Molecular Mechanism of The Effect of Huanglian Jiedu Decoction On Ulcerative Colitis Based On Network Pharmacology And Molecular Docking

Huanglian jiedu decoction (HLJDD) is a heat-clearing and detoxifying agent composed of four kinds of Chinese herbal medicine. Previous studies have shown that HLJDD can improve the inammatory response of ulcerative colitis (UC) and maintain intestinal barrier function. However, its molecular mechanism is not completely clear. In this study, we veried the bioactive components (BCI) and potential targets of HLJDD in the treatment of UC by means of network pharmacology and molecular docking, and constructed the pharmacological network and PPI network. Then the core genes were enriched by GO and KEGG. Finally, the bioactive components were docked with the key targets to verify the binding ability between them. A total of 54 active components related to UC were identied. Ten genes are considered to be very important to PPI network. Functional analysis showed that these target genes were mainly involved in the regulation of cell response to different stimuli, IL-17 signal pathway and TNF signal pathway. The results of molecular docking showed that the active components of HLJDD had good anity with Hub gene. This study systematically elucidates the "multi-component, multi-target, multi-pathway" mechanism of anti-UC with HLJDD for the rst time, suggesting that HLJDD or its active components may be candidate drugs for the treatment of ulcerative colitis.


Introduction
Ulcerative colitis (UC) is a chronic idiopathic colonic in ammatory bowel disease characterized by intestinal in ammation, mucosal injury and brosis, which can cause different degrees of super cial mucosal in ammation from rectum to oral cavity. The pathological features of UC are in ammatory and ulcerative lesions of mucosa and submucosa [18]. The pathogenesis is multifactorial, involving genetic susceptibility, epithelial barrier defects, immune response disorders and environmental factors [56]. The main manifestations of ulcerative colitis are hematochezia, abdominal pain and diarrhea. Endoscopic biopsy can con rm the diagnosis. The best goal of UC treatment is a sustained and lasting non-steroid remission period, reducing hospitalization and surgery, and minimizing cancer risk [21; 50].
Clinical treatment of ulcerative colitis is di cult, including 5-aminosalicylic acid, steroids, immunosuppressants and biological agents [44]. However, these drugs have many side effects, including headache, nausea, vomiting, diarrhea, nephrotoxicity and increased risk of infection, which affect the normal work and life of patients. Even some patients did not respond to these drugs [49]. Therefore, it is particularly urgent to nd more reliable and effective therapeutic drugs.
Huanglian jiedu decoction (HLJDD), which is composed of four kinds of Chinese herbal medicines, such as Huanglian (HL), Huangbai (HB), Zhizi (ZZ)and Huangqin (HQ), is a representative medicine for heatclearing and detoxi cation of traditional Chinese medicine (TCM). It has been widely used in the treatment of gastrointestinal diseases, sepsis and Alzheimer's disease in China [9; 11; 25; 26]. In addition, recently, associated pharmacological studies show that Huanglian jiedu decoction could function on type II diabetes, in ammatory disease and ischemic stroke [27; 31; 38; 48; 58; 60]. Huanglian jiedu decoction has signi cant anti-in ammatory effects. In 1999, studies found that HLJDD can reduce experimental colitis in rats [68]. Subsequently, some studies have con rmed that HLJDD can reduce in ammation and intestinal mucosal injury by down-regulating the expression of JAK2 and STAT3, and alleviated DSSinduced colitis in mice [41]. In addition, it can also repair intestinal barrier damage and alleviate ulcerative colitis induced by DSS in mice by regulating NF-κB and NRF2 signal pathway [63]. Due to the multicomponent, multi-pathway and multi-target characteristics of traditional Chinese medicine prescription, the mechanism of HLJDD in the treatment of UC remains to be revealed systematically.
Network pharmacology was rst proposed by Andrew Hopkins in 2007 [29]. It is a new cross-discipline based on the theory of systems biology to study the mechanism of drug action and design multi-target drug molecules at the system level. TCM Network Pharmacology is a new research paradigm, which aims to transform traditional Chinese medicine from experience-based medicine to evidence-based medicine.
We can analyze the interaction network of "traditional Chinese medicine-compound-protein / genedisease" from the perspective of system biology, so as to understand the effect of traditional Chinese medicine on disease. TCM Network Pharmacology has also been successfully used to identify the elucidating mechanism of HLJDD in the treatment of pneumonia [37]. Molecular docking is an established method based on computer simulation structure, which helps to predict the interaction between molecules and biological targets. Through molecular docking, we can verify the binding ability between active compounds and key targets and improve the accuracy of the target network [47].
In this study, we used the methods of network pharmacology and molecular docking to study the potential mechanism of HLJDD in the treatment of UC. Firstly, HLJDD bioactive chemical components (BCI) and related target genes were searched and screened in TCMSP, and UC speci c genes were obtained in GEO database. Then, the UC speci c genes related to the BCI of HLJDD were screened by compound-target interaction analysis. Differential expression analysis was used to verify the expression of key target genes in UC and normal samples. Construction of TCM-BCIs-targets pharmacological network and PPI network, screening of top10 gene in PPI network as an important core target of this study. Then the target genes were analyzed by GO and KEGG enrichment analysis. Finally, the binding ability of bioactive components to key targets was veri ed by molecular docking with AutoDock Tools.
The ow chart of the analysis program we studied is as follows (Fig. 1 in supplementary table S1). The corresponding targets of HB, HL, HB and ZZ ltered out by TCMSP were 77,159,96 and 171 respectively. After removing the repeated targets, a total of 205 human target proteins were obtained. Draw the traditional Chinese medicine-active ingredients-targets network of HLJJD (Fig. 2). The network consists of 274 nodes (including 205 target nodes, 65 compound nodes and 4 traditional Chinese medicine nodes) and 1337 edges, in which the square node represents the active compound and the traditional Chinese medicine, and the circular node represents the related target. The solid blue lines represent traditional Chinese medicine and the corresponding active ingredients. The gray dotted line represents the interaction between the active compound and the target protein. Among these targets, quercetin (MOL000098), kaempferol (MOL000422) and wogonin (MOL000173) correspond to 50 and 38 targets respectively. Beta-sitosterol (MOL000358) and stigmasterol (MOL000449) correspond to 26 targets. In addition, quercetin, beta-sitosterol and stigmasterol are shared BCI of many drugs. This shows that there are many active ingredients in different drugs, the same ingredients may exist in many kinds of traditional Chinese medicine, and the compounds in the formula of HLJDD may play a role in some common goals, which is an important basis for the multi-component and multi-target action of traditional Chinese medicine, so that HLJDD can play a pharmacological role in UC and other diseases.
UC speci c genes related to BCI of HLJDD In this study, 108 mucosal biopsy samples were included, including 87 ulcerative colitis tissue samples and 21 normal tissue samples. The GSE87466 is analyzed by Rstudio software, and then the DEG set is determined. According to the truncation criteria, a total of 424 genes were identi ed, of which 279 were up-regulated and 145 were down-regulated (Supplementary table S2). As shown in the gene volcano map, the differential genes in the disease samples showed normal distribution, and the genes with signi cant differences were highlighted in the map (Fig. 3A). After crossing 205 BCI-related targets with 424 UCspeci c genes, 21 genes were obtained as UC-speci c HLJDD target genes (Fig. 3B). Among these genes, MMP3, DUOX2, CXCL8, IL1 β, CXCL2, CXCL11, MMP1, MMP9, PTGS2, CXCL10, SPP1, CCL2, PLAU, ICAM1, THBD, NOS2 and SELE were up-regulated in UC samples, while ABCG2, PPARG, CYP2B6 and ADH1C were down-regulated in UC samples, and the difference was signi cant (Tabel1, Fig. 3C). After excluding the BCI which is not targeted at the speci c target of UC, 54 effective BCI are nally used in the construction of pharmacological network.
Construction of TCM-BCIs-targets Pharmacological Network, PPI Network and screening of Core targets By comparing the targets of HLJDD and UC, we get a total of 21 targets of HLJDD and UC. According to the common targets, 54 kinds of active components were found (Supplementary table S3). The TCMcompounds-common targets network was constructed by Cytoscape software (Fig. 4A). The network consists of 79 nodes (including 21 target nodes, 54 compound nodes and 4 traditional Chinese medicine nodes) and 181 edges, in which the square node represents the active compound and the traditional Chinese medicine, and the circular node represents the related target. Four key active components with degree ≥ 6 were identi ed as quercetin (MOL000098), wogonin (MOL000173), kaempferol (MOL000422) and stigmasterol (MOL000449).
The protein-protein interaction network (PPI) analysis of ulcerative colitis target was carried out by using String database online service platform, and all the protein-protein interaction relationships were obtained, and then imported into Cytoscape software for visualization (Fig. 4B). Then the Cytohubba plug-in is used to identify the Hub gene, and the top 10 nodes in the network generated by the MNC method are regarded as the Hub genes: CXCL8, CCL2, ICAM1, IL-1β, MMP9, PTGS2, MMP1, SELE, CXCL10and MMP3 (Fig. 4C). The PPI network with 10 Hub genes has 10 nodes and 45 edges, the average node degree is 9, and the p value is 5.41E-07. The higher the con dence score, the larger the node size and the darker the color. According to the active components related to hub gene, a disease Hub genescompounds-TCM network was constructed (Fig. 4D). The most active ingredient of traditional Chinese medicine in the action of HLJDD on UC is HQ, which contains 28 active components that can target UCspeci c genes. PTGS2 is targeted by 54 BCI. As an important gene, it is highlighted in the gure that the key active components in the network are quercetin (MOL000098), wogonin (MOL000173), kaempferol (MOL000422). A total of 35 pathways were enriched by 10 target genes (supplementary table S7). The top 20 KEGG analyses were shown by bubble chart (Fig. 6C). Loop maps are used to represent signal pathways and related gene self-networks (Fig. 6D). KEGG enrichment analysis showed that the key targets were mainly related to immune response, in ammatory reaction and disease process. In terms of immune and in ammatory responses, KEGG pathways are classi ed as IL-17 signaling pathway (hsa04657), TNF signaling pathway (hsa04668) and NF-κ B signaling pathway (hsa04064). In terms of involvement in the disease process, the categories of KEGG pathways are lipid and atherosclerosis (hsa05417), rheumatoid arthritis (hsa05323), malaria (hsa05144), age-RAGE signaling pathway in diabetic complications (hsa04933) and coronavirus disease-COVID-19 (hsa05171).

Molecular docking
Quercetin, wogonin and kaempferol in the Hub genes-compounds-TCM network are selected as receptors, these active components target Hub gene most, and their targeted Hub gene is used as ligands. Then the receptors and ligands are docked to verify their binding activity. The stability of binding between receptors and ligands depends on the binding energy. If the binding energy is less than − 5kJ/mol, it indicates that the target has a certain binding activity with the compound, and the lower the binding energy, the more stable the binding conformation of the receptor and ligand [20]. The docking results were visualized by Pymol software (Fig. 7). The docking of hub targets and active components is mainly realized by Hydrogen Bond, Hydrophobic Interaction, π-Cation Interaction and π-Stacking (parallel). The results of docking showed that the active component of HLJDD had good binding ability to hub gene (Tabel2).

Discussion
Ulcerative colitis (UC)is a chronic idiopathic colonic in ammatory bowel disease. Because its pathogenesis is multi-factor and has the characteristics of recurrence and remission, it makes the treatment of ulcerative colitis more di cult. Clinical treatment is mainly drug therapy, including 5aminosalicylic acid, steroids, immunosuppressants and biological agents. However, the unnecessary side effects of these drugs affect the normal work and life of patients. Even some patients do not respond to these drugs. Due to the characteristics of natural sources, multi-components, multi-pathways and multitargets, traditional Chinese medicine prescription has advantages in the treatment of complex diseases, especially in the treatment of diseases with poor response to western medicine alone. Huanglian jiedu decoction (HLJDD) is composed of HL, HB, ZZ and HQ, which is the representative medicine of heatclearing and detoxi cation of traditional Chinese medicine. HLJDD has signi cant anti-in ammatory effects, and many other scholars' studies are consistent with our ndings as early as 1999, when it was found that HLJDD can reduce experimental colitis in rats [68]. Some studies have also con rmed that HLJDD can improve DSS-induced colitis and enhance intestinal barrier function in mice [41; 63]. In our study, we systematically discussed the molecular mechanism of HLJDD in the treatment of UC by using the methods of network pharmacology and molecular docking, so as to provide a more powerful theoretical basis for clinical treatment.
We collected 85 active components of HLJDD through TCMSP database, of which 54 were related to UC targets. Nine BCI are associated with multiple target genes in the pharmacological network. Quercetin (MOL000098), wogonin (MOL000173) and kaempferol (MOL000422) were identi ed as active components related to most targets. Among the 21 HLJDD targets for UC, 18 targets are involved in quercetin (MOL000098), and kaempferol (MOL000422) and wogonin (MOL000173) correspond to 7 and 5 targets respectively. The results of molecular docking also proved that they have good binding ability with key target genes. Quercetin is a plant-derived polyphenol compound, and its anti-in ammatory, antioxidant and anti-tumor activities have been reported [10]. Quercetin can reduce the secretion of a variety of in ammatory cytokines in mouse bone marrow-derived dendritic cells (BMDC) [15]. Some studies have shown that quercetin can inhibit LPS-induced in ammation in wild-type organoids and spontaneous in ammation in ulcerative colitis-organoids. Organoids were produced from Winnie, a mouse model of ulcerative colitis [17].In addition, the anti-in ammatory effect of quercetin is also associated with the decreased expression of Chand EBP-β, a transcriptional factor that can trigger the expression of various in ammatory mediators [1]. Wogonin is a naturally occurring avonoid, and there has been a lot of evidence that wogonin has the role of anti-in ammatory and antioxidant stress [12; 69]. Cartilage protection is achieved by inhibiting molecular events involved in oxidative stress, in ammation and matrix degradation in osteoarthritis chondrocytes and cartilage explants [32]. Wogonin has been shown to inhibit in ammation-related colorectal cancer by inhibiting NF-kappa B and activating Nrf2 signaling pathway [61]. Kaempferol is an effective anti-in ammatory agent, and it has been shown that kaempferol can protect mouse colonic mucosa from DSS-induced UC [45]. These ndings indicate the potential of HLJDD in the treatment of UC.
Network pharmacological analysis showed that 21 target genes in UC were regulated by HLJDD, including MMP3, DUOX2, CXCL8, IL1 β, CXCL2, CXCL11, MMP1, MMP9, PTGS2, CXCL10, SPP1, CCL2, PLAU, ICAM1, THBD, NOS2, SELE, PPARG, ADH1C, CYP2B6 and ABCG2. We screened the top 10 Hub genes: CXCL8, CCL2, ICAM1, IL-1β, MMP9, PTGS2, MMP1, SELE, CXCL10, MMP3 through the Cytohubba plug-in of Cytoscape software. CXCL8 secreted by immune cells is a chemokine that attracts neutrophils that produce CXCR1 [+] CXCR2 [+] IL-23 to in ltrate and accumulate in in amed colon tissue [34]. Studies have con rmed that there is excessive secretion of chemokine CXCL8 in colitis and colitis-associated cancer (CAC) [54]. It has been reported that the increase of in ammatory macrophages in IBD patients who did not respond to anti-TNF-α therapy is associated with the upregulation of the TREM-1/CCL7/CCR2 axis [22]. In addition, ICAM1 has been shown to be a candidate biomarker of UC activity [59]. Matrix metalloproteinases (MMP) can activate cytokines and release isolated growth factors to initiate, amplify or down-regulate signaling cascades involved in growth and in ammation. Overexpression and activation of MMP can cause colonic mucosal injury and in ammation [57]. Studies have shown that MMP1 overexpression is associated with the initial steps of mucosal in ammation and ulcers in UC[4; 16], MMP9 activity is associated with the production and persistence of UC in ammatory state[6; 33], and MMP3 has been shown to play an important role in T cells and TNF-α mediated intestinal injury [46]. Studies have found that IL-1β, MMP9 and CXCL10 may be candidate biomarkers of UC-related cancer [40; 53; 66]. In the pharmacological network, PTGS2 (Cyclooxygenase-2, COX-2) is targeted by 54 BCI. Studies have shown that PTGS2 is involved in the development of colitis and CAC[28].
Inhibition of COX-2 can reduce the expression of pro-in ammatory mediators and reduce the symptoms and pathological features of UC in mouse models [39].
We did KEGG enrichment analysis of these 10 Hub genes, and the results showed that the key targets were mainly related to immune response, in ammatory reaction and disease process. The therapeutic effect of HLJJD on UC may be achieved by regulating the immune system and in ammation-related pathways. IL-17signalingpathway is one of the most important signal pathways. IL-17 is a T cell-derived cytokine produced by memory CD4 + and CD8 + T cells. The pro-in ammatory property of IL-17 is the key to its host protection ability, but the unrestricted IL-17 signal transduction is related to immunopathology, autoimmune diseases and cancer progression [3]. IL-17 cytokines play a key role in the pathogenesis of IBD. IL-17 can stimulate the expression of pro-in ammatory cytokines in human cells, and the expression of mucosal IL-17 and serum IL-17 levels in patients with active IBD are increased [19]. In the mouse colitis model induced by acute trinitrobenzenesulfonic acid (TNBS), it was found that IL-17 was produced in 24hour and 48-hour colon tissue. IL-17R KO mice showed less severe in ammation in response to acute TNBS treatment and signi cantly protected against TNBS-induced weight loss [67]. A meta-analysis showed that serum IL-17 levels were signi cantly correlated with the severity of UC [35]. Tumor necrosis factor (TNF) is a pro-in ammatory mediator that can up-regulate the production of reactive oxygen species (ROS) and reactive nitrogen (RNS) and aggravate cell damage [7]. Proin ammatory cytokine TNFα plays a key role in coordinating the in ammatory cascade of chronic intestinal in ammation. After binding to type 1 TNF receptor (TNFR1) and type 2 (TNFR2) receptor, TNF-α activates MAPK and NF-κB pathway and initiates pro-in ammatory signal [14]. MAPK and NF-κB pathway can induce cell proliferation, differentiation and up-regulation of many proin ammatory cytokines (including TNF-α) during in ammation. In addition, the combination of TNF-α and TNFR1 can also induce intestinal epithelial cell apoptosis [51]. NF-κB is the core regulator of in ammatory response. NF-κB is signi cantly induced in IBD patients. NF-κB signal pathway can induce the expression of pro-in ammatory cytokines and lead to in ammation-related tissue damage [5; 43]. Some studies have shown that both classical and

Conclusion
Generally speaking, this study systematically elucidates the "multi-component, multi-target, multipathway" mechanism of HLJDD against UC for the rst time. quercetin, wogonin and kaempferol are the main active ingredients of HLJDD in the treatment of UC. CXCL8, CCL2, ICAM1, IL-1β, MMP9, PTGS2, MMP1, SELE, CXCL10 and MMP3 may be potential therapeutic targets of HLJDD in UC. The therapeutic effect of HLJJD on UC may be achieved by regulating the immune system and in ammation-related pathways. The method of network pharmacology can provide a new and systematic analysis method for the research of Chinese herbal medicine. But our research also has limitations. Focusing on validated targets may rule out some unveri ed potential targets, and our research needs to be veri ed by further clinical and basic research.

Acquisition and screening of HLJDD bioactive compounds
Clinical drug treatment is usually oral administration. Human oral bioavailability (OB) ≥ 30% of the compounds have good absorption and slow metabolism after oral administration. The compounds with drug-likeness (DL) ≥ 0.18 are chemically suitable for drug development. By using "Huanglian", "Huangbai", "Huangqin" and "Zhizi" as keywords, we screened the bioactive compounds of HLJDD in TCMSP (https://old.tcmsp-e.com/tcmsp.php). According to the drug screening criteria recommended by the TCMSP database, if OB ≥ 30% and DL ≥ 0.18, BCIs retained [37].

Targets related to bioactive compounds
According to the active component of HLJDD obtained by TCMSP, the target corresponding to this active component was searched. Using the UniProt database (https://www.uniprot.org/), all target proteins were converted into corresponding gene symbols of "Homo sapiens" species, and the corresponding target information was obtained . First of all, the Affy package in R studio software (version 4.0.5) is used to preprocess the original data [23]. Then, the background adjustment, standardization and logarithmic conversion of the expressed data are carried out by using the the robust multiarray average (RMA) algorithm [30]. Paired t-test based on microarray data linear model (LIMMA) packet in R was used to identify differentially expressed genes (DEG) between UC and normal samples. P value criteria < 0.01 and a | log2 fold-change (log2FC) | ≥ 1.5 were used as truncation criteria for follow-up analysis.
UC speci c genes related to BCI of HLJDD According to the related targets of active components of HLJDD obtained by TCMSP and UC speci c genes, the genes that are not only HLJDD related targets but also UC speci c genes are obtained. Then differential expression analysis was used to verify the expression of key target genes in UC and normal samples.

Network construction
The construction of pharmacological network Page 10/23 The corresponding active components are found according to the four traditional Chinese medicines of HLJDD, and these components are related to speci c genes of UC. UC speci c genes were obtained by GEO database. A TCM-compounds-targets network was constructed using Cytoscape software (version3.8.2) to illustrate the anti-UC regulation mechanism between the BCI of HLJDD and their speci c targets [52].
Construction of protein-protein interaction (PPI) network and screening of core targets Import HLJDD BCI-related UC speci c targets into the String database (https://string-db.org/). The species was set as "Homosapiens", and all protein-protein interaction relationships were obtained, and the con dence score was ≥ 0.4 [55]. Then it was imported into Cytoscape software to generate the PPI network diagram of UC regulated by the active components of Huanglian jiedu decoction. The Cytohubba plug-in was used to identify the hub gene, and the top 10 genes generated by the maximum neighborhood component (MNC) method were regarded as the hub gene [13; 37]. For each node in an interactive network, the degree represents the number of edges between one node and other nodes in the network, which measures the number of connections to other nodes and re ects the importance of the node [42].
Construction of disease core target-active ingredientstraditional Chinese medicine network According to the active components related to hub gene, a Hub targets-compounds-TCM network was constructed. It can be seen that HLJDD acts on the most active traditional Chinese medicine and active components of the core genes of the disease.

Gene Ontology (GO) Functional Annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway Analysis
The hub gene selected by MNC method was analyzed by GO functional enrichment analysis, that is, biological process (BPs), cellular component (CCs) and molecular function (MFs). In order to clarify the role of target proteins in signal transduction, KEGG pathway analysis was carried out. The enrichment analysis was carried out through the clusterPro ler software package of R platform, and corrected by Bonferroni method, the adjusted GO term and adjusted KEGG pathway were considered to be of great signi cance[62].

Compounds -Targets Molecular Docking
Semi-exible docking was carried out by AutoDock Tools (version4.2.6) to verify the binding ability of bioactive components to key targets. The speci c operations are as follows: (1) preparation of receptor molecules: the 3D structures of 10 Hub gene receptors were obtained from RCSBPDB database (https://www.rcsb.org/) and saved in pdb format. Then the ligand small molecules and water molecules of the protein were removed by PyMOL software (Version2.4.0), and the pdb format was saved. (2) preparation of ligand small molecules: select the top three active components with the largest number of targeted Hub gene in Hub genes-compounds-TCM network, download the 3D structure of ligand small molecules from ZINC (http://zinc.docking.org/) website, and save the mol2 format le. (3) Molecular docking: polar hydrogen and Gasteiger charges were added to the above receptors and ligands by AutoDock Tools (ADT) and stored in PDBQT format. Then use AutoGrid tool to set the parameters of the docking frame, select Lamarckian genetic algorithm (LGA) to nd the best docking conditions, and record the docking position and binding energy of receptors and ligands. Finally, the docking results are visualized by Pymol software.   Figure 1 The owchart of the analysis procedures of the study.   Circles are used to represent the nodes of the target protein, using red and blue to represent up-regulated and down-regulated genes in UC tissue, respectively. The edge is expressed according to the con dence score of the protein-protein interaction relationship, and the higher the score, the darker the color. (C) PPI networks of Hub gene. The higher the con dence score is, the larger the node size is, and the darker the color will be. (D) Hub gene-compounds-TCM network. Hub genes, active components, nodes and edge representations are shown in gure 4A. PTGS2 is highlighted as the gene with the most active components in the network.