Fig 1.
Uninflamed Inflammatory bowel disease (IBD; Ulcerative colitis (UC) and Crohn’s disease (CD)) and healthy control (HC) samples were all taken from the hepatic flexure, while inflamed samples were taken from maximally inflamed, intact colon mucosa. All samples were evaluated by an experienced gastrointestinal pathologist to verify status as inflamed/uninflamed. An epithelial monolayer area corresponding to approx. 10 000 cells was isolated from each sample using laser capture microdissection. Total RNA was isolated, and mRNA sequenced. Results were then used to perform differential gene expression analysis and weighted gene co-expression network analysis (WGCNA) in parallel. Using significant modules identified by WGCNA in combination with differential expression identified by LIMMA voom, MetaCore was used to characterize each module with regards to process and pathway enrichment. For a selection of genes, epithelial protein expression was verified by immunohistochemistry. Figure created in and republished from BioRender under a CC BY license, with permission from Biorender, 2021.
Table 1.
Characteristics of IBD patients and controls included in the study.
Fig 2.
Epithelial gene expression in IBD—Quality control and initial analyses.
(A) Principal component analysis reveals a clear distinction in gene expression signatures between inflamed and uninflamed samples. There was no separation between UCa and CDa (not shown), and no visual distinction between uninflamed IBD (IBDu) and healthy controls. (B) Unsupervised cluster analysis of top 1000 genes selected on variance as captured by RNA-Seq analysis. The heatmap shows a separation between inflamed and uninflamed samples, but no clustering of UCa or CDa. Counts were transformed by variance stabilizing transformation (VST) as described in the Methods section. (C-E) Top 10 up- and downregulated genes (by adj. p) in (C) IBDa vs. HC, (D) UCa vs. HC and (E) CDa vs. HC. (F) Venn diagram of upregulated (left) and downregulated (right) genes in CDa vs. HC and UCa vs. HC.
Fig 3.
Matrix showing the correlation between WGCNA modules (n = 10), and the sample categories IBDa, UCa, CDa, UCu and CDu. Undesignated genes were collected in the Grey module. Significant correlation between module and trait suggests that genes within the module display expression levels that correlate with sample trait. For each contrast, correlation coefficient is shown, with p-values in parentheses.
Fig 4.
Top differentially expressed genes (by adj. p) for WGCNA modules where expression pattern was significantly correlated with disease status.
Using data from the differential expression analysis coupled with module memberships identified through WGCNA, the most significant differentially expressed genes for the contrasts UCa vs. HC (left) and CDa vs. HC (right) were selected. For contrasts with > 40 DEGs, top 40 is presented. For each module, results for both contrasts are shown. The fold-change was max/min-normalized within each module for visualization purposes. No fold-change is visualized (marked with X) for genes where the adjusted p-value was > 0.05. Fold-change and adj. p was found using LIMMA linear models with least squares regression and corrected for multiple testing using Benjamini–Hochberg FDR-adjustment.
Table 2.
Log2 fold-change (LFC) and adjusted p-values (adj. p) for differentially expressed genes within each WGCNA module for the contrasts IBDa vs. HC, UCa vs. HC and CDa vs. HC.
Fig 5.
Immunohistochemistry confirms epithelial expression.
IHC was used to evaluate the protein expression pattern of a small subset of differentially expressed genes. The proteins evaluated were HLA-DR/DP/DQ, IRF1, JunB, RARβ, CIITA, PI3Kγ, ATF3 and RARα. All staining’s showed positivity in epithelial cells. Scale bars (50μm) indicated.
Fig 6.
Immunohistochemical ATF3-quantitation confirmed increased nuclear ATF3 positivity in UCa surface epithelium compared to HC.
(A) Representative ATF3 staining from HC (left), UCa (middle) and CDa (right). Scale bars (20μm) indicated. (B) A series of sections from HC (n = 12), UCa (n = 11) and CDa (n = 8) were quantified using QuPath. For each section, surface epithelial ATF3-positivity is presented as the ratio of ATF3+ nuclei and total nuclei count. One-way ANOVA followed by Dunnett’s multiple comparisons test showed that nuclear ATF3 positivity was significantly higher in UCa than HC (adj. p = 0.013), but not in CDa vs. HC (adj. p = 0.521).
Table 3.
Characteristics of IBD patients and controls included for immunohistochemical ATF3-quantitation.
Table 4.
Description of primary antibodies and immunohistochemistry parameters.