Intercellular interaction between FAP+ fibroblasts and CD150+ inflammatory monocytes mediates fibrostenosis in Crohn’s disease

Crohn’s disease (CD) is marked by recurring intestinal inflammation and tissue injury, often resulting in fibrostenosis and bowel obstruction, necessitating surgical intervention with high recurrence rates. To elucidate the mechanisms underlying fibrostenosis in CD, we analyzed the transcriptome of cells isolated from the transmural ileum of patients with CD, including a trio of lesions from each patient: non-affected, inflamed, and stenotic ileum samples, and compared them with samples from patients without CD. Our computational analysis revealed that profibrotic signals from a subset of monocyte-derived cells expressing CD150 induced a disease-specific fibroblast population, resulting in chronic inflammation and tissue fibrosis. The transcription factor TWIST1 was identified as a key modulator of fibroblast activation and extracellular matrix (ECM) deposition. Genetic and pharmacological inhibition of TWIST1 prevents fibroblast activation, reducing ECM production and collagen deposition. Our findings suggest that the myeloid-stromal axis may offer a promising therapeutic target to prevent fibrostenosis in CD.


Comparison of IBD data sets
For Seurat label transfer, we used our integrated mesenchymal data and myeloid data as references for the corresponding subsets of the 3 published IBD data sets (3)(4)(5)(6).To transfer cell type labels, we employed the Seurat label transfer method.The "FindTransferAnchors" function was applied to identify anchors between the reference and each of the published datasets, utilising the CCA dimensionality reduction method.Subsequently, the "TransferData" function was used to transfer the labels from the reference dataset to the published datasets based on these anchors.Cell type classification probability was visualised with heatmaps.
To create a shared gene signature for FAP fibroblasts and Inflammatory monocytes in Inflammatory Bowel Disease (IBD), individual datasets were divided into two groups: myeloid cells (monocytes, macrophages, and dendritic cells) and mesenchymal cells (fibroblasts, myofibroblasts, and pericytes).The data were then analysed to identify clusters of cells that were similar to specific cell types such as FAP fibroblasts or Inflammatory monocytes.Next, significantly upregulated genes (with an adjusted p-value <0.05) in the Inflammatory monocytes or Inflammatory fibroblasts were identified within the respective compartments of each dataset using the FindMarkers function from Seurat.The genes common in all these lists of upregulated genes in inflammatory fibroblasts in each data set formed the common gene signature for Inflammatory monocytes or FAP fibroblasts in IBD (Supplemental data).Finally, a single Seurat object with raw counts was created for each cell compartment (myeloid or mesenchymal), retaining the original annotation and meta-data, and normalised.
The gene signature was then scored for each cell compartment using the AddModuleScore function from Seurat.

RNA isolation and gene expression
RNA was extracted utilising the micro plus Kit (Qiagen), following the manufacturer's guidelines.cDNA was then synthesised the High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific), as per the provided instructions.The RT-PCR process was carried out using the LightCycler 480 SYBR Green I Master (Roche) on a Light Cycler 480 instrument (Roche).The 2 −ΔCT method was employed to quantify the results, and gene expression levels of interest were normalised relative to the reference gene RPLP0.

Protein quantification
Supernatants from sorted myeloid cells were processed as described above.The Meso Scale Discovery (MSD) ELISA-based assay platform employed U-PLEX Custom Biomarker (hu) Assays (K15067M-1, LOT 473953) for quantifying protein concentrations of IL-1α, IL-1β, IFNγ, and TNFα, following the manufacturer's manual.Additionally, the U-PLEX Human TGF-β1 Assay (K151XWK-1, LOT 473954) was utilized to quantify the protein concentration of TGF-β1.The analyses were conducted using a 4PL fitting method to calculate concentrations.Immunofluorescent staining was imaged on the Operetta CLS high-content analysis system (PerkinElmer) and analysed on Harmony 4.5 software (PerkinElmer) at the VIB Leuven.

Immunofluorescence staining of IOs
IOs were washed with PBS and fixed with 4 % paraformaldehyde (PFA) for 20 minutes at room temperature and then washed three times with PBS.The IOs were dehydrated in 15 % sucrose at 4°C overnight.The IOs were frozen in Tissue Freezing Medium (Leica Biosystems).

Chronic DSS colitis
Animal studies (project P188/2019) were performed after approval by the KU Leuven Ethical Committee for Animal Experimentation.Mice (aged 10 weeks) used for this study were group-housed under controlled temperature (22°C) and photoperiod (12∶12-hour light-dark cycles) conditions and given unrestricted access to standard rodent diet and water (or DSS drinking solution).The mice were subjected to 3 cycles of 2% DSS (w/v) (40 kDa, DS001, TdB Labs) administration (7 days of DSS administration followed by 14 days of regular drinking water was defined as one cycle).The mice were monitored for signs of colitis each day (i.e., body weight loss, diarrhoea score and blood in the faeces).Hemoccult II® Fecal Occult Blood Test Kit (61200, Beckman Coulter) was used to detect blood in faeces.B6;129S7-Twist1 tm2Bhr /Mmnc mice were acquired from the Mutant Mouse Resource and Research Center (MMRRC_016842-UNC) (10).The COL1A2 Cre-ER mice were acquired from the Jackson Laboratory (IMSR_JAX:029567) (11).TWIST1 Δ/ΔCOL1A2 mice were generated by crossbreeding from Twist1 fl/fl mice and COL1A2 Cre-ER mice.Genotype were confirmed by PCR with the following sequence primer shown in Supplemental data.
Tamoxifen (3mg/mouse/day, T5648-5G, Sigma) in sesame seed oil was oral gavaged to Twist1 Δ/ΔCol1a2 mice and littermates daily for 5 consecutive days one week prior of every DSS cycle to induce Cre activity.Male wild type C57BL/6J mice were aquired from Envigo (Harlan) for evaluating anti-fibrotic properties of Harmine.Harmine (10 mg/kg) was dissolved in DMSO and Tween-80 and injected intraperitoneally from day 0 to day 10 during each DSS cycle.The groups without Harmine treatment received DMSO and Tween-80 intraperitoneally.
Mice were euthanised using CO 2 and colons collected for further analysis.

Picrosirius red staining of the mouse colons
Mouse colon was fixed in 4% formaldehyde, and biopsies were embedded in paraffin and cut into 5µm thick sections for histological analysis.In brief, the tissue slides were dewaxed by HistoChoice clearing agent (H2779-1L, Sigma-Aldrich) and rehydrated by the gradient decrease of ethanol solution and water.Then the tissue was stained in 0.1% of Direct Red 80 (365548-5G, Sigma-Aldrich) in picric acid solution (P6744-1GA, Sigma-Aldrich) for an hour, followed by two changes of acidified water and 100% ethanol.To quantify the relative histologic area of collagen on Masson's trichrome stained slides, we took an average of ten images in each sample on Leica DM2500 M and quantified them with Image J.

Histology score of mouse colon
A haematoxylin and eosin (H&E) staining was performed to determine the rate of damage and inflammation in the colon.Each section was blindly scored using a validated scoring system (12).Briefly, samples were checked for submucosal infiltration (0; none up to 3; large infiltrate), goblet cell loss (0; none up to 3; >50%), crypt density (0; normal up to 3; decreased by >50%) and crypt hyperplasia (0; none up to 3; >3-fold increase in crypt length).
The total score was calculated as 1x Goblet cell loss + 2x Crypt Density + 2x Hyperplasia + 3x Submucosal infiltrate and it is reflected in the Mouse Colitis Histology Index (MHCI).

Supplemental figures
Similarity was established based on Seurat label transfer (Figure S2I,).These subsets were annotated as Inflammatory fibroblasts/Activated fibroblasts or Inflammatory monocytes in various studies.Fibroblast subsets similar to FAP fibroblast were annotated as Inflammatory fibroblasts in Smillie et al., activated fibroblasts in Martin et al. and Inflammatory fibroblasts IL11 CHI3L1 in Kong et al.Myeloid subsets similar to inflammatory monocytes were as annotated as Inflammatory monocytes in Smillie et al., Inflammatory macrophages in Martin et al. and Monocytes S100A8 S100A9 in Kong et al.
3, eBioscience), and anti-E-Cadherin (1:300, 24E10, Cell Signaling Technology) antibodies at 4 °C overnight.The samples were washed three times with PBS and then incubated with secondary antibodies, including Alexa Fluor 555-conjugated goat anti-mouse IgG antibody (A32727, Invitrogen), Alexa Fluor 488-conjugated donkey anti-rat IgG antibody (A-21208, Invitrogen), and Alexa Fluor 647-conjugated goat anti-rabbit IgG antibody (A-21245, Invitrogen) for one hour at room temperature.The samples were washed with PBS three times and stained with DAPI for twenty minutes at room temperature.The samples were washed with deionised water and embedded in Mowiol.The samples were acquired with Zeiss LSM 780 confocal microscope with 10x magnification and analysed with Fiji ImageJ.

Figure S1 .
Figure S1.Single-cell profiling of fibro-stenotic ileal from CD and control ileum from CRC, related to Figure 1.(A) H&E and Masson's trichrome stain showing signs of inflammation and fibrosis.(B) the plot of histologic scores and collagen quantification in different lesions of terminal ileum.Data are shown as box and whisker plots.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (*p <0.05, ** p <0.01, *** p <0.005, ****p <0.001).(C) Pie plot showing proportions of cells from different lesions of terminal ileum.(D) Bar plot showing cell proportions in different lesions in CD's terminal ileum.(E) Bar plot showing percentage of cell clusters among the patients.(F) Heatmap showing core ECM genes in each cell type.

Figure S2 .
Figure S2.Heterogeneity of fibroblast in fibro-stenotic CD, related to Figure 2. (A) Bar plot showing proportions of mesenchymal subtypes in each terminal ileal lesion sample from each CD patient in the scRNA-seq data.(B) Contour plot showing gating strategy of fibroblast subsets by flow cytometry.(C) and (D) Percentage of fibroblast subsets in different lesions of terminal ileum are shown as bar plot with SEM.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (*p <0.05, **p <0.01).(E) Before-after plot showing FAP activity in different lesions of terminal ileum.Data is shown as a before-after plot.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (*p <0.05, **p <0.01).(F) Immunofluorescence staining for PDPN, ADAMDEC1, CD34 and FAP expression in healthy ileum and CD diseased ileum.Original image composed of stitched 25× images.(G) Heatmap depicting expression of chemokine and chemokine ligand genes in fibroblast subsets.(H) ssGSEA scores depicted as ridgeplots for selected terms, related to inflammation in fibroblast subsets.(I) Cross-dataset cell type prediction score heatmap showed similarity of stromal cell subset among published human IBD atlas.(J) Contour plot showing gating strategy of fibroblast subsets in colon by flow cytometry.(K) Percentage of fibroblast subsets in different lesions of colon from CRC (n=3), CD (n=3) and UC patients (n=4) are shown as bar plot with SEM.Statistically significant differences were determined using a oneway ANOVA test corrected with Tukey's multiple comparisons test (***p <0.005).(L) ssGSEA scores depicted as ridgeplots for selected terms, related to EMT in fibroblast subsets.(M) ssGSEA scores depicted as ridgeplots for selected terms, related to cellular senescence in fibroblast subsets.(N) Heatmap showing relative expression of senescence-associated secretory phenotype (SASP) markers.

Figure S3 .
Figure S3.Fibroblast-myeloid cell interaction modulates fibro-stenosis, related to Figure 3. (A) Partition-based graph abstraction (PAGA) analysis on fibroblast subsets showing most likely trajectories.(B) Heatmap showing gene expression change along cell Monocle3 pseudotime.(C) Cellphone DB dot plot showing ligand-receptor interactions between mesenchymal compartment and immune cells.First and second interacting molecules correspond to first and second cell types on the y axis respectively.Black circles indicate significant interactions (D)ssGSEA scores depicted as ridgeplots for selected terms in fibroblast subsets.

Figure S4 .
Figure S4.Heterogeneity of myeloid cells in fibro-stenotic CD, related to Figure 4. (A) Bar plot showing proportion of myeloid subsets in each terminal ileal lesion sample from each CD patient in the scRNA-seq data.(B) Selected terms from the Reactome biological pathway enrichment analysis for differentially upregulated genes in Inflammatory monocytes (logFC,>0.5;FDR, <0.1).(C) Heatmap showing number of interactions (Ligand-Receptor pairs) between myeloid and mesenchymal.(D) SCENIC showing relative transcription factor activity in each myeloid cell subset.(E) and (F) Pseudo-time trajectory projected onto a UMAP of selected myeloid subsets.(G) Contour plot showing gating strategy of myeloid subsets by flow cytometry among the different disease states.(H) Bar plots showing cell proportion of myeloid subsets in different lesions of terminal ileum.Data are shown as bar plots with SEM.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (*p <0.05, ** p <0.01, *** p <0.005, **** p <0.001).(I) Immunofluorescence staining for CD68, SLAMF1 (CD150) and FAP expression in healthy ileum and CD diseased ileum.Original image composed of stitched 25× images.(J) Crossdataset cell type prediction score heatmap showed similarity of myeloid cell subset among published human IBD atlas.(K) Contour plot showing gating strategy of myeloid subsets by flow cytometry among the colon from CRC (n=3), CD (n=3) and UC patients (n=4).(L) Bar plots showing cell proportion of CCR2 monocytes and CD150 Inflammatory monocytes in different lesions of colon.Data are shown as bar plots with SEM.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (**p <0.01, *** p <0.005, **** p <0.001).(M) Bar plots showing CD150 - monocytes and CD150 + monocytes in PBMC of healthy donor and CD patients.Data are shown as mean with SEM.Statistically significant differences were determined using a t-test.

Figure S5 .
Figure S5.Gene expression of FAP fibroblasts and Inflammatory monocytes in inflamed and stenotic ileum of fibro-stenotic CD patients, related to Figure 5. (A) Heatmap showing gene expression of cell clusters based on Molecular Cartography of transmural CD ileum (n=3).

Figure S6 .
Figure S6.CD150 + monocytes-derived cytokines promote FAP fibroblast activation and extra-cellular matrix protein deposition, related to Figure 6.(A) Experimental workflow showing FACS-sorting gating strategy and setup for in vitro ECM production.(B) Bar plot showing the protein expression of IL-1a, IL-1b, IFNg, TNFa and TGF-b1.Data are shown as bar plots with SEM.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (*p <0.05, ** p <0.01, *** p <0.005, **** p <0.001).(C) Heatmap showing relative expression of FAP, TWIST1 and type III collagen in CCD-18Co fibroblast treated with supernatant from FACS sorted myeloid subsets (n= 5 individual patients).(D) Experimental workflow.(E) Heatmap showing extracellular deposition of FAP, type I collagen and type III collagen in monocyte-stimulated CCD-18Co fibroblasts.(F) Heatmap showing relative protein expression of FAP, TWIST1 and type II collagen in primary ileal fibroblasts after being stimulated by selected cytokine combinations, predicted by NicheNet.(G) Experimental workflow.(H) Immunofluorescence staining (10× image) and (I) Bar plot showing relative expression of fibronectin and type IV collagen in pro-fibrotic cues-stimulated primary ileal fibroblasts after TWIST1 inhibition.Data are shown as bar plots with SEM.Statistically significant differences were determined using a one-way ANOVA test corrected with Tukey's multiple comparisons test (*p <0.05, ** p <0.01, *** p <0.005, **** p <0.001).

Figure S7 .
Figure S7.TWIST1 inhibition attenuates intestinal fibrosis, related to Figure 7. X-Y plots showing (A) disease activity index (DAI) (B) body weight change (C) stool blood score and (D) stool consistency through chronic DSS colitis in Twist1 Δ/ΔCol1a2 mice.Data are shown as mean with SEM.Statistically significant differences were determined using a two-way ANOVA test.(E) Picrosirius red staining and (F) Violin plot showing collagen deposition in mouse colon of chronic DSS colitis in Twist1 Δ/ΔCol1a2 mice.Data are shown as mean with SEM.Statistically significant differences were determined using a t-test.(G) Violin plot showing colon tissue size post chronic DSS colitis in Twist1 Δ/ΔCol1a2 mice.Data are shown as mean with SEM.Statistically significant differences were determined using a t-test.Violin plot showing (H) mouse colitis histology index (I) goblet cell loss score (J) crypt density (K) crypt

Figure S8 .
Figure S8.Previously published datasets using mucosal biopsies do not show consistent presence of FAP fibroblasts and inflammatory monocytes across IBD patients, related to Figure S2 and S4.(A) Dotplot showing gene signature score for common markers of inflammatory fibroblasts in 4 independent IBD data sets.The area between the two horizontal lines indicates the interquantile range of the gene signature score for the cells in Ke & Abdurahiman et al. (B) Dotplot showing gene signature score for common markers of inflammatory monocytes in 4 independent IBD data sets.The area between the two horizontal lines indicates the interquantile range of the gene signature scores for the cells in Ke & Abdurahiman et al.

Figure S1 .Figure S2 .
Figure S1.Single-cell profiling of fibro-stenotic ileal from CD and control ileum from CRC, related to Figure 1.

Figure S8 .
Figure S8.Previously published datasets using mucosal biopsies do not show consistent presence of FAP fibroblasts and inflammatory monocytes across IBD patients, related to Figure S2 and S4.A