All procedures and assays were approved by the Institutional Animal Care and Use Committee of Jining Medical University.

Using Traditional Chinese Medicine Systems Pharmacology database and Analysis Platform (TCMSP) (Ru et al., 2014), the active components were obtained according to the suggested parameter information. The cutoff was greater or equal to 0.18 (Drug likeness, DL) and 20% (Oral bioavailability, OB). The targets of each active component were transformed into gene symbols of Homo sapiens species by the UniProt knowledge database (www.uniport.org).

We collected colitis-related genes from five sources with the keyword “ulcerative colitis,” including GeneCards (Rebhan et al., 1997; Safran et al., 2010), DrugBank (Wishart et al., 2018), Online Mendelian Inheritance in Man (OMIM) (Hamosh et al., 2002), PharmGkb (Whirl-Carrillo et al., 2012), and Statistics of Therapeutic Target database (TTD) (Wang et al., 2020b) (Supplementary Table S1).

The shared genes between GCZX capsule and colitis were selected to build the HIT interaction network utilizing Cytoscape software (Su et al., 2014).

Utilizing the STRING database (Search Tool for Retrieval of Interacting Genes/Proteins) (http://string-db.org/) (Szklarczyk et al., 2019), a PPI network based on the shared genes was computed with a confidence score ≥0.7 for significance and without disconnected nodes. Network nodes refer to proteins; edges do not mean the physically binding of two proteins, but the protein-protein interactions that they jointly facilitate a function. The thicker the edge is, the higher the confidence is.

Hub genes with a high-degree node were calculated by CytoNCA according to six indices, including closeness centrality, betweenness centrality, degree centrality, eigenvector centrality, local average connectivity, and network centrality (Tang et al., 2015). Genes above the median value of each index were selected per calculation. These indices were calculated three times to establish the final sub-network.

Gene Ontology (GO) divides genetic functions into cellular component (CC), molecular function (MF), and biological process (BP) (Ashburner et al., 2000). Kyoto Encyclopedia of Genes and Genomes (KEGG) is a reference knowledge base for systematic interpretation of genes functions (Kanehisa and Goto, 2000). GO and KEGG enrichment analysis was executed by Bioconductor (R). Adjusted p-value below 0.05 was considered significantly enriched genes.

A murine model of experimental colitis was established utilizing C57 male mice that had free access to 4% DSS water for 5 days, and was gavaged twice a day with GCZX treatment (Figure 1F). HE staining showed that GCZX treatment dramatically relieved the epithelial structural collapse in UC mouse models (Figure 1A). Moreover, the severity of fecal blood and albumin, shortened colon length in UC mice were considerably alleviated after GCZX administration (Figures 1B–D); GCZX also reversed DSS-induced IL-17 and CXCL1 levels in mouse serum (Figure 1E), suggesting the intervention effect of GCZX capsule against experimental colitis.

In virtue of the TCMSP database, we obtained 138 active compounds targeting 2,482 genes. Zr yields five components targeting 54 genes, CPr yields 26 components targeting 400 genes, Ar yields 35 components targeting 797 genes, Mf yields nine components targeting 308 genes, CRr yields 49 components targeting 164 genes, Pp yields 14 components targeting 374 genes (Supplementary Table S1). Based on the acquired 5811 UC-relevant genes (Figure 2A, Supplementary Table S2), Venn diagram demonstrated that GCZX formula shared 208 putative targets with UC (Figure 2B). With the aim of exploring the pharmacological mechanisms of GCZX, a HIT network was constructed (Figure 2C). The light blue rectangle nodes forming the outside circle represented UC-related genes and the circle nodes inside were GCZX active components; each color indicated one herb. As illustrated, all the herbs not only worked synergistically targeting some common genes but also shared some active components.

As shown in Figure 3A, CRr , CPr, and Mf were responsible for 69% of targeted genes and considered the major components of GCZX formula. GO and KEGG analysis was performed with these genes to predict the mechanism of each herb combating UC (Figure 3B). Mf, the core herb in the formula, correlated with the regulation of oxidative stress and response to reactive oxygen species that both play a pivotal role in the pathogenesis of IBD; and with the response to cadmium ion that causes microbiota dysbiosis hence increases intestinal permeability (Liu et al., 2020); and with histone deacetylase activity that is involved in tumorigenesis. Moreover, Mf influenced virus infection and the neoplastic progression of colorectal cancer. CRr and CPr shared many common components and their targeted genes were similarly enriched in immune response to lipopolysaccharide and bacterial as well as virus, all of which could induce inflammation and infection. Other seemingly minor but vital herbs were Ar and Zr as well as Pp. Pp and Zr as well as Ar worked with neurological processes, including neuroactive ligand-receptor interaction and neurotransmitter receptor activity as well as postsynaptic membrane, which pointed towards the modulation of intestinal innervation by these herbs. Additionally, Ar, Pp, CRr, and CPr were associated with catecholamine activity including dopamine and adrenaline, both of which innervate gastrointestinal muscles whereby regulating gastrointestinal motility. Specifically, Zr and Ar are also related to regulation of vascular processes such as blood vessel size and diameter, indicating a favorable role in hemostasis. Furthermore, GCZX represented a strong capacity to suppress neoplastic progression including prostate cancer, small cell lung cancer, and colorectal cancer. Altogether, these herbs function synergistically to alleviate rectal bleeding, orchestrate enteric nervous system, and control host immune response.

Utilizing the intersected genes between UC and GCZX, we construct a full PPI network with a PPI enrichment p-value (<1.0e-16) based on the STRING database (Figure 4A). In agreement with the individual function of each herb, GO analysis showed that these genes were enriched in responses to oxygen, hypoxia, lipopolysaccharide, bacterial, nutrient levels as well as cellular response to drug. KEGG further corroborated the role of GCZX in cancer, inflammation, and infection.

Sub-network was established with 10 hub-genes with higher than the median values of the indices (Figure 5A). The median values of betweenness centrality, closeness centrality, degree centrality, eigenvector centrality, local average connectivity, and network centrality in the PPI network were 291.9072, 0.408736, 15.35052, 0.049987, 6.280781, 8.980268, respectively. The median values in the final subnetwork (right) were 10.77105, 0.880509, 21.5, 0.223418, 14.83785, 18.95284, respectively. As shown in Figure 5B, we selected the common-shared components of the herbs for subsequent in-vitro experiments, composed of (R)-Canadine, palmatine, coptisine, palmatine, quercetin, beta-sitosterol, Stigmasterol, EIC (-)-alpha-cedrene, and fumarine. To validate the existence of these compounds and their metabolites, the serum isolated from DSS-GCZX mice was collected and subjected to LC-MS analysis (Supplementary Figure S1; Supplementary Table S3).

Network pharmacology analysis showed the antibacterial effect of GCZX capsule, so we examined changes in microbiota composition and its metabolites. With the aim of identifying whether the sequencing amount was sufficient to evaluate the diversity of the original microbiota, α-diversity was calculated by coverage meaning the true state of the microbes, Chao and Ace reflecting the bacterial richness and species abundance (Grice et al., 2009), and Shannon and Simpson indices reflecting the diversity of original microorganisms (White et al., 2009).

As illustrated in Table1, all sample libraries covered above 99%, which indicated the library size of this study sufficed to cover the majority of the microbes. Rarefaction analysis illustrated that the number of OTUs in both groups reached saturation (Figure 6B). Among the observed OTUs, GCZX treatment restored DSS-reduced bacterial diversity of UC mice without influencing the bacterial richness (Figure 6A). β-diversity reflecting between-habitat diversity was analyzed by binary Jaccard. Both PCoA and Nonmetric multidimensional scaling analysis (NMDS) indicated a clear distinction between DSS group and DSS-GCZX group (Figures 6C,D). Stress value <0.2 suggests the validity of NMDS. Furthermore, based on the distance matrix acquired from β-diversity, UPGMA was utilized to establish hierarchical clustering of these samples with the aim of evaluating the similarity of bacterial composition after GCZX therapy. As shown in Figure 6E, it manifested that DSS group and DSS-GCZX group clustered clearly in their own groups. Species annotation analysis showed only bacterial of the top 10 abundance at the genus and phylum levels with remaining species merged into “others.”

At the phylum level, the relative abundance of Bacteroidetes that belongs to normal gut microbiota (Le Chatelier et al., 2013) was higher in GCZX group, while Fusobacteria and Proteobacteria, reported as UC-related microorganisms (Petersen et al., 2020), were more abundant in the DSS group. At the genus level, the normal flora of the mouse gut such as Lachnospiraceae and Muribaculaceae (Chung et al., 2020) were most abundant in GCZX group (Figure 6F). Cladogram and Linear discriminal analysis coupled with effect size measures (LEFSe) revealed that Ruminiclostrdium and Lachnospiraceae, Muribaculaceae, and Clostridia were enriched in the GCZX group (Figures 6G,H).

Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRust2) (Langille et al., 2013) based on 16S rRNA sequencing data was utilized to predict functional profiles of microbial communities. GCZX treatment suppressed the incidence of cancer and infectious disease and digestive system (Figure 7). Specifically, GCZX inhibited microbial metabolism, suppressed UC-related metabolic pathways including arachidonic acid and glycerophospholipid metabolism (de Silva et al., 2010; Yuan et al., 2020) and abrogated pro-inflammatory responses such as lipopolysaccharide biosynthesis, restored colitis-reduced arginine (Coburn et al., 2016; Singh et al., 2019) and propanoate (Chambers et al., 2019) levels by impeding their metabolisms whereas contributed to methionine metabolism the derivatives of which favor reconstruction of destroyed epithelium during UC (Roediger et al., 1996; Stavsky and Maitra, 2019). Additionally, it facilitated antibiotics and secondary bile acid biosynthesis that both facilitate the remission of UC (Sinha et al., 2020). Tax4Fun (Aßhauer et al., 2015) also corroborated a strong inhibition of inflammatory pathways such as IL-17 and estrogen signaling pathways as well as antigen processing presentation, and the suppressed incidence of infectious diseases (Supplementary Figure S3).

As predicted in the network pharmacology analysis, the therapeutic mechanism of GCZX capsule against colitis involves oxidative stress that enables stem cells to take on an “over-active” state undergoing differentiation or apoptosis (Tothova et al., 2007; Zhang et al., 2018). The intestinal stem niche environment maintains homeostasis of epithelial renewal by orchestrating the balance between stemness and differentiation. IO culture directly resembles intestinal epithelial organization and thereby recapitulates the dynamic cellular processes of re-epithelization under pathological settings, such as the stimulation of pro-inflammatory cytokines (TNFα).

An ex-vivo UC cellular model was constructed utilizing IOs in the presence of TNFα (20 ng/ml, 24 h), and GCZX solution was administered for another 24 h. Mitochondria consume oxygen and produce reactive oxygen species (ROS), which can be detrimental resulting in intrinsic apoptosis or initiate acute responses to external stimuli such as pro-inflammatory interleukins and pathogens (El-Osta and Circu, 2016). As illustrated in Figure 8A, TNFα treatment enhanced IO ROS production and inflated its size, an effect that was restored by GCZX administration. In line with the findings above, immunofluorescence showed TNFα-induced apoptosis was suppressed by GCZX therapy (Figure 8B). Therefore, GCZX treatment could suppress apoptosis of intestinal stem cells, thereby favoring mucosal healing and combating UC progression.

Peritoneal Mφs isolated from GCZX mice showed stronger phagocytic capacity and higher expression levels of M2 markers compared with DSS group, as evidenced by flow cytometry and western blot (Figures 9A,B), corroborating the anti-inflammatory effect of GCZX capsule.

To mimic the microenvironment in proximity to epithelial cells, we established a co-culture system utilizing BMDMs and NCM460 cells. Interestingly, the migration of BMDMs was subdued with GCZX-pretreated NCM460 cells, indicating that GCZX treatment weakens chemokine secretion from NCM460 cells and hence inhibits the recruitment of BMDMs (Figure 9C).

Given the anti-tumor role of GCZX capsule predicted by network analysis, we examined the metastatic expansion of disseminated cancer cells in the presence of GCZX solution. KI67 represents proliferation, and VEGF and Vimentin refer to the invasion of cancer cells. Neither migration nor invasion as well as proliferation of CW-2 cells was affected by GCZX treatment, as shown by scratch assay and ICC experiments (Supplementary Figure S4), which suggested that GCZX capsule might influence tumor microenvironment, but not directly suppress tumor invasion.