2.1 Chemicals and reagents

Formononetin reference (Batch number: MUST-21033005, purity ≥ 99.03 %) and Ononin reference (Batch number: MUST-21080311, purity ≥ 98.48 %) were provided by Chengdu Must Biotechnology Co., Ltd. (Sichuan, China). HepG2 cells was obtained from Binzhou medical university (Yantai, China). Kits measuring the levels of bicinchoninic acid (BCA), total cholesterol (TC), total triglycerides (TG), alanine transaminase (ALT), aspartate transaminase (AST), superoxide dismutase (SOD), and glutathione peroxidase (GSH) were purchased from Jiancheng Institute of Biotechnology (Nanjing, China). Palmitic acid / oleic acid (PA/OA) were obtained Xi'an Kunchuang Technology Development Co., Ltd. (Xian, China). Oil Red O and Cell Counting Kit-8 (CCK-8) were purchased from Beyotime Biotechnology (Shanghai, China).

HPLC grade methanol, acetonitrile, and formic acid (FA) were purchased from Thermo Fisher Scientific (Fair Lawn, NJ, USA), and pure water for analysis was purchased from Watson Group Co., Ltd. (Jinan, China). All the other chemicals of analytical grade were available at the workstation, Shandong Academy of Chinese Medicine (Jinan, China). Grace Pure^TM SPE C18-Low solid phase extraction cartridges (200 mg·3 mL⁻¹, 59 μm, 70 Å) were purchased from Grace Davison Discovery Science (Deerfield, IL, USA). Male Sprague Dawley (SD) rat liver microsomes (1 mL, Batch number: 20210305), Magnesium chloride (MgCl₂), β-nicotinamine adenine dinucleotide phosphate (NADPH), and uridine 5′-diphosphoglucuronic acid trisodium salt (UDPGA) were all purchased from NEWGAINBIO Co., Ltd. (Wuxi, China).

2.2 In vivo animal experiment

2.3 In vitro liver microsomes incubation

Firstly, incubation mixtures were prepared in PBS buffer and divided into Drug Group and the Control Group. The buffer for the Drug Group contained rat liver microsomes (1 mg·mL⁻¹), formononetin (0.1 mg·mL⁻¹), and MgCl₂ (3 mM, final concentration). The buffer for the Control Group did not contain formononetin. 900 μL of the above incubation mixture, 100 μL of NADPH (25 mg·mL⁻¹) and 100 μL of UDPGA (25 mg·mL⁻¹) were added to each of the 6-well plates to start the reaction. The reaction was incubated at 37 °C for 5, 10, 15, 30, 45, 60, 120, 240, and 480 min, and then 100 μL of incubation solution was removed and the reaction was terminated by adding 200 μL of cold acetonitrile. Finally, the supernatant was purified by acetonitrile precipitation.

2.4 Instruments and analytical conditions

UHPLC analysis was conducted on the Dionex Ultimate 3000 UHPLC system (Thermo Fisher Scientific, MA, USA). Separation was performed on a Waters ACQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm). The column temperature was kept at 30℃, the flow rate was 0.3 mL·min⁻¹. The mobile phase was composed of water containing 0.1 % formic acid (A) and acetonitrile (B). The elution gradient condition was as follows: 0–5 min, 5 %-30 % B; 5–10 min, 30 %-50 % B; 10–27 min, 50 %-90 % B; 27–27.1 min, 90 %-5% B; 27.1–30 min, 5 % B.

HRMS and MS/MS spectra were obtained using Q-Exactive Focus Orbitrap MS (Thermo Fisher, Waltham, MA, USA) equipped with HESI. The ion source parameters were set as follows: Scan range, m/z 80–1,200; Aux gas flow rate, 10 arbitrary units; Aux gas heater temperature, 320 °C; Sheath gas flow rate, 45 arbitrary units; Capillary temperature, 320 °C; Spray voltage, 3.8/3.5 kV (+/-); Scan modes, full MS resolution, 70,000; dd-MS² resolution, 17,500; Collision energy, 15 %, 30 %, 45 %.

2.5 Data processing and peak selections

In order to obtain accurate metabolites of formononetin, the parameters of the Thermo Xcalibur 2.1 workstation were set within a reasonable range. The compositional structural formulae were set to C[0–30], N[0–5], O[0–15], H[0–60], S[0–1]. The mass error range was set to within ± 5 ppm. The ring double bond (RDB) equivalent was set to [0–15].

2.6 Cells and treatments

Logarithmic growth phase HepG2 cells were inoculated into 96-well plates at a density of 5 × 10³ per well for cell viability. Subsequently, HepG2 cells were seeded into 6-well plates at a density of 2 × 10⁵ per well. The cells were stimulated with PA (187.5 μM) / OA (375 μM) and formononetin (5, 10, and 50 μg·mL⁻¹) for 24 h. After fixation with 4 % paraformaldehyde, Oil Red O staining was added for 20 min, and the images were captured using an Echo Laboratories RevolveFL imaging system (San Diego, USA). Finally, HepG2 cells were collected by inoculating 6-well plates in the same way to determine protein content. The levels of TC, TG, AST, ALT, SOD, and GSH were also determined according to the kit instructions.

2.7 Network pharmacology

2.8 Statistical analysis

Cell experiments were repeated at least three times. GraphPad Prism 8.0 was used to analyze the results of biochemical indices and related graphs. Two-tailed Student’s t-test was used for statistical analysis. P < 0.05 was considered significant.

3.1 The establishment of a recursive tree-based analytical strategy

We focused on a “recursive tree” analysis strategy to systematically screen and characterize the metabolites of drugs, using formononetin as an example (Fig. 1). Firstly, we defined the concept of “recursive tree”. The drug is compared to a tree trunk, the metabolites resulting from the addition or subtraction of a single functional group are the primary branches (main branches), the metabolites resulting from the addition or subtraction of a single functional group from the primary branches are the secondary branches (lateral branches), and so on. Among the addition and subtraction of a single functional group include methylation (demethylation), hydroxylation (dehydroxylation), hydrogenation (dehydrogenation), methoxylation (demethoxylation), decarbonylation, decarbonization, etc. During the recursion process, two noteworthy points aroused our interest. One was the emergence of duplicate metabolites as the recursion progresses, and it was essential to remove duplicate metabolites when summarizing. Secondly, after the screening of the tertiary branching metabolites, fewer and fewer metabolites could be identified, so we mainly performed the identification of the primary, secondary, and tertiary branching metabolites.

Close

Fig. 1

The summary diagram of recursive tree analysis strategy.

Export to PPT

Not to be ignored, the diagnostic product ions (DPIs) played a key role in the “recursive tree” analysis strategy. The structural backbone of the prototype drug is largely preserved during metabolism, thus allowing rapid and precise structural identification of the metabolite using DPIs based on the mass fragmentation behavior of atomoxetine. In addition, the seamless integration of primary branching metabolites, which are used as effective forms of drugs due to their higher blood concentrations and well-defined structural formulas, with network pharmacology is key to the application of “recursive tree” analysis strategies.

3.2 Dpis construction based on formononetin mass cleavage pattern

Formononetin belongs to the class of isoflavones, and the DPIs are mainly produced by cleavage at different positions on the C-ring. It includes Retro-Diels-Alder (RDA) cleavage, deA-ring cleavage, and deB-ring cleavage, often accompanied by neutral loss of CO₂, CO, and H₂O. The prototype with the chemical formula C₁₆H₁₂O₅ was rapidly identified based on the retention time of formononetin standard at 9.65 min and the ions at m/z 267.06 ([M−H]^-) and m/z 269.08 ([M + H]⁺). In order to more conveniently analyze the metabolites of formononetin, we list its six cleavage pathways (Fig. 2). In the negative ion mode, four typical ions with m/z 251, m/z 239, m/z 223, and m/z 209 were obtained after the sequential loss of O, CO, CO₂, and CH₂, respectively. Three characteristic ions with m/z 252, 236, and 208 were generated after the successive loss of CH₃, OCH₃, and CO, respectively. Moreover, six product ions were observed at m/z 107, 159, 146, 120, 135, and 132 through the cleavage pathway (a)\(b)\(c). Among them, m/z 135 and m/z 132 were typical RDA cleavage ions. According to the cleavage pathways (d)\(e)\(f), the ions were produced at m/z 119, 147, 91, 175, 159, and 107. Among them, m/z 91 and m/z 175 were the key ions of flavonoid cleavage mode II, and the product ion with high abundance at m/z 161 was produced by demethylation of m/z 175. Thus, the DPIs of formononetin were summarized as m/z 91, 107, 119, 120, 132, 135, 146, 147, 159, 175, 208, 209, 223, 236, 239, 251, 252 in negative ion mode. Similarly, the DPIs in positive ion mode were m/z 93, 109, 121, 122, 134, 137, 148, 149, 161, 177, 210, 211, 225, 238, 241, 253, 254.

Close

Fig. 2

ESI-MS spectrum (A) and typical cleavage pathways (B) of formononetin in negative ion mode.

Export to PPT

3.3 Dpis construction of primary, secondary, tertiary, and multi-level branching metabolites

Based on the “recursive tree” analysis strategy and the general rules of drug metabolism, formononetin was susceptible to methylation (demethylation), oxidation (deoxidation), glucuronidation, sulfation, and their composite reactions. With the addition or subtraction of a single functional group in the parent nucleus as the core of the analysis strategy, we classified these metabolites into primary, secondary, tertiary, and multi-level branching metabolites, which resulted in the generation of regular DPIs. For example, daidzein, a demethylated metabolite of formononetin (primary branching metabolite), might yield negative ions at m/z 253 (m/z 267 - CH₂), m/z 237 (m/z 251 - CH₂), m/z 215 (m/z 239 - CH₂), m/z 209 (m/z 223 - CH₂), m/z 145 (m/z 159 - CH₂), m/z 118 (m/z 132 - CH₂), m/z 161 (m/z 175 - CH₂), and m/z 93 (m/z 107 - CH₂). Another one, glucuronidation metabolites of daidzein (secondary branching metabolite) might produce negative ions at m/z 429 (m/z 253 + GluA), m/z 413 (m/z 237 + GluA), m/z 391 (m/z 215 + GluA), m/z 385 (m/z 209 + GluA), m/z 321 (m/z 145 + GluA), m/z 294 (m/z 118 + GluA), m/z 337 (m/z 161 + GluA), and m/z 269 (m/z 93 + GluA). By analogy, the DPIs ± nX [n indicated the presence of multiple functional groups, X indicated molecular weight of substituents, e.g. 14 (CH₂), 16 (O), 18 (H₂O), 30 (OCH₂), 80 (SO₃), 176 (GluA), etc.] could be used to identify formononetin metabolites if multi-level branching reactions occurred.

3.4 Identification of formononetin metabolites

This study identified 131 metabolites of formononetin, with 53, 27, 46, 9, and 31 metabolites detected in plasma, urine, faeces, liver, and liver microsomes samples, respectively. Based on the “recursive tree” analysis strategy, we divided the metabolites into four groups, including primary, secondary, tertiary, and multi-level branching metabolites group. Table 1 provided detailed information on the formononetin metabolites. Fig. 3 illustrated the structure of the metabolites. Representative metabolites of formononetin were listed and the identification analysis of all metabolites was detailed in the Supplementary Material.

Close

Fig. 3

Structural identification and proposed metabolic pathway of formononetin metabolites. A: primary branching metabolites of formononetin (A1-A20); B: secondary branching metabolites of formononetin (B1-B42); C: tertiary branching metabolites of formononetin (C1-C31); D: Multi-level branching metabolites of formononetin (D1-D37).

Export to PPT

3.5 Effects of formononetin on HepG2 cells induced by PA/OA

Literatures reported that PA/OA stimulation of HepG2 cells can mimic fatty hepatocytes for NAFLD studies. We successfully established a NAFLD cell model by stimulating HepG2 cells with PA (187.5 μM) / OA (375 μM). The effect of formononetin on HepG2 cells viability was assessed by CCK-8, and the results showed that cells viability was significantly reduced at formononetin concentrations greater than 100 μg·mL⁻¹ (Fig. 4A). To ensure the anti-NAFLD efficacy of formononetin, the concentrations of 5, 10, and 50 μg·mL⁻¹ were selected for subsequent experiments.

Close

Fig. 4

Effects of formononetin on NAFLD cell models. (A) Results of cell viability; (B) Results of Oil Red O staining (×200); (C–H) Results of TG, TC, AST, ALT, SOD, and GSH levels. ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 vs the Con group; *P < 0.05, ^**P < 0.01, ^***P < 0.001 vs the Mod group.

Export to PPT

After 24 h of PA/OA stimulation, HepG2 cells showed severe lipid accumulation. However, the number of lipid droplets in the cells was significantly reduced after formononetin treatment (Fig. 4B), and the levels of TC and TG decreased, indicating that formononetin has a lipid-lowering effect (Fig. 4C, D). At the same time, the levels of AST and ALT were also significantly reduced (Fig. 4E, F), indicating that formononetin has hepatoprotective effects. Moreover, lipid accumulation easily leads to oxidative stress. Formononetin dose-dependently reversed the PA/OA-induced increase in SOD levels and decrease in GSH levels (Fig. 4G, H). Our data preliminarily demonstrated the ameliorative effect of formononetin in NAFLD.

3.6 Network pharmacology analysis

Primary branching metabolites play an important role in drug metabolism because they retain similar parent nucleus structures (Li et al., 2009). Moreover, in our study, the structures of the primary branching metabolites of formononetin were completely characterized. Therefore, we selected formononetin and its primary branching metabolites for a network pharmacological analysis to explore the mechanism of anti-NAFLD.

3.6.1

3.6.1 Relevant targets of formononetin and primary branching metabolites against NAFLD

Fig. 5A illustrates the metabolic process of the primary branching metabolites of formononetin. The targets of 21 (including formononetin) metabolites were gathered from the Swiss Target Prediction database, and a total of 566 targets were obtained after removing duplicate targets. Additionally, using the GeneCards database, 1409 targets associated with NAFLD were acquired. Through Venny 2.1 software, the metabolites and NAFLD targets were overlapped, and 104 common target genes were determined (Fig. 5B).

Close

Fig. 5

(A) Metabolic process of the primary branching metabolites of formononetin; (B) Common targets of NAFLD and 21 metabolite targets (including formononetin); (C) PPI network diagram; (D) The 20 targets with the highest degree valu.

Export to PPT

3.6.3

3.6.3 GO and KEGG enrichment analysis

GO functional annotation and KEGG pathway enrichment were performed on 104 targets using the Metascape database. The top 10 enriched BP (biological process), MF (molecular function), and CC (cellular compound) were graphically analyzed (Fig. 6A). BP was mainly involved in response to the hormone, positive regulation of phosphorylation, protein phosphorylation, and cellular response to the organonitrogen compound. MF was mainly involved in protein kinase activity, phosphotransferase activity, alcohol group as acceptor, kinase activity, and protein serine/threonine kinase activity. CC was mainly involved in the lytic vacuole, lysosome, vesicle lumen, and receptor complex. KEGG enrichment analysis showed that many targets in pathways in cancer, lipid and atherosclerosis, and MAPK signaling pathways were enriched. Fig. 6B provided the top 20 important signaling pathways.

Close

Fig. 6

(A) Circle and dot bubble plots for GO analysis; (B) Sankey dot and category summary plots for KEGG.

Export to PPT

3.6.4

3.6.4 Building a “disease-pathways-targets-metabolites” network

The “Disease-Pathway-Target-Metabolite” network diagram constructed by Cytoscape 3.9.1 software includes 20 pathways, 104 targets, 21 metabolites, and NAFLD. In the diagram, 20 primary branching metabolites were surrounded by formononetin as the center. With NAFLD as the center, 20 related pathways were encircled. Among them, NAFLD, formononetin, pathways, metabolites, and targets were represented by yellow triangles, green triangles, purple circles, pink circles, and baby blue quadrilaterals, respectively. Both metabolites and pathways were connected to the targets with lines (Fig. 7).

Close

Fig. 7

Disease-Pathways-Targets-Metabolites Network Diagram. Note: yellow triangles: NAFLD; Green triangles: formononetin; Purple circles: pathways; Pink circles: primary branching metabolites of formononetin; Baby blue quadrilaterals: targets.

Export to PPT

4.1 Feasibility analysis of recursive tree strategy establishment

Metabolite identification is a crucial step in drug metabolism research, as well as the material basis for pharmacology and pharmacodynamics research. Drug metabolites have the characteristics of complex background matrix, diverse species, and low content, which require strong detection capability of their analytical instruments. LC-HRMS has become a powerful technical tool for the analysis of drug metabolite due to its high sensitivity, high accuracy, and high separation capacity (Cui et al., 2022). By combining various data processing methods, it can be used for the characterization of metabolites, as well as the detection and structural identification of unknown trace metabolites. However, the spectral information included in the existing chemical standards and databases is very limited, and the complex and diverse metabolite structure identification cannot be satisfied by LC-HRMS technology alone. Therefore, more new techniques and research strategies are needed to meet this challenge.

In the face of this bottleneck, we have proposed the first “recursive tree” analysis strategy. Based on the addition or subtraction of a single functional group, the metabolites were classified into primary, secondary, tertiary, and multi-level branching metabolites. Taking formononetin as an example, formononetin metabolites were identified and analyzed using the recursive tree analysis strategy in vivo and in vitro. A total of 131 metabolites were identified, of which 106 were detected in vivo and 31 were detected in vitro. From the identification results, the implementation of the recursive tree analysis strategy could expand the scope of metabolite search and obtain as many metabolites as possible. In addition, the use of the recursive tree strategy and the analysis of parameters in the mass spectrometry profiles ensure the accuracy of metabolite identification.

Most importantly, the recursive tree strategy is simple, accurate, and convenient, and can be widely applied to the study of drug metabolites. The primary branching metabolites identified based on this strategy are regarded as the material basis for the treatment of diseases due to their high blood concentrations and well-defined molecular structures. In the present study, the recursive tree strategy was combined with network pharmacology, which not only comprehensively identified the metabolites of formononetin, but also improved the study of the mechanism of action of formononetin against NAFLD. The innovative initiative is of great significance as it lays the foundation for the identification of drug metabolites and their pharmacological activity studies.

4.2 Comparison of formononetin metabolites in vivo and in vitro

Biotransformation is a featured segment of drug metabolism research. Most of the metabolism studies in the crowd are mainly based on rats. Of interest is the rapidity and simplicity of microsome metabolism assays in vitro. The selective interaction between enzymes and substrates can be directly observed, reducing the interference of many internal factors (Knights et al., 2016). Therefore, microsomes can serve as a reliable source for in vivo and in vitro overall research. Among them, liver microsomes are the most effective means because they contain a variety of hepatic drug enzymes (Chen et al., 2023). At present, there have been no reports on the in vivo and in vitro metabolism of formononetin. Thus, a recursive tree analysis strategy and LC-HRMS were used to compare the metabolism of formononetin in rats and liver microsomes with a view to retrieve the metabolites of formononetin more comprehensively.

In the in vivo study, 53, 27, 46, and 9 metabolites were detected in rat plasma, urine, faeces, and liver, respectively. Although there were fewer metabolites in the liver, it might be the result of rapid metabolism by multiple enzymes. In addition, plasma, urine, and faeces have higher metabolic activity, and most of the metabolites were excreted through urine and faeces. Genistein (B9) was detected in plasma and urine as one of the important metabolites of formononetin. It has various pharmacological properties, including tyrosine and topoisomerase inhibition (Cheng et al., 2023; Jaiswal et al., 2019). In addition, the demethylated metabolite Daidzein (A2) was detected in plasma, urine, and faeces. The pharmacological effects of Daidzein, a typical isoflavone, have been extensively studied, including anti-inflammatory, hypolipidemic, antitumor, and neuroprotective effects (Pyo et al., 2009; Zhu et al., 2021; Grisley et al., 2022). It was worth mentioning that only 9 metabolites, mostly tertiary branching metabolites, were detected in the liver, even though it was the most important metabolic organ in the body. As a consequence, preliminary speculation suggested that formononetin was rapidly metabolized by hepatic drug enzymes, prone to multi-level reactions, and excreted from the body by urine and faeces via internal circulation.

In vitro study, liver microsomes also exhibited strong metabolism, and although fewer formononetin metabolites were found in liver microsomes than in rats, most phase II metabolites were shown. More importantly, Ononin (A1) was found only in liver microsomes. It indicated that the metabolic conversion of ononin to other metabolites occurred rapidly in the body. Currently, there were more studies on ononin, which was widely used in clinical practice and has clear pharmacological effects such as anti-inflammatory (Yu et al., 2023), anti-cancer (Ye et al., 2022), and anti-angiogenic (Fedoreyev et al., 2008). Overall, the metabolism of formononetin was much more complex in vivo than in vitro, but metabolism in vitro enriches the lack of biotransformation in vivo and provides a comprehensive database for the study of formononetin metabolism.

4.3 Comparison of the different biological treatment methods

The aim of sample preparation is to eliminate the interference of impurities with an appropriate recovery rate. Plasma samples are the most commonly used samples in metabolic assays and usually provide a good response to the overall level of the organism. It contains inorganic salts, oxygen, hormones, enzymes, antibodies, and cellular metabolites (Frampton et al., 2023). Due to the large number of proteins in plasma, precipitation is required prior to mass spectrometry analysis. Currently, researchers often use methanol and acetonitrile as protein precipitation reagents. The mechanism is to use organic solvents to disrupt the protein structure in biological samples, which reduces the solubility of proteins in aqueous solutions and thus converts them into solids for separation. Moreover, SPE is a sample preparation technique developed from chromatographic columns for extraction, separation, and concentration, using the difference in the interaction of different substances in the solid–liquid phases to achieve separation (Jordan et al., 2009). Hence, the plasma samples were first adsorbed onto the stationary phase and then eluted step by step using solvents with different elution capacities to achieve separation, purification, and enrichment.

The number and type of metabolites obtained by the three preparation methods used for the plasma samples were different. 49, 39, and 34 metabolites were obtained by the SPE method, methanol precipitation method, and acetonitrile precipitation method, respectively. Obviously, the SPE method yielded the highest number of metabolites and might be the best method for metabolite acquisition. Surprisingly, a large number of identical metabolites were obtained in both methanol and acetonitrile treatments. This also suggests that methanol can be chosen for protein precipitation in metabolite studies as it is cheap and easily available. The results in Table 1 showed that the samples prepared by SPE were mostly secondary and tertiary branching metabolites, specifically demethylated, sulfated, and glucuronidated metabolites. In contrast, arginine conjugation metabolites were predominantly captured in methanol and acetonitrile precipitations. (See Table 1a-1d.).

The differences in the metabolites obtained by the three methods were due to a variety of reasons, including the influence of the matrix, the strength of the purification ability, and the divergence in separation options. A series of factors resulted in different intensity and quantity signal peaks of formononetin metabolites after UPLC-HRMS detection.

4.4 Potential metabolic pathways of formononetin

The elucidation of the drug metabolic pathways can help the in-depth study of the pharmacological mechanisms. In the present study, the metabolites of formononetin were identified for the first time based on the recursive tree analysis strategy. There were 20, 42, 32, and 37 metabolites with primary, secondary, tertiary and multistage branches, respectively. The analysis revealed that there was a universal phase I and phase II metabolic process for formononetin, and the metabolic pathways included demethylation, decarbonylation, methylation, hydroxylation, sulfation, glucuronidation, arginine conjugation, glutathione conjugation, and their complex reactions, etc. The metabolic profile of formononetin was shown in Fig. 3. Significantly, the most metabolites were identified in plasma, followed by faeces and urine, while the fewest metabolites were identified in the liver. We speculated that after oral administration, formononetin might be biotransformed with the participation of gastric acid, digestive enzymes, and intestinal flora. After absorption into the blood, it is rapidly metabolized by the kidneys and liver and then excreted in the faeces and urine.

Understanding the metabolic properties of drugs is of great significance for drug development and clinical treatment. Formononetin is a typical isoflavone component with low polarity. Therefore, it mostly exists in organisms as hydroxylation products and conjugation products such as sulfation and glucuronidation, increasing the polarity to facilitate absorption and excretion. For example, Ononin (A1), Daidzein (A2), Formononetin-7-O-sulfate (A3), Formononetin-7-O-GluA (A4), Calycosin (A8 or A9), Biochanin A (A12), Dihydroformononetin (A14), Formononetin-7-O-Arg (A15), Genistein (B9), etc.

4.5 Formononetin attenuated PA/OA-induced lipid accumulation and oxidative stress

NAFLD is a disease caused by excessive fat deposition in the liver in addition to alcohol and other liver damage factors, and is recognized as the most common metabolic liver disease worldwide (Nian et al., 2023). NAFLD can progress from simple steatosis at the beginning to hepatitis and finally to cirrhosis and liver cancer, with a prevalence of about 25 % in adults (Jiang et al., 2023). Studies have proven that NAFLD has a long course and complex pathogenesis, which is still unclear. During the development of NAFLD, lipid metabolism disorders cause lipid accumulation. This process results in a high accumulation of ROS inducing oxidative stress in the liver. In turn, the mitochondrial dysfunction induced by oxidative stress may cause oxidative damage, which triggers a series of cascade signals leading to inflammation, cell death, and fibrosis (Loh et al., 2023). Therefore, inhibition of the inflammatory response and antioxidant may be effective in the treatment of NAFLD. At present, only some hepatoprotective and lipid-lowering drugs can be used clinically, and there is no specific drug on the market. The search for safe and effective anti-NAFLD drugs is imminent.

Our in vitro experiments preliminarily demonstrated that TG and TC levels were significantly elevated and lipid droplet aggregation was obvious in HepG2 cells induced by PA/OA. Formononetin treatment significantly reduced lipid accumulation. Oxidative stress is considered to be a key pathogenic link in the development of NAFLD (Fang et al., 2023). Dysregulation of SOD and GSH predisposes to mitochondrial dysfunction and the production of pro-inflammatory factors such as IL-6 and TNF-α (Meca et al., 2021). However, with the accumulation of lipids in the liver, the increased metabolism of free fatty acids and lipopolysaccharides leads to increased ROS production, accelerating oxidative stress and inflammatory responses in the liver (Gao et al., 2022). As stated, AST, ALT, and GSH levels were significantly elevated in PA/OA-induced HepG2 cells, whereas SOD levels were significantly decreased. However, the above indicators were reversed after treatment with formononetin. Our data suggest that formononetin ameliorated PA/OA-induced NAFLD.

4.6 Prediction of the mechanism of formononetin against NAFLD

To investigate the possible mechanisms of formononetin intervening in NAFLD, we evaluated the mechanism of action of formononetin and its primary branching metabolites against NAFLD using network pharmacology. As a result, it was indicated that AKT1, ALB, TNF, STAT3, and EGFR were more important core targets, which played pharmacological roles in NAFLD. Among them, AKT1 is involved in insulin resistance and FoxO signaling pathways, and its high level of expression contributed to the severity in patients with NAFLD (Abolfazli et al., 2022). Members of the tumor necrosis factor protein family are among the most typical triggers of hepatocyte death, and overexpression of TNF-α triggers pro-apoptotic signaling in hepatic adipocytes (Shirakami et al., 2023). Furthermore, STAT3 is a promising therapeutic target. Activation of phosphorylated STAT3 is one of the predisposing factors for the development of NAFLD to hepatocellular carcinoma (Krizanac et al., 2023). In addition, EGFR is a cell membrane protein that can be activated by its ligand, epidermal growth factor. It has been shown that epidermal growth factor receptor deletion in mice not only reduced the accumulation of macrophages in the liver, but also reduced the release of inflammatory factors (Li et al., 2023). The amelioration of NAFLD was verified by EGFR knockout mice (Shao et al., 2023). Thus, reducing the expression of EGFR may be an effective breakthrough point for the treatment of NAFLD.

To further explore the mechanism of action of formononetin and its metabolites against NAFLD and provide a scientific basis for related experimental studies, the researchers performed GO and KEGG pathway enrichment analysis. GO analysis enriched these targets in the categories such as “response to hormone”, “positive regulation of phosphorylation”, “protein kinase activity”, and “lytic vacuole”. KEGG analysis pooled targets across multiple pathways, and signaling pathways with high gene counts included: pathways in cancer, lipid and atherosclerosis, MAPK signaling pathway, PI3K-Akt signaling pathway, and prostate cancer, etc. Among these, regulation of the PI3K-AKT pathway ameliorates hepatic oxidative stress and further inhibits NF-κB mediated inflammatory responses in NAFLD. Fang et al. (2023) found that the MAPK signaling pathway has a regulatory role in oxidative stress, mitochondrial dysfunction, and hepatic steatosis. Inhibiting the activation of the MAPK signaling pathway, abnormal metabolism, and oxidative stress could be alleviated to improve NAFLD. Lipid autophagy was considered a novel mechanism to regulate cellular lipid metabolism and lipid deposition (Liu et al., 2023). Activation of the AMPK/mTOR signaling pathway enhanced hepatic macrophage autophagy, which could further inhibit IL-17 mediated inflammatory response and thus significantly improved NAFLD-related liver injury (Meng et al., 2021). In addition, TNF-α has been found to play an important role in lipid accumulation. It first promotes lipid synthesis in the liver and activates adipokines leading to obesity. At the same time, it regulates the activity of lipid metabolizing enzymes and inhibits the uptake of free fatty acids leading to lipid accumulation in the liver (Peng et al., 2023). Therefore, the TNF signaling pathway has been concluded to be a predisposing pathway in the development of NAFLD and contributes to the advancement of NAFLD to NASH.

The results of the study revealed the potential mechanism of formononetin against NAFLD from the perspective of preliminary in vitro experiments and network pharmacology, but bioinformatics has limitations. We will next validate the network pharmacology results through pharmacological experiments and molecular biology techniques.