Creeping fat-derived mechanosensitive fibroblasts drive intestinal fibrosis in Crohn's disease strictures

Abstract

A significant complication of Crohn's disease (CD) is intestinal fibrosis, which narrows the bowel lumen to form a stricture. Creeping fat (CF) is the wrapping of mesenteric adipose tissue around diseased bowel, of which the role in CD stricture progression is unclear. By constructing a human single-cell CD fibroblast atlas, we identified CF-derived, CTHRC1+ fibroblasts enriched for Yes-associated protein (YAP)/transcriptional co-activator with PDZ-binding motif (TAZ) signatures and localized to a fibrotic CF-bowel wall interface within the stricture. We further showed that analogous Cthrc1+ mouse fibroblasts derive from mesenteric adipose tissue stromal cells, infiltrate fibrotic bowel, and deposit extracellular matrix in a YAP/TAZ-dependent manner in a mouse model of intestinal fibrosis. Our findings identify CF as a key source of pro-fibrotic fibroblasts and raise the possibility of improving future clinical management of stricture progression by targeting not only the bowel but also CF.

Keywords: Crohn’s disease; Yes-associated protein; creeping fat; fibroblasts; fibrosis; inflammatory bowel disease; meta-analysis; scRNA-seq; strictures.

Figure 1 |. Mechanosensitive fibroblasts are present and expand in human creeping fat.

A) Schematic (top) and gross histology (bottom) of uninvolved and stricture regions and mesenteric adipose tissue (MAT) wrapping (arrows) (left). Scatter plot of bowel wall (BW) thickness and the fraction of MAT wrapping over BW circumference in seven surgical samples from uninvolved and stricture regions within three pediatric strictures (right). B) Schematic showing the identification of pro-fibrotic, creeping fat (CF)-derived fibroblasts. C) Uniform manifold approximation and projection (UMAP) of 101,189 independently re-clustered and integrated fibroblasts. Conserved fibroblasts across bowel and MAT outlined. D) Dot plot of the scaled average expression of ECM and YAP signatures in fibroblast subpopulations. CTHRC1+ fibroblasts highlighted (red box). E) Schematic of comparison disease states for Crohn’s Disease (CD) and non-CD patients from full-thickness small bowel and MAT samples. F) Bar plot of fibroblast subpopulation enrichment in comparison disease states in global (left), bowel only (middle), and MAT only (right) subsets. G) Dot plot of the scaled average expression of ECM and YAP signatures in CTHRC1+ fibroblasts across comparison disease states. CF CTHRC1+ fibroblasts highlighted (red box). H) Schematic of fibroblast heterogeneity in CD strictures. Red and black arrows indicate increased and decreased enrichment in strictures, respectively. Gray, wavy lines indicate high ECM production.

Figure 2 |. The interface between bowel and MAT is a highly fibrotic region.

A) Schematic of the integration of ECM ultrastructure analysis with Visium spatial transcriptomics. B) Gross histology (left) and a representative PicroSirius Red-stained section (right) from uninvolved (left) or stricture (right) bowel from a pediatric stricture. Histology sections were divided into bowel wall (BW), mesenteric adipose tissue (MAT), and BW/MAT interface (IF). C) Separate UMAPs (left) of ECM ultrastructure from uninvolved BW (UI BW) (left), stricture BW (S BW) (middle), and IF (right) (n = 3). Representative images of PicroSirius red staining from normal, intermediate, and fibrotic ECM (top left insets). Overall UMAP of ECM ultrastructure with Monocle 3 pseudotime using normal ECM as the origin (right). Normal (black) and fibrotic (red) ECM are outlined. D) Violin plot showing ECM ultrastructure pseudotime in each domain. E) Schematic of section locations (left), H&E (middle), and final domain classification (right) for spatial transcriptomics on two sections from a single pediatric patient. Inner bowel (lumen through muscularis propria (MP) to superficial IF) (top). Outer bowel (MP through IF to CF) (bottom). IF is outlined for H&E and domains (black) F) Spatial feature plots of module scores for fibroblast signatures. IF is outlined (black). Scale bars: 2 mm (E), 1 mm (B). One-way ANOVA with Tukey’s test for multiple comparisons using median pseudotime values from biological replicates (D). P value: P < 0.05 (*). Dots represent individual 20X images (C-D).

Figure 3 |. CTHRC1+ mechanosensitive fibroblasts localize to the interface and creeping fat in pediatric strictures.

A) Schematic of the deconvolution of Xenium hi-plex in situ hybridization data by Niche Covariation (NiCo) using the global scRNA-seq CD atlas combined with adipocytes as the reference. B) Representative NiCo-predicted mapping of CTHRC1+ fibroblasts of uninvolved (left) and stricture (right) Xenium sections. Submucosa (blue), interface (IF) (red), and creeping fat (CF) (green) are outlined. C) Representative immunofluorescent region of DAPI (blue) and ATP1A1/CD45/E-Cadherin (magenta) displaying predictive mapping of CTHRC1+ fibroblasts (yellow) present in representative areas within CF (left) and the IF (right). Dots represent CTHRC1 (light blue), CCN1 (purple), and POSTN (orange) transcripts. D) Bar plot of predicted fibroblast subpopulation enrichment in uninvolved (n = 3) and stricture (n = 5) samples. E) Schematic of stricture bowel tissue segmentation labeling for BW, IF, superficial CF (SF CF), and deep CF (left). Immunostaining of representative Xenium section (right). DAPI (blue) and ATP1A1/CD45/E-Cadherin (magenta). Predicted CTHRC1+ fibroblasts (yellow). F) Density plot of YAP target genes CCN1 and SERPINE1 transcripts overlaid with predicted CTHRC1+ fibroblasts (yellow) in low magnification (first column) and high magnification (insets, last three columns). Images are taken from IF (*), SF CF (**), and deep CF (***). Immunostaining for DAPI (blue) and ATP1A1/CD45/E-Cadherin (magenta). BW (light blue), IF (red), and CF (green) are outlined in low magnification image. G) Violin plot of CCN1 (left) and SERPINE1 (right) transcript counts in predicted CTHRC1+ fibroblasts from pediatric stricture samples containing the highest number of CTHRC1+ fibroblasts by CF subdivisions (n = 3). Scale bars: 1000μm (B, C, E, F (lower magnification)), 50μm (F (higher magnification)), 10μm (C). One-way ANOVA with Tukey’s test for multiple comparisons using individual cells (G). P values: P > 0.0001 (****).

Figure 4 |. CTHRC1+ fibroblast-niche cross-talk primarily occurs via the extracellular matrix.

A) Representative immunofluorescent region of DAPI (blue) and ATP1A1/CD45/E-Cadherin (magenta) with predicted cell type labeling in CTHRC1+ neighborhoods in creeping fat (CF) (left), interface (IF) (middle), and bowel wall (BW) (right). B) Line graph depicting maximum normalized interaction scores between predicted CTHRC1+ fibroblasts and neighbor cells in CF (green circles), IF (red triangles), and BW (blue squares) in pediatric strictures (n = 5). C) Cord diagram of differentially upregulated ligand-receptor interactions in CTHRC1+ neighborhoods from CD atlas by CellChat. D) Chord diagrams showing significantly upregulated interactions between CTHRC1+ fibroblasts and neighboring fibroblasts (top left), pericytes (top right), plasma cells (bottom left), and macrophages (bottom right) from CD atlas by CellChat. Relevant ECM (red), cytokine (blue), and angiogenesis (green) interactions highlighted. E) Representative immunofluorescent region of DAPI (blue) and ATP1A1/CD45/E-Cadherin (magenta) displaying predicted mapping of CTHRC1+ fibroblasts (yellow), macrophages (blue), and pericytes (green). Dots represent ITGB1 (purple), TGFB1 (orange), and COL4A1 (red) transcripts. F) Schematic summarizing major interactions between CTHRC1+ fibroblasts and neighbors. Scale bars: 50μm (A), 10μm (E).

Figure 5 |. A mouse colotomy model recapitulates features of human strictures and creeping fat.

A) Schematic of the creation of colotomies under tension to generate fibrosis and creeping fat (CF) viewed from above (top) and in cross-section (below). B) Gross image of sham (left) and colotomy (right) bowel isolated at postoperative day (POD) 14. Two colotomy sites (black arrows) and adjacent regions (grey arrows). Mesenteric adipose tissue (MAT) (outlined). C) Masson’s trichrome of sham, uninvolved (UI), and colotomy at POD 14. Cross-section schematic (left) and low (middle) and high (inset, right) magnifications for each region. Bowel wall (BW) (blue), MAT-BW interface (IF) (red), and MAT (yellow) are outlined. D-F) Scatter plots of average BW thickness (D) and normalized trichrome-positive area (E) in BW adjacent to MAT and fraction of MAT-covered bowel (F) from the timecourse, sham and UI (POD 14). D-F) Scatter plots of average BW thickness (D) and normalized trichrome-positive area (E) in BW adjacent to MAT and fraction of MAT-covered bowel (F) from the timecourse, sham and UI (POD 14). G) Separate UMAPs of ECM ultrastructure from the timecourse from bowel with and without MAT. Overall UMAP of ECM ultrastructure with Monocle 3 pseudotime using normal ECM as the origin (bottom row, right), sham (POD 14), UI (POD 7–90). H) Violin plot of ECM ultrastructure pseudotime from the timecourse from bowel with MAT only. I) UMAP of ECM ultrastructure of bowel only (right) or bowel with MAT (left) from POD 7–90 images. J) Violin plot of ECM ultrastructure pseudotime from bowel only and bowel with MAT. Scale bars: 1cm (B), 200μm (C, (lower magnification), 50μm (C (higher magnification)). One-way ANOVA with Tukey’s test for multiple comparisons using mean values (D-F) or median values (H) from biological replicates. Student’s t-test (two-tailed) using median pseudotime values from biological replicates (J). P values: P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), P < 0.0001 (****). (*) represents significance relative to sham. (#) represents significance relative to uninvolved. n = 3 biological replicates per timepoint. One dot represents one histological section (D-F) or image (G-J). Bars represent standard deviations. Normal (black) and fibrotic (red) ECM is outlined in UMAPs.

Figure 6 |. MAT-derived mechanosensitive fibroblasts are conserved in mouse colotomies.

A) Schematic of the separate dissection and dissociation of bowel and associated mesenteric adipose tissue (MAT) from sham and colotomy mice at post-operative day (POD) 14 for scRNA-seq analysis. B) UMAP of 5,028 mouse fibroblasts integrated by sample and independently re-clustered. C) Bar plot of fibroblast enrichment in sham and colotomy bowel (left) and MAT (right). D) Dot plot of the scaled average expression of YAP and ECM signatures in fibroblast subpopulations. E) Bar plot of pathway analysis (GO Biological Process) on Cthrc1+ fibroblasts. ECM pathways highlighted (red). F) Dot plot of the scaled average expression of YAP signatures and ECM in Cthrc1+ fibroblasts in different conditions between bowel and MAT. G) Dot plot of scaled anchor label transfer scores for fibroblast subpopulations between mouse query dataset and human CD atlas reference. Higher values signify greater similarity between subpopulations. H) Immunostaining for YAP1 (green) and POSTN (red) in uninvolved (left) and colotomy (right) bowel wall (*BW), interface (**IF) and mesenteric adipose tissue (***MAT) (n = 3) in low (left) and high (inset, right four columns) magnification. MAT is outlined (white). Nuclear YAP1+POSTN+ cells are highlighted (arrowheads). Scale bars: 200μm (H (lower magnification). 20μm (H (higher magnification)). Cthrc1+ fibroblasts are highlighted (red boxes)

Figure 7 |. MAT-derived Cthrc1+ fibroblasts promote intestinal fibrosis via YAP/TAZ signaling.

A) Schematic showing the local administration of 4-hydroxytamoxifen-encapsulated liposomes into the mesentery in Col1a2-CreERT2; YAP^lox/+; TAZ^lox/+; ZsGreen+ (YAP/TAZ Heterozygous (Het)) and Col1a2CreERT2; YAP^lox/lox; TAZ^lox/lox; ZsGreen+ (YAP/TAZ Knockout (KO)). B) Gross histology of YAP/TAZ Het (top) and YAP/TAZ KO (bottom). Mesenteric adipose tissue (MAT) is outlined (dotted line). Colotomies are indicated (arrowheads). C) Masson’s trichrome of YAP/TAZ Het (left) and YAP/TAZ KO (right). Total bowel (green) and bowel adjacent to mesenteric adipose tissue (MAT-ADJ) (orange) are outlined in low magnification (top). Bowel wall (BW) (blue), interface (IF) (red), and MAT (yellow) are outlined in high magnification (bottom). D-F) Scatter plots of average BW thickness (D) and normalized trichrome-positive area (E) in BW adjacent to MAT and fraction of MAT-covered bowel (F) in YAP/TAZ Het (n = 5) and KO (n = 6) colotomies. Measurements in BW adjacent to MAT (D-E). G) Separate UMAPs (left) of ECM ultrastructure from YAP/TAZ Het (n = 5) and KO (n = 6) uninvolved and colotomy. Overall UMAP of ECM ultrastructure with Monocle 3 pseudotime using normal ECM as the origin (right). H) Violin plot of ECM ultrastructure pseudotime in YAP/TAZ Het and KO uninvolved and colotomy. I) Immunostaining for GFP (green), YAP1 (red), and mechanosensitive marker POSTN (white) in Het (left) and KO (right) colotomies (n = 3) in low (left) and high (right three columns) magnification of BW (*), IF (**), and MAT (***). Yellow arrowheads denote nuclear YAP+GFP+POSTN+ cells. MAT, IF, and BW outlined. Scale bars: 200μm (C, I (lower magnification)). 20μm (C, I (higher magnification)). Student’s t-test (two-tailed) using mean values from biological replicates (D-F). One-way ANOVA with Tukey’s test for multiple comparisons using median pseudotime values from biological replicates (H). P values: P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), P < 0.0001 (****). One dot represents one histological section (D-F) or image (G-H). Normal (black) and fibrotic (red) ECM are highlighted in UMAPs (G). Bars represent standard deviations.