Network Pharmacology Prediction and Metabolomics Validation of the Mechanism of Fructus Phyllanthi against Hyperlipidemia

Summary

The present protocol describes an integrated strategy for exploring the key targets and mechanisms of Fructus Phyllanthi against hyperlipidemia based on network pharmacology prediction and metabolomics verification.

Abstract

Hyperlipidemia has become a leading risk factor for cardiovascular diseases and liver injury worldwide. Fructus Phyllanthi (FP) is an effective drug against hyperlipidemia in Traditional Chinese Medicine (TCM) and Indian Medicine theories, however the potential mechanism requires further exploration. The present research aims to reveal the mechanism of FP against hyperlipidemia based on an integrated strategy combining network pharmacology prediction with metabolomics validation. A high-fat diet (HFD)-induced mice model was established by evaluating the plasma lipid levels, including total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C). Network pharmacology was applied to find out the active ingredients of FP and potential targets against hyperlipidemia. Metabolomics of plasma and liver were performed to identify differential metabolites and their corresponding pathways among the normal group, model group, and intervention group. The relationship between network pharmacology and metabolomics was further constructed to obtain a comprehensive view of the process of FP against hyperlipidemia. The obtained key target proteins were verified by molecular docking. These results reflected that FP improved the plasma lipid levels and liver injury of hyperlipidemia induced by a HFD. Gallic acid, quercetin, and beta-sitosterol in FP were demonstrated as the key active compounds. A total of 16 and six potential differential metabolites in plasma and liver, respectively, were found to be involved in the therapeutic effects of FP against hyperlipidemia by metabolomics. Further, integration analysis indicated that the intervention effects were associated with CYP1A1, AChE, and MGAM, as well as the adjustment of L-kynurenine, corticosterone, acetylcholine, and raffinose, mainly involving tryptophan metabolism pathway. Molecular docking ensured that the above ingredients acting on hyperlipidemia-related protein targets played a key role in lowering lipids. In summary, this research provided a new possibility for preventing and treating hyperlipidemia.

Introduction

Hyperlipidemia is a common metabolic disease with serious impacts on human health, and is also the primary risk factor for cardiovascular diseases¹. Recently, there has been a downward age-related trend for this disease, and younger people have become more susceptible because of long-term irregular lifestyles and unhealthy eating habits². In the clinic, various drugs have been used to treat hyperlipidemia. For example, one of the most commonly used drugs for patients with hyperlipidemia and related atherosclerotic disorders is statins. However, long-term use of statins has side effects that can't be neglected, which lead to a poor prognosis, such as intolerance, treatment resistance, and adverse events^3,4. These shortcomings have become additional pains for hyperlipidemia patients. Therefore, novel treatments for stable lipid-lowering efficacy and fewer side effects should be proposed.

Traditional Chinese Medicine (TCM) has been widely used to treat diseases because of its good efficacy and few side effects⁵. Fructus Phyllanthi (FP), the dried fruit of Phyllanthus emblica Linn. (popularly known as amla berry or Indian gooseberry), is a famous medicine and food homologous material of traditional Chinese and India medicines^6,7. This medicine has been used for clearing heat, cooling blood, and promoting digestion, as per TCM theories⁸. Modern pharmacological studies have shown that FP is rich in bioactive compounds such as gallic acids, ellagic acids, and quercetin⁹, which are responsible for a range of multifaceted biological properties, by acting as an antioxidant, an anti-inflammatory, liver protection, an anti-hypolipidaemic, and so on¹⁰. Recent research has also showed that FP could effectively regulate the blood lipids of patients with hyperlipidemia. For example, Variya et al.¹¹ have demonstrated that FP fruit juice and its main chemical ingredient of gallic acid can decrease plasma cholesterol and reduce oil infiltration in the liver and aorta. The therapeutic efficacy was related to FP's regulation in increasing the expression of peroxisome proliferator-activated receptor-alpha and decreasing hepatic lipogenic activity. However, the underlying mechanism of FP in improving hyperlipidemia should be further investigated, because its bioactive ingredients are quite extensive. We sought to explore the potential mechanism of FP's therapeutic efficacy, which may be beneficial for the further development and utilization of this medicine.

Currently, network pharmacology is regarded as a holistic and efficient technique to study the therapeutic mechanism of TCM. Instead of looking for single disease-causing genes and drugs treating solely an individual target, a complete drug-ingredients-genes-diseases network is constructed to find the multi-target mechanism of the multi-ingredient drug regarding their comprehensive treatment¹². This technique is especially suitable for TCM, as their chemical compositions are massive. Unfortunately, network pharmacology can only be used to forecast targets affected by chemical ingredients in theory. The endogenous metabolites in the disease model should be observed to validate the effectiveness of network pharmacology. The metabolomics method, which emerges with the development of systems biology, is an important tool for monitoring the changes in endogenous metabolites¹³. The changes in metabolites reflect the steady state changes of the host, which is also an important indicator for studying the internal mechanism. Some researchers have successfully integrated network pharmacology and metabolomics to explore the interaction mechanism between drugs and diseases^14,15.

This article explores the mechanistic basis of FP against hyperlipidemia by integrating network pharmacology and metabolomics techniques. Network pharmacology was applied to analyze the relationship between the main active ingredients in FP and molecular targets for hyperlipidemia. Subsequently, metabolomics was performed to observe the change of endogenous metabolites in the animal model, which can explain the medicine actions at the metabolic level. Compared with the application of network pharmacology or metabonomics alone, this integrated analysis provided a more specific and comprehensive research mechanism. Additionally, the molecular docking strategy was used to analyze the interaction between active ingredients and key proteins. In general, this integrated approach could compensate for the lack of experimental evidence for network pharmacology and the lack of an endogenous mechanism for the metabolomics method, and can be used for the therapeutic mechanism analysis of natural medicine. The main schematic flowchart of the protocol is shown in Figure 1.

Protocol

All procedures involving the handling of animals were conducted in accordance with the Chengdu University of Traditional Chinese Medicine Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Ethics Committee of the Chengdu University of Traditional Chinese Medicine (Protocol number 2020-36). Male C57BL/6 mice (20 ± 2 g) were used for the present study. The mice were obtained from a commercial source (see Table of Materials).

1. Network pharmacology-based prediction

NOTE: The network pharmacology is used to predict the active ingredients and their key targets of FP against hyperlipidemia.

1. Selection of active ingredients and key targets
   1. Search the keyword "Phyllanthi Fructus" on the Traditional Chinese Medicine system's pharmacology database (TCMSP; http://tcmspw.com/tcmsp.php) to obtain the list of the candidate active ingredients and targets of FP.
      NOTE: Normally, only ingredients with oral bioavailability (OB) ≥30% and drug-like (DL) values ≥0.18 in the database are included as active ingredients.
   2. Search the keyword "hyperlipidemia" in the GeneCards database (https://www.genecards.org/), the Online Mendelian Inheritance in Man database (OMIM; https://omim.org/), and the therapeutic target database (TTD; http://db.idrblab.net/ttd/) to obtain the respective candidate targets of hyperlipidemia. Download the spreadsheets of disease targets. Delete the repeated targets to obtain the hyperlipidemia targets list.
   3. Copy these lists from steps 1.1.1 and 1.1.2 into a new spreadsheet. Use the "Data - Identify Duplicates" function in the toolbar to get intersection targets. Import the intersection target list into UniProtKB (http://www.uniprot.org/) to standardize the gene and protein names.
      NOTE: These targets are related to both FP and hyperlipidemia. Therefore, predict these intersection targets as the targets of FP against hyperlipidemia.
2. Construction of a protein-protein interaction network
   1. Open the STRING database (https://string-db.org/) 11.5. Paste the intersection target list of FP against hyperlipidemia in the "List of Names" dialog box. Select Homo sapiens in "Organisms" and click on the SEARCH > CONTINUE.
      NOTE: Humans and mice have highly similar genes. Therefore, further experimental verification is carried out with mice.
   2. When the results are available, tick the hide disconnected nodes in the network in "Advanced Settings". Set the highest confidence (0.900) in "minimum required interaction score", then click on the UPDATE button.
   3. Click on Exports in the title bar, and download the short tabular text of the protein-protein interaction (PPI) network in PNG and TSV format.
3. Construction of a drug-component-disease-target network
   1. Open Cytoscape 3.9.1 (see Table of Materials). Import the TSV format file of step 1.2.3. Optimize the color, font, and side of the network nodes through the style bar in the control panel.
   2. Use the "Analyze Network" function for network topology analysis. Obtain hub genes by CytoHubba in Cytoscape. Establish the drug-ingredient-target-disease network.
4. GO and KEGG enrichment analysis
   1. Open DAVID Bioinformatics Resources (https://david.ncifcrf.gov/home.jsp). Click on Start Analysis and paste the target list into the left dialog box. Select OFFICIAL GENE SYMBOL in "Select Identifier". Select Homo sapiens in "Select species". Tick Gene List in "List Type". Click on Submit List.
   2. When the results are available, click on Analyze above gene list with one of DAVID tools. Tick GOTERM_BP_DIRECT, GOTERM_CC_DIRECT, GOTERM_MF_DIRECT in "Gene Ontology" for GO function enrichment analysis. Tick KEGG_Pathway in "Pathways" for KEGG pathway enrichment analysis.
   3. Click on Functional Annotation Chart to display the results.
      ​NOTE: Set the statistical significance threshold of the enrichment analysis at p < 0.05.

2. Experimental design

1. The FP aqueous extract preparation
   NOTE: FP is processed in the laboratory of Professor Lina Xia at the Chengdu University of TCM⁸.
   1. Soak the dried powder of FP (90 g) in 1 L of pure water in a clean 2 L volumetric flask. Use ultrasonic treatment (in a 4 °C water bath, power: 250W, frequency: 35 kHz) to help dissolve for 30 min. Filter the solution to obtain the extract with a double-layer, 1 mm x 1 mm sterile medical gauze. Repeat the above operation three times to ensure the complete dissolution of FP.
   2. Use the rotary evaporation method for further concentration. Set the rotation speed to 50 rpm with a temperature of 60 °C for 4 h. Concentrate the aqueous extract to 100 mL.
   3. Divide the crude extract of FP (0.9 g/mL) evenly into two parts (50 mL). One part is used as the high-dose FP liquid (0.9 g/mL). Add 50 mL of pure water into another part, and consider it as the low-dose FP liquid (0.45 g/mL). Use the high- and low-dose FP aqueous solutions for administration. Store the liquid at -20 °C until use.
2. Animal preparation
   1. House 50 male C57BL/6 mice (20 ± 2 g) in a well-ventilated room at room temperature, with a 12 h light-dark cycle and free access to food and pure water.
   2. Randomly assign the mice to two groups: feed 10 mice with a normal diet and 40 mice with a high-fat diet (see Table of Materials) to induce hyperlipidemia.
      NOTE: After feeding for 8 weeks, the mice were screened for further drug intervention.
   3. In the 8th week, withdraw about 200 µL of blood from each mouse orbit. Centrifuge the blood for 10 min at 5,733 x g at 4 °C to obtain plasma samples. Determine the TC and TG levels with commercially available assay kits (see Table of Materials).
   4. Select six mice with the most normal lipid levels as the no-treatment control (NC) group. Select 24 mice with a significantly higher lipid level as the high-fat diet group, and randomly divide them into four groups: high-fat diet (HFD) group, low-dose FP (FP_L) group, high-dose FP (FP_H) group, and positive control (PC) group.
   5. Administer gastric irrigation to the FP_L and FP_H groups with two dosages of FP (low dose, 4.5 g/kg and high dose, 9 g/kg), respectively; gastric irrigation to the PC group with simvastatin tablets (5 mg/kg; see Table of Materials); and gastric irrigation to the NC and HFD groups with the same volume of physiological saline once a day for 4 weeks.
      NOTE: The current study used the aqueous solutions of FP and simvastatin for treatment.
   6. In the 12th week, after anesthesia by 1% pentobarbital sodium (30 mg/kg), sacrifice the mice of all the groups. Collect ~400 µL blood samples from each mouse's orbital vein.
      NOTE: Stimulate the toes and soles of the mice with tweezers. If there is no reaction, it proves adequate anesthesia.
   7. Centrifuge the blood for 10 min at 5,733 x g at 4 °C to obtain plasma samples, and determine the TC, TG, LDL-C, and HDL-C levels with commercially available assay kits (see Table of Materials). Obtain liver tissue samples¹⁶ and subject them to histopathological analysis. Use the remaining plasma and liver samples for metabolomics analysis (step 3).
      NOTE: All samples are stored at -80 °C until use.
3. Liver histopathological examination
   1. Fix fresh liver tissues with 4% paraformaldehyde solution for more than 24 h. Take out the tissue from the fixative and smooth the target tissues with a scalpel. Place the tissue and the corresponding label into the dehydrator.
   2. Dehydrate in an ethanol gradient: 75% alcohol for 4 h, 85% alcohol for 2 h, 90% alcohol for 2 h, 95% alcohol for 1 h, absolute ethanol for 1 h, xylene for 30 min. Place the tissue cassette in a tissue mold in paraffin wax for three washes, 30 min each¹⁶.
   3. Put the wax-soaked tissues into the tissue embedder (see Table of Materials). Before the wax solidifies, remove the tissues from the dehydrator, put them into the embedded box, and attach the corresponding label.
   4. Cool the wax blocks in a -20 °C freezing table, remove them from the embedded frame, and trim the wax block.
   5. Cut the trimmed wax blocks into 3 µm thick sections using a microtome (see Table of Materials). Float the sections in 40 °C water, remove them from the slides, and bake in a 60 °C oven. After baking with water and dry wax, take it out and keep it at room temperature.
   6. Successively place the sections in xylene I for 10 min, xylene II for 10 min, xylene III for 10 min, absolute ethanol I for 5 min, absolute ethanol II for 5 min, 75% alcohol for 5 min, and wash in water¹⁶.
   7. Stain the sections with hematoxylin staining solution for 4 min, 1% hydrochloric acid alcohol solution (75% alcohol) for differentiation, 1% ammonia water solution back blue, and wash them with water.
   8. Stain the sections with eosin staining solution for 2 min and wash them with water.
   9. Observe the sections using an optical microscope with a magnification of 200x and 400x.
4. Liquid chromatography-mass spectrometry )LC-MS) analysis
   1. Ingredient identification of FP
      NOTE: The analysis is performed using ultra-high-performance liquid chromatography coupled with hybrid quadrupole-orbitrap high-resolution mass spectrometry (UPLC-Q-Orbirap HRMS, LC-MS; see Table of Materials).
      1. Precisely measure 1 g of dried powder of FP and put it into a clean 50 mL volumetric flask.
      2. Add 25 mL of 70% methanol into the volumetric flask and accurately weigh. Use ultrasonic treatment (in a 4 °C water bath, power: 250 W, frequency: 35 kHz) for 30 min to help dissolution. Accurately weigh again to precisely determine the loss after dissolution, and use 70% methanol to make up the loss.
         NOTE: Don't measure the volume, as the scale of the volumetric flask is not accurate, especially after the 4 °C water bath.
      3. Shake up to mix fully. Use a 0.22 µm microporous membrane to filter.
   2. Plasma sample preparation
      1. Precisely add 100 µL of plasma (step 2.2.7) into a double volume (200 µL) of acetonitrile in a 1.5 mL centrifuge tube, and vortex it with a vortex vibrator for at least 30 s. Follow this procedure for all samples.
      2. Centrifuge all samples at 17,200 x g for 10 min at 4 °C. Transfer the supernatants after centrifugation into a new 1.5 mL centrifuge tube. Dry the supernatants under nitrogen. Reconstitute with 200 µL of extraction solvent (acetonitrile:water = 4:1 [v/v]).
      3. Vortex the reconstituted solution for at least 30 s and use ultrasonic treatment for 10 min (in a 4 °C water bath, power: 250 W, frequency: 35 kHz). Centrifuge at 17,200 x g for 10 min at 4 °C.
      4. Filter the supernatants with 0.22 µm filter membranes and keep them at 4 °C for analysis.
   3. Liver sample preparation
      1. Homogenize 90 mg of liver tissue (step 2.2.7) for 1 min in ice-cold methanol-water (1:1, v/v, 1 mL) and centrifuge them at 21,500 x g for 10 min at 4 °C. Transfer the supernatant into 1.5 mL centrifuge tubes. Follow this procedure for all samples.
      2. Extract the precipitates again following the same procedure, and pool the supernatants together into new 1.5 mL centrifuge tubes. Dry the supernatants under nitrogen. Reconstitute with 300 µL of the extraction solvent (methanol:water = 4:1 [v/v]).
      3. Vortex the reconstituted solution for at least 30 s and use ultrasonic treatment for 10 min (in a 4 °C water bath, power: 250 W, frequency: 35 kHz). Centrifuge at 17,200 x g for 15 min at 4 °C.
      4. Filter the supernatants with 0.22 µm filter membranes and keep them at 4 °C for analysis.
         NOTE: The pooled quality control (QC) samples were prepared by mixing 10 µL aliquots from each plasma and liver sample (one per six samples).
   4. Analysis parameters of LC-MS
      NOTE: The mobile phase consists of 0.1% formic acid (solvent A) and acetonitrile (solvent B). Transfer these solvents to a clean glass bottle and connect them with the LC-MS system.
      1. Set the gradient programs of plasma samples in the "inlet file" of the LC-MS system as follows: 1% B (0-1.5 min), 1%-60% B (1.5-13.0 min), 60%-99% B (13.0-20.0 min), maintain at 99% B (20.0-25.0 min), 99%-1% B (25.0-25.1 min), and maintain at 1% B until 27 min.
      2. Set the autosampler conditions of the plasma samples in the "inlet file" of the LC-MS system as follows: the volume of injection, 2 µL; and the flow rate, 0.3 mL/min, for each analysis.
      3. Set the gradient program of liver samples in the "inlet file" of the LC-MS system as follows: 1% B (0-1 min), 1%-53% B (1-15 min), 53%-70% B (15-30 min), 70%-90% B (30-32 min), 90%-95% B (32-40 min), 95%-1% B (40-42 min), and maintain at 1% B until 45 min.
      4. Set the autosampler conditions of the liver samples in the "inlet file" of the LC-MS system as follows: volume of injection, 5 µL; and the flow rate, 0.3 mL/min, for each analysis.
      5. Set the MS detection conditions of both the plasma and liver samples in the "MS tune file" of the LC-MS system. Perform the MS acquisition using both positive and negative ionization modes.
         ​NOTE: The heated electrospray ionization parameters are as follows: spray voltage: 3.5 kV for positive ionization and 3.8 kV for negative ionization; sheath gas flow: 55 arb; auxiliary gas flow: 15 arb; probe heater temperature: 300 °C; and capillary temperature: 350 °C.
      6. Import the collected raw data into Compound Discoverer software, and set the method template following the manufacturer's instructions (see Table of Materials).

3. Metabolomic validation

NOTE: The metabolomic profiling data of plasma and liver metabolites are imported into Compound Discoverer software to perform the metabolic feature extraction by adopting a molecular feature extraction algorithm. Set the parameters as follows: mass deviation, 5 x 10-6; mass range, 100-1,500; signal to noise ratio (SNR) threshold, 3; and retention time deviation, 0.05. Evaluate the stability and repeatability of metabolomics by the relative standard deviation (RSD) of QC peak areas.

1. Use SIMCA-P software (see Table of Materials) for multivariate statistical analysis of the integral values obtained from LC-MS findings. Use orthogonal partial least squares discriminant analysis (OPLS-DA) for the mean-centered data and the modeling of sample classes.
2. After the OPLS-DA test, consider the metabolites, with integral with variable importance in the projection (VIP) values of >1 and a p-value of <0.05 from Student's t-test as the potential differential metabolites.
3. Identify the disturbed metabolites and metabolic pathways by open database sources, including Human Metabolome (HMDB; http://www.hmdb.ca/), Kyoto Encyclopedia of Genes and Genomes (KEGG; https://www.kegg.jp/), and MetaboAnalyst5.0 (https://www.metaboanalyst.ca/).
4. Visualize the result views by MetaboAnalyst5.0 and the 'Wu Kong' platform (https://www.omicsolution.com/wkomics/main/).

4. Molecular docking

1. Download the 3D structure of the selected FP ingredients from the TCMSP database, respectively. Search the ingredient names in the 'Chemical name' search box and download the corresponding 3D structure files in mol2 format.
2. Download the crystal structures of the key targets from the AlphaFold Protein Structure Database (Alphafold DB;, https://alphafold.ebi.ac.uk/). Search the target names in the search box and download the corresponding crystal structures files in pdb format.
3. Import ingredients and target structures file into AutoDockTools software. Click on Edit > Delete Water to delete water molecules. Click on Edit > Hydrogens > Add to add hydrogens. Set the ingredients as the 'ligand' and perform blind docking by selecting the whole targets as the 'receptor'¹⁷.
4. Enter a value in the box behind "center" and "size" to adjust the newly-developed space, making it possible to fully encompass the ligand and the receptor. Save the ligand and receptor files in pdbqt format.
5. Use AutoDock Vina to perform molecular docking. Set the "Receptor" bar to the name of the 'receptor.pdbqt', and the "Ligand" bar to the name of the 'ligand.pdbqt'. Obtain the optimal location for ligand binding to the receptor. Record the binding energy value at the optimal position.
   NOTE: The docking process was calculated by the Genetic Algorithm¹⁴. All docking run options were default values. Docking frames will be automatically ranked from the highest to the lowest binding energy.
6. Import the docking files into PILP (Protein-ligand Interaction Profiler' https://plip-tool.biotec.tu-dresden.de/plip-web/plip/index) to get the visual system model. Download the model files in pse format, and import them into PyMOL software (see Table of Materials) to construct further visualization.

5. Statistical analysis

NOTE: Use SPSS statistical software (see Table of Materials) for data analysis. Consider the value of p < 0.05 as statistically significant.

1. Expresse the values as means ± standard deviation (SD).
2. Perform a one-way ANOVA followed by post hoc least significant difference (LSD), Dunnett (in case of equal variance), or Dunnett's T3 (in case of unequal variance) to test statistical significance among groups.

Representative Results

Network pharmacology
A total of 18 potential ingredients in FP were screened according to their pharmacokinetic and pharmacodynamic properties from the database and LC-MS analysis (the total ion chromatograms are shown in Supplementary Figure 1). Through relevant literature, the content of gallic acid is much higher than other ingredients and is effective in lowering lipids^9,11. Therefore, this ingredient was considered a potential ingredient too. In total, 19 ingredients and 134 ingredient-related targets of FP have been founded. All 19 ingredients are shown in Table 1. To select the most representative ingredients for further analysis, these ingredients were imported into Bioinformatics Analysis Tool for Molecular mechANism of Traditional Chinese Medicine database (BATMAN-TCM; http://bionet.ncpsb.org/batman-tcm/). According to the ingredient-target-pathway-disease network, some bioactive ingredients, such as gallic acid, quercetin, and beta-sitosterol, were identified as the most important ingredients of FP related to hypercholesterolemia and coronary atherosclerosis (Supplementary Figure 2). Among these, gallic acid is one of the most widely studied phenolic acids; it is the main bioactive ingredient presented in FP¹⁸. Meanwhile, gallic acid is also the highest content ingredient in FP; its concentration is usually 1% to 3%. El-Hussainy et al.¹⁹ have revealed that gallic acid can limit cardiac injury, improve lipid profile, and downregulate cardiac inflammatory markers. The contents of quercetin and beta-sitosterol are lower, but some studies have proved their effect on lowering lipids. Quercetin, as an important flavonoid that widely exists in plants, has various properties, such as antioxidant, anti-inflammatory and cardiovascular protection effects²⁰. Lu et al.²¹ have studied that the quercetin-enriched juice can attenuate TC, LDL-C, and HDL-C levels in healthy individuals with mild hypercholesterolemia. As for beta-sitosterol, clinical studies have shown that plant sterol can significantly prevent hypercholesterolemia and cardiovascular disease^22,23. Althwab et al.²⁴proved that beta-sitosterol could improve lipid profile and atherogenic index in HFD rats. It can be seen that the lipid-lowering effect of FP may be related to these three ingredients.

Additionally, 1,552 targets of hyperlipidemia-related from the Genecards, OMIM, and TTD databases were collected. After matching the 134 FP-related targets with the hyperlipidemia-related targets, 62 targets were identified as potential targets for FP against hyperlipidemia (Figure 2A). All the intersected targets were normalized to their official symbols, according to the UniProt database. Subsequently, the PPI network was constructed by STRING (Figure 2B) and Cytoscape (Figure 2C). Combining the scores of computational methods, the top 10 targets were ESR1, RELA, FOS, EGFR, HIF1A, AR, CCND1, IL6, MAPK8, and MYC. The details are presented in Supplementary Figure 3. All of these 62 targets are the basis of further analysis, which is integrated with the results of metabonomics.

GO and KEGG pathways were performed by enrichment analysis. The top 15 pathways, according to the number of targets, were selected for analysis according to the p-value. The GO enrichment results suggested that the biological processes and the molecular function of FP against hyperlipidemia were mainly related to gene expression and protein binding (Figure 2D). The KEGG enrichment proved that FP could intervene in the process of lipid metabolism and atherosclerosis (Figure 2E), which means FP relieves hyperlipidemia through affecting the lipid metabolism.

The effect of FP on plasma lipid levels and liver index
To test the improved effect of FP on hyperlipidemia, the changes in TC, TG, LDL-C, HDL-C, and liver index (the ratio of liver weight to body weight) were first measured. Compared with the NC group, mice in the HFD group showed a significant increase in the plasma levels of TC (p < 0.001), LDL-C (p < 0.001) and TG (p < 0.05), indicating that long-term HFD intervention can increase lipid levels and induce hyperlipidemia (Figure 3).

After administering FP aqueous extract, the levels of TC in FP_L and FP_H groups were significantly (p < 0.05) reduced by 18.8% and 12.4%, respectively (Figure 3A). The levels of LDL-C in FP_L and FP_H groups were significantly (p < 0.05) reduced by 13.7% and 21.8%, respectively (Figure 3B). Regarding HDL-C level, the FP_H group was significantly (p < 0.01) increased from 1.81 ± 0.08 mmol/L to 2.65 ± 0.16 mmol/L, compared with the HFD group (Figure 3C). Although the TG level remained insignificant after FP intervention, it was reduced compared with the HFD group (Figure 3D). Recent studies have indicated that the index of LDL-C/HDL-C ratio is a better index than LDL-C or HDL-C alone in predicting cardiovascular disease^25,26. Compared with the HFD group, the LDL-C/HDL-C ratio was significantly (p < 0.01) reduced by 46.3% in the FP_H group (Figure 3E), meaning that FP intervention reduced bad cholesterol and increased good cholesterol levels. As the main fat metabolic organ, the liver weight reflects fat storage in mice to a certain extent²⁷. After 12 weeks, the liver indexes in the FP_L group and FP_H group were significantly (p < 0.01) reduced compared with the HFD group (Figure 3F). The PC group also showed different degrees of reduction in these indicators above, demonstrating that FP had similar effects to statins, and the protective effect showed a dose-response relationship.

Various clinical studies revealed that, after taking either extract or whole FP for a while, TC and LDL-C levels were significantly decreased. Meanwhile, the HDL-C level was remarkably raised upon long-term administration of FP^28,29. Nambiar and Shetty³⁰ found that FP juice could reduce oxidized low-density lipoproteins, therefore greatly reducing the risk of atherosclerosis. Gopa et al.³¹ evaluated the hypolipidemic effect of FP in patients with hyperlipidemia and compared it with simvastatin. Treatment with FP resulted in a considerable reduction in TC, LDL-C, and TG, and a significant increase in HDL-C levels, similar to that of simvastatin. In this research, FP and simvastatin also had similar therapeutic effects, and the LDL-C-lowering effect and hepatic repairing action of FP were superior to simvastatin.

Liver histopathological observation
The effect of FP on hepatic steatosis in HFD mice is shown in Figure 4. The liver pathological sections in the NC group expressed regular hepatocyte morphology, clearly defined cell borders, and no obvious fat vacuoles (Figure 4A,B). Comparatively, the HFD group had fat vacuoles of different sizes around the blood vessels and showed obvious hepatic damage, as characterized by cell swelling, fatty degeneration, loss of cellular boundaries, cellular contraction, and hepatocyte necrosis (Figure 4C,D). As shown in Figure 4E,F, FP intervention could improve liver steatosis, especially in the FP_L group. Compared with the HFD group, the FP_H (Figure 4G,H) and PC group (Figure 4I,J) had a certain degree of recovery of the liver cell structure, fat degeneration, and fat vacuole reduction. This means that FP intervention can protect liver tissue from HFD-induced hepatic injury.

Metabolomics profiling
According to plasma lipid level and liver histopathological observation, high-dose FP had a better effect on hyperlipidemia than low-dose FP. Therefore, NC, HFD and FP_H groups were chosen to analyze their change in the metabolism level. The total ion chromatograms of QC samples were shown in Supplementary Figure 4. To ensure the accuracy of the data, the features with RSD values >30% were removed from all the QC samples. PCA and ion chromatograms reflected that QC samples were stable during the process (Supplementary Figure 5). A total of 626 and 562 features in the plasma and liver were determined after the data preprocessing. Among them, 120 and 124 metabolites in the plasma and liver, respectively, were identified based on the KEGG database. OPLS-DA analysis was used to explore the separation among the NC, HFD, and FP_H groups. OPLS-DA showed that the same group samples clustered together and different group samples distinguished well (Figure 5A,B). These results indicated that the HFD and FP interventions caused obvious metabolic variations.

To identify the potential differential metabolites that contributed to the metabolic distinction, further OPLS-DA and t-test analyses of NC versus HFD and HFD versus FP_H were performed, respectively. The OPLS-DA results distinguished well, and showed significant differences between different groups of models¹⁴ (Supplementary Figure 6). Based on VIP (Variable important in projection) >1 and p < 0.05, 32 metabolites in the plasma showed differentiation between the NC and HFD group, and 72 metabolites showed differentiation between the HFD and FP_H group. In the liver, 38 metabolites showed differentiation between the NC and HFD group, and 17 metabolites showed differentiation between the HFD and FP_H group. Finally, a total of 16 and 6 metabolites were identified as differential metabolites in FP-affecting HFD mice in the plasma and liver, respectively (Supplementary Figure 7). The information on these metabolites is shown in Table 2.

To visualize the variation in metabolites among the three groups, heat maps were plotted by MetaboAnalyst 5.0. All of the differential metabolites in the plasma and liver were changed in the HFD group and most of them were reversed in the FP group, indicating that FP intervention can improve metabolic disorder (Figure 5C,D). Further, differential metabolites were imported into MetaboAnalyst 5.0 to explore the metabolic pathways of FP in HFD mice. Based on p < 0.05 and a pathway impact >0.10, tryptophan metabolism was affected significantly in the plasma, and the metabolites related to this pathway were D-tryptophan and L-kynurenine (Figure 5E). Jung et al.³² studied that prolonged hyperlipidemia may lower the serum levels of kynurenine. Taurine and hypotaurine metabolism was affected significantly in the liver, and the related metabolite related was taurine (Figure 5F). Taurine is an important and necessary amino acid in the animal body; Dong et al.³³ studied that taurine could mildly reduce the damage of blood lipids and lower the atherosclerosis risk caused by HFD. In this research, FP intervention increased the content of L-kynurenine and taurine, which is positively related to the reduction of lipid levels, supporting the effectiveness of FP against hyperlipidemia.

Integrated analysis of network pharmacology and metabolomics
An integrated strategy of network pharmacology combined with metabolomics has become more and more indispensable in studying disease mechanisms and intervention strategies. The relevance between network pharmacology and metabolomics with limited evidence was established. To obtain a comprehensive view of the mechanism of FP against hyperlipidemia, the interaction networks based on network pharmacology and metabolomics were constructed. Differential metabolites were imported into the MetScape plugin in Cytoscape and matched the hub genes identified in network pharmacology to collect the compound-reaction-enzyme-gene networks (Figure 6). As shown in Table 3, in plasma metabolites, L-kynurenine and corticosterone were related to CYP1A1, which can catalyze lipid peroxidation and induce non-alcoholic fatty liver disease^34,35; the affected pathways were tryptophan metabolism and steroid hormone biosynthesis, respectively. Acetylcholine was related to AChE and affected glycerophospholipid metabolism. In liver metabolites, MGAM and raffinose were related to galactose metabolism. Several studies have demonstrated that the intake of raffinose family oligosaccharides could improve metabolic disorders in HFD mice³⁶.

Further, the ingredients-targets-metabolites-pathways network has been constructed (Figure 7). In the ingredients, quercetin connected the most edges, indicating that quercetin of FP plays the most important role in lowering lipids. The above-integrated analysis revealed the key targets, metabolites, and pathways of FP against hyperlipidemia, which could be the foundation of further study of this medicine's therapeutic mechanism and clinical application.

Molecular docking
To further investigate the possibility of interaction between the selected ingredients and the key targets, molecular docking was used to analyze their ligand-active site interactions. AutoDock Vina software (see Table of Materials) was used to perform molecular docking, and the first docking pose was outputted according to the rank of the scoring function. The docking results are shown in Figure 8.

In the integrated analysis, CYP1A1, AChE, and MGAM were related to differential metabolites; they built bridges between targets and metabolites. Further molecular docking was performed to verify the relation between the target and ingredients. The results of ingredient docking with CYP1A1 were as follows: gallic acid formed four hydrogen bonds through the amino acid residues Asn-185, Tyr-187, Asn-219, and His-500, and formed π-π stacking interaction through the amino acid residue Tyr-187 (Figure 8A); quercetin formed three hydrogen bonds through Asn-185, Asn-219, and His-500, hydrophobic interaction, and π-π stacking interaction through Tyr-187 (Figure 8B); beta-sitosterol formed four hydrogen bonds through Arg-362, Ser-363, Leu-365, and Arg-464, and hydrophobic interaction through Glu-369 and Ile-439 (Figure 8C). The binding energies were 5.3, 7.0, and 7.3 kcal/-mol, respectively. In the interaction with AChE, gallic acid was stabilized by hydrogen bonds with Arg-237, Arg-238, and Arg-480 (Figure 8D); quercetin was stabilized by hydrogen bonds with Arg-237 and Phe-474, by hydrophobic interaction with Phe-157, and by π-π stacking interaction with Tyr-478 (Figure 8E); beta-sitosterol was stabilized by hydrophobic interaction with Phe-157, Val-244, Ile-248, Phe-474, Ala477, and TYR478 (Figure 8F). The binding energies were 5.0, 6.5, and 8.0 kcal/- mol, respectively. In the interaction with MGAM, gallic acid was stabilized by hydrogen bonds with Ile-1716, Gly-1747, and Trp-1749, and by hydrophobic interaction with Tyr-1715 and Trp-1749 (Figure 8G); quercetin was stabilized by hydrogen bonds with Arg-1311, Thr-1726, Gln-1731, and Trp-1752, by hydrogen bonds with Arg-1730, and by π-π stacking with His-1727 (Figure 8H); beta-sitosterol was stabilized by hydrophobic interaction with Pro-1159, Trp-1355, Phe-1427, and Phe-1560, The binding energies were 5.9, 8.1, and 6.9 kcal/mol, respectively. Detailed information on the interactions and binding affinities is exhibited in Table 4. Multiple binding sites and high binding energies explain the high affinities between ingredients and protein targets, verifying that these ingredients play the role of lowering lipids by acting on hyperlipidemia-related targets.

Figure 1: Schematic flowchart of the integrated strategy. Hub ingredients and genes were extracted by network pharmacology (Part 1). Differential metabolites of FP against hyperlipidemia were analyzed by plasma and liver metabolomics (Part 2). Key targets, metabolites, and pathways were identified and linked based on an integrated analysis of Part 1 and Part 2 (Part 3). Please click here to view a larger version of this figure.

Figure 2: Target screening, network construction, and enrichment analysis of the effect of FP against hyperlipidemia. (A) Venn diagram of the FP-hyperlipidemia targets. (B) Potential active drug-ingredients-targets-disease network: different color symbols as mentioned here: disease (red), drug (blue), ingredients (green), and targets (yellow). (C) PPI network by STRING. (D) GO pathway enrichment analysis. (E) KEGG pathway enrichment analysis. Please click here to view a larger version of this figure.

Figure 3: The effect of FP on plasma lipid levels and liver index in mice with HFD-induced hyperlipidemia (n = 6). (A) Levels of TC. (B) Levels of LDL-C. (C) HDL-C level. (D) TG level. (E) The LDL-C/HDL-C ratio. (F) Liver indexes.*p < 0.05, **p < 0.01, ***p < 0.001. Statistically significant differences were evaluated using a one-way ANOVA followed by Dunnett's multiple comparisons test or post hoc analysis. Please click here to view a larger version of this figure.

Figure 4: The effect of FP on liver tissue in mice with HFD-induced hyperlipidemia (H&E staining). (A,B) NC group, (C,D) HFD group, (E,F) FP_L group, (G,H) FP_H group, (I,J) PC group (n = 6). Scale bar: (A,C,E = 200 µm; B,D,F = 50 µm). Please click here to view a larger version of this figure.

Figure 5: The OPLS-DA score plots, heat maps, and metabolic pathways of differential metabolites. The OPLS-DA score plots of FP on HFD mice in the plasma (A) and liver (B). The heat maps of differential metabolites in the plasma (C) and liver (D). The metabolic pathways of differential metabolites in the plasma (E) and liver (F). Please click here to view a larger version of this figure.

Figure 6: The compound-reaction-enzyme-gene networks of the key metabolites and targets. Low-degree nodes have been removed. The red hexagons, blue circles, round green rectangles, and grey diamonds represent the active compounds, genes, proteins, and reactions, respectively. The key targets and metabolites were magnified. The pathways with the white background are significantly regulated in the plasma. The pathway with the grey background is significantly regulated in the liver. Please click here to view a larger version of this figure.

Figure 7: The ingredients-targets-metabolites-pathways network. The darker the color, the more the connected edges, signifying the node is more important in this network. Please click here to view a larger version of this figure.

Figure 8: The interaction diagrams of FP ingredients and the key targets. (A) Gallic acid acting on CYP1A1. (B) Quercetin acting on CYP1A1. (C) Beta-sitosterol acting on CYP1A1. (D) Gallic acid acting on AChE. (E) Quercetin acting on AChE. (F) Beta-sitosteroling act on AChE. (G) Gallic acid acting on MGAM. (H) Quercetin acting on MGAM. (I) Beta-sitosterol acting on MGAM. Please click here to view a larger version of this figure.

Figure 9: Overview of FP against hyperlipidemia result. Please click here to view a larger version of this figure.

Table 1: The selected ingredients of FP aqueous extract. Please click here to download this Table.

Table 2: The differential metabolites between the three groups. Please click here to download this Table.

Table 3: The information on key targets, metabolites, and pathways. Please click here to download this Table.

Table 4: Binding sites and action forces between FP ingredients and target proteins. Please click here to download this Table.

Supplementary Figure 1: The positive and negative ion chromatograms of FP aqueous extract. Please click here to download this File.

Supplementary Figure 2: The FP ingredient-target-pathway-disease network by BATMAN-TCM. Please click here to download this File.

Supplementary Figure 3: The frequency analysis of hub genes in network pharmacology. Please click here to download this File.

Supplementary Figure 4: Ion chromatograms of plasma and liver QC samples. The representative positive (A) and negative (B) ion chromatograms of plasma QC samples. The representative positive (C) and negative (D) ion chromatograms of liver QC samples. Please click here to download this File.

Supplementary Figure 5: The PCA score plots of plasma (A) and liver (B) QC samples. Please click here to download this File.

Supplementary Figure 6: The OPLS-DA score plots of plasma (A and B) and liver (C and D) samples. Please click here to download this File.

Supplementary Figure 7: Venn diagrams of the differential metabolites in plasma (A) and liver (B) samples. Please click here to download this File.

Discussion

In recent years, the incidence rate of hyperlipidemia has been increasing, mainly due to long-term unhealthy eating habits. TCM and its chemical ingredients have various pharmacological activities, which have been widely studied in recent years^37,38. FP is a kind of fruit resource, used both as medicine and food, and has an important potential for treating hyperlipidemia. However, the potential therapeutic mechanism of FP against hyperlipidemia needs further study.

Network pharmacology evaluates drug polypharmacological effects at a molecular level, and predicts the interaction of natural products and proteins to determine the major mechanism³⁹. The first step is to select the active ingredients and key targets of the drug. In this research, nine active ingredients and 62 hub genes were found. To further understand the molecular mechanism of FP on hyperlipidemia, PPI and ingredient-target networks were established based on network pharmacology analysis. To narrow the scope of key ingredients and targets, three key ingredients (gallic acid, quercetin, and beta-sitosterol) related to hypercholesterolemia and coronary atherosclerosis have been founded by BATMAN-TCM. All these ingredients could reduce LDL-C levels or increase HDL-C levels, validating the specific effects of FP on hyperlipidemia. Besides, according to KEGG enrichment analysis, the function of FP on hyperlipidemia is related to the activity of the lipid and atherosclerosis pathway. Although this method depends too much on the database and lacks experimental verification, it has theoretical value and provides ideas for subsequent experimental verification research.

For further experimental validation, mice were fed with a fat-supplemented diet for 8 weeks to induce hyperlipidemia. The results showed that plasma TC, LDL-C, and TG levels were significantly increased. Although the level of HDL-C decreased significantly, the ratio of LDL-C to HDL-C increased significantly. The histopathological observations showed that the liver tissue of HFD mice was severely damaged, but there was no significant increase in liver index; it may be that changes in body weight and visceral weight take longer. The lipids and liver changes adequately showed the intervention effect of FP on hyperlipidemia. However, the internal mechanism of the intervention effect still needs further exploration.

Metabolomics provides a list of potential metabolites and related pathways, which aim to explore the mechanism of metabolic diseases and the action of therapeutic drugs⁴⁰. The result of metabolomics can be affected by the type of sample. Considering the pathogenic characteristics of hyperlipidemia, plasma and liver samples were chosen for metabonomic analysis in this research. According to OPLS-DA results, the NC, HFD, and FP_H groups' metabolites were discriminated well. A total 16 differential metabolites were found in the plasma, and 6 differential metabolites were found in the liver. There were more affected metabolites in the plasma than in the liver, proving that blood is the main place of metabolic disturbance induced by hyperlipidemia. FP intervention can reverse the change of these metabolites under the influence of HFD. Furthermore, these differential metabolites were imported into the KEGG database. The significant metabolic pathways of differential metabolites in the plasma were tryptophan metabolism, and in the liver were taurine and hypotaurine metabolism. In this research, FP intervention increased the content of L-kynurenine of tryptophan metabolism and taurine content of taurine and hypotaurine metabolism, meaning FP could be effective in favorably adjusting metabolic disorders and hyperlipidemia. The metabolomics analysis revealed which metabolites were related to hyperlipidemia or FP intervention, and determined the downstream mechanism of the FP effect.

By combining the result of network pharmacology with metabolomics, three key targets (CYP1A1, AChE, and MGAM) were identified in the compound-reaction-enzyme-gene networks. According to molecular docking analysis, these targets showed high affinities with FP ingredients (gallic acid, quercetin, and beta-sitosterol). Four metabolites (L-kynurenine, corticosterone, acetylcholine, and raffinose) and four related pathways (tryptophan metabolism, steroid hormone biosynthesis, glycerophospholipid metabolism, and galactose metabolism) were identified as the key metabolites and metabolic pathways. Among these, quercetin was associated with the most targets, and tryptophan metabolism appeared in both metabonomics and integrated results. They play the most essential role in the therapeutic effect of FP against hyperlipidemia. Molecular docking result showed that CYP1A1, AChE, and MGAM have high affinities with ingredients. The above results prove that these screened targets are closely related to the therapeutic effect of FP.

In the present research, gallic acid, quercetin, and beta-sitosterol were identified as FP active ingredients toward anti-hyperlipidemia, and tryptophan metabolism is the main metabolic pathway of FP therapy in HFD mice. The overview of the result is shown in Figure 9. This research offered data and theoretical support for further studies of mechanisms and provided a foundation for the clinical application of FP medicine. It also proved that natural food might be a promising option with great prospects in clinical practice. However, there are still some shortcomings in this research. The therapeutic effect of the active ingredient alone on hyperlipidemia has not been verified. In addition, the pathway of key targets has not been studied; it also needs further systematic molecular biology experiments to verify the accurate mechanism.

Materials

Name	Company	Catalog Number	Comments
101-3B Oven	Luyue Instrument and Equipment Factory	\
80312/80302 Glass Slide	Jiangsu Sitai Experimental Equipment Co., LTD	\
80340-1630 Cover Slip	Jiangsu Sitai Experimental Equipment Co., LTD	\
AccucoreTM C18 (3 mm × 100 mm, 2. 6 μm)	Thermo Fisher Scientific	\
Acetonitrile	Fisher Chemical	A998	Version 1.5.6
ACQUITY UPLC HSS T3 Column (2.1 mm × 100 mm, 1.8 μm)	Thermo Fisher Scientific	\
Aethanol	Fisher Chemical	A995	Version 3.0
Ammonia Solution	Chengdu Cologne Chemicals Co., LTD	1336-21-6	Version 3.9.1
AutoDockTools	Scripps Institution of Oceanography	\
BS-240VT Full-automatic Animal Biochemical Detection System	Shenzhen Mindray Bio-Medical Electronics Co., Ltd.	\
Compound Discoverer	Thermo Fisher Scientific	\
Cytoscape	Cytoscape Consortium	\
DM500 Optical Microscope	Leica	\
DV215CD Electronic Balance	Ohaus Corporation ., Ltd	T15A63
Ethyl Alcohol	Chengdu Cologne Chemicals Co., LTD	64-17-5
Formic Acid	Fisher Chemical	A118
HDL-C Assay Kit	Nanjing Jiancheng Bioengineering Institute	A112-1-1
Hematoxylin Staining Solution	Biosharp	BL700B
High Fat Diet	ENSIWEIER	202211091031
Hitachi CT15E/CT15RE Centrifuge	Hitachi., Ltd.	\
Homogenizer	Oulaibo Technology Co., Ltd	\
Hydrochloric Acid	Chengdu Cologne Chemicals Co., LTD	7647-01-0
Image-forming System	LIOO	\
JB-L5 Freezer	Wuhan Junjie Electronics Co., Ltd	\
JB-L5 Tissue Embedder	Wuhan Junjie Electronics Co., Ltd	\
JK-5/6 Microtome	Wuhan Junjie Electronics Co., Ltd	\
JT-12S Hydroextractor	Wuhan Junjie Electronics Co., Ltd	\
KQ3200E Ultrasonic Cleaner	Kun Shan Ultrasonic Instruments Co., Ltd	\
LDL-C Assay Kit	Nanjing Jiancheng Bioengineering Institute	A113-1-1
Male C57BL/6 Mice	SBF Biotechnology Co., Ltd.	\	Version 2.3.2
Neutral Balsam	Shanghai Yiyang Instrument Co., Ltd	10021190865934
Pure Water	Guangzhou Watson's Food & Beverage Co., Ltd	GB19298
PyMOL	DeLano Scientific LLC	\	Version 14.1
RE-3000 Rotary Evaporator	Yarong Biochemical Instrument Factory ., Ltd	\
RM2016 Pathological Microtome	Shanghai Leica Instruments Co., Ltd	\	Version 26.0
SIMCA-P	Umetrics AB	\
Simvastatin	Merck Sharp & Dohme., Ltd	14202220051
SPSS	International Business Machines Corporation	\
TC Assay Kit	Nanjing Jiancheng Bioengineering Institute	A111-1-1
TG Assay Kit	Nanjing Jiancheng Bioengineering Institute	A110-1-1
UPLC-Q-Exactive Quadrupole Electrostatic Field Orbital Hydrazine High Resolution Mass Spectrometry	Thermo Fisher Scientific	\
Vortex Vibrator	Beijing PowerStar Technology Co., Ltd.	LC-Vortex-P1
Xylene	Chengdu Cologne Chemicals Co., LTD	1330-20-7