A metabolomics-binding network pharmacology study of Zuojin Pill intervention in chronic atrophic gastritis induced by Helicobacter pylori

Background Zuojin Pill (ZJP) is widely used for the treatment of gastrointestinal diseases, while its specic mechanism has not been systematically investigated. The aim of this study was to explore the mechanism of intervention of ZJP in chronic atrophic gastritis (CAG) through metabolomics combined with network pharmacology. Materials and methods Potential metabolites and possible pathways for ZJP treatment of CAG were explored using a UPLC-Q-TOF/MS-based metabolomics technique. The key targeting mechanism of ZJP for CAG was explored by combining the analysis with network pharmacology. Results ZJP signicantly reduced serum levels of IL-1β, IL-6, IL-10 and iNOS, and improved pathological characteristics. Metabolomic results indicated that the therapeutic effect of ZJP was mainly related to ten metabolites, including choline, L-threonine, hydroxypyruvic acid, creatine, taurine, succinic acid, cis-aconitic acid, citric acid, succinic acid semialdehyde and uric acid. Pathway analysis showed that the treatment of CAG by ZJP was associated with taurine and hypotaurine metabolism, glyoxylate and dicarboxylate metabolism, glycine, serine and threonine metabolism, glycerophospholipid metabolism, citrate cycle (TCA cycle), alanine, aspartate and glutamate metabolism, butanoate metabolism and purine metabolism. Validation of potential metabolic markers and key targets of network pharmacology by RT-PCR analysis showed that ZJP signicantly down-regulated a series of inammatory markers, such as MAPK1, PKIA, RB1, SCN5A, RXRA, E2F1, PTGS1, IGF2, ADRB1, ADRA1B, PTGS2, and GABRA1. Conclusion For the rst time, a combination of metabolomics and network pharmacology has been used to clarify the therapeutic effects of ZJP on CAG and its relationship to the regulation of multiple metabolic pathways. and urine data of the control group and model group for ESI+ and ESI- modes could explain the 0.999, 0.993, 0.991 and 0.999 variance of the response variable (R 2 Y) and the cumulative explained variance for modeling in cross-validations (Q 2 ) were 0.975, 0.938, 0.857 and 0.92, respectively. For serum and urine data of the model group vs 2.52 g/kg ZJP group in ESI+ and ESI-modes, R 2 Y values were 0.998, 0.987, 0.98 and 0.989 with Q 2 0.979, 0.81, 0.774 and 0.91, individually. These parameters indicate that the models have good explanatory and predictive capabilities. The S-plot was used to investigate the inherent clustering variables. Variables with an average VIP value above 1 and an absolute P(corr) value above 0.58 in the corresponding S plot can be considered as potential biomarkers. The S-curve plots show the apparent differences between the endogenous metabolites in the control group, model group and 2.52 g/kg ZJP group. The variables located at the ends of the S-curve made an important contribution to the separation of the two groups. The predictive power of the established OPLS-DA model is then veried by a permutation test (n=100). The results showed that the R 2 and Q 2 values of serum and urine data were lower than the permutation tests, indicating the superior t and better predictive ability of the OPLS-DA model.


Introduction
Helicobacter pylori (H. pylori) was de ned as a type 1 carcinogen by the World Health Organization's International Agency for Research and Cancer (IARC) in 1993, and it is widely accepted that H. pylori infection induces chronic atrophic gastritis (CAG) and even the development of gastric cancer, and several studies have now con rmed this association [1][2]. CAG is a common clinical digestive disorder that is the result of the in ammatory process of H. pylori infection and is commonly characterized by thinning of the mucosal epithelium and atrophy and loss of glands. Although triple and quadruple therapies are commonly used in clinical practice to treat CAG, long-term use enhances bacterial resistance and decreases eradication rates and can lead to a number of side effects, such as gastrointestinal reactions and liver dysfunction [3]. Therefore, there is an urgent clinical need for an effective drug with less side effects to treat H. pylori induced CAG.
Zuojin Pill (ZJP) is a famous Chinese herbal formula, rst recorded centuries ago in a ancient medical treatise "Danxi New Law". ZJP consists of two traditional Chinese medicines Coptidis Rhizoma (CR) and Euodiae Fructus (EF) according to a weight ratio of 6:1 (w/w) and has been o cially listed in the Chinese Pharmacopoeia. The main ingredients in CR are alkaloids, including berberine, palmatine, coptisine and epiberberine. Studies have shown that CR has a wide range of pharmacological effects, such as antibacterial, anti-atherosclerotic, antiviral, antioxidant, anti-hepatic steatosis, anti-in ammatory and antitumor effects [4]. EF is the dried near-mature fruit of Euodia rutaecarpa (Juss.) Benth., and the main chemical constituents of EF include alkaloids, and terpenes which have antibacterial, hypotensive and anti-hypoxic pharmacological activities. ZJP has been shown to have anti-in ammatory, anti-ulcer, and anti-acid pharmacological effects, including inhibition of H. pylori growth and proliferation [5]. Currently, ZJP is clinically used to treat a variety of gastrointestinal disorders such as gastritis, cholecystitis, and peptic ulcers [6]. Previous studies have shown that ZJP and its constituent active components of herbal CR and EF exhibit multiple pharmacological effects on cancer through multiple molecular mechanism [7][8]. ZJP plays a gastroprotective role by modulating the NF-кB signaling pathway to regulate in ammatory cytokines and thereby reduce the risk of gastric ulcers [9]. ZJP treats gastrointestinal disorders by restoring the rhythm of the stomach as well as improving gastrointestinal peristalsis [10].
The ethanolic extract of ZJP inhibited in ammation by regulating the expression of iNOS, COX-2, IL-6, IL-1β, and TNF-α [11]. These studies provide a basis for the therapeutic effect of ZJP on gastric ulcers, gastritis and other clinical in ammatory diseases.
As an important branch of systems biology, metabolomics conducts a systematic research of the relationships between, medicine, disease and metabolites. Metabolomics not only characterizes the overall physiological and pathological state after exogenous stimuli, but also systematically identi es endogenous small molecule metabolites. Modern analytical technologies, including nuclear magnetic resonance (NMR), gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) are combined with multivariate data analysis, such as principal component analysis (PCA), partial least-squares discriminate analysis (PLS-DA) and orthogonal projection to latent structures-discriminant analysis (OPLS-DA) [12]. In these methods, ultra-high performance liquid chromatography with quadrupole time-of-ight (UPLC-Q-TOF) has the advantages of high selectivity, resolution and accuracy and is particularly suitable for traditional Chinese medicine (TCM) [13]. Due to the easy collection of blood and urine samples and their close association with diseases, choice of these two biological uids may be a novel and effective methods for metabolomics research [14]. Serum metabolite pro les are considered to be important indicator of physiology and pathology and can help to understand the interventions, metabolic pathways and mechanisms of disease progression. Urine metabolite pro les is important for the discovery of disease biomarkers, particularly excreted metabolites.
Serum samples typically represent low-polarity metabolites, while urine samples consist of high-polarity metabolites. These two metabolite pro les complement each other and re ect overall characteristics, demonstrating the mechanism, intervention and metabolic pathways [15]. Nevertheless, the variety of metabolites and their wide range of concentrations limits the ability to understand disease progression using metabolomics alone, metabolomics analysis of interference with metabolic pathways is still far from completely elucidating the mechanism of TCM. In recent years, network pharmacology has been used to investigate and anticipate mechanisms of drug-target interactions in biological metabolism and related pathways [16]. It provides a systems level approach to understanding the interactions between disease features, physiologically active substances and drug targets [17]. A growing number of studies have shown that network pharmacology, as a holistic and effective tool for studying the effects of TCM, can better understand the complexity of TCM and elucidate its underlying mechanisms through understanding its active chemical components and therapeutic targets [18]. In addition, integrated metabolomics and network pharmacology have been successfully used to explore the interactions between TCM and disease targets, which provide important directions for the therapeutic mechanisms of TCM.
In order to elucidate the intervention of ZJP in the treatment of CAG, this study explored the pathological changes and serum biochemical parameters of ZJP through pharmacological experiments based on metabolomics. A rat model of CAG was established by H. pylori, and the success of the model was assessed by a rapid urease kit. UPLC-Q-TOF/MS serum metabolomics and urine metabolomics were used to determine the changes of endogenous metabolites in urine and serum caused by ZJP and to screen relevant targets in combination with network pharmacology techniques to elucidate possible biomarkers and potential metabolic pathways for ZJP treatment of CAG, providing new strategy and feasibility analysis.

Reagents
Interleukin-6 (IL-6, Lot: 202007), interleukin-10 (IL-10, Lot: 202007), interleukin-1β (IL-1β, Lot: 202007) and Thirty-six rats were randomly divided into 6 groups. Rats in the control group were gavaged with regular saline, and rats in the other ve groups were gavaged with H. pylori suspension (1.5×10 8 colony forming units (CFU)/ml, 1.5mL) four times a week for 8 weeks. Urine was collected from rats 12 hours after the last treatment for 12 hours. Then, except for the control and model groups, ZJP (0.63, 1.26, 2.52 g/kg/day) and omeprazole (1.8 mg/kg/day, as a positive group) were administered by gavage for 4 weeks. Twelve hours after the last treatment, the rats were executed and stomach and blood were collected. The blood was centrifuged at 3,500 rpm for 15 m and the serum was separated. The serum and stomach tissues were stored at -80°C for further experiments.
Serum levels of IL-6, IL-10, IL-1β and iNOS were measured according to the kit instructions. Stomach tissues were xed in 10% neutral formalin buffer for 24 hours. The xed gastric tissues were embedded in para n wax and then cut into slices using a slicer. Hematoxylin-eosin (HE) was used to highlight gastric sinus injury and in ammatory cell in ltration.
Sample preparation and UPLC-Q-TOF/MS testing Mix 600 μL of methanol with 200 μL of serum. The mixture was allowed to stand at 4°C for 20 minutes and then centrifuged at 12,000 rpm for 10 minutes. Finally, the supernatant was aspirated and ltered through a 0.22 μm micropore lter. The ltrate was collected and analyzed in the next step, and the urine sample was prepared in the same procedure as the serum sample.
The analysis of serum metabolic pro le and urine metabolic pro le was performed by Agilent 6550 iFunnel Q-TOF LC/MS (Agilent Technologies, USA) system. On a ZORBOX RRHD C 18 analytical column (2.1 mm i.d. × 100 mm, 1.8 μmi.d., Agilent Technologies, USA), 4μL of each sample was taken and injected into the system, and samples were separated at 30°C. Solvent (water containing 0.1% formic acid) and solvent (acetonitrile containing 0.1% formic acid) were used as the mobile phase and separated at a ow rate of 0.30 mL/min for 25 min with a linear gradient of 100% over 0-1.0 min, over 100-60% over 1.0 -9.0 min, 60-10% over 9.0-19.0 min, 10-0% over 19.0-21.0 min, and 100% over 21.0-25.0 min.
Positive and negative mode electrospray source parameters were set as follows: electrospray capillary voltage of 3.5 kV in negative mode and 4 kV in positive mode, mass range of m/z 50-1200, gas temperature and ow rate were 225°C and 13 L/min, atomizer setting of 20 psi, sheath gas temperature and ow rate were 275°C and 12 L/min with a nozzle voltage of 2000 V in both positive and negative ion mode.

Data extraction and multivariate analysis
Data were extracted, peaks detection and comparison performed by using MassHunter Pro nder software (Agilent, CA, USA). Full scan mode was applied to the mass range m/z 80-1000 and the initial and nal retention times for data collection were set. Data were standardized using the MetaboAnalyst website (https://www.metaboanalyst.ca) and then analyzed by principal component analysis (PCA) and orthogonal-partial least squares-discriminant analysis (OPLS-DA) using SIMCA-P 14.1 software (Umetrics, Umea, Sweden).

Potential biomarker identi cation and pathway enrichment analysis
Biomarkers were identi ed by metabolic differential metabolite screening. In the OPLS-DA analysis, the screening conditions for differential metabolites were set to VIP > 1.0, | p(corr)| ≥ 0.58, P < 0.05, and the obtained metabolites were identi ed as potential biomarkers [19]. Metabolites were identi ed based on the precise molecular weight of the Human Metabolome Database (HMDB) (molecular weight error < 20 ppm). The identi ed compounds were resubmitted to MetaboAnalyst to enrich for potential signaling pathways.

ZJP drug target identi cation and network pharmacology techniques
The chemical composition and associated targets of CR and EF were collected in the TCMSP database (https://tcmspw.com/tcmsp.php). Potential metabolite-associated protein targets were collected via the MBROLE 2.0 database (http://csbg.cnb.csic.es/mbrole2/index.php). The Genecards database (https://www.genecards.org/) is used to collect CAG-associated protein targets. Uniprot ID was used to convert different types of protein IDs, and then a network of component-metabolite-target interactions was established using protein interaction information. Finally, the component-metabolite-target interaction network was visualized and analyzed using Cytoscape 3.7.1 software.

Real-time polymerase chain reaction (RT-PCR) detection
Total RNA was extracted from gastric tissues using the RNA-Quick Extraction Kit and the total RNA was reverse transcribed to cDNA using the PrimerScript RT kit (Promega, Madison, USA). The cDNA was subsequently ampli ed by PCR by ABI Step One Plus. The cDNA was subsequently PCR ampli ed by ABI Step One Plus. The data were analyzed by the 2 -ΔΔCT method, and the list of primers is shown in Table 1.

Statistical Analysis
All data were presented as mean ± standard deviation (SD) and analyzed using the SPSS software program (version 25.0; SPSS Inc., Chicago, IL, USA). The differences were considered to be statistically signi cant when P < 0.05 and highly signi cant when P < 0.01.

ZJP reduces H. pylori-induced CAG pathological damage and serum in ammatory factors
As shown in Figure 1, the control group rats had normal gastric histology, and the model group rats showed pathological features of CAG, such as epithelial rupture of the gastric mucosa, atrophy and misarrangement of glands, accompanied by in ammatory cell in ltration. ZJP administration was effective in improving gastric mucosal lesions, glandular loss, and reducing in ammatory in ltration, especially in the 2.52 g/kg ZJP group. Moreover, the serum levels of IL-1β, IL-6, IL-10 and iNOS were signi cantly increased in the model group rats compared with the control group (P < 0.01). serum concentrations of IL-1β, IL-6, IL-10 and iNOS were decreased after ZJP administration. IL-1β, IL-6, IL-10, and iNOS levels were signi cantly reduced after administration of 2.52 g/kg ZJP compared to the model group (P < 0.01).

Multivariate data analysis
Two analytical methods, PCA and OPLS-DA, are powerful statistical tools that provide insight into the separation between different groups based on the results of different spectra [20]. After Pareto normalization of data from serum and urine samples from the control group, model group and 2.52 g/kg ZJP group, the three groups were distinguished by PCA using SIMCA-P 12.0 software. As shown in Figure  3, there was a good separation between the control group, model group and 2.52 g/kg ZJP group in both ESI+ and ESI-modes, indicating a signi cant difference in the metabolic distribution of the three groups. As shown in Figure 3B, D, there was a clear separation between the serum and urine samples of the control group and model group in ESI-mode, and the 2.52 g/kg ZJP group was located between the control group and model group, indicating that ZJP was regulating the metabolically abnormal rats to a normal state. The ESI+ model has a poor trend in the 2.52 g/kg ZJP group therefore further analysis is required.
OPLS-DA models were then developed for the ESI+ and ESI-modes of serum and urine samples. Given that OPLS-DA was used to screen for differentially expressed metabolites between the two groups [21], we identi ed the differential metabolites under OPLS-DA analysis for the control group and model group, and for the model group and 2.52 g/kg ZJP group, respectively. As shown in Figure 4  These parameters indicate that the models have good explanatory and predictive capabilities. The S-plot was used to investigate the inherent clustering variables. Variables with an average VIP value above 1 and an absolute P(corr) value above 0.58 in the corresponding S plot can be considered as potential biomarkers. The S-curve plots show the apparent differences between the endogenous metabolites in the control group, model group and 2.52 g/kg ZJP group. The variables located at the ends of the S-curve made an important contribution to the separation of the two groups. The predictive power of the established OPLS-DA model is then veri ed by a permutation test (n=100). The results showed that the R 2 and Q 2 values of serum and urine data were lower than the permutation tests, indicating the superior t and better predictive ability of the OPLS-DA model.
Identi cation of potential metabolites in ZJP treatment For more accurate identi cation, potential metabolites were selected based on the principles of |VIP|> 1 and |P(corr)| ≥ 0.58 in the S-plots extracted between the control group and model group, model group and 2.52 g/kg ZJP group. At this threshold, variables with signi cant differences were screened using ANOVA analysis and t-test for multivariate and univariate analyses, respectively. Matching the m/z of metabolite candidates to online databases, including the HMDB database and the Metaboanalyst database, identi es candidates with signi cant changes as biomarkers. The mass tolerance value and the exact mass of the measured m/z are de ned as less than 20 ppm.
For serum samples, a total of six endogenous metabolites were screened in ESI+ mode and ESI-mode (see Table 2 for details). For urine samples, four endogenous metabolites were identi ed based on the above procedure (please see Table 2). These ten potential biomarkers are choline (C00114), L-threonine (C00188), hydroxypyruvic acid (C00168), creatine (C00300), taurine (C00245), succinic acid (C00042), cis-aconitic acid (C00417), citric acid (C00158), succinic acid semialdehyde (C00232) and uric acid (C00366). The levels of L-threonine, hydroxypyruvic acid, taurine, succinic acid and cis-aconitic acid were signi cantly lower in the model group compared to the control group (p < 0.05 and p < 0.01). In addition, levels of choline, creatine, citric acid, succinic acid semialdehyde and uric acid were signi cantly higher in the model group compared to the control group (p < 0.01). The mean peak area of these metabolites tended to normalize in the 2.52 g/kg ZJP group compared with the model group (P < 0.05 and p < 0.01), showing good therapeutic e cacy. The mean peak area of these metabolites is shown in Figure 6 to visualize the effects of ZJP. Alterations in these parameters indicate that metabolic abnormalities occur in control group rats following gastric mucosal in ammatory injury and that metabolic markers return to normal levels after ZJP treatment.

Metabolic pathway analysis
To explore the potential mechanism of ZJP's effect on CAG, potential metabolites identi ed from serum and urine samples were introduced into MetaboAnalyst to construct metabolic pathways, and twentyeight metabolic pathways were obtained. Of these twenty-eight metabolic pathways, seven of them play an important role, including Taurine and hypotaurine metabolism, Glyoxylate and dicarboxylate metabolism, Glycine, serine and threonine metabolism, Glycerophospholipid metabolism, Citrate cycle (TCA cycle), Alanine, aspartate and glutamate metabolism, Butanoate metabolism and Purine metabolism. The match status, P value, -log(P) and impact of each pathway are listed in Table 3.

Network Pharmacology
To visualize the interactions between potential metabolites, disease targets and ZJP components, components and potential CAG disease targets were collected and combined with potential metabolites via the TCMSP database and Genecards database to construct a potential metabolite-CAG targetcomponents network, as shown in Figure 8. MBROLE was used to extend this potential metabolite to related metabolic pathways to further explore the relationship between metabolite-associated proteins and disease targets. Information of the 12 proteins of relevance in compounds-disease targets-metabolites is presented in Table 4, which are directly regulated by 124 chemical components and collected from ZJP ( Figure 8B).
The effect of ZJP on the mRNA expression in gastric tissue of CAG rats CAG is a critical step in the development of gastric cancer, and with increased in ammation comes a progressively higher chance of cancer. A large amount of clinical evidence suggests that ZJP inhibits in ammatory lesions including gastric mucosal injury, neutrophil in ltration and intestinal epithelial chemosis in CAG patients. Thus, in ammation related genes MAPK1, PKIA, RB1, SCN5A, RXRA, E2F1, PTGS1, IGF2, ADRB1, ADRA1B, PTGS2, and GABRA1 may play a more important role in ZJP treatment of CAG. To verify the validity of the network pharmacology predictions, RT-PCR was used to detect the mRNA expression of these genes in gastric tissues. As shown in Figure 9, the relative mRNA expression levels of MAPK1, PKIA, RB1, SCN5A, RXRA, E2F1, PTGS1, IGF2, ADRB1, ADRA1B, PTGS2, and GABRA1 in the model group were signi cantly elevated (P<0.01). The expression levels of the above genes in the ZJP groups were signi cantly decreased (P<0.05 and P<0.01), indicating the effectiveness of ZJP in treating CAG through anti-in ammation.

Discussion
CAG is a common clinical disorder of the gastrointestinal tract and is considered to be a precancerous condition of gastric cancer. Current conventional treatments for CAG are mainly focus on symptomatic treatment, however, symptomatic treatment is often associated with some degree of side effects, poor e cacy and high relapse rates [22]. Metabolomics combined with network pharmacology can effectively overcome these problems by extensively mining data to identify targets of metabolism and disease [23]. Metabolomics analyzes small molecule levels in serum and urine to explore compounds with signi cant abnormalities and reveal mechanisms of disease development. Network pharmacology reveals the molecular mechanisms of drug treatment for diseases by determining the interactions between chemical components and disease protein targets. Therefore, in this study, rat serum metabolomics and urine metabolomics based on the UPLC-Q-TOF/MS method combined with network pharmacology were used to elucidate the therapeutic mechanism of ZJP on CAG, which also provides a molecular level exploration of the ZJP mechanism.
H. pylori is the most commonly infected bacterium worldwide and is a signi cant risk factor for the development of gastric cancer. Numerous studies have been published showing that H. pylori can colonize and grow in an extremely acidic gastric environment, establishing a persistent infection and decreasing host immune function, leading to the development of gastritis and gastric cancer [24]. Currently, H. pylori is widely used to establish rodent models of CAG [25]. Therefore, in the present study, the CAG rat model was established by gavage of H. pylori. HE staining con rmed that ZJP was effective in reducing gastric mucosal damage and maintaining epithelial structural integrity, and improving monocyte and lymphocyte in ltration. Serum biochemical parameters showed that ZJP signi cantly reduced serum levels of IL-1β, IL-6, IL-10 and iNOS, and restored the integrity of gastric mucosa and effectively reduced the expression of in ammatory factors. The above results con rm that ZJP shows a therapeutic effect on CAG, which is consistent with clinical e cacy.
Next the changes in serum metabolomic and urine metabolomic characteristics of ZJP in CAG treatment were identi ed. ZJP protects the normal structure of the gastric mucosa by restoring potential metabolites to normal levels. The screened biomarkers can be effective targets for disease diagnosis, treatment, and prevention [26]. The PCA model revealed signi cant differences in metabolite distribution between the control group, model group and 2.52 g/kg ZJP group. The distribution of potential metabolic biomarkers was identi ed in the OPLS-DA model. There was a signi cant separation of biomarkers between the control group and model group, indicating a signi cant effect of H. pylori on the serum metabolomic pro le and the urine metabolomic pro le. There were also signi cant differences between the 2.52 g/kg ZJP group and model group, suggesting a therapeutic effect of ZJP on CAG. Finally, six differential metabolites were screened in serum samples and four metabolic differentials in urine samples to reveal the regulatory mechanism of ZJP in the treatment of CAG. These metabolites interact and regulate the disease process in different ways. The results suggest that the occurrence and development of CAG is regulated by changes in many physiologically and pathologically relevant biomarkers, most of which in the organism are mutually in uential. ZJP restores normal levels of expression by modulating these ten potential biomarkers, suggesting that ZJP can treat H. pylori-induced CAG through multiple targets and multiple pathways.
In total, ten differential metabolites were identi ed and were involved in seven major metabolic pathways.
These metabolites involve seven major metabolic pathways, including taurine and hypotaurine metabolism, glyoxylate and dicarboxylate metabolism, glycine, serine and threonine metabolism, glycerophospholipid metabolism, TCA cycle, alanine, aspartate and glutamate metabolism, butanoate metabolism and purine metabolism. Taurine and hypotaurine metabolism is important metabolic pathways. Taurine is involved in multiple biological pathways and has a variety of pharmacological activities, such as eye and brain development, immune function, osmoregulation, as well as antioxidant and anti-in ammatory activities. Previous studies have shown that it is protective against oxidative stress-induced gastrointestinal injury [27]. Taurine is one of the most abundant free amino acids in animal tissues and plays an important role in a variety of important physiological processes. Upregulation of taurine in serum by ZJP treatment may be a marker of in ammatory response. Elevated amino acid levels are frequently observed in patients with gastritis, which helps to distinguish CAG patients from normal people [28]. In addition, metabolic pathway analysis identi ed several amino acidrelated pathways in CAG, including the metabolism of glycine, serine, and threonine and the metabolism of alanine, aspartate, and glutamate. The results of the present study showed that the expression of choline, creatine, and succinic acid semialdehyde associated with amino acid metabolism was elevated and L-threonine expression was decreased in the model group, and this trend was ameliorated by ZJP administration. The involvement of glyoxylate and dicarboxylate metabolism further supports the involvement of ZJP in energy metabolism. Gastric wall cells contain a large number of mitochondria [29]. Mitochondria provide energy to the cells through oxidative phosphorylation, which plays an important role in maintaining the steady state and function of the gastric mucosa. ZJP further elevates hydroxypyruvate expression by restoring energy metabolism. Purines are essential components of nucleotides in cell proliferation, thus impaired purine metabolism is associated with in ammation and the development of cancer [30]. The expression of uric acid was elevated in CAG rats, which is consistent with what has been reported in the research [31]. The TCA cycle is the nal metabolic pathway for carbohydrates, lipids, and amino acids and is the most important metabolic pathway for energy supply. Succinic acid and cis-aconitic acid belong to the TCA cycle and mediate the formation and metabolism of intracellular ATP [32]. The citric acid is a key intermediate of the TCA cycle [33]. The results showed that succinic acid and cis-aconitic acid contents were decreased and citric acid contents were signi cantly increased in CAG rats. These results indicate the importance of the TCA cycle in the pathological process of CAG.
In an attempt to further understand the mechanism of H. pylori-induced CAG by ZJP treatment and the association between chemical composition and metabolites, we established a "compounds-disease targets-metabolites" interaction network by combining network pharmacology. The results showed that 124 chemical components originating from ZJP acted directly with 12 potential protein targets, suggesting the advantages of multiple components in ZJP in regulating endogenous metabolic changes in the body, showing the advantages of ZJP in the clinical treatment of CAG. Twelve targets were strongly associated with in ammation, including MAPK1, PTGS2, PTGS1, RXRA, ADRA1B, RB1, ADRB1, E2F1, SCN5A, PKIAIGF2, and GABRA1 [34][35][36][37]. Therefore, to verify the accuracy of the network pharmacology predictions and to further elucidate the potential mechanism of ZJP, we further tested the expression of these genes in gastric tissues of CAG rats. The expression of MAPK1, PTGS2, PTGS1, RXRA, ADRA1B, RB1, ADRB1, E2F1, SCN5A, PKIAIGF2, and GABRA1 was signi cantly elevated in the CAG model group.
However, ZJP signi cantly reduced the relative mRNA expression of these genes. These results suggest that ZJP may play a role in the treatment of CAG by further ameliorating gastric mucosal in ammatory lesions through anti-in ammation.

Conclusion
This study systematically investigated the e cacy of ZJP for the treatment of H. pylori-induced CAG and its molecular mechanisms through a combination of metabolomics and network pharmacology. This

Consent for publication
All authors agree to publish this paper.

Competing interests
The authors declare that they have no competing interests.           The

Figure 5
The OPLS-DA score plots, S-plots and 100-permutation test generated of urine samples between the control group, model group and 2.52 g/kg ZJP group. OPLS-DA score plots were the pair-wise comparisons in ESI+ mode, control group vs model group (A), model group vs 2.52 g/kg ZJP group (D) as well as in ESI-mode, control group vs model group (G), model group vs 2.52 g/kg ZJP group (J)