Background
Inflammatory bowel disease (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC), has become a global public health challenge owing to its increasing incidence [
1]. Intestinal fibrosis is a frequent long-term complication of IBD that often results in functional damage and bowel stenosis requiring surgical intervention, particularly in patients with CD [
2]. To date, a poor understanding of its pathogenesis has hampered the clinical management of patients and development of effective anti-fibrotic therapies [
2]. Moreover, the presence and degree of fibrosis or stenosis cannot be predicted by cross-sectional imaging, endoscopy, or histology [
3]. Thus, there remains a pressing need to elucidate the dynamic pathogenesis of intestinal fibrosis. However, a major obstacle for understanding the pathogenetic mechanisms of intestinal fibrosis in IBD patients is that its early and ensuing time-dependent phases cannot be tracked [
2]. Therefore, well-characterized animal models of intestinal fibrosis have been widely used to reproduce the fibrogenic pathological process to reveal pathogenesis [
4].
Intestinal fibrosis is a heterogeneous process involving multiple intricate and interacting mechanisms that include aberrant immune and non-immune responses, host-microbiome interactions, mesenteric adipocytes, and genetic susceptibility [
5,
6]. Studies have indicated that genetic variants encoding immunomodulatory proteins, pro- and anti-inflammatory cytokines, and fibrogenic factors are associated with the fibrostenotic phenotype of CD [
3]. Some studies have suggested that the abnormal expression of certain genes, such as nucleotide-binding oligomerization domain 2 (
NOD2), toll-like receptors-4 (
TLR4), and tumor necrosis factor-like ligand 1A (
TL1A), might be related to the development of intestinal fibrosis [
5,
7]. These genes play a role in intestinal fibrogenesis via regulating certain important processes, including epithelial-mesenchymal transition (
EMT) initiation and progression, fibrogenic signal transduction, and macrophage-fibroblast program [
7‐
9]. However, current studies have failed to identify the relationship between the dynamic fluctuation of gene expression and colitis-related intestinal fibrosis. Hence, we sought to profile the dynamic gene expression during intestinal fibrosis by time-course RNA-sequencing (RNA-seq) in a murine model, which could provide important information for studying the underlying mechanisms of colitis-related intestinal fibrosis.
The close and complex connections between genes and metabolites have contributed to the development of multiple diseases. Subtle changes in protein-coding genes (particularly those encoding metabolic enzymes) can lead to 10,000-fold changes in metabolite abundance [
10]. Mechanistic studies have revealed that metabolism-related genes can promote fibrosis in other organs by regulating metabolic processes that are strongly linked with fibrosis development, such as fatty acid metabolism [
11]. Metabolites can also serve as active modulators of gene activity by controlling transcription factors and performing post-transcriptional modifications, to modulate biological processes and phenotypes [
12]. Moreover, metabolites participate in the pathogenesis of multiple diseases by acting as signaling molecules, immune modulators, endogenous toxins, and environmental sensors [
10]. In particular, changes in tricarboxylic acid cycle intermediates, amino acid and lipid metabolism products, as well as oxidative metabolites are closely associated with energy metabolism, intestinal barrier, immune system, and disease activity in patients with IBD [
13‐
15]. Additionally, emerging evidence suggests that alterations in metabolism are not only a feature of fibrosis but may also play an influential role in its pathogenesis [
16‐
18]. Recently, in vivo studies have shown that leptin and trimethylamine N-oxide can promote the fibrosis process in various organs, including the liver, lung, heart, and kidney [
19‐
21]. Macias et al. [
22] reported increased serum succinate levels and colonic succinate receptor (
SUCNR1) expression in CD patients and demonstrated the role of
SUCNR1 in murine colitis and intestinal fibrosis. However, the dynamically altered metabolites and their effects on intestinal fibrosis remain obscure. Thus, a comprehensive genetic and metabolic analysis might elucidate the potential roles of altered genes and metabolites in the development of intestinal fibrosis.
In this study, the 2,4,6-trinitrobenzene sulfonic acid (TNBS) model, a classic murine model [
23], was employed to reproduce and evaluate the fibrogenic pathological process at different stages of chronic colitis-related intestinal fibrosis model. RNA-seq was conducted to identify the underlying genetic changes in colonic samples. Simultaneously, fecal widely-targeted metabolomics were performed to investigate the dynamic metabolite disturbances and screen for associated metabolic markers. Integrated analysis combining RNA-seq and targeted metabolomics was then performed to identify correlations between host genes and metabolites that were associated with morbid conditions. Our findings provide new insights into the pathogenesis of intestinal fibrosis.
Methods
Induction of intestinal fibrosis
The animal experimental protocol was approved by the Institutional Animal Care and Use Committee of the Zhujiang Hospital of Southern Medical University (Guangzhou, China). We randomly divided 15 male 8-week-old C57BL/6 mice into three groups (Control, TNBS-4W, and TNBS-6W groups,
n = 5 per group). All mice received weekly intra-rectal administration of TNBS solution (Sigma-Aldrich, USA) or vehicle for 6 weeks as previously described [
24,
25]. Briefly, after fasting for 12 h, mice were anesthetized and treated with 0.1 mL of an increasing dose of TNBS solution (in 45% ethanol) or 45% ethanol. The TNBS enema concentrations from the first to sixth week were 0.75% (
w/v), 1.0% (
w/v), 1.5% (
w/v), 2.0% (
w/v), 2.0% (
w/v), and 2.5% (
w/v), respectively. The feces of mice were collected on day 2 after the fourth and sixth enemas and frozen in liquid nitrogen for metabolites detection. Mice were anesthetized and then sacrificed on day 3 after the fourth and sixth doses, and the colons were harvested for RNA extraction and histological staining.
Histological assessment and quantitative polymerase chain reaction (qPCR)
After sample collection, colons were frozen in liquid nitrogen for RNA extraction, or fixed in 4% paraformaldehyde. The fixed colons were then dehydrated in gradient ethanol, embedded in paraffin, sliced into 4-μm-thick sections, and subjected to Masson’s trichrome and hematoxylin and eosin (H&E) staining. Subsequently, collagen deposition (blue staining) was quantified using the ImageJ software (National Institutes of Health).
The expression of fibrotic indicators in colons was determined by qPCR. Briefly, total RNA was extracted from colons using the TRIzol reagent (Takara, Japanese) and then converted to cDNA using the reverse transcriptase kit (Accurate Biology, China). Subsequent qPCR was performed on the CFX Connect real-time PCR detection system (Bio-Rad Laboratories, Hercules, CA, USA) with a SYBR Green Pro Taq HS Premix (Accurate Biology, China). The primer sequences are shown in the Additional file
1: Table S1.
RNA-seq and data processing
RNA integrity and purity were assessed prior to cDNA library preparation. Samples with concentrations > 50 ng/μL, RIN > 7.0, and total RNA > 1 μg were used in downstream experiments. Briefly, mRNA was isolated according to the polyA selection method using oligo (dT) beads (Cat. 25–61005, Thermo Fisher Scientific, USA) and then fragmented using a magnesium ion interruption kit (Cat. E6150S, USA). Double-stranded cDNA was synthesized using reverse transcriptase (Cat. 1896649, CA, USA), E. coli DNA polymerase I (Cat. m0209, USA), and RNase H (Cat. m0297, USA). After end repair, 3′ adenylation, adaptation ligation, and UDG enzyme treatment, the double strands were pre-denatured by PCR to form a 300 ± 50 bp chain library. The paired-end RNA-seq library was sequenced using an Illumina Novaseq 6000 PE150 platform (LC Bio Technology, Hangzhou, China).
Raw sequencing data were processed using the FastQC software, and clean reads were aligned to mouse reference genomes (GRCm38) using the HISAT2 software with default parameters. The mRNA levels were quantified as the value of fragments per kilobase of exon per million mapped reads (FPKM). The edgeR package [
26] was used to identify differentially expressed genes (DEGs) among the three experimental groups according to the screening criteria
q < 0.05. The Short Time-series Expression Miner (STEM) [
27] was used to classify the identified DEGs expression patterns. The online platform AnimalTFDB v4.0 [
28] was used for transcription factors (TFs) and transcription cofactors (TcoFs) analysis. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using the pathview [
29] and clusterProfiler [
30] packages to determine the functions of the DEGs(species: Mus musculus). The enrichment analysis results with P < 0.05 were further analyzed. Correlations between the screened genes and fibrotic indicators were then assessed using Spearman’s correlation analysis.
After thawing on ice, 20 mg of fecal sample was mixed with 400 μL of 70% methanol–water internal standard extractant, vortexed for 3 min, and sonicated for 10 min in an ice water bath. Homogenization and sonication cycles were repeated twice, followed by incubation at − 20 °C for 30 min and centrifugation at 12,000 ×
g for 10 min at 4 °C. The supernatants were transferred to LC–MS vials and stored at − 80 °C until UPLC–MS/MS analysis. The sample extracts were analyzed using an LC-ESI–MS/MS system xionLC AD (UPLC, E,
https://sciex.com.cn/; MS, QTRAP® System,
https://sciex.com/). Based on the self-built target database of Matteville Biotechnology Co., Ltd. (Wuhan, China), qualitative analysis was performed according to the retention time, parent ion pair information, and secondary spectrum data. Quantification was performed using the multiple reaction monitoring mode of triple quadrupole mass spectrometry. The identified metabolites were annotated using the KEGG compound database. Mass spectrometry data were processed using the Analyst 1.6.3 software.
The raw data was log transform (log2) and mean centering before partial least squares discriminant analysis (PLS-DA). A permutation test (200 permutations) was performed to avoid overfitting. The significance of metabolites was determined using the variable importance projection (VIP) value of the PLS-DA model combined with the fold-change (FC) value of univariate analysis. Metabolites with VIP ≥ 1 and absolute log2FC (|log2FC|) > 0.58 were considered to be significantly altered metabolites (SAMs). Venn diagram and classification pie chart were used to identify the common SAMs. The SAMs content was standardized, and k-means clustering analysis and cluster heatmap were then performed to study the metabolite trends between samples. A random forest algorithm in machine learning was completed to select the characteristic material, and its performance was subsequently evaluated using the receiver operating characteristic (ROC) curve analysis module of the online platform MetaboAnalyst5.0 (
https://www.metaboanalyst.ca). Correlations between the screened metabolites and fibrotic indicators were determined using Spearman’s correlation analysis.
Integrated analysis of transcriptomics and metabolomics
Integrated metabolic pathway analysis of transcriptomics and metabolomics data was performed on the MetaboAnalyst5.0 platform. After identifying the DEGs enriched in lipid metabolism-related biological processes, the STRING database and Cytoscape plug-in CytoHubba were used to perform protein–protein interaction (PPI) network prediction and screen the top15 hub genes. Correlations between the screened genes and metabolites were determined using Spearman’s correlation analysis.
Statistical analysis
Statistical analyses were conducted using the GraphPad Prism 9.0.1 (GraphPad Software, CA, USA). The results were assessed using one-way analysis of variance (ANOVA) and expressed as the mean ± standard error of mean (SEM). In datasets operated by ANOVA, the Bonferroni post-hoc test was adopted. The parameters of all the clustering correlation heatmap with signs were set as followed: cluster methods: complete, cluster distance: Euclidean, correlation methods: Spearman. Statistical significance was set at P < 0.05.
Discussion
In this study, we investigated the changes of gene expression and metabolite abundance in a TNBS-induced intestinal fibrosis mouse model at different time points using colonic RNA-seq and fecal targeted metabolomics techniques. Tendency and enrichment analyses showed that 679 DEGs with enduring changes were primarily enriched in immune response-related signaling pathways and metabolism-related biological processes. Noticeably, 15 hub genes related to lipid metabolism were found to closely correlate with the development of intestinal fibrosis. We found that the formation of intestinal fibrosis was accompanied by marked fecal metabolic disturbances. Among the 48 SAMs, 14 lipid metabolites were dramatically altered during the development of intestinal fibrosis. Integrated analysis of the transcriptomics and metabolomics implied that perturbations of lipids were strongly associated with enduring alterations in lipid metabolism-related genes. In addition, six metabolites were considered as potential biomarkers of intestinal fibrosis in mice. Collectively, our results indicate that intestinal fibrosis in colitis mice might be related to dysregulated lipid metabolism and host-metabolism interactions.
Various signaling pathways are involved in fibrotic formation through myriad complex interactions and signaling cascades [
35]. In this study, we observed that the DEGs were functionally annotated in immune response-related KEGG pathways, including the PI3K-Akt, MAPK, NF-kappa B, and JAK-STAT signaling pathways. A number of studies have shown that persistent or dysregulated signaling pathways, such as the MAPK, PI3K-Akt, and JAK-STAT pathways, contribute to renal and pulmonary fibrosis via regulating cell proliferation, differentiation, apoptosis, and extracellular matrix accumulation [
35‐
37]. Additionally, specific blockade of the NF-kappa B and PI3K-Akt signaling pathways could potently protect against colitis-associated intestinal fibrosis [
38,
39]. These studies confirm that dysregulated immune response-related signaling pathways might play an important role in the onset of intestinal fibrosis. Mechanistic studies have shown that pharmacologic targeting of immune-related signaling pathways could effectively alleviates colitis in animal models and individuals with IBD [
40,
41]. Therefore, we can develop drugs that target interventions for these important specific pathways to prevent and treat intestinal fibrosis.
Moreover, the 15 hub genes that were identified in the PPI network might play a key role in causing intestinal fibrosis. Among the eight continuously up-regulated genes, the well-known pro-fibrotic factors
Tgfb1 and
Il1b, exhibited the strongest positive correlations with the degree of intestinal fibrosis. Studies had proved that the transcriptional levels of
Il1b and
Tgfb1 are up-regulated in the mucosa, submucosa, and muscle of narrow intestines in IBD patients [
34]. However, the precise mechanisms of
Tgfb1 and
Il1b in intestinal fibrosis warrant further investigation. Among the seven genes that exhibited a continuous downward trend,
Vdr,
Bmp2, and
Ppargc1a showed the most marked negative correlation with the degree of intestinal fibrosis.
Vdr, a member of the nuclear receptor superfamily, is a key molecule in genetic regulation, immunomodulation, inflammation control, and microbiota regulation [
42]. Both in vitro and in vivo,
Vdr activation can alleviate intestinal fibrosis by inhibiting abnormal fibroblast activation and migration, as well as epithelial mitochondria-mediated EMT [
43‐
45].
Bmp2, a member of the TGF-β superfamily, is involved in fibrosis development of a variety of tissues and cells through the Smad signaling pathway [
46,
47]. Additionally,
Bmp2 functions as a negative regulator of organ fibrogenesis by antagonizing TGF-β1-induced profibrogenic signals [
48,
49]. Hence,
Bmp2 might also play an important role in intestinal fibrosis.
Ppargc1a induction is beneficial in maintaining mitochondrial integrity, enhancing intestinal barrier function, and decreasing colitis [
50,
51], which might help to prevent chronic colitis-associated intestinal fibrosis. Additionally, mechanistic studies have revealed that the restoration of
Ppargc1a activity protects against kidney fibrosis by restoring mitochondrial viability and dynamics and reversing fatty acid oxidation defects [
52,
53]. Accordingly, we speculated that continually down-regulated
Ppargc1a might be an important factor in the formation of intestinal fibrosis.
Lipid plays vital importance in affecting cell membranes, metabolic processes and signaling pathways, and acting as energy storage sources. Lipid metabolism disorders have been reported in the serum, plasma, urine, feces, and colonic mucosa samples of IBD patients [
54]. Nevertheless, few studies have focused on the metabolite profiles of IBD patients with intestinal fibrosis. In this study, we found that metabolite disturbances in mouse feces, especially lipid and organic acid metabolites, were closely associated with intestinal fibrosis progression. We found that Phytosphingosine continually decreased during the progression of intestinal fibrosis, which was previously reported to perform anti-inflammatory activity in cell-based assays and ameliorate acute colitis in mice [
55,
56]. We also found arachidonic acid metabolism products such as 11,12-EET and prostaglandin E2, was decreased in the fibrosis phase, which have exhibited certain anti-fibrotic activities in other organ [
57‐
60]. Therefore, we speculated that supplementation with substances with anti-inflammatory and anti-fibrotic activities during the inflammation phase may help to mitigate or prevent intestinal fibrosis. However, the relationship between the lipid metabolites identified in our study and fibrotic diseases have rarely been reported, possibly because previous studies have only focused on differential metabolites after fibrotic formation. Our study explored changes in metabolites during the progression of inflammation-associated intestinal fibrosis, and these significantly altered metabolites might favor the early recognition of fibrosis.
The close connections between lipid metabolism-related genes and metabolites might contribute to the development of intestinal fibrosis. On the one hand, bioactive metabolites especially the lipid metabolites could drive key modification processes for DNA, RNA and proteins to regulate fundamental biological processes of IBD development, such as signal transduction, protein balance, and gene expression regulation [
10,
12]. On the other hand, recent studies have revealed that alterations in lipid metabolic processes, especially the fatty acid metabolism, are common mechanisms and central pathophysiological pathways for the development of various fibrotic diseases [
11]. We have identified four DEGs strongly correlated with lipid metabolites, including
Bmp2,
Ppargc1a,
Pik3r3, and
Snai1.
Ppargc1a plays a pivotal role in lipid and metabolic regulation in many vital organs, including adipose tissue, skeletal muscle, heart, liver, and kidney [
61].
Bmp2 is likely to induce adipogenesis by promoting the expression of lipoxygenase (
LOX) and PPAR gamma (
PPARγ) in preadipocytes [
62]. Yang et al. [
63] reported that
Pik3r3 regulates PPAR alpha (
PPARα) expression to stimulate fatty acid β-oxidation. Studies have suggested that adipose
Snail1 acts as an epigenetic rheostat that governs lipid metabolism and partitioning between tissues [
64,
65]. Furthermore, genetic alterations or pharmacologic targeting of altered lipid metabolic processes have great potential to inhibit fibrosis development [
11]. Thus, more mechanistic studies are required to investigate the role of the interaction between these DEGs and lipid metabolism in the pathogenesis of intestinal fibrosis.
We have conducted time series analyses to better identify the dynamic characteristics of gene regulatory and metabolites fluctuate network models, which largely compensates for the fact that clinical samples cannot dynamically track intestinal fibrosis. To the best of our knowledge, this is the first comprehensive dynamic transcriptomics and metabolomics analysis of colitis-associated intestinal fibrosis. However, we acknowledge certain limitations in our study. First, we currently conduct fecal metabolomics studies with limited samples, but many factors can influence the abundance of metabolites. So, adding more biological replicates for each group and detecting serum metabolites might greatly help to increase the reliability and persuasiveness of our findings and better explore metabolic changes associated with fibrosis. Second, we can’t cross-validate our findings with external data, because similar longitudinal studies for chronic inflammation-associated intestinal fibrosis are currently lacking. Third, we did not characterize the precise underlying mechanism for the development of intestinal fibrosis, which must be addressed in more detailed and comprehensive studies. Finally, we need to perform colonic transcriptomic and fecal metabolomic of IBD patients before and after intestinal fibrosis to further confirm our findings. However, it is inconvenient to dynamically track the progression of intestinal fibrosis in patients because the fibrogenesis processes have already been established when fibrosis is detected. Thus, in a future study, we will try to collect intestinal biopsies and fecal samples from patients with intestinal fibrosis to verify our results.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.