Introduction

UC is a chronic, nonspecific intestinal inflammatory disease that belongs to category of “diarrhea” and “dysentery” in TCM,1 with typical clinical symptoms of recurrent diarrhea, mucopurulent bloody stools, and abdominal pain.2 The etiology is unknown, mainly related to the interaction of multiple factors such as environment, genetics and intestinal microecology, which lead to an abnormal immune imbalance in the intestinal tract.3 Currently, the drug treatments of UC mainly include 5-aminosalicylic acid drugs, steroids, immunomodulators and biological agents.3–6 However, complementary and alternative therapies, containing TCM, dietary supplements, probiotics, and mind/body interventions,7 are becoming a novel strategy for treating UC, due to the side effects and adverse reactions such as drug dependence, drug resistance and decreased immune function. Among them, TCM has been manifested to have potential advantages owing to its clinical application and low toxicity.1,2,8–11

SYD, a well-known traditional Chinese medicine formula, possesses the functions of clearing heat and drying dampness, as well as regulating Qi and blood, and mainly focus on the treatment of damp-heat dysentery clinically.3 SYD consists of nine herbs, containing Paeoniae Radix Alba (Chinese name: Shaoyao), Angelicae Sinensis Radix (Chinese name: Danggui), Coptidis Rhizoma (Chinese name: Huanglian), Scutellariae Radix (Chinese name: Huangqin), Glycyrrhizae Radix et Rhizoma (Chinese name: Gancao), Arecae Semen (Chinese name: Binglang), Aucklandiae Radix (Chinese name: Muxiang), Rhei Radix et Rhizome (Chinese name: Dahuang), Cinnanmomi Cortex (Chinese name: Rougui), in a ratio of 30:15:15:15:6:6:6:9:5.3

Many studies have shown that these herbs contained many bioactive compounds with anti-inflammatory and antioxidant activities.10 A study has indicated that glycyrrhizic acid extract and its active ingredients are able to inhibit the free of pro-inflammatory mediators NO, IL1β and IL6 by mouse macrophages.12 Berberine, the active ingredient of Coptidis Rhizoma, can reduce the inflammatory response in UC and protect intestinal epithelial cells from damage.13 One of the main components of Rhei Radix et Rhizome, rhein, can eliminate the enhanced intestinal permeability caused by uric acid, regulate the intestinal microbiota, and relieve intestinal inflammation.14 SYD protects gut barrier function by adjusting the mucus layer genes MUC1, MUC2, MUC4 and TFF3, along with the epithelial barrier genes zonula occludens-1 (ZO-1) and Occludin.3

It is noteworthy that the key active components and targets of SYD are still unknown. Moreover, despite the known therapeutic effects of SYD, its molecular mechanism has not been fully elucidated and requires further study.7

Based on the complex chemical compositions and multi-target as well as multi-level characteristics of TCM, new methods such as network pharmacology and metabolomics have been created to explore the mechanism of drugs on diseases.

On the one hand, metabolomics is a new discipline that simultaneously conducts qualitative and quantitative analysis of all low-molecular-weight metabolites of a certain organism or cell in a specific physiological period. Ultra-performance liquid chromatography-quadrupole-time of flight-mass spectrometry (UPLC-Q-TOF-MS) has been widely used in drug multicomponent analysis and metabolite research owing to its high resolution, precision and sensitivity,15 which can provide primary and secondary fragment ion information of small molecule metabolites, with the ability to help us quickly analyze metabolite structures, laying the foundation for metabolite identification. Through metabolomic analysis, we can understand which metabolic pathways are mainly restored by SYD, such as glycerophospholipid metabolism and sphingolipid metabolism. Notably, plasma, feces, urine, et cetera can all be detected by metabolomics.

It has been confirmed that UC pathogenesis is associated with metabolic disorders,13,16 the gut microbiota also produces many important metabolites containing short-chain fatty acids, biotin and vitamin K.17 However, there are few studies on metabolic changes during the treatment of colitis with SYD. Therefore, it is beneficial to study the metabolite changes during the pathogenesis and treatment of UC to help us further understand the pathogenesis of UC and the underlying therapeutic mechanisms of SYD.

On the other hand, network pharmacology is a new discipline, which discusses the mechanism of drug therapy by constructing a component-target network, suitable for the study of Chinese medicine compounds.

Molecular docking is a method to predict the binding affinity between protein receptors and small molecule ligand.18 Binding energy lower than 0 indicates that the two molecules have the ability to bind spontaneously, and the smaller the binding energy, the more stable the conformation.2

In this study, UC model was established in mice, and untargeted LC-MS-based metabolomic approach and network pharmacology were used to evaluate the protective effect of SYD on UC, and to explore its mechanism. The flow chart of the research process is displayed in Figure 1.

Figure 1 Flow chart of the research process.

Materials and Methods Main Chemicals and Reagents

DSS was purchased from MP Biomedicals (Irvine, CA, USA). LC-MS grade methanol and acetonitrile for extraction or analysis were purchased from Merck KGaA (Darmstadt Germany). Ultrapure water was prepared by a Milli‐Q water purification system (Millipore, Bedford, MA, USA). LC-MS grade formic acid was obtained from Shanghai Aladdin Biochemical Technology (Shanghai, China).

Composition and Preparation of SYD

After mixing Paeoniae Radix Alba (30g), Angelicae Sinensis Radix (15g), Coptidis Rhizoma (15g), Scutellariae Radix (15g), Glycyrrhizae Radix et Rhizoma (6g), Arecae Semen (6g), Aucklandiae Radix (6g), Rhei Radix et Rhizome (9g), Cinnanmomi Cortex (5g), we soaked them in distilled water (1:10, w/v) for 30 minutes, boiled them for 1h, filtered them with gauze, collected the filtrate, decocted the residue with distilled water (1:6, w/v) in the above method, then combined the two filtrates, and concentrated them to 1 g/mL with a rotary evaporator at 60 ℃.3 Finally, the SYD oral solution was stored in the refrigerator at 4 ℃ before intragastric administration to mice. All herbs were purchased from the Second Affiliated Hospital of Zhejiang Chinese Medical University (Hangzhou, Zhejiang, China).

Animals and Groups

A total of 40 8-week-old healthy male C57BL/6 mice (no specific pathogen, SPF) weighing 16–20g were bought from Zhejiang Chinese Medical University Laboratory Animal Research Center. All mice were fed with standard pelleted feed and free drinking water ad libitum under the standard laboratory conditions of controlled temperature (22 ± 2 ℃), humidity (50 ± 10%) and 12-hour light/dark cycle.6 In addition to the above, they were acclimatized for 7 days before the experiments we started. Moreover, the study was approved by the Ethics Committee of Zhejiang Chinese Medical University (IACUC-20210531-02).

Forty mice were randomly divided into three groups: normal control (NC), model (UC), and SYD group, with 10 mice in the normal control group and 15 mice in the model group and SYD group. Then mice in the model group and SYD group were given 3% DSS in drinking water for 7 days, respectively, to establish the DSS-induced UC model, while mice in the normal control group were treated with water.5 After the UC model was established, mice in the SYD group were gavaged with 0.02mL/g SYD, whereas those in the control and model groups were gavaged with the same volume of normal saline for 7 consecutive days, and all mice returned to normal drinking water.

Specimen Collection

Fecal samples of mice were gathered and stored at −80 ℃ on four time periods: Day 0, Day 7, Day 10 and Day 14. At the end of the experiment, urine samples were collected after a 12-hours fast, after which all mice were sacrificed and liver, serum samples were collected and stored at −80 ℃ for subsequent metabolomic analysis. The mouse colon tissues were taken for histopathological analysis. During the experiment, the daily body weight and disease activity index (DAI) of mice in each group were recorded.

During the experiment, the percentage of daily weight loss, stool consistency and occult blood or gross bleeding of mice in each group were recorded to calculate the disease activity index (DAI). The scoring criteria of DAI are shown in Table 1.9,13

Table 1 Criteria for Scoring Disease Activity Index

Table 2 Evaluation of Pathological Score

Table 3 The Gradient Elution Process of SYD

Table 4 The Gradient Elution Process of Fecal, Liver, Serum, Urine Samples

Table 5 The Mass Spectrum Conditions of All Samples

Table 6 The Molecular Docking Parameters

Table 7 Validation Parameters of PCA, PLS-DA and OPLS-DA Models for Mice Feces, Livers, Serum and Urine

Table 8 Differential Metabolites in Fecal Tissues

Table 9 Differential Metabolites in Liver Tissues

Table 10 Differential Metabolites in Serum Tissues

Table 11 Differential Metabolites in Urine Tissues

Table 12 Metabolic Pathways of SYD During the Treatment of UC

Table 13 Degree and Basic Information of 13 Active Components

Evaluation of Hematoxylin-Eosin (H-E) Staining

The colon tissues were fixed in polyoxymethylene solution. After fixation, dehydration, paraffin infiltration and embedding, the sections were stained with H-E, and the histological score was calculated according to the degree of mucosal epithelial injury and inflammatory cell infiltration. The pathological scoring criteria are displayed in Table 2.9

Preparation of Samples

Part of the prepared SYD samples were centrifuged at 12000g for 10 minutes, the supernatant was diluted to 100mg/mL with methanol, sonicated for 15 minutes and centrifuged at 12000g for 10 minutes. Transfer the supernatant (80 μL) to autosampler vials for subsequent analysis, and repeated for 3 times.

Prior to analysis, 100 μL of each collected serum sample per group was thawed at 4℃, and subsequently aliquoted with 400 μL methanol to precipitate proteins. Stir the mixture ultrasonic for 15 minutes and centrifuge at 12,000 g for 10 minutes at 4℃. The supernatant (400 μL) was transferred to the EP tube and evaporated to dryness in a nitrogen stream at 4 ℃. Dissolve the residue with 100 μL methanol followed by vortexing for 60s and centrifuge at 12,000 g for 10 minutes. Transfer the supernatant (80 μL) to autosampler vials and run the instrument. Urine was processed in the same way as serum.

100mg of each fecal sample was collected and thawed at 4℃, extracted with 1mL methanol, sonicated for 15 minutes, and then was centrifuged at 12,000g for 10 minutes, and the supernatant was obtained for UPLC-Q-TOF-MS analysis. The liver samples were ground into powder separately in liquid nitrogen, and 100mg of each sample was collected, following the same procedure as stool samples.

Standardization correction adopts the method of quality control (QC) sample standardization, and 10 μL of each prepared sample was mixed as QC samples.13 QC samples have the ability to display the signal offset law in the process of data acquisition, which is convenient to assess the constancy of the system in the whole process.

UHPLC-Q-TOF-MS Conditions

All samples were analyzed in both negative and positive mode on an ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 µm; Waters Corporation, Ireland) at a flow rate of 0.3 mL/min, sample injection volume of 2.0 μL, the autosampler temperature of 4 ℃ and the column temperature of 40 ℃.

Mobile phase A was 0.1% (V/V) formic acid/water and mobile phase B was 0.1% (V/V) formic acid/acetonitrile. The linear gradient elution processes of SYD, fecal, liver, serum and urine samples were displayed in Tables 3 and 4. The mass spectrum conditions of SYD, fecal, liver, serum and urine samples in positive and negative modes, respectively, are shown in Table 5.

Before the start of the analysis, a blank sample was injected three times, after which a blank sample was injected every 8 times and a QC sample every 10 times. All samples were randomly sequenced.

Untargeted Metabolomics Analyses

All data were obtained on MassLynx V4.1 (Waters Corporation, Milford, MA, USA) software, and then imported into Progenesis QI V2.0 (Waters Corporation, Milford, MA, USA) software for data preprocessing, including baseline filtering, peak identification, integration, retention time correction, peak alignment and normalization. Furthermore, on the basis of accurate mass number, secondary fragments and isotopic distribution, the Human Metabolome Database (HMDB, http://www.HMDB.ca), Lipid Maps (http://www.lipidmaps.org), NIST (https://www.nist.gov/) and ChemSpider (http://www.chemspider.com/) were used to speculate and identify the compounds.

Later, import the data stabilized as normal distributions (Figures S1 and S2)13 into SIMCA-p14.1 software, select pareto scaling for multivariate statistical analysis, including Principal Component Analysis (PCA), Partial Least Squares Discriminant Analysis (PLS-DA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA). The quality of PLS-DA and OPLS-DA models were assessed by 10-fold cross-validation, analysis of variance of the cross-validated residuals (CV-ANOVA), 200 permutation tests and Q2 value.

Differential metabolites were screened by the combination of univariate analysis and multivariate analysis. Metabolites with variable importance in the projection (VIP) >1 in OPLS-DA analysis and q-value less than 0.05 in two sample t-test followed by false discovery rate (FDR) were selected as the differential compounds between the two groups. Furthermore, metabolites with fold-change (FC) >2 or FC <0.5 were selected from the above differential metabolites as significantly differential metabolites in this study. Receiver operating characteristics (ROC) curves were applied to verify the accuracy of potential biomarkers.

These significantly altered metabolites were analyzed by MetaboAnalyst 5.0 (http://www.metaboanalyst.ca/)19 for pathway analysis.20,21

Network Pharmacology Analysis

In the three databases of the gene map of the Online Mendelian Inheritance in Man (OMIM, https://OMIM.org/), Therapeutic Target Database (http://db.idrblab.net/TTD/) and GeneCards (https://www.genecards.org/, score>20), “ulcerative colitis” was used as the search term to screen candidate targets.

Through Traditional Chinese Medicine Systems Pharmacology (TCMSP, https://tcmspw.com/tcmsp.php) database, take the names of various traditional Chinese medicines in SYD as the search words, drug‐like properties (DL) ≥0.18 and oral availability (OB) ≥30% as the screening conditions to screen the effective components of every traditional Chinese medicines.22 The predicted targets of the screened compounds were acquired from the DrugBank (https://www.drugbank.ca/) database and verified literature.7 After summarizing the targets, import them into UniProt database (https://www.uniprot.org/) to standardize the names of genes and proteins.

Combine UC-related targets with SYD-related targets in Venny (htttps://bioinfogp.cnb.csic.es/tools/venny/), choose the intersection genes as the key genes. In addition, use STRING 11.5 (https://STRING-db.org/) and Cytoscape 3.8.2 to structure the protein–protein interaction (PPI) network. At the same time, the relevant active components were screened by using the intersection target, the active component-target network was drawn by Cytoscape, and the connection number between each node was calculated by NetworkAnalyzer, an analytical tool in Cytoscape. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were ran by R 4.1.0.

The identified metabolomic differential metabolites and targets were introduced into Integrated Molecular Pathway Level Analysis (IMPaLA, http://impala.molgen.mpg.de/) and Cytoscape, then a compound-pathway-target network was constructed to manifest the interaction between metabolites, pathways and genes.23

Molecular Docking

Through the active component-target network, the four active ingredients with the largest degree were selected as the ligands for molecular docking, and the three-dimensional structures of these active ingredients were obtained in PubChem (https://www.ncbi.nlm.nih.gov/). The crystal structures of targets were acquired from the RCSB Protein Data Bank (https://www.rcsb.org/). Altogether five targets were selected in this study, containing AKT1 (PDB ID: 1h10), IL1B (PDB ID: 2nvh), IL2 (PDB ID: 1m48), IL6 (PDB ID: 4ni9) and STAT3 (PDB ID: 6njs).

First of all, active ingredients were converted to pbdqt format using The Open Babel GUI 3.1.1 and AutoDockTools 1.5.7. Next, the target protein was imported into Pymol 2.5.2 and AutoDockTools for structural optimization, including removal of water molecules, removal of small molecule ligands, and addition of hydrogen atoms and charges. Then, molecular docking was performed using Autodock Vina, and the docking parameters are indicated in Table 6. Finally, the highest scoring docking results were visualized in Pymol.24

Immunohistochemistry

Colon tissue was taken to make paraffin sections (4um) for immunohistochemical staining. After deparaffinization to water, the sections were repaired by microwave in citrate buffer (10mM citric acid, pH6.0). The sections were cooled naturally and then transferred to Tris buffered saline (TBS; pH 7.4). Endogenous peroxidase activity was then blocked by treatment with 3% hydrogen peroxide for 20 minutes and incubated with 5% BSA at room temperature for 30 minutes. Next, the sections were treated with p-PI3K (YP0224, dilution ratio of 1:100, immunology), p-AKT (YP0864, dilution ratio of 1:100, immunology), p-mTOR (YP0176, dilution ratio of 1:200, immunology) overnight at 4℃, respectively, and the next day with HRP*Goat Anti Rabbit IgG (H+L) (RS0002, dilution ratio of 1:500, immunology) for 1 hour at room temperature. Finally, color was developed with DAB and counterstained with hematoxylin.25,26

Statistical Analysis

All data were analyzed by SPSS 21.0 statistical software (IBM, USA). Two sample t-test was used for the comparison between the two groups and one-way analysis of variance (ANOVA) was used for the comparison between the three groups followed by a least significant difference (LSD) test. Kruskal–Wallis statistical test was used for the data of skew distribution, and Bonferroni method was used to correct the p-value. The p-value less than 0.05 was considered statistically significant.

Results Components Analysis of SYD

For the purpose of identifying the main chemical components of SYD, the samples were analyzed by UPLC-Q-TOF-MS. The total ion chromatograms (TIC) of SYD sample in positive and negative modes demonstrated the chemical composition of all compounds.27 The results manifested that SYD contains an ocean of components. As shown in Figure 2, the 12 compounds were distinguished: arecoline, trans-ferulic acid, baicalin, glycyrrhizin, gallic acid, liquiritin, berberine, chlorogenic acid, licoisoflavone A, quercetin, formononetin and kaempferol.

Figure 2 Identification of chemical components of SYD. SYD samples were examined by UPLC-Q-TOF-MS. Total ion chromatography (TIC) in positive (A) and negative (B) modes for SYD samples are shown. (C) Molecular structure of compounds.

The Effect of SYD on DSS-Induced UC

The weight of DSS-induced UC mice decreased significantly from the day 4 (P-value < 0.001). After SYD treatment, the weight loss trend of mice was slightly improved (Figure 3A). In addition, the DAI score (Figure 3B) in the model group was significantly higher than normal group and SYD group (P-value < 0.05), and the colon length (Figure 3C and D) in the model group was significantly shorter than that in the normal group and SYD group (P-value < 0.05), accompanied by a certain degree of edema.

Figure 3 SYD ameliorated UC mice induced by DSS. (A) Body weight change. (B) Disease activity index. (C) Colon length. (D) Representative colon images from each group. **p<0.01, ***p<0.001, compared to normal group; #p<0.05, ##p<0.01, ###p<0.001, compared to model group.

The results of H-E staining (Figure 4A) displayed that inflammatory cell infiltration and crypt loss or deformation appeared in the colon tissue of the model group, and the pathological injury score (Figure 4B) of the model group was significantly higher than that of the normal group (P-value < 0.05). The related symptoms in the colon of SYD mice were alleviated compared with the model group (P-value < 0.05).

Figure 4 Histopathological changes. (A) H-E staining images of colonic tissues in normal group, model group and SYD group. Arrows refer to more inflammatory cell infiltration and crypt loss. (B) Histological score. ***p<0.001, compared to normal group; ###p<0.001, compared to model group.

Taken together, SYD is capable of effectively treating UC in mice and improve the symptoms of UC related bloody stool and mucosal damage.

Untargeted LC-MS-Based Metabolomic Analysis of Fecal, Liver, Serum and Urine Samples of Mice After SYD Treatment

After sample collection, untargeted metabolomics analysis was conducted on fecal, liver, serum and urine samples, respectively, by LC-MS in both positive and negative ion mode. Subsequently, 15838, 7853, 11534, 9088 features were detected in positive ion mode, as well as 13435, 8601, 5452, 11682 features in negative ion mode, respectively.

Figure 5 shows that the change trend of fecal samples on Day 0, Day 7, Day 10, and Day 14 was consistent with each node of this experiment. Before modeling (Day 0, Figure 5A and E), the PCA images of the three groups were clustered together without obvious separation; after modeling (Day 7, Figure 5B and F) and during the SYD treatment (Day 10, Figure 5C and G), the UC group and the SYD group gradually separated, and they were obviously separated from the NC group. After treatment (Day 14, Figure 5D and H), the three groups were well separated, suggesting that there were significant differences in metabolites among the three groups.

Figure 5 PCA score plots of feces metabolic profiling between normal, model and SYD group during the day 0, day 7, day 10 and day 14. (A) Fecal samples collected in the day 0 in positive mode. (B) Fecal samples collected in the day 7 in positive mode. (C) Fecal samples collected in the day 10 in positive mode. (D) Fecal samples collected in the day 14 in positive mode. (E) Fecal samples collected in the day 0 in negative mode. (F) Fecal samples collected in the day 7 in negative mode. (G) Fecal samples collected in the day 10 in negative mode. (H) Fecal samples collected in the day 14 in negative mode.

Surprisingly, our experimental results showed that Day 7 (Figure 5B and F) had better between-group separation than Day 10 (Figure 5C and G). As far as we are concerned, it may be because ulcerative colitis tends to heal on its own. On the Day 7 (Figure 5B and F), the content of some metabolites in the feces of ulcerative colitis mice had begun to turn to normal, with similar changes with the Day 10 (Figure 5C and G), resulting in a less obvious separation trend on the 10th day. However, on the Day 14 (Figure 5D and H), the therapeutic effect of the SYD exceeded the self-healing effect of ulcerative colitis, leading to a very obvious separation trend.

QC samples were used to evaluate the stability and reproducibility of metabolomics, and the results revealed that the instrument had high stability and the method was reproducible. Simultaneously, the PCA images of the SYD group at Day 0, Day 7, Day 10 and Day 14 (Figure 6) indicated that the samples at the four time points were well separated as well.

Figure 6 PCA score plots of feces metabolic profiling in SYD group during the day before the experiment, day 7, day 10 and day 14. (A) Fecal samples of SYD group detected in positive mode. (B) Fecal samples of SYD group detected in negative mode.

The PCA plots (Figure 7) show that the serum samples in positive ion mode, liver and urine samples in each group can be separated, suggesting that the metabolites in the above tissues in UC mice were different from those in normal tissues and can be regulated by SYD. It can be seen in the figure that most of the QCs were clustered together, indicating that the instrument ran stably and the data was reliable during the analysis process, and the differences in the metabolic profiles obtained in the experiment can reflect the biological differences between the samples. It is worth noting that the three groups of PCA plots in serum negative ion mode did not manifest a clear separation trend. Figure S3 indicates the representative TIC plots of each group, respectively. These results indicated that the endogenous metabolic spectrum of serum, urine, feces and liver in mice with ulcerative colitis has significantly changed compared with that of normal mice, and has obvious changes after SYD treatment.24

Figure 7 PCA score plots of livers, serum and urine metabolic profiling between normal, model and SYD group. (A) Liver samples detected in positive mode. (B) Liver samples detected in negative mode. (C) Serum samples detected in positive mode. (D) Serum samples in negative mode. (E) Urine samples detected in positive mode. (F) Urine samples detected in negative mode.

A supervised multivariate statistical analysis was performed using OPLS-DA plot to screen for differential metabolites. In the OPLS-DA diagrams of most samples, there was a clear separation between NC group and UC group, as well as UC group and SYD group (Figures 8–11). P-value was tested with the CV-ANOVA method. The PLS-DA plots are shown in Figures S4–S7.

Figure 8 Metabolite changes in fecal samples. (A and B) OPLS‐DA plot of the normal and model groups in positive mode. (C and D) OPLS‐DA plot of the SYD and model groups in positive mode. (E and F) OPLS‐DA plot of the normal and model groups in negative mode. (G and H) OPLS‐DA plot of the SYD and model groups in negative mode. (B, D, F and H) Validation of the corresponding OPLS‐DA model by the 200‐time permutation test.

Figure 9 Metabolite changes in liver tissues. (A and B) OPLS‐DA plot of the normal and model groups in positive mode. (C and D) OPLS‐DA plot of the SYD and model groups in positive mode. (E and F) OPLS‐DA plot of the normal and model groups in negative mode. (G and H) OPLS‐DA plot of the SYD and model groups in negative mode. (B, D, F and H) Validation of the corresponding OPLS‐DA model by the 200‐time permutation test.

Figure 10 Changes in sera metabolites. (A and B) OPLS‐DA plot of the normal and model groups in positive mode. (C and D) OPLS‐DA plot of the SYD and model groups in positive mode. (E and F) OPLS‐DA plot of the normal and model groups in negative mode. (G and H) OPLS‐DA plot of the SYD and model groups in negative mode. (B, D, F and H) Validation of the corresponding OPLS‐DA model by the 200‐time permutation test.

Figure 11 Changes in urine metabolites. (A and B) OPLS‐DA plot of the normal and model groups in positive mode. (C and D) OPLS‐DA plot of the SYD and model groups in positive mode. (E and F) OPLS‐DA plot of the normal and model groups in negative mode. (G and H) OPLS‐DA plot of the SYD and model groups in negative mode. (B, D, F and H) Validation of the corresponding OPLS‐DA model by the 200‐time permutation test.

From Table 7, P-value was less than 0.05 in most of the models, demonstrating that the model is valid. However, the P value of the serum samples of the SYD group vs UC group was more than 0.05 in the positive ion mode in both PLS-DA model and OPLS-DA model, which may be due to the fact that the sample size was so small as to reduce the test efficiency. It can be known from the 200 permutation tests that the R2 and Q2 values of different groups (Table 7) demonstrated that the PLS-DA model and the OPLS-DA model were not overfitted, and the results had high accuracy, indicating that this analysis was reliable and effective.

Next, in order to dig out the different metabolites in each tissue, we combined the positive and negative ion mode to analyze each sample separately.

A total of 124 metabolites with significant difference (VIP>1, Q-value<0.05, FC(NC/UC)>2 or <0.5, Table S1) were found in fecal samples between NC group and UC group, among which 58 metabolites were up-regulated (FC(NC/UC)<0.5) and 66 metabolites were down-regulated (FC(NC/UC)>2). These altered metabolites were subjected to pathway analysis in MetaboAnalyst (Figure 12A, Table 8), based on impact>0.1, we discovered that 2 metabolic pathways were significantly affected, including glycerophospholipid metabolism and sphingolipid metabolism.

Figure 12 Summary of pathway analysis of differentially expressed metabolites in feces (A), livers (B), serum (C) and urine (D). (A) a-Glycerophospholipid metabolism; b-Sphingolipid metabolism. (B) a-Pentose phosphate pathway. (C) a-alpha-Linolenic acid metabolism; b-biosynthesis of unsaturated fatty acids (P<0.05); c-linoleic acid metabolism (P<0.05). (D) a-Phenylalanine, tyrosine and tryptophan biosynthesis; b-Tyrosine metabolism; c-Arachidonic acid metabolism; d-Pantothenate and CoA biosynthesis.

There were 59 significantly different metabolites (VIP>1, Q-value<0.05, FC(SYD/UC)>2 or <0.5) in fecal samples between UC group and SYD group, of which 17 were up-regulated (FC(SYD/ UC)>2), 42 down-regulated (FC(SYD/UC)<0.5).

By comparing the fecal differential metabolites between the NC/UC and the SYD/UC, 46 metabolites were found to be common to these groups, of which 24 were first up-regulated (FC(NC/UC)<0.5) and then down-regulated (FC(SYD/UC)<0.5). Four metabolites were first down-regulated (FC(NC/UC)>2) and then up-regulated (FC(SYD/UC)>2). At the same time, among the 124 differential metabolites in NC/UC, 11 metabolites had FC(SYD/UC) between 0.5 and 2, and 10 metabolites had the same trend as NC group. It was suggested that these metabolites also had the potential to play a role in SYD therapy.

Majority (65.8%) of these 38 differential metabolites (Table 9) belonged to lipids and lipid-like molecules, and a heat map (Figure 13A) was used to visualize the changes of these metabolites. The metabolic profiles of most candidate metabolites were reversed in SYD group, suggesting that SYD can alleviate metabolic disorders, and this effect was mainly through the down-regulation of metabolites.

Figure 13 The heat maps of the results of differential metabolites in feces (A), livers (B), serum (C), and urine (D) analyzed by untargeted metabolomics with UPLC-Q-TOF-MS.

The main metabolites related to pathways were

PC(18:0/20:4(5Z,8Z,11Z,14Z)), sphingosine 1-phosphate (S1P) and sphinganine, whose contents were all down-regulated after SYD treatment. PC(18:0/20:4(5Z,8Z,11Z,14Z)) was related to glycerophospholipid metabolism, linoleic acid metabolism, α-linolenic acid metabolism and arachidonic acid metabolism, while S1P and sphinganine were involved in sphingolipid metabolism.

In liver samples, 8 significantly different metabolites were observed between NC and UC groups (VIP>1, Q-value<0.05, FC(NC/UC)>2 or <0.5, Table S2, Table 10), 5 were up-regulated and 3 were down-regulated. The heat map was displayed in Figure 13B. Pathway analysis (Figure 12B, Table 8) indicated that these differential metabolites mainly affected pentose phosphate pathway (impact>0.1). Two of the above metabolites were reversed (VIP>1, Q-value<0.05) after SYD therapy. Among them, PC(16:1(9Z)/18:2(9Z,12Z)) belonged to glycerophospholipids in lipids and lipid-like molecules, mainly related to linoleic acid metabolism, α-linolenic acid metabolism, arachidonic acid metabolism and glycerol phospholipid metabolism.

In serum samples, there were 26 significantly different metabolites between NC group and UC group (VIP>1, Q-value<0.05, FC(NC/UC)>2 or <0.5, Table S3), of which 19 were up-regulated and 7 were down-regulated. Pathway analysis (Figure 12C, Table 8) revealed that these differential metabolites mainly affected α-linolenic acid metabolism (impact>0.1, P<0.05), biosynthesis of unsaturated fatty acids (P<0.05) and linoleic acid metabolism (P<0.05). After SYD therapy, 14 of the above metabolites were reversed (VIP>1, P-value<0.05, Table 11), of which 10 were down-regulated (1 with FC(SYD/UC)<0.5), 4 were up-regulated. The heat map is shown in Figure 13C, and the vast majority (71.4%) of them belonged to lipids and lipid-like molecules. The metabolite related to the pathway was 8,11,14-Eicosatrienoic acid, which belonged to fatty acyls in lipids and lipid-like molecules, and mainly affects biosynthesis of unsaturated fatty acids.

In the urine samples, 244 metabolites (VIP>1, Q-value<0.05, FC(NC/UC)>2 or <0.5, Table S4) were significantly different between NC group and UC group, among which 82 were up-regulated and 162 were down-regulated. Pathway analysis (Figure 12, Table 8) indicates that, based on impact>0.1, a total of 4 metabolic pathways were significantly affected, involving phenylalanine, tyrosine and tryptophan biosynthesis, tyrosine metabolism, arachidonic acid and pantothenate and CoA biosynthesis.

There were 163 significantly different metabolites in UC group and SYD group (VIP>1, Q-value<0.05, FC(SYD/UC)>2 or <0.5), of which 155 were up-regulated (FC(SYD/UC)>2), 8 down-regulated (FC(SYD/UC)<0.5).

By comparing the significantly different metabolites between NC/UC and SYD/UC in the urine samples, 19 metabolites were found to be common to these groups, of which 4 were up-regulated first (FC(NC/UC)<0.5) and then down-regulated (FC(SYD/UC)<0.5), 4 metabolites were down-regulated first (FC(NC/UC)>2) and then up-regulated (FC(SYD/UC)>2). The heat map (Figure 13D) was applied to visualize the changes of these 8 metabolites, among which 3 belonged to lipids and lipid-like molecules, 4 belonged to organic acids and derivatives, and 1 belonged to phenylpropanoids and polyketides. Their specific information is displayed in Table 12. The sensitivity and specificity of the above differential metabolites were evaluated by ROC curve, and the area under curve (AUC) was calculated to test the accuracy of ROC diagnostic test results. When AUC ≥ 0.7, the metabolite had moderate or higher diagnostic value, and AUC ≥ 0.9 indicates that the metabolite has high diagnostic value. Figures 14 and 15 are the ROC curve analysis results and histograms of some differential metabolites.

Figure 14 Histogram and the corresponding ROC curves of sphingosine 1-phosphate (S1P) (A–C) and sphinganine (D–F), PC(18:0/20:4(5Z,8Z,11Z,14Z)) (G–I) and PC(16:1(9Z)/18:2(9Z,12Z)) (J–L). (B) ROC curve between model group and normal group for S1P. (C) ROC curve between model group and SYD group for S1P. (E) ROC curve between model group and normal group for sphinganine. (F) ROC curve between model group and SYD group for sphinganine. (H) ROC curve between model group and normal group for PC(18:0/20:4(5Z,8Z,11Z,14Z)). (I) ROC curve between model group and SYD group for PC(18:0/20:4(5Z,8Z,11Z,14Z)). (K) ROC curve between model group and normal group for PC(16:1(9Z)/18:2(9Z,12Z)). (L) ROC curve between model group and SYD group for PC(16:1(9Z)/18:2(9Z,12Z)). **p<0.01, ***p<0.001, ****p<0.0001, compared to normal group; #p<0.05, ##p<0.01 ###p<0.001, ####p<0.0001, compared to model group.

Figure 15 Histogram and the corresponding ROC curves of 8,11,14-Eicosatrienoic acid (A–C), PI(16:0/16:1(9Z)) (D–F) and apigenin (G–I). (B) ROC curve between model group and normal group for 8,11,14-Eicosatrienoic acid. (C) ROC curve between model group and SYD group for 8,11,14-Eicosatrienoic acid. (E) ROC curve between model group and normal group for PI(16:0/16:1(9Z)). (F) ROC curve between model group and SYD group for PI(16:0/16:1(9Z)). (H) ROC curve between model group and normal group for apigenin. (I) ROC curve between model group and SYD group for apigenin. #p<0.05, ##p<0.01 ###p<0.001, compared to model group.

To sum up, the metabolomics results proved metabolic alterations in the feces, livers, serum and urine of mice with UC. SYD mainly relieves UC symptoms and reduces metabolic disorders by down-regulating metabolites. There were 61 potential endogenous markers, involving 38 in feces, 2 in liver, 14 in serum and 8 in urine, of which 7-Ketodeoxycholic acid coexisted in fecal and serum samples. Most of these differential metabolites belong to lipids and lipid-like molecules. Furthermore, glycerophospholipid metabolism, sphingolipid metabolism and biosynthesis of unsaturated fatty acids were the main metabolic pathways for SYD to play a therapeutic role.

Network Pharmacology

To further explore the mechanism by which SYD treats UC, we collected 393 UC-related primary targets from the OMIM, TTD, and GeneCards databases using a network pharmacology approach. Later, in the TCMSP database, SYD contains 129 active ingredients in total, involving 4 in Paeoniae Radix Alba, 2 in Angelicae Sinensis Radix, 9 in Coptidis Rhizoma, 28 in Scutellariae Radix, 83 in Glycyrrhizae Radix et Rhizoma, 4 in Arecae Semen, 3 in Aucklandiae Radix, 6 in Rhei Radix et Rhizome and 0 in Cinnanmomi Cortex. The targets were predicted by the DrugBank database and UniProt database.2 Eventually, 2541 targets, containing 101 in Paeoniae Radix Alba, 56 in Angelicae Sinensis Radix, 239 in Coptidis Rhizoma, 431 in Scutellariae Radix, 1545 in Glycyrrhizae Radix et Rhizoma, 35 in Arecae Semen, 39 in Aucklandiae Radix and 95 in Rhei Radix et Rhizome.

After matching UC targets and active ingredient targets, a total of 15 potential targets related to SYD in the treatment of UC were identified (Figure 16A). All intersecting target names were converted into gene symbols in the UniProt database.

Figure 16 Venn diagram and various networks. (A) Venn diagram showing the intersection of UC-related genes and SYD-related genes. (B) The compound-target network. The circles represent the active components, in which red represents Paeoniae Radix Alba, Orange represents Rhei Radix et Rhizome, yellow represents Glycyrrhizae Radix et Rhizoma, green represents Scutellariae Radix, purple represents common component, and blue squares represent genes. The line between two nodes represents the interaction, and the size of each node indicates the number of connections. (C) The PPI network of 15 potential targets. The line width between two nodes represents the interaction. (D) The target-pathway network. Orange squares represent genes and red arrows represent KEGG pathways. This network shows the relationship between the enriched 20 pathways and 13 genes, and the size of the graph shows the number of pathways or genes connected.

There were 13 active components corresponding to potential targets. In order to better show the complex relationship between SYD active components and potential UC targets, we constructed an active component-target network (Figure 16B). According to degree, the top four were quercetin, wogonin, kaempferol, baicalein, and the basic information of these components is manifested in Table 13.

Furthermore, in order to identify the key targets of SYD in the treatment of UC, we constructed PPI network (Figure 16C) by using String database and Cytoscape. Taking degree greater than the average as the screening standard,28 STAT3, IL1B, IL6, IL2, AKT1, IL4, ICAM1 and CCND1 were screened as key targets (Table 14).

Table 14 The Potential Hub Proteins

For the sake of understanding the ability of potential targets to resist UC symptoms, we performed KEGG enrichment analysis (Figure 17A) and GO enrichment analysis (Figure 17B) on these targets. The pathways affected significantly in KEGG enrichment analysis were JAK-STAT signaling pathway, Th17 cell differentiation, C-type lectin receptor signaling pathway and TNF signaling pathway. These results indicated that SYD probably acted on UC in a multi-target, multi-pathway, and overall integrative mode. In addition, the signaling pathways related to UC were extracted from KEGG database, and target-pathway network was generated by using Cytoscape software (Figure 16D).

Figure 17 GO and KEGG analysis. (A) The top 20 KEGG pathways. The size of each node indicates enriched counts, the abscissa represents the enriched gene ratio, and color means the enrichment degree according to the -log10(Q value). (B) GO enrichment analysis of SYD targets in treating UC. The horizontal axis represents the number of genes enriched in each item, and the color represents BP and MF respectively.

In terms on GO analysis (Figure 17B), there were several pathways involving cytokine activity, cytokine receptor binding, receptor ligand activity and signaling receptor activator activity in molecular function (MF), as well as adaptive immune response, leukocyte cell–cell adhesion, regulation of leukocyte cell–cell adhesion, T cell activation, positive regulation of cell–cell adhesion, positive regulation of leukocyte cell–cell adhesion and extrinsic apoptotic signaling pathway in biological process (BP).

Molecular Docking

To further study the possibility of the interaction between the active components of SYD and the key targets of UC, we screened 4 key components, involving quercetin, wogonin, kaempferol, baicalein by degree in Cytoscape, and obtained 5 key targets by searching the RCSB Protein Data Bank database.24

Through molecular docking, the binding energy between small ligand molecules and receptor protein macromolecules is calculated to predict the affinity between them.2 In this study, molecular docking confirmed that the active compounds of SYD had significant regulatory effects on STAT3 (Figure S8A–D), IL1B (Figure S8E–H), IL6 (Figure S9A–D), IL2 (Figure S9E–H) and AKT1 (Figure S