The GEO database is a database for storing chips, second-generation sequencing, and other high-throughput sequencing data. The GEO database was searched to find the datasets (GSE107499, GSE75214, and GSE59071) containing differential genes in colon tissues of normal people and IBD patients. The differentially expressed genes were analyzed and found by the GEO2R tool.

SEA version 3.0 is an online database that can provide information about super-enhancers. By collecting and analyzing the public Chip-Seq data, the SEA website contains information on the super-enhancers and their related genes found in different cells and tissues. In SEA, we identified the super-enhancers in the human sigmoid colon using H3k27ac Chip-sequencing and collected the information on the super-enhancer, including super-enhancer ID, genomic site, length, related genes, and transcription factors. Then, the super-enhancer–related genes and differentially expressed genes were cross-linked to obtain the differentially expressed genes driven by the super-enhancers. The interaction network between transcription factors, super-enhancers, and related genes was constructed using Cytoscape software.

The overlapping genes were analyzed by the KEGG pathway. The KEGG pathway data were uploaded to the Hiplot website to form a bubble chart.

The Caco-2 cell line was purchased from Wuhan Procell Life Technology Co., Ltd (Wuhan, China) and cultured in DMEM high-glucose medium (Solarbio, China) supplemented with 20% fetal bovine serum (FBS, Yeasen, China). All of the media contained 50 U/mL penicillin and 50 U/mL streptomycin. All cells were incubated with 5% CO₂ at 37°C. After pre-protecting the cells for 2 h by adding the drug or an equal volume of vehicle, cells were treated with 1 μg/ml LPS to stimulate Caco-2 cells for 24 h, and the protein and mRNA expression levels were detected.

Total RNA was extracted from Caco-2 cells and colon tissues by TRIzol reagent (Tiangen, Beijing, China) and quantified by using a NanoDrop 8000 spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, United States). RNA was reverse transcribed into cDNA with the first-strand cDNA synthesis Supermix kit (Yeasen, Shanghai, China), and the PCR system was prepared with the Yeasen qRT-PCR kit to amplify cDNA and performed with CFX96 touch (BIORAD, United States) according to the protocol of the manufacturer. The reaction conditions were as follows: predenaturation at 95°C for 30 s, denatured at 95°C for 10 s, annealed at 60°C for 20 s, extended at 70°C for 20 s, and repeated 40 cycles. GAPDH was used as the standard control. The forward and reverse primer sequences of CEBPB, PCK1, EFNA1, and GAPDH are as follows. The cycle threshold was automatically generated, and the relative expressions of CEBPB, PCK1, and EFNA1 mRNA were calculated by the 2^-△△Ct method.

For humans:

CEBPB forward primer: ACG​GGC​CGC​CCT​TAT​AAA​T; CEBPB reverse primer: CAGGCCACCAGGCGTTG; PCK1 forward primer: CGG​AAA​GAA​ACC​TGT​GGA​TCT​C; PCK1 reverse primer: CAG​ATG​TGG​ATG​TGA​TCA​GGC​T; EFNA1 forward primer: GCT​ATG​GAG​TTC​CTC​TGG​GC; EFNA1 reverse primer: ACG​TAG​TCA​TTC​AGC​TGC​ACA; GAPDH forward primer: GAC​AGT​CAG​CCG​CAT​CTT​CT; GAPDH reverse primer: GCG​CCC​AAT​ACG​ACC​AAA​TC;

For mice:

CEBPB forward primer: TTA​TAA​ACC​TCC​CGC​TCG​GC; CEBPB reverse primer: TTC​CAT​GGG​TCT​AAA​GGC​GG; PCK1 forward primer: ACA​CAC​GCA​AAC​TAC​CAA​GC; PCK1 reverse primer: GTC​CTC​GTA​ATG​TGG​GCA​GA; EFNA1 forward primer: GGA​AGA​ACA​AGG​AGT​GGA​GAC; EFNA1 reverse primer: CAG​GCA​GGG​TCA​ATA​ATG​GG; GAPDH forward primer: GGA​GAG​TGT​TTC​CTC​GTC​CC; GAPDH reverse primer: CCG​TTG​AAT​TTG​CCG​TGA​GT;

After grouping Caco-2 cells and colon tissues, the RIPA lysate was used to extract the protein in the cells and tissues, and the protein content was measured using the BCA kit (Beyotime, Beijing), and 30 μg protein was taken for detection. The protein samples were added into polyacrylamide gel (8–15%) and then electrophoretic gel, then transferred to the PVDF membrane (Millipore, MA, United States), sealed for 2 h with 5% skim milk powder, and incubated overnight with 1–1,000 diluted antibody at 4°C. Subsequently, PVDF membranes were incubated with rabbit antibodies diluted 1:1 000 for 2 h, and protein bands were obtained after development using a biospectral gel imaging system (UVP, CA, United States). Finally, Image J software was used for gray analysis.

Male C57BL/6J mice (SPF grade) aged 6–8 weeks, weighing 18–22 g, came from the animal experiment center of the Jinnan campus of Nankai University. Fifty SPF mice were randomly divided into normal group, dextran sodium sulfate (DSS) group, sulfasalazine (SASP) group, nintedanib low-dose group (Nin, 50 mg/kg), and nintedanib high-dose group (100 mg/kg), with 10 mice in each group. After adaptive feeding for 1 week, the mice in the normal group were free to eat and drink water, and for the mice in other groups, the daily drinking water volume of each mouse was calculated to be 6 ml, and the DSS solution was added to the daily drinking water volume for 8 consecutive days to establish the model of IBD in mice. While drinking 5% DSS solution daily, SASP (200 mg⋅kg⁻¹⋅Day⁻¹) and nintedanib (50 and 100 mg⋅kg⁻¹⋅Day⁻¹) were administered by gavage at 1–7 days, respectively. 0.5% CMC-Na was administered by gavage in the normal group and the DSS group for 7 days. The body weight, stool consistency (degree of diarrhea), and blood in stool were recorded every day. On the 8th day, after fasting for 12 h, 10% chloral hydrate was injected intraperitoneally. The colon and rectum of mice were washed with the PBS solution. 1 cm tissue from the anus was selected and placed in 10% formalin. The remaining intestinal tissue was put into a cryopreservation tube and stored at −80°C. The intestinal tissue was fixed in formalin for 24 h, dehydrated, embedded, sliced, stained with hematoxylin-eosin and Alcian blue (PH = 2.5, Beyotime, Beijing), and observed under the microscope.

The characteristics of feces were observed, and the situation of fecal occult blood was determined and scored accordingly. DAI scoring criteria are as follows (Peng et al., 2019):

1) Body mass score: no change or increase, 0 point; reduce 1%–5%, 1 point; reduce 5%–10%, 2 points; reduce 10%–15%, 3 points; the reduction is greater than 15%, 4 points.

The sections were stained with hematoxylin-eosin and observed under the microscope. The histopathological evaluation criteria are as follows (Stillie and Stadnyk, 2009):

1) Degree of inflammation: none, mild, moderate, and severe;

A total of 3 ulcerative colitis tissues and 3 colonic ulcer distal control tissues were collected from healthy volunteers (1 male and 2 females, aged 49–63 years) and UC patients (2 males and 1 female, aged 37–58 years) during endoscopy at the Department of Digestive Diseases, Tianjin Medical University General Hospital (Tianjin, China). Patients with ulcerative colitis were untreated preoperatively. Samples were independently identified by two pathologists. These tissues were obtained from patients who underwent colonoscopy at Tianjin Medical University General Hospital after signing informed consent. The entire study was approved by the ethics committee of Tianjin Medical University General Hospital.

The tissue sections were dewaxed, hydrated, and antigen repaired. After incubation with endogenous peroxidase blocker and immune serum (rabbit) with the immunohistochemical kit (Maixin, Fuzhou, China), the CEBPB Rabbit antibody (1: 50 dilution; Affinity) was incubated overnight at 4°C. The next day, nonspecific secondary antibody and streptavidin peroxidase were used for incubation, DAB (Maixin, Fuzhou, China) color development, hematoxylin staining, and the slices were sealed after dehydration. Finally Image J software was used for IHC scoring.

Caco-2 cells were inoculated into 96-well plates and cultured in a 5% CO₂ incubator at 37°C for 24 h. When the cell density reached 50–70%, si-CEBPB (Gene Pharma, shanghai) and the negative control group (NC) were transfected into the cell line according to the instructions of the lipo 8000 reagents (Beyotime, Beijing). 48 h after transfection, the cells were collected for the Western blot analysis. CEBPB siRNA sequences were sense 5′-CGCCUGCUUUAAAUCCOUTT-3′ and antisense 5′- AUG​GAU​UUU​AAA​GGC​AGC​GTT-3'. The CEBPB negative control sequences were sense 5′-UUC​UCC​GAA​CGU​GUC​ACG​GUT​T-3′ and antisense 5′-ACG​UGA​CAC​GUU​CGG​AGA​ATT-3'.

The feces of mice in the normal group, the DSS group, and the nintedanib (100 mg/kg) group were collected and stored at −80°C. For 16S rDNA gene sequencing, fecal samples were sent to the gene and genome analysis center (Microread gene, Beijing, China) for sequencing under dry ice conditions. Primers corresponding to the bacterial 16S rRNA hypervariable region (V3–V4) were selected for amplification. The sequencing results of all samples and the statistical results of sequence data are based on sequencing reads and operational taxonomic units (OTUs).

All the data were analyzed using GraphPad Prism 8.4.2 (GraphPad Software, Inc., San Diego, CA) and expressed as mean ± SD. Two and multiple groups were compared using the independent sample t-test and ANOVA, respectively. p < 0.05 was considered statistically significant.

We first analyzed the GEO database (GSE107499\GSE75214\GSE59071), and the DEGs volcano is shown in Figure 1A. Super-enhancers usually positively regulate gene expression, so we analyzed the upregulated DEGs in the database. A total of 1,366 upregulated DEGs in IBD patients were found using three different GEO datasets (Table 1). 863 super-enhancer–related genes were identified in the human sigmoid colon by the SEA website. We cross-linked these super-enhancer–related genes with upregulated DEGs to predict potential super-enhancer–driven pathogenic genes in IBD. The Venn diagram shows that there are 5 gene overlaps (Figure 1B), which are considered potential super-enhancer–driven IBD pathogenic genes. The information on potential pathogenic SEs of IBD is shown in Table 2. Cytoscape software was used to construct the interaction network of transcription factors, super-enhancers, and related genes (Figure 1C). Subsequently, we performed KEGG network analysis of these five potential super-enhancer–driven pathogenic genes and found that PCK1 and EFNA1 genes were enriched in the PI3K–Akt pathway (Table 3, Figure 1D). Subsequently, we constructed the interaction networks of these two genes with super-enhancers and TFs. Surprisingly, we found that the transcription factor CEBPB existed in the relationship between the two interaction networks at the same time (Figure 1E). Therefore, we believe that CEBPB/PCK1 and CEBPB/EFNA1 may be the key pathways for the potential treatment of IBD.

To verify whether CEBPB regulates PCK1 and EFNA1 to improve colitis, we designed the interfering RNA of CEBPB. As shown in Figure 2A, after silencing the transcription factor CEBPB, the protein expression levels of PCK1 and EFNA1 decreased significantly. This indicates that PCK1 and EFNA1 are regulated by the transcription factor CEBPB. At the same time, silencing the transcription factor CEBPB also found that the intestinal barrier proteins ZO-1, occludin, and claudin-1 increased (Figure 2B), indicating that inhibiting CEBPB can restore the damaged intestinal barrier proteins. The aforementioned results indicate that PCK1 and EFNA1 pathways regulated by CEBPB may participate in the pathogenesis of IBD through epithelial barrier proteins. Subsequently, we detected the mRNA levels of CEBPB, PCK1, and EFNA1 in LPS-stimulated Caco-2 cells after being treated with the marketed drugs commonly used for the treatment of IBD (sulfasalazine, mesalazine, olsalazine, and tofacitinib). The results showed that LPS significantly promoted the expression of CEBPB, PCK1, and EFNA1 mRNA in Caco-2 cells, while sulfasalazine, mesalazine, olsalazine, and tofacitinib significantly inhibited the expression of CEBPB, PCK1, and EFNA1 mRNA stimulated by LPS (Figure 2C).

The aforementioned four marketed drugs have different degrees of side effects, so it is necessary to develop new safe and effective drugs. Some small-molecule drugs of tyrosine kinase inhibitors (such as nintedanib and anlotinib) have been proved to have certain anti-inflammatory effects (Wollin et al., 2014). We wanted to determine the efficacy of tyrosine kinase inhibitors such as anlotinib, lenvatinib, axitinib, and nintedanib in treating IBD. As shown in Figure 3A, compared with the normal group, the mRNA levels of CEBPB, PCK1, and EFNA1 in the LPS group increased significantly. Similar to tofacitinib, sulfasalazine, oxalazine, and mesalazine, mRNA levels of CEBPB, PCK1, and EFNA1 in the nintedanib group were significantly reduced after administration, while the mRNA levels in anlotinib, lenvatinib, and axitinib groups did not change significantly after administration. After treating LPS-induced Caco-2 cells with nintedanib, it was found that the nintedanib group could significantly reduce the protein expression levels of CEBPB, PCK1, and EFNA1. At the same time, nintedanib can repair the damage to intestinal barrier protein caused by LPS (Figure 3B,C). These results suggest that nintedanib may alleviate IBD by inhibiting CEBPB/PCK1 and CEBPB/EFNA1 pathways.

To further verify the therapeutic effect of nintedanib on colitis, we used 5% DSS to induce colitis in mice. As shown in Figure 4A, the murine experimental colitis animal model was established. Compared with the control group, DSS-treated mice significantly reduced body weight, increased DAI, and shortened colon weight length (Figure 4B,C,D). Significant tissue damage was found in the histological examination of the colon of DSS-treated mice, with a high microdamage score (Figure 4E). The positive control SASP (100 mg/kg) and nintedanib (50, 100 mg/kg) could significantly recover the experimental colitis-related symptoms of mice caused by DSS, such as weight loss, increased DAI, shortened colon length, and colonic tissue injury.

Alcian blue staining was used to observe the expression of intestinal mucins. Intestinal mucin is stained blue by Alcian blue, and nuclei are stained red by nuclear fast red fuel. As shown in Figure 4F, DSS caused damage to the intestinal mucosa, with a significant decrease in the mucin expression. The intestinal mucosal injury caused by DSS was alleviated by intragastric administration [SASP, nintedanib (50, 100 mg/kg)].

We conducted the Western blot test on colonic tissue and found that the intestinal barrier protein in the DSS group was seriously damaged. The positive drug group and low- and high-dose groups (50 and 100 mg/kg) of nintedanib could significantly recover the damaged intestinal barrier protein, and the high-dose group of nintedanib had a better effect (Figure 4G).

As shown in Figure 5A,B, the mRNA and protein expression of CEBPB, PCK1, and EFNA1 in the DSS group increased significantly, and the mRNA and protein expression of CEBPB, PCK1, and EFNA1 decreased significantly after nintedanib administration. IHC results showed that CEBPB increased significantly after DSS treatment, while CEBPB immunostaining decreased significantly after nintedanib treatment (Figure 5C). These results suggested that nintedanib may inhibit DSS-induced experimental colitis in mice by inhibiting CEBPB/PCK1 and CEBPB/EFNA1 pathways. IHC of tissue sections from healthy individuals and IBD patients similarly demonstrated a significant increase in CEBPB in tissue sections from IBD patients (Figure 5D). The patient information is shown in Table 4.

Since the gut microbiota is an important factor contributing to IBD, we utilized bacterial 16S rRNA sequencing to investigate the modulation of gut microbiota composition by nintedanib in mice. As shown in Figure 6A, we first used the Venn diagram to visualize the results of OTU cluster analysis between different groups. In total, 2931 OTUs were obtained from the feces of control, DSS-, and nintedanib-treated mice, 802 of which were shared by all three groups. The number of unique OTUs in the control, DSS, and nintedanib groups was 651, 302, and 182, respectively. Next, as shown in Figure 6B, we used linear discriminant analysis effect size (lefse) to identify differentially expressed OTUs between the feces of different groups of mice. Principal coordinate analysis (PCOA) was used to analyze the community structure of the mouse gut microbiota in the three groups (Figure 6C). On the PCOA plot, each symbol represents one gut microbiota. The more similar the abundance and composition of species, the closer the distance between samples. The results showed that the gut microbiota was significantly different between the different groups. The relationship between the gut microbiota from the three groups of fecal samples was investigated using a gut microbiota tree generated by the UPGMA algorithm. The dendrogram (Figure 6D) described and compared the similarity and difference relationship between the three samples intuitively through the dendritic structure. Figure 6E showed that the composition of the gut microbiota in different groups was significantly different at the genus level. After DSS treatment, the relative abundances of Lachnospiraceae, Rikenellaceae, and Oscillospira significantly decreased and that of Sutterella significantly increased compared with those of the control group (Figure 6F). After nintedanib administration, the gut microbiota of the mice was significantly altered compared with the DSS group (Figure 6G). The relative abundances of Sutterella and S24-7 decreased, and those of Bacteroides, Bilophila, and Parabacteroides increased significantly, with Bacteroides becoming the dominant genus. Collectively, the gut microbiota of the control, DSS, and nintedanib groups were significantly different, and nintedanib significantly changed the composition of the intestinal microbiota in mice with DSS-induced experimental colitis.