Elucidating the fundamental fibrotic processes driving abdominal adhesion formation

Abstract

Adhesions are fibrotic scars that form between abdominal organs following surgery or infection, and may cause bowel obstruction, chronic pain, or infertility. Our understanding of adhesion biology is limited, which explains the paucity of anti-adhesion treatments. Here we present a systematic analysis of mouse and human adhesion tissues. First, we show that adhesions derive primarily from the visceral peritoneum, consistent with our clinical experience that adhesions form primarily following laparotomy rather than laparoscopy. Second, adhesions are formed by poly-clonal proliferating tissue-resident fibroblasts. Third, using single cell RNA-sequencing, we identify heterogeneity among adhesion fibroblasts, which is more pronounced at early timepoints. Fourth, JUN promotes adhesion formation and results in upregulation of PDGFRA expression. With JUN suppression, adhesion formation is diminished. Our findings support JUN as a therapeutic target to prevent adhesions. An anti-JUN therapy that could be applied intra-operatively to prevent adhesion formation could dramatically improve the lives of surgical patients.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1. JUN promotes adhesions and upregulates PDGFRA expression.

a Representative samples of hematoxylin and eosin (H&E) stained abdominal adhesion tissue specimen from JUN+/+ (right panel), JUN+/− (middle panel) and wild-type mice (left panel). Green dotted lines outline adhesion interface, structures as labeled in figure. n > 10 biological replicates. Scale bars, 100 μm. b Application of an objective histologic adhesion rating score by blinded pathologists (based on to the gross score used by Tsai et al. (2018) and the histologic score used by Linsky et al., (1987)) quantifies relative adhesion severity in wild-type, JUN+/−, and JUN+/+ mouse specimens. n = 10 biological replicates. c Schematic of the PDGFRAGFP mouse construct. d Fluorescent imaging of PDGFRAGFP mouse uninjured (control) visceral and parietal peritoneum. Structures as labelled in figure, white dotted lines outline area of potential adhesion interface, POD postoperative day. n = 10 biological replicates. Scale bar, 100 um. e Fluorescent imaging of PDGFRAGFP mouse adhesion tissue at POD 7 (top panel), POD 14 (middle panel—visceral-parietal adhesion, bottom panel—visceral-visceral adhesion). Structures as labelled in figure, white dotted lines outline adhesion interface. GFP green fluorescent protein. n > 5 biological replicates. Scale bars, 100 μm. f Quantitation of GFP+(PDGFRAGFP) cells per high power field (HPF) in the adhesion interface in mouse adhesions (from Fig. 1d, e). n = 5 biological replicates. g Immunofluorescent staining of representative samples shows PDGFRA and phospho (p)-JUN colocalization (left panels) and independent expression (right panels) within the adhesion interface. Individual panels at top, merge in bottom row, white dotted lines highlight cells of interest. Scale bars, 25 μm. Data and error bars represent means ± standard deviation (SD). *P = 0.0009 (one-way Anova), **P = 0.0001 (one-way Anova). Source data are provided as a Source Data file.

Fig. 2. Local fibroblasts proliferate polyclonally to form adhesions.

a Parabiosis schematic. b H&E and IF data showing the presence of an abscess identified along the abdominal wall adjacent to the adhesion interface in a mouse parabiont, which is strongly GFP+ (green fluorescent protein) secondary to the presence of circulating immune cells in the abscess. c No GFP+ cells were identified in the adhesion interface. Green (GFP) represents circulating cells, purple (CD45) stains for immune cells. n = 3 biological replicates. Scale bars, 50 μm. d Schematic of the ActinCreER::ROSA26VT2/GK3 Rainbow mouse construct. e Schematic showing ActinCreER::ROSA26VT2/GK3 Rainbow mice locally induced with activated tamoxifen liposomes (LiTMX) at time of adhesion formation, using a published protocol, ERT2 estrogen receptor T2. f ActinCreER::ROSA26VT2/GK3 Rainbow mouse uninjured control (top panel) and adhesion tissue harvested at POD 14 (bottom panel). Clonal proliferation of fibroblasts are visualized along the adhesion interface. Representative samples, structures as labelled in figures, white dotted lines outline adhesion, white asterisk marks adhesion interface, confocal imaging. n = 5 biological replicates. Scale bars, 25 μm. g Schematic of the PDGFRACreER::ROSA26VT2/GK3 Rainbow mouse construct. h Confocal imaging of representative PDGFRACreER::ROSA26VT2/GK3 Rainbow mouse adhesion samples showing cellular clonality in the adhesion interface at postoperative day (POD) 7 (top panels, Imaris rendering at right), and POD 14 (bottom panels, Imaris rendering at right). Structures as labelled in figures, thick white dotted lines outline adhesion interface, thin white dotted lines outline individual clones, confocal imaging. n = 5 biological replicates. Scale bars, 50 μm.

Fig. 3. Adhesion fibroblasts derive primarily from the viscera.

a Schematics of the PDGFRAGFP::ROSA26mTmG mouse “donor” model (left panel) and PDGFRAGFP “recipient” model (right panel). b Schematic for abdominal wall transplant model using abdominal wall from PDGFRAGFP::ROSA26mTmG mice, transplanted into PDGFRAGFP mice, followed by adhesions surgery. c Confocal imaging of representative abdominal wall transplant model mouse adhesion tissue shows a prominence of GFP+ cells within the adhesion interface, relative to GFP-mTomato+ cells, harvested at POD 14. White dotted lines mark structures as labelled in images, confocal imaging. n = 5 biological replicates. Scale bar, 100 μm. d Confocal imaging of the abdominal wall transplant model representative tissue shows PDGFRA/ASMA co-expression (PDGFRA labelled with GFP using the mTmG mouse model, IF staining for ASMA) among the cells migrating from the visceral peritoneum into the adhesion interface. White dotted lines mark structures as labelled in images, confocal imaging. n = 5 biological replicates. Scale bars, 100 μm.

Fig. 4. Adhesion fibroblasts upregulate EMT and show heterogeneity.

a Principal component analysis (PCA) plot comparing bulk RNA-seq gene expression for adhesion (n = 4) and control peritoneum (n = 4) FACS-isolated mouse fibroblasts. Colors as labelled, variances noted on plot. b Heatmap of mouse adhesion-forming fibroblast bulk RNA-seq data shows significant differential gene expression between adhesion and control peritoneum (sham surgery) cohorts. Upregulated EMT-pathway genes noted at right. Gene enrichment as noted in figure, color key, and histogram at far right. c Quantitation of qPCR for vimentin (VIM) and collagen 1a2 (COL1A2) shows upregulation of gene expression in the context of mouse abdominal adhesions. Data and error bars represent means ± SD; P-values noted in figure, unpaired two-tailed t-test. n = 3 replicates per condition, datapoints represent average of technical replicates. d Uniform manifold approximation and projection (UMAP) plot showing single-cell (sc) RNA-seq data from mouse adhesion fibroblasts FACS-isolated using an unbiased, lineage-negative sort strategy at POD 2 (n = 4) and POD 7 (n = 4) following adhesion induction. Three unique clusters of fibroblasts are identified. Colors as labelled in the figure panel. e UMAP plot showing distribution of mouse scRNA-seq fibroblasts in terms of harvest timepoint relative to the clusters in panel d. Cells isolated at both timepoints are represented in all clusters. Colors as labelled in the figure panel. f Pseudotime analysis (Monocle 2) of mouse scRNA-seq data: Pseudotime analysis (left panel), representation of scRNA-seq clusters across the pseudotime analysis shows a clear progression from cluster 1 to clusters 0 and 2 (middle panel) and relative to timepoints, the cells follow a logical time progression that mirrors the pseudotime with the largest representation of POD 2 cells in cluster 1 and more POD 7 cells in clusters 0 and 2 (right panel). Arrows indicated direction of pseudotime progression. g Violin plots showing expression of STAT5 and ASMA within the scRNA-seq data. Colors and numbering on x-axis match cluster colors assigned in panel d. h Additional violin plots showing expression of JUN, STAT3, FSP1, IL6, MCP1, and PDGFRA relative to the scRNA-seq data clusters seen in panel d. Colors and numbering on x-axis match cluster colors assigned in panel d. Source data are provided as a Source Data file.

Fig. 5. JUN is an early promotor of abdominal adhesion fibrosis.

a Schematic illustrating the targeting construct used in the doxycycline (dox)-inducible JUN mouse model. In this construct, rtTA is expressed at the endogenous ROSA26 promotor. With dox induction, rtTA undergoes nuclear translocation to activate the Tet-responsive element (minCMV-tet(o)) driving expression of JUN. SA splice acceptor, pA poly(A) sequence. b Representative phospho-flow-cytometry analysis for phosphorylated (phospho-) JUN (left and middle panels), and phospho-STAT5 (right panel) expression in abdominal adhesion fibroblasts isolated from JUN mice 24 h after adhesion surgery (with local induction with doxycycline at the time of adhesion formation) compared with vehicle control. n = 3 biological replicates. c Immunofluorescent assessment of phospho-JUN and phospho-STAT5 co-expression in mouse abdominal adhesion tissue. Thick white dotted line indicates edge of adhesions interface, co-expressing cells highlighted with thin white dotted lines. n = 3 biological replicates. Scale bars, 25 μm. d Quantitation of phospho-flow-cytometry analysis showing a significant decrease in phospho-JUN (top panel) and phospho-STAT5 (bottom panel) expression with application of JUN inhibitor versus vehicle control in JUN mice at 24 h. n = 3 biological replicates per condition, datapoints represent averages of technical replicates. e Representative gross images of mouse adhesions at POD 3 treated with (vehicle, DMSO) control (left panels) versus JUN inhibitor (right panels). Adhesion interface highlighted in green, structures as indicated in figures, black sutures (circled with blue dotted line) visible on inhibitor specimens are nidus for adhesion formation (these are not visible on control specimens as they are covered with bowel that is adhesed to the abdominal sidewall). n ≥ 3 biological replicates per condition per treatment. f Gross assessment (using an adhesion severity grading score established by Tsai et al. 2018) of adhesion severity following in vivo inhibition of JUN using JUN inhibitor (T-5224) versus vehicle control in JUN (JUN expression induced with doxycycline in all JUN mice used) and wild-type mice at POD 3. n ≥ 3 biological replicates per condition per treatment. Data and error bars represent mean ± SD. *P = 0.01, **P = 0.0001, ***P = 0.004, ****P = 0.003, *****P = 0.01, ******P = 0.009, unpaired two-tailed t-test. Source data are provided as a Source Data file.

Fig. 6. Functional modulation of JUN regulates adhesion formation.

a, b Representative H&E sections for vehicle control (left panels) and JUN-inhibitor-treated (right panels), in JUN (JUN expression induced with doxycycline in all JUN mice used) (a) and wild-type (b) mice. Adhesion interfaces outlined with green dotted lines. Structures as labelled in figure. Scale bars, 100 μm. n = 5 biological replicates per condition per treatment. c Histologic scoring (as used in Fig. 1b) of adhesion tissue following in vivo inhibition of JUN using JUN inhibitor versus vehicle control in JUN (JUN expression induced with doxycycline in all JUN mice used) and wild-type mice. n = 5 biological replicates per condition per treatment. d Representative images of trichrome staining of JUN mouse vehicle control (top panel) and JUN-inhibitor-treated adhesion specimen (bottom panel). JUN expression induced with doxycycline in all JUN mice used. Adhesion interface outlined with black dotted lines. Structures as labelled in figure. n = 5 biological replicates per treatment. Conditions and structures at noted in figure panels. Scale bars, 100 μm. e Representative figures showing IF staining for phospho-JUN and PDGFRA in adhesion tissue following in vivo inhibition of JUN (using JUN inhibitor, T-5224) in JUN (JUN expression induced with doxycycline in all JUN mice used) mice. Conditions and structures at noted in figure. Scale bars, 50 μm. f Quantification of p-JUN+ cells in e. HPF high power field. n = 3 biological replicates assessed per treatment condition, datapoints represent average of multiple measures per replicate. Data and error bars represent mean ± SD. *P = 0.001, **P = 0.0001, ***P = 0.0001, unpaired two-tailed t-test. Source data are provided as a Source Data file.

Fig. 7. Human adhesions recapitulate biology and gene expression.

a Representative human abdominal adhesion tissue (n = 24 human adhesion specimens) histology (H&E—left, trichrome—middle, picrosirius red—right). Adhesion interface bounded by yellow dotted lines and labeled. Scale bars, 100 μm. b On unbiased-FACS analysis, PDGFRA expression is significantly upregulated in human abdominal adhesion fibroblasts, compared with control peritoneum. CD26 expression is also upregulated, although not significantly. Conditions as labelled in figure. n = 5 biological replicates per condition. c Representative IF staining of human abdominal adhesion tissue for JUN and PDGFRA, right panel is zoom of indicated region in white. Adhesion interface outlined with thick white dotted lines (left panel), colocalization of PDGFRA and JUN expression highlighted with thin white dotted lines (right panel). n = 3 biological replicates. Scale bars, 50 μm. d Quantitation of p-JUN+ cells in human abdominal adhesion tissue pictured in panel c, compared with human control peritoneum tissue. n = 3 biological replicates. e Representative IF staining of human abdominal adhesion tissue shows colocalization of PDGFRA with collagen 1 (COL1) and collagen 3 (COL3). Adhesion interface outlined with white dotted lines. Structures as labelled in figure. n = 5 biological replicates. Scale bar, 50 μm; ×10 zoom at right. f Quantitation of cytokine production (including IL6, MCP-1, PDGF-AA, and IL8) by fluorescent assessment of the cell supernatant from primary human abdominal adhesion fibroblasts in vitro, measured 24 and 48 h after isolation. Values normalized to cell-free media for each cytokine assessed. MFI median fluorescence intensity. Conditions as labelled in figure. n = 3 replicates analyzed per condition per timepoint. g PCA plot of human bulk RNA-seq data shows distinct clustering of human fibroblast specimens FACS-isolated from human abdominal adhesion (n = 6) and control peritoneal tissues (n = 3) (colors as indicated, variances noted on plot). h Heatmap of human adhesion-forming fibroblasts shows significant differential gene expression between conditions. Highly expressed EMT and JUN kinase GSEA pathway genes highlighted in yellow and blue panels, respectively, at right. Color key and histogram at far right. Data and error bars represent means ± SD.*P = 0.04, **P = 0.0078, unpaired two-tailed t-test. Source data are provided as a Source Data file.

Fig. 8. Human adhesion fibroblasts are heterogeneous and JUN dependent.

a Uniform manifold approximation and projection (UMAP) plots showing single-cell (sc)RNA-seq data from human adhesion fibroblasts FACS-isolated using an unbiased, lineage-negative sort strategy from three unique human specimens. Four unique clusters are identified. b UMAP plot showing representation of individual patient samples (arbitrarily numbered) across the cluster presented in 1. Human hashtag labels indicated at right, 2 hashtag antibodies were used for patient 33 as an internal control, as noted in the figure panel legend. c Pseudotime analysis of human abdominal adhesion fibroblast scRNA-seq data. Colors match cluster colors assigned in panel a. d Quantitation of qPCR analysis for JUN, STAT5, STAT3, and SPP1 of vehicle control versus JUN CRISPR Cas9-knockdown human abdominal adhesions fibroblasts. P-values noted in figure. n = 3 replicates per condition, datapoints represent average of technical replicates. e Quantitation of Ki67 expression using ICC of primary human abdominal adhesions fibroblasts treated with CRISPR Cas9 JUN knockdown compared with vehicle control. n = 4 biological replicates assessed per condition. f Quantitation of collagen type 1 expression using ICC of primary human abdominal adhesions fibroblasts treated with virally mediated JUN overexpression, vehicle control, or CRISPR Cas9 JUN knockdown. n = 3 biological replicates assessed per condition. g Schematic (based on published KEGG pathways) displays proposed JUN-relevant signaling pathways identified in this study. The left panel shows the acute phase response following tissue injury by which JUN is initially activated. The right panel shows the chronic profibrotic state that is established in adhesion fibroblasts. Blue circle highlights AP-1, red indicates the role of JUN inhibitor. Colors as labelled in the figure. ECM extracellular matrix. Data and error bars represent means ± SD. HPF high power field. A.U. arbitrary units. *P = 0.04, **P = 0.03, ***P = 0.03, unpaired two-tailed t-test. Source data are provided as a Source Data file.