It can reflect the dependency of each common peak to the efficiency in terms of the variable importance in projection (VIP) value, and values greater than 1 indicate significant correlation [30]. Similarly, the regression coefficient is also an important parameter of PLSR used in the spectrum-effect relationship study to reflect the qualitative trend [31].
e PLSR between chemical spectrum and bioactivity was modeled using SIMCA-P 13.0 (Umetrics), and the regression coefficient and VIP values were calculated.

Molecular Docking.
e MOE software was used for the docking between the ligands and the protein FXR. Before docking, the 2D structure of the small molecule compounds were converted to 3D structures through energy minimization. en, the 3D X-ray structure of the protein FXR was downloaded from RCSB Protein Data Bank (PDB ID: 6HL1). In the MOE-Dock program, the force field of AMBER10: EHT and the implicit solvation model of Reaction Field (R-field) were selected for the docking. e side chains of the protein pocket were allowed to move according to ligand conformations during the docking process, and the weight used for tethering side chain atoms was set as 10. e docked positions were ranked by London dG scoring, and the top 10 positions were rescored by GBVI/WSA dG. e best binding conformation was one with least free energy.

UPLC-MS Spectrums.
For the first step of spectrumeffect analysis, nine ARD samples with AR and AMR in the ratio (w/w) of 3 :1, 5 : 2, 2 :1, 3 : 2, 1 :1, 2 : 3, 1 : 2, 2 : 5, and 1 : 3 were prepared and analyzed under the optimized UPLC-Q-TOF/MS analytical conditions. Nine MS spectra which represented the chemical information of ARD with different compatibility ratios were constructed ( Figure 1). en, the spectra were imported into the idiomatic software to automatically match the common peaks. Finally, based on retention time and the MS/MS spectra, a total of twenty common peaks were obtained as shown in Table 1. It was found that the nine samples possessed similar chemical profiles, but there were still some significant differences among these ARD samples due to the variations in the AR/AMR ratio, as values of the common peak areas fluctuated greatly in nine samples, reflecting unequal contents of the corresponding compounds. is result provided the essential content "spectrum" of the spectrum-effect study, and the calculated common peaks could be utilized to precisely pinpoint active ingredients in ARD.

ARD Mitigated Lipid Accumulation in HepG 2 Cells.
As the chemical constituent basis is closely related to the pharmacological activity, examinations of the toxicities and activities of the nine ARD samples were performed as the second stage of the spectrum-effect study. As shown in Table 3, all nine ARD extracts at the concentration of 160 μg/ mL did not show significant impact on the cell viability of HepG 2 cells, and this concentration was selected for the subsequent pharmacological assay.
An oleic acid-induced lipid accumulation model in HepG 2 cells was utilized to evaluate the lipid-lowering ability of the nine ARD extracts, and Nile red staining was performed to quantify the intracellular lipid contents by detecting the fluorescence intensity. e lower the value of the fluorescence, the better was the effect of the sample. When the nine ARD samples were added to the lipid accumulated model system, different efficacy values were obtained because of the different chemical substance basis ( Retention time (min) Absorption intensity Figure 1: UPLC-MS spectra of the 9 ARD samples.  e common peaks and the inhibition on lipid accumulation of ARD were, respectively, used as X matrix and Y matrix in the GRA to find the active compounds corresponding to the lipid-lowering efficacy. e result of the GRA was shown in Figure 2. e prominent peaks in terms of correlation coefficients were X20 (0.82), X18 (0.79), X9 (0.77), and X11 (0.72), which were respectively attributed to alisol B 23-acetate, alisol B, alisol C 23-acetate, and atractylenolide I, showing that these four peaks played major role in the activity in ARD. Furthermore, it could be seen that the whole four peaks had large peak areas, which were indicative of high contents in ARD. us, peak X20, X18, X9, and X11 were clearly pointed to major biological roles for the lipid-lowering effect.

PLSR.
In addition to GRA, the relationship between the twenty common peaks (X-variables) and the impact on lipid accumulation (Y-variables) was also evaluated by using a PLSR model. e R2X and Q2 values of the PLSR model were 0.997 and 0.986, respectively, demonstrating  that the model was credible for the spectrum-effect analysis. As shown in Figure 3, the peaks X3 (aetradecylcitric acid), X6 (16-oxo-11-anhydro-alisol A), X12 (alisol G), and X20 (alisol B 23-acetate) correlated strongly with the lipid-lowering effect with high positive correlation coefficients. Furthermore, the VIP values (Figure 4) of peaks X20 (alisol B 23-acetate), X9 (alisol C 23-acetate), and X18 (alisol B) were over 1.5, showing these peaks possessed the most important effects on the Y-variable. Also, the peaks X11 (atractylenolide I), X6 (16-oxo-11-anhydro-alisol A), X3 (aetradecylcitric acid), and X15 (alisol O) showed VIP values greater than 1, which were also indicative of biological significance. Based on the entire analysis of PLSR, peaks X20, X9, X18, X3, and X6 scored high in terms of VIP values and correlation coefficients and were eventually identified as the lipid-lowering active compounds in ARD.
Taken together the results from both the GRA and PLSR models, alisol B 23-acetate, alisol B, and alisol C 23-acetate were confidently identified as the major core lipid-lowering compounds in ARD.

Docking between the Effective Compounds and FXR.
e MOE software was used for the docking between compounds alisol B 23-acetate, alisol B, alisol C 23-acetate, and protein FXR, and the binding sites and the docking scores were obtained. e docking scores of the above compounds are summarized in Supplementary  Table S1. The common peaks Grey relational coefficiens X19 X20 Figure 2: Grey relational analysis of the fingerprint and bioactive effect. e grey relational coefficients were used to assess the correlation between the common peaks and the lipid-lowing efficiency. e common peaks and the grey relational coefficients were respectively used as X matrix and Y matrix to find the potentially active compounds. The common peaks X1  X2  X3  X4  X5  X6  X7  X8  X9  X10  X11  X12  X13  X14  X15  X16  X17  X18  X19  X20 -10 The correlation coefficients 0 Figure 3: e correlation coefficients of the PLSR model. 6 Journal of Analytical Methods in Chemistry Alisol B 23-acetate had the most negative docking score, which corresponded to best possible binding to FXR. e binding affinities of alisol C 23-acetate and alisol B 23-acetate to FXR were similar. As shown in Supplementary Figure S1, alisol B 23-acetate was sterically compatible with the binding site of FXR and formed hydrogen bonds via the oxygen atom in the ring ether at C-24 with the side chain N atom of His294 and Arg331 in FXR. Other hydrogen bonds were observed between the C-11 hydroxyl groups and the backbone O atom of Leu287 and the O atom in the carbonyl groups with the side chain N atom of His447 and the side chain O atom of Tyr361. Van der Waals (VDW) interactions were also observed between alisol B 23-acetate and protein FXR, which likely contributed to the binding energy.
As shown in Supplementary Figure S2, the binding mode of alisol C 23-acetate with FXR protein was similar to that of alisol B 23-acetate, except the absence of the hydrogen bond with His294. Alisol B had relatively poor affinity to FXR with a docking score of −10.38 kcal/mol and formed hydrogen bonds with Arg331, His447, and Tyr361 (Supplementary Figure S3). Overall, alisol B 23-acetate, alisol C 23-acetate, and alisol B can bind to FXR via the hydrogen bond and VDW interactions. e results of docking suggested the three core active compounds might present the lipid-lowering effect by affecting the activity of FXR.

Discussion
TCM formulas are promising in treating complex pathologies for their characteristics of multicomponent and multipathway.
eir extremely complex chemical composition, however, poses challenges to quality control.
us, TCM formulas can hardly be broadly accepted by the international medical community, especially the western academia [37]. Spectrum-effect study is considered as an effective technique to find active ingredients in TCM and contributes a lot for the quality control [38]. e premise of performing spectrum-effect analysis is that there must be a certain amount of samples to supply the diversity chemical spectrums and pharmacodynamic information. For the single Chinese herbal medicine, different varieties, regions, and different harvest times all contribute to provide sufficient samples with different material basis for the analysis [8]. When it comes to TCM prescription, screening simple and appropriate methods to guarantee the required samples is an imperative and challenging step needed to be overcome. In the previously reported literature, the changing therapeutic effects brought about by altering the ratio of AR and AMR from 5 : 2 to 2 : 5 or 1 : 1 were memorably observed [39]. is result yielded a creative hint to us that it might be feasible to conduct the spectrumeffect study from the perspective of changing the compatibility of AR and AMR in ARD. erefore, we made a new attempt by setting the ratio of AR and AMR as 3 : 1, 5 : 2, 2 : 1, 3 : 2, 1 : 1, 2 : 3, 1 : 2, 2 : 5, and 1 : 3 to gain dissimilar ARD samples for the analysis, and our result proved this idea could be successfully applied into the spectrum-effect analysis and was easy to be operated to explore pharmacological ingredients in ARD. Specifically, qualitative trends (positive or negative) were obtained through PLSR [31], and quantitative explanations were from GRA and VIP values [16,40]. Peaks of X20 (alisol B 23-acetate), X9 (alisol C 23-acetate), and X18 (alisol B) showed positive correlation coefficients in Figure 3, and they got high VIP values as well as grey relational coefficients compared with other peaks at the same time. Accordingly, the three ingredients were more likely to have better lipid-lowering efficiency than others in ARD. erefore, our work offered valuable insights into the research of efficacy components in TCM formula such as ARD. Furthermore, to ensure the accuracy of the spectrum-effect analysis, the chromatographic and mass spectrometric conditions were optimized, and the water content in AR and AMR was determined using near-infrared spectroscopy [41].
ARD has been shown to be of clinical value in treating hyperlipidemia.
e lipid-lowering effect of ARD is generally considered to correspond with a flurry of triterpenoids and lactones present in ARD, such as alisol A, alisol B, alisol The common peaks

Journal of Analytical Methods in Chemistry
A 23-acetate, alisol A 24-acetate, and alisol B 23-acetate [6,[42][43][44]. Related investigations, however, focused more on the lipid-lowering effect of monomeric compounds. e monomeric compounds were administered in vivo and in vitro, which was not equivalent to administration of TCM formulas for different contents of various compounds, let alone their combinations. us, the core lipid-lowering components present in the prescription ARD are actually unclear.
In previous studies, a large number of compounds were identified in AR and AMR [35,45], which could contribute greatly to their quality control and pharmacological research. However, identification of compounds in ARD was scanty. It was reported that alisol A, alisol B, alisol B 23acetate, atractylode I, atractylode II, and atractylode III were determined in ARD [46]. Moreover, we determined twentyfour compounds in the ethanol and aqueous extracts of ARD [21]. Basing on these results, we further identified twenty common peaks in ARD, of which 15 compounds were of definitive structure. Alisol B 23-acetate, alisol C 23-acetate, and alisol B were considered to be the core lipid-lowering components of ARD through the mathematical models of the spectrum-effect study. Molecular docking further showed that these compounds can bind to the FXR protein through hydrogen bonding and VDW interactions, and the key binding sites were the carbonyl group at C-3, hydroxyl groups at C-11, and the ring ether structure at C-24. A previous study showed that alisol B 23-acetate reduced hepatic lipogenesis and increased lipolysis via FXR activation, resulting in anti-inflammatory, anti-fibrotic, and hepatoprotective effects [17,47].
us, it is rational to surmise that alisol B and alisol C 23-acetate may also exert their lipid-lowering effects as FXR agonists. Since alisol C 23acetate exhibited almost the same binding affinity to FXR as alisol B 23-acetate, its biological activity is worth further exploration.

Conclusion
In the present study, the relationship between the spectrum and efficacy of 9 samples of ARD with different ratios of AR and AMR was elucidated. With the help of chemometric methods, alisol B 23-acetate, alisol C 23-acetate, and alisol B were found might be the core bioactive components of ARD. e three compounds can bind to the FXR protein via hydrogen bonding and VDW interactions, indicating that they may exert lipid-lowering effects as FXR agonists. is study provided new impetus for the exploration of the active constituents and further molecular mechanism studies of TCM such as ARD.
Data Availability e original contributions presented in the study are included in the article/supplementary material; further inquiries can be directed to the corresponding author.

Conflicts of Interest
e authors declare no conflicts of interest.

Authors' Contributions
Xiao-Yan Chang and Jia-Shuo Wu contributed equally to this work. Yue Shi and Xiao-Yan Chang designed the study. Xiao-Yan Chang and Jia-Shuo Wu conducted the experiments. Zhuang-Zhuang Li, Wei-Yi Jin, and Fang-Qing Zhang refined the framework of the present study according to the relevant research. Yue Shi and Xiao-Yan Chang analyzed the data and wrote the manuscript. Jia-Shuo Wu polished and revised the manuscript.