A. macrocephala Koidz rhizomes constituted the raw material of the collected AMR samples (RAMRs). All fresh RAMRs of A. macrocephala Koidz were picked in December 2019 (No. 20191215) in Shiliang Town, Tiantai County, Zhejiang Province, China (E121° 9′15″, N29° 13′26″). Dried immature young fruit of Citrus aurantium L. was the raw material of the collected AFI samples (RAFIs, No. 20190621), which were purchased from Jiangxi Gu Han Refined Chinese Medicine Decoction Pieces Co., Ltd. (Nanchang, China). According to our previous research and processing technology optimization (Wang et al., 2018), we generated Rhizoma Atractylodis macrocephalae processed by Aurantii Fructus Immaturus decoction (PAMR, No. 20191225). In brief, 500 mL of AFI decoction (obtained by adding 100 g of AFI to 1430 mL of pure water, extracting at 100°C for 90 min, filtering, and adding to 1200 mL of pure water for extraction at 100°C for 90 min; the extracts were combined and concentrated to 500 mL with occasional turning) was added to 1000 g of AMR. After soaking for 24 h, the sample was stir-fried at 110°C for 6 min and then dried at 40°C for 24 h. All materials were identified by Professor Lingyun Zhong of Jiangxi University of Chinese Medicine and stored in the Department of TCM Processing, Jiangxi University of Chinese Medicine.

Naringenin (lot no. CHB180916), hesperidin (lot no. CHB180523), diosmin (lot no. CHB180125), hesperetin (lot no. CHB180524), tangeretin (lot no. CHB190125), neohesperidin (lot no. CHB180316), 4′,5,7-trimethoxyflavone (lot no. CHB181127), 5,7,3′-trihydroxy-6,4′,5′-trimethoxyflavone (lot no. CHB190311), 5-hydroxy-3′,4′,6,7,8-pentamethoxyflavone (lot no. CHB180321), narirutin (lot no. CHB180917), neoeriocitrin (lot no. CHB180311), naringenin-7-O-glucoside (lot no. CHB181105), naringin (lot no. CHB180914), 3,3′,4′,5,6,7,8-heptamethoxyflavone (lot no. CHB180120), limonin (lot no. CHB180124), eriocitrin (lot no. CHB190219), poncirin (lot no. CHB180625), and vicenin II (lot no. CHB180322) were purchased from Chengdu Chroma-Biotechnology Co., Ltd. (Chengdu, China). Synephrine (lot no. MUST19101716), atractylenolide Ⅰ (lot no. MUST19030221), atractylenolide II (lot no. MUST19101113), atractylenolide III (lot no. MUST19101114) and atractylone (lot no. MUST20070305) were purchased from Chengdu Manst Biotechnology Co., Ltd. (Chengdu, China). The purities of all chemical standards were ≥98%. Water for HPLC and UPLC–Q–TOF–MS/MS analyses was purified using a Milli-Q system (Millipore, Bedford, MA, United States). Iodoacetamide (IAA) was purchased from Sigma‒Aldrich Chemicals (St. Louis, Missouri, United States). Domperidone was obtained from Xi’an Janssen Pharmaceutical Ltd. (Xi’an, China). All other chemicals were of analytical or chromatographic grade.

To determine the contents of different standard metabolites in RAMR, RAFI, and PAMR, the samples were crushed and passed through a 200-mesh sieve. In a 50-mL centrifuge tube, 1.00 g of each sample powder was accurately weighed, and 10 mL of methanol solution was added. Each mixture was then extracted for 10 min at 25°C with ultrasonication and sequentially centrifuged for 5 min at 8000 r·min⁻¹ at 25°C. The extraction procedure was repeated once. The supernatants were combined, concentrated, and diluted to 10 mL with methanol. The final extraction solution was filtered through a 0.22-μm nylon 66 microfiltration membrane before HPLC analysis. Six replicates of each sample were prepared in parallel.

To explore the metabolites in the different samples, 1.00 g of sample powder was extracted for 30 min using 25 mL of methanol with ultrasonication. After the samples were cooled to room temperature (25°C), methanol was added to compensate for the weight loss before centrifugation for 10 min at 8000 r·min⁻¹. Subsequently, 1 mL of the supernatant was collected, diluted to 10 mL with methanol and filtered through a 0.22-μm nylon 66 microfiltration membrane. Ten microliters of each sample was placed in a vial to generate a mixed quality control (QC) sample to validate the reproducibility and stability of the UPLC–Q–TOF–MS/MS method.

UPLC analysis was conducted using an Agilent 1290 LC system (Agilent Technologies, Palo Alto, CA, United States) equipped with an automatic degasser, an infinity binary pump, and an autosampler. A Waters ACQUITY UPLC CSHTM C18 column (2.1 × 100 mm, 1.7 μm; Waters Corp.) was utilized for chromatographic separation. The mobile phases consisted of 0.1% formic acid/water (A) and acetonitrile (B). The mobile phase gradient was as follows: 0–1 min, 5% B; 1–5 min, 5%–10% B; 5–10 min, 10%–18% B; 10–22 min, 18%–24% B; 22–30 min, 24%–45% B; 30–41 min, 45%–95% B; 41–41.5 min, 95%–5% B; and 41.5–45 min, 5% B. The flow rate was set to 0.30 mL·min⁻¹, the injection volume was 2 μL, the column oven was set to 40°C, and the automatic injector temperature was set to 4°C.

MS/MS detection was conducted using a time-of-flight mass spectrometer (Triple TOF^TM 5600+ system) equipped with a Duo Spray source for detecting ions in both the positive and negative modes with high resolution (AB SCIEX, Foster City, CA, United States). In the positive mode, electrospray ionization was applied with the following parameters: ion spray voltage, 5500 V; ion source temperature, 500°C; curtain gas, 40 psi; nebulizer gas (GS 1), 50 psi; heater gas (GS 2), 50 psi; and declustering potential (DP), 100 V. In the information-dependent acquisition (IDA) experiment, the collision energy (CE) was set to 35 eV, and the collision energy spread (CES) was ±10 eV. In the negative mode, electrospray ionization was applied with the following parameters: ion spray voltage, −4500 V; ion source temperature, 550°C; curtain gas, 40 psi; GS 1, 50 psi; GS 2, 50 psi; and DP: −100 V. In the IDA experiment, the CE was set to −35 eV, and the CES was ±10 eV. In both the positive and negative ion modes, the mass ranges were m/z 50–1250 Da for the TOF−MS and MS/MS scans. MS/MS fragmentation was selected for the eight ions with the most intense signals from each TOF−MS scan. Dynamic background subtraction (DBS) was applied to match the IDA tests for UPLC−Q−TOF−MS/MS. All gases were high-purity nitrogen.

To enhance the analysis for precise metabolite identification, UPLC-Q-TOF-MS/MS analysis was performed using 23 standard metabolites. The same LC−MS-related parameters were used to ensure consistent ion cleavage.

To ensure the reliability of the method used, verification was conducted by analyzing QC samples prior to sample analysis. Twenty-three chromatographic peaks were randomly selected from the chromatograms of the QC samples, and the relative standard deviations (RSDs) of the peak area and retention time (RT) were calculated to evaluate the precision, repeatability, and sample stability of the system. Each QC sample was tested 6 times to determine the precision of the system. The same QC sample was prepared six times in parallel, and the repeatability of the method was evaluated after all injections were complete. In addition, the stability of the samples was investigated by injecting and analyzing the QC samples at 0, 6, 12, 18, and 24 h.

The samples were quantitatively analyzed using a Waters 2695 HPLC system equipped with a 2996 PDA detector (Waters, Milford, MA, United States). For the separation of synephrine, a Titank C18 column (250 × 4.6 mm, 5 μm; Waters, Milford, MA, United States) was used. The HPLC settings for synephrine were as follows: injection volume, 10 μL; column oven, 30°C; flow rate, 1.0 mL·min⁻¹; and detection wavelength, 275 nm. Mobile phases A and B were methanol and potassium phosphate dibasic aqueous solutions (0.6 g of potassium phosphate dibasic, 1.0 g of sodium laurylsulfonate, and 1.0 mL of acetate dissolved in water and diluted to 1000 mL), respectively, and elution was conducted isocratically (50%–50%) for 20 min. For analysis, synephrine was precisely weighed and dissolved in methanol. The synephrine standard solution was then diluted to different concentrations (1000, 500, 250, 100, 50, and 25 μg·mL⁻¹) before HPLC analysis. The HPLC settings for the other 7 metabolites (atractylenolide Ⅰ, atractylenolide II, atractylenolide III, atractylone, neohesperidin, hesperidin and naringin) were as follows: injection volume, 10 μL; column oven temperature, 30°C; flow rate, 1.0 mL·min⁻¹; and detection wavelengths, 275 nm (atractylenolide II, atractylenolide III and atractylone), 275 nm (atractylenolide Ⅰ), and 283 nm (neohesperidin, hesperidin and naringin). Mobile phases A and B were methanol and water, respectively. Gradient elution was performed using the following program: 0–50 min, 5% A-50% A; 50–85 min, 50% A-80% A; and 85–120 min, 80% A-95% A. For analysis, these 7 metabolites were precisely weighed and dissolved in methanol. The standard solutions were diluted to different concentrations (100, 50, 25, 10, 5, and 2.5 μg·mL⁻¹) before HPLC analysis.

Using UPLC-Q-TOF-MS/MS analysis, a database of PAMR metabolites considered potentially active metabolite candidates was established. Subsequently, the activity of these metabolites was screened. Based on previous experience from relevant studies (Yang et al., 2020), two classic ADME parameters, namely, oral bioavailability (OB) and drug likeness (DL), were used as screening strategies in this study. OB refers to the percentage of unmodified drug that enters the circulation after oral administration (Liu et al., 2013). In this study, a nonlinear support vector regression (SVR) mathematical model was constructed to predict the OB (Xu et al., 2012). The final expression of the SVR function is as follows: f α = ∑ i = 1 n A i − A i * K α , α i + B . The Lagrange multipliers A _i and A _i ^* are obtained by minimizing the canonical risk function. B is the regression parameter. Moreover, its kernel function K(α, α _i ) is obtained from the linear dot product of the nonlinear mapping using the following formula: K α , α i = ϕ α ⋅ ϕ α i . A higher value indicates a higher possibility that the metabolite could be used in clinical practice (Alam et al., 2015). DL can predict the likelihood of a metabolite becoming a drug by comparing the functional groups of the target metabolite with the active functional groups that exist in known drugs (Walters and Murcko, 2002). In this study, the Tanimoto coefficient was used to evaluate the DL index of the PAMR metabolites with the following specific formula: T α , β = α × β α 2 + β 2 − α × β , where α was calculated using Dragon V7.0 software (chm.kode-solutions.net/products_dragon.php) based on the molecular properties of metabolites in the prepared samples of PAMR and β is the average molecular properties of all drugs in the DrugBank database (www.drugbank.ca/) (Mauri et al., 2006). Metabolites with OB ≥ 15% and DL ≥ 0.10 were considered to have relatively good pharmacological properties.

Because TCMs have modes of action that occur through multiple metabolites and targets, the prediction of potential targets is particularly important. Computer models based on large datasets, which are currently relatively mature, were utilized to predict potential interactions between the drugs and targets in this study. The model integrated chemical, genomic and pharmacological information with a core composed of random forest (RF) and support vector machine (SVM) methods. Because these are well-known and widely used theories, only brief descriptions are given here (Breiman, 2001; Burbidge et al., 2001; Zhao et al., 2006). This study used the R randomForest function in the software package (http://cran.r-project.org/web/packages/randomForest/index.html, Publisher on 25 March 2018) to perform classification and regression, optimize the tuning parameters ntree and mtry to eliminate redundant descriptors that would seriously affect the performance of the model, and use out-of-bag (OOB) errors to obtain feature importance estimates (Hao et al., 2010; Hao et al., 2011). For nonlinearly separated data, an SVM method based on the radial basis function (RBF) was used for classification according to the following formula: f x = s i g n ∑ i = 1 n α y i k x , x i + β , where x_i is a set of descriptors; y_i is the input class label with a value of −1 or 1; f(x) is the decision function; and k (x,x_i) is a kernel function that shows the similarity between two vectors, which can be obtained from the linear dot product of nonlinear maps, k(x, x _i ) = φ(x)·φ(x _i ). The parameters α and β can be obtained by solving the quadratic programming problem. According to the positive or negative value of the decision function, the relationships between the metabolites and proteins were classified as positive or negative, respectively. A positive sample indicates that a metabolite interacts with a protein; a negative sample indicates that a metabolite interacts with a protein in a nonproductive manner. The R kernlab function in the software package (https://cran.rstudio.com/web/packages/kernlab/index.html, 12 November 2019) was then used to establish the SVM classification model.

The keyword “functional dyspepsia” was used to collect the related targets from two sources in this study. The first source was GeneCards®: The Human Gene Database V5.0.0 (www.genecards.org/, updated on 22 September 2020), and the second source was the Online Mendelian Inheritance in Man (OMIM) Database (www.omim.org/, updated on 28 February 2019). In addition, to collect the relevant targets of FD as comprehensively as possible, we also reviewed the literature.

The possible association between PAMR and FD was analyzed by crossing the predicted potentially active metabolites with the acquired disease-related targets, resulting in a Venn diagram of the common targets. The BMTD network was then established and visualized using Cytoscape software V3.7.1 (www.cytoscape.org/).

The STRING online database (https://string-db.org/; the organism was set to Homo sapiens) was used to obtain the PPI data of the molecular targets of PAMR. Subsequently, the PPI network was established, and a topological analysis was performed.

To more deeply analyze why PAMR can treat FD, we performed GO and KEGG pathway enrichment analyses using Bioconductor (R) V3.8 bioinformatics software (http://bioconductor.org/). GO items (p.adjust ≤ 0.05) were collected for functional annotation clustering. The KEGG database was used for pathway enrichment analysis to verify statistically significant gene functional categories (p.adjust ≤ 0.05).

RAMR, RAFI, and PAMR, each weighing 1000 g, were soaked in a tenfold volume of water for 30 min before decoction. The mixture was brought to a boil and maintained at a low boil for 30 min. The aqueous extracts were then filtered out, and the process was repeated once more (maintaining a low boil for 30 min after boiling). All aqueous extracts were combined, filtered, concentrated and dried with a Freeze Dryer (EYELA, Japan). The yields of the powdered aqueous extracts were 67.3% for RAMR, 38.1% for RAFI and 68.61% for PAMR.

Ten-day-old specific pathogen-free (SPF) male Sprague‒Dawley (SD) rat pups were obtained from the Anshengmei Pharmaceutical Research Institute Co., Ltd. (SYXK 2018-0004, Changsha, China). These animals were randomly divided into a normal control (NC) group and an FD model group. The ten-day-old rat pups in the NC group were given 0.2 mL of 2% sucrose solution daily for 6 days, and the ten-day-old rat pups in the FD model group were given 0.2 mL of a mixed solution of 0.1% IAA and 2% sucrose daily for 6 days to establish the FD model (Wei et al., 2019). When the IAA-treated rats were 6 weeks of age, they were fed every other day for 2 weeks of alternate-day fasting. Starting from the eighth week, the model rats were further divided into treatment groups and administered the appropriate combination of aqueous extracts as follows: low-dosage RAMR (RAMR-L, 0.63 g·kg⁻¹ daily, n = 10), medium-dosage RAMR (RAMR-M, 1.26 g·kg⁻¹ daily, equivalent to a clinical daily dosage of 12 g for a 60-kg adult, n = 10), and high-dosage RAMR (RAMR-H, 2.52 g·kg⁻¹ daily, n = 10); low-dosage PAMR (PAMR-L, 0.63 g·kg⁻¹ daily, n = 10), medium-dosage PAMR (PAMR-M, 1.26 g·kg⁻¹ daily, equivalent to a clinical daily dosage of 12 g for a 60-kg adult, n = 10), and high-dosage PAMR (PAMR-H, 2.52 g·kg⁻¹ daily, n = 10); low-dosage RAFI (RAFI-L, 0.063 g·kg⁻¹ daily, n = 10), medium-dosage RAFI (RAFI-M, 0.126 g·kg⁻¹ daily, equivalent to a clinical daily dose of 1.2 g for a 60-kg adult, n = 10), and high-dosage RAFI (RAFI-H, 0.252 g·kg⁻¹ daily, n = 10); and domperidone (3.15 mg·kg⁻¹ by intragastric administration, n = 10). These doses adhere to the guidelines stipulated by the Chinese Pharmacopoeia, which specify a clinical dosage of 6–12 g per day. The Pharmacopoeia also suggests that with a doctor’s guidance, the maximum effective dose can be twice the usual maximum dose. As such, the dosages used in our study align strictly with these stipulations and mimic the therapeutic dosing applied in clinical practice. We also strictly followed these principles when administering domperidone. All our drug doses adhere to the guiding principle of “from the clinic, to the clinic”. The NC rats (n = 20) and FD model rats (n = 20) were given the same volume of normal saline. The drugs were administered continuously for 28 days, during which normal feeding was conducted. The body weight of each rat was recorded daily for subsequent evaluation. Ten 8-week-old rats each from the NC and FD model groups were randomly selected for the gastric emptying experiment, which was conducted according to the literature (Liu et al., 2008). At the end of the experiment and after 24 h of fasting, the rats were anesthetized with 5% chloral hydrate, and serum samples and stomach tissues were harvested. The full thickness of the stomach tissue, including the stomach body and the stomach wall, was stained with hematoxylin and eosin (H&E). The other tissue samples were frozen at −80°C for further analysis, whereas the serum samples were frozen at −20°C for analysis. All animal protocols were approved by the Ethics Committee of Jiangxi Provincial People’s Hospital (no. LLB20200518033) and were performed following the institution’s guidelines for the care and use of experimental animals.

ELISA kits for interleukin 6 (IL-6), interleukin 1β (IL-1β), motilin (MTL), gastrin (GAS), acetylcholine (ACh) and acetylcholinesterase (AchE) supplied by Elabscience Biotechnology Co., Ltd. (Wuhan, China) were used to detect the content of the corresponding protein.

Total RNA was extracted using TRIzol® Reagent (Thermo Scientific, Waltham, MA, United States) and reverse-transcribed with oligo-DT using HiScript^TM Reverse Transcriptase (Vazyme, Beijing, China) according to the manufacturer’s instructions. The primers for PTGS1, PTGS2, DRD2, stem cell factor (SCF) and c-kit were synthesized by Tsingke (Beijing, China). GAPDH was used as an internal reference gene. The 2^−ΔΔCT method was used to determine the relative expression of each gene. The primers used in this study are provided in Supplementary Table S1.

qRT‒PCR was performed using SYBR^TM Green Master Mix (Vazyme) with a QuantStudio 6 Flex system (Applied Biosystems, Foster City, CA, United States). The PCR cycling profile was as follows: one cycle of 50°C for 2 min and 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. Fluorescence signals were detected using the QuantStudio 6 Flex system. Gene expression was normalized to that of the endogenous control GAPDH. The 2^−ΔΔCT method was used to determine the relative expression of each gene.

According to our previously described method (Wang et al., 2020b), the stomach tissue protein levels of DRD2 were determined by Western blotting. GAPDH was used as the internal control. DRD2 and GAPDH antibodies were obtained from Affinity Biosciences Pty Ltd. (Affinity, Melbourne, Australia). Their corresponding secondary antibodies and protein markers were supplied by Boster Biological Technology Co., Ltd. (Boster, Wuhan, China).

All samples were qualitatively analyzed by UPLC–Q–TOF–MS/MS, and the raw data were processed using MarkerView v1.2.1 software (AB SCIEX). The data are presented as the means ± standard deviations (SDs). The significance of the results was determined by one-way analysis of variance using Prism 8.0.1 (GraphPad, San Diego, CA, United States). Tukey’s test was used for multiple comparisons if the variance was uniform, whereas the Tamhane test was used if the variance was not uniform. p < 0.05 was considered to indicate statistical significance.

The samples are shown in Supplementary Figure S1. The results from the UPLC–Q–TOF–MS/MS method validation are shown in Supplementary Table S2. Regarding precision, the RSDs of RT and peak area were 0.51% and 1.78%, respectively. The method repeatability yielded RSDs of the RT and peak area that were less than 2.13% and 4.59%, respectively. Moreover, the RSDs of the RT and peak area in terms of sample stability were less than 1.07% and 4.35%, respectively. These validation results suggested that the system precision, method repeatability, and sample stability were all in agreement with the requirements of qualitative analysis (RSD < 15%) (Fan et al., 2022).

To identify the metabolites of RAMR, raw AFI (RAFI), and PAMR, a combination of various methods was utilized for chemical analysis. First, the most reliable method was the identification of the target metabolites by comparison with the standards. In addition, the mass fragment ions (MS/MS ions) were analyzed to obtain the typical fragmentation pattern, and the structure was inferred based on the available literature data. Overall, a total of 128 metabolites were preliminarily identified from the analyzed samples. PAMR contained 127 of these metabolites, as shown in Figure 1. The details of the identified metabolites are shown in Supplementary Table S3. The asterisks in the table indicate that the metabolite was compared with a standard. Representative total ion chromatograms (TICs) of the standard metabolites and samples in the positive and negative ion modes are presented in Supplementary Figure S2. Supplementary Figure S3 shows the mass spectral properties of the standard metabolites in the positive and negative ion modes. As shown in the table, after processing with the RAFI decoction, the PAMR sample showed the presence of 5-hydroxymethyl furfural (5-HMF), which was not detected in RAMR. Moreover, unlike RAFI, PAMR included quercetin-7-O-rutinoside, whereas the content of 5,7,3′-trihydroxy-6,4′,5′-trimethoxyflavone was greatly reduced and below the detection limit. As the main metabolites of AMR, many lactones are considered the key differential marker metabolites, including atractylenolides and their isomers, especially the atractylenolide Ⅰ isomer.

Based on our previous studies (Wang et al., 2018; Yang et al., 2020), we determined the contents of 8 potentially active metabolites in this study (including synephrine, atractylenolide Ⅰ, atractylenolide II, atractylenolide III, atractylone, neohesperidin, hesperidin and naringin) by HPLC to investigate the influence of PAMR processing. The established quantitative HPLC method was validated based on its linearity, intraday and interday precision, repeatability, stability, recovery, limit of detection (LOD) and limit of quantitation (LOQ), as shown in Supplementary Table S4. Thus, the results revealed that the established method was precise for quantitative determination of the 8 selected metabolites.

To understand the changes in the 8 metabolites in the different sample groups that were present at different amounts, the samples were analyzed using validated HPLC analysis methods. The HPLC chromatograms are shown in Figure 2. The concentrations of the 8 metabolites were recorded, and their contents were calculated and are shown in Table 1.

PAMR, as the processed product of RAMR and RAFI, contained synephrine, naringin, hesperidin and neohesperidin at different amounts after processing. Moreover, these results implied that the ancient processing method has an inconsistent effect on the changes in the metabolites of AMR. Classical AMR processing methods, such as frying, could significantly reduce the content of atractylone and increase the contents of atractylenolides Ⅰ, II, and III due to the poor stability of atractylone, which is easily oxidized at room temperature and then converted into the abovementioned atractylenolides (Hao et al., 2007), and the decomposition of atractylone is accelerated during heat-mediated processing (Wen et al., 1999). Moreover, atractylenolide III is dehydrated and decomposes into atractylenolide II after heat treatment (Wen et al., 1997). However, the metabolites in PAMR were changed in a different manner. This ancient and special processing method had little effect on increasing the contents of atractylenolides II and III but slightly reduced the content of atractylenolide Ⅰ. Of course, this method has some similarities with the classical processing method; that is, it can reduce the content of atractylone, although the reduction was small. We therefore speculated that the method presented here can transform atractylone into isomers of these lactones, but further extraction and separation are needed.

The UPLC–Q–TOF–MS/MS results were utilized as the database for identifying potentially active metabolites. The screening criteria of OB ≥ 15% and DL ≥ 0.10 resulted in the identification of 25 metabolites that met the criteria. Although 7 metabolites did not meet the criteria, they were still listed as potentially active metabolites due to their clear biological activities to ensure that subsequent research was comprehensive and objective. First, although its values were lower than the set thresholds, synephrine (OB = 7.90%, DL = 0.04) has been suggested by the Chinese Pharmacopoeia to be the key metabolite of AFI. This a natural product derived from citrus has a structure that is similar to that of adrenergic receptor agonists, acts on some adrenergic and serotonergic receptors (Rossato et al., 2011) and regulates gastric intestinal motility (Fang et al., 2009). Synephrine has good anti-inflammatory properties and can affect the STAT-6 signaling pathway to inhibit the expression of eotaxin-1 (Roh et al., 2014). Moreover, the other 6 metabolites that were identified are flavonoids with strong biological activities. For example, hesperidin (OB = 13.33%, DL = 0.67) can regulate GI motility in both directions through H1 histamine receptors (Fang et al., 2009). Moreover, hesperidin also exerts anti-inflammatory (Lorzadeh et al., 2019), antitumor, antioxidant and analgesic effects (Aggarwal et al., 2020). Considering these factors, the screening conditions were expanded to include these metabolites. As a result, a total of 32 potentially active metabolites were identified in PAMR, as shown in Supplementary Table S5.

Among the 32 potentially active metabolites identified in PAMR, a total of 158 proteins were identified as their respective targets. Detailed information is provided in Supplementary Table S6. Additionally, a total of 2,452 targets related to FD were collected from the GeneCards database, and 24 related targets were obtained from the OMIM database. The targets were collated, and after removing duplicates, the targets were normalized using the UniProt database (www.UniProt.org), which ultimately yielded 2,357 therapeutic targets for FD. To find potential associations between PAMR and FD, we analyzed related targets of PAMR and FD, and a total of 105 overlapping targets were identified among these metabolites, as illustrated in Figure 3A. This finding indicates that PAMR is likely to treat FD through these 105 targets, as shown in Supplementary Table S7. To explore the potential mechanism underlying the therapeutic effect of PAMR on FD, a BMTD network based on the 32 potentially active metabolites and 105 prominent targets was constructed and visualized using Cytoscape software, as shown in Figure 3B. To further explore the relationship between the targets and analyze the mechanism of action underlying the efficacy of PAMR in treating FD, a PPI network was established. The network consisted of 105 nodes and 1570 edges, as shown in Figures 4A, B (the top 35 proteins). To unravel the biological processes and potential signaling pathways associated with the treatment of FD by PAMR, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. The top 20 GO terms and KEGG pathways based on the p.adjust values are presented in Figures 5A, B.

Eight-week-old rats in the normal control (NC) and FD model groups were randomly chosen for gastric tissue removal, and pathological sections were obtained. Section staining showed that the gastric mucosa morphology in the model group did not become abnormal compared with that in the NC group (Figure 6A). In the NC group, the gland arrangement was also normal, and no abnormalities were found. This result indicated that there were no organic lesions in the gastric tissue of the rats in the FD model group, which was consistent with the fact that FD is a nonorganic disease. The gastric emptying experiment with the 8-week-old rats showed that the gastric emptying rate of the rats in the FD model group was significantly lower than that of the rats in the NC group (p < 0.01), as displayed in Figure 6B. This finding demonstrates that the FD model rats exhibited significantly hindered gastric emptying compared with those in the NC group. Based on the above evaluation results and referring to the relevant literature (Liu et al., 2008; Wei et al., 2019), the FD model was successfully established. The body weights of the rats were recorded at each predetermined time point. The success of FD model establishment was verified by analyzing the recorded weight results. On Day 10, IAA intragastric administration was performed to create the FD model. At the end of modeling on Day 56 (week 8), the rats were treated, and treatment was completed on Day 84 (week 12). Before the administration of IAA for modeling (Day 10), no significant difference in body weight was found between the different groups (Figure 6C). After modeling was completed (Day 56), the body weights of the rats in the model group were significantly lower than those of the rats in the NC group (p < 0.01), indicating that the FD model was successfully established (Figure 6D). After treatment (Day 84), the body weights of the rats in the NC, domperidone, and PAMR-H groups were significantly higher than that of the rats in the FD model group (p < 0.01) (Figure 6E). Moreover, the body weight of the rats in the RAMR-H group were significantly higher than that of the rats in the FD model group (p < 0.05). The body weights of the rats in the other treatment groups were also higher than that of the rats in the FD model group, but the differences were not significant. However, a positive correlation between body weight and dosage was found in all the treatment groups. These results implied that domperidone, high-dose PAMR and high-dose RAMR can treat FD. Although no significant difference was detected, PAMR was slightly more effective than RAMR; thus, this result indicates that PAMR can treat FD more effectively than RAMR.

First, the anti-inflammatory effects of PAMR were verified. The results of the in vivo experiments showed that different doses of PAMR reduced the concentrations of IL-6 and IL-1β in the serum of FD model rats (Figures 7A, B). Different doses of each drug caused significant decreases in the mRNA expression levels of PTGS1 and PTGS2 (p < 0.01), as shown in Figures 7C, D. In addition, these anti-inflammatory effects were dose dependent. Notably, the results also showed that RAMR and RAFI had clear anti-inflammatory effects when used alone. However, the anti-inflammatory effect of PAMR was slightly greater than those of RAMR and RAFI, although the difference was not significant. These results also prove the results from the network analysis. Therefore, it can be concluded that multiple potentially active metabolites of both AMR and AFI can regulate these inflammatory targets, indicating that the anti-inflammatory effect of RAMR is enhanced after processing.

Further exploration was conducted to examine the effects of PAMR on gastric motility in the context of treating FD (Figure 8). MTL is thought to be a unique hormone capable of acting during the digestive interphase rather than postprandially (Ohno et al., 2010). Many studies have shown that the motilin receptor exists in smooth muscle cells (Strunz et al., 1975; Adachi et al., 1981), where it exerts its direct contractile effect. GAS is secreted by G cells in the gastric antrum, functions as an acidic secretor and regulates various biological processes as a cell growth factor (Dockray et al., 2001). Therefore, MTL and GAS are considered important indicators of GI motility. As shown in Figures 8A, B, the MTL concentrations in the serum of rats in the domperidone group and the groups receiving high dosages of various drugs were significantly higher than those in the serum of rats in the FD model group (p < 0.01). Similarly, the concentrations of GAS in the serum of rats in the domperidone, RAMR high-dose and RAFI high-dose groups were significantly higher than those found in the FD model group (p < 0.01). However, the concentration of GAS in the PAMR group was higher than that in the FD model group, but the increase was not significant (p > 0.05). Additionally, PAMR did not regulate the MTL and GAS concentrations as effectively as RAMR and RAFI, but no significant differences were observed (p > 0.05).

The concentrations of Ach and AchE in the serum of each treatment group increased following drug administration (Figures 8C, D). However, the regulation by PAMR was still weaker than that by RAMR and RAFI, although no significant difference was found (p > 0.05). Ach can induce muscle contraction and promote intestinal peristalsis and gland secretions, ultimately accelerating intestinal emptying. However, the mechanism of PAMR is different from that of traditional AchE inhibitors, which improve GI motility by inhibiting AchE secretion and increasing Ach secretion. The hypothesis suggests that PAMR may directly enhance the secretion of Ach and thereby promote GI motility. The increased Ach secretion also increases the secretion of AchE. Nonetheless, further verification is needed to explain this mechanism.

Modern research has found that interstitial cells of Cajal (ICCs) with electrical rhythmicity can produce slow waves and electrical excitation conduction (Langton et al., 1989). ICCs also form a network with the GI nervous system and GI smooth muscle cells, which is known as the ENS–ICC–SMC network. This network is distributed throughout the GI system and can regulate GI motility. Modern medical studies have proposed that ICCs are closely related to GI motility disorder (Yin and Chen, 2008), which is one of the most important causes of FD. C-kit is specifically expressed by ICCs and can regulate the proliferation and differentiation processes of ICCs when combined with its specific ligand SCF (Lin et al., 2016; Zhang et al., 2016). The mRNA expression levels of SCF and c-kit in the gastric tissue of rats were increased by the middle and high doses of each drug (Figures 8E, F). The mRNA expression levels of SCF and c-kit in the PAMR-H group were higher than those in the RAMR and RAFI groups and similar to those in the domperidone group.

High DRD2 expression could lead to the colon being underpowered and the generation of movement disorders. The mRNA and protein expression levels of DRD2 in the gastric tissue of rats in the FD model group were significantly higher than those found in the NC group (p < 0.01), as shown in Figures 8G, H. The mRNA expression levels of DRD2 in all treatment groups, with the exception of the RAMR-L group, were significantly decreased (p < 0.01) in a dose-dependent manner compared with those in the FD model group. However, PAMR regulated the DRD2 mRNA levels slightly better than RAMR and RAFI, but the difference was not significant (p > 0.05). The analysis of DRD2 protein expression supported this finding.

The results from this series of analyses suggests that the processing mechanism of PAMR primarily enhances the regulation of certain targets associated with promoting gastric motility rather than enhancing the regulation of all related targets. Specifically, the processing of PAMR can weaken the regulation of MTL, GAS, Ach, and AchE and enhance the impact on DRD2, SCF, and c-kit.