Heterocellular OSM-OSMR signalling reprograms fibroblasts to promote pancreatic cancer growth and metastasis

Abstract

Pancreatic ductal adenocarcinoma (PDA) is a lethal malignancy with a complex microenvironment. Dichotomous tumour-promoting and -restrictive roles have been ascribed to the tumour microenvironment, however the effects of individual stromal subsets remain incompletely characterised. Here, we describe how heterocellular Oncostatin M (OSM) - Oncostatin M Receptor (OSMR) signalling reprograms fibroblasts, regulates tumour growth and metastasis. Macrophage-secreted OSM stimulates inflammatory gene expression in cancer-associated fibroblasts (CAFs), which in turn induce a pro-tumourigenic environment and engage tumour cell survival and migratory signalling pathways. Tumour cells implanted in Osm-deficient (Osm^-/-) mice display an epithelial-dominated morphology, reduced tumour growth and do not metastasise. Moreover, the tumour microenvironment of Osm^-/- animals exhibit increased abundance of α smooth muscle actin positive myofibroblasts and a shift in myeloid and T cell phenotypes, consistent with a more immunogenic environment. Taken together, these data demonstrate how OSM-OSMR signalling coordinates heterocellular interactions to drive a pro-tumourigenic environment in PDA.

Fig. 1. OSMR expression is associated with tumour-promoting inflammation and poor outcome in PDA.

a Heatmap displaying relative expression levels of membrane receptors of stellate cells adopting a quiescent, inflammatory cancer-associated fibroblasts (iCAF), or myofibroblastic CAF (myCAF) phenotype (dataset: GSE93313). b Violin plot of normalised Osmr expression levels in myCAF, iCAF, and apCAFs in murine PDA (dataset: GSE129455) with mean expression z-scores shown below. c Violin plots of normalised OSMR/Osmr expression levels for individual cell types in human (left) and murine (right) PDA (top) and mean expression z-scores shown below (datasets: GSE129455, phs001840.v1.p1). d Representative in situ mRNA hybridisation (ISH) of OSMR multiplexed with immunofluorescence of VIM and PanCK in human PDA (n = 3). 20× magnification. Upper left quadrant, full overlay; upper right, OSMR; lower left, OSMR with VIM; lower right, OSMR with PanCK. Scale bar = 100 μm. e OSM (left) and OSMR (right) expression in human PDA (TCGA PanCancer PAAD dataset, n = 178) and normal pancreatic tissue (GTEx, n = 171). Mantel–Cox test. T tumour, N normal, TPM transcripts per million. f Kaplan–Meier survival curves showing overall survival of 179 PDA patients from TCGA PanCancer PAAD dataset. Patients were stratified by high (Top 50%, n = 89) and low (bottom 50%, n = 90) OSMR expression levels. Analysis by log-rank test and cox-proportional hazard regression. HR hazard ratio. g Violin plots of inflammatory gene expression in Osmr^pos and Osmr^neg CAFs in murine PDA tumours (datasets: GSE129455, MTAB-8483) with mean expression z-scores shown below. h Violin plots of normalised OSM/Osm expression across cell types of human (top) and murine (bottom) PDA (datasets: GSE129455, phs001840.v1.p1). Mean expression z-scores shown. i Top 14 predicted interactions between macrophage-derived ligands and fibroblast expressed receptors using CellPhoneDB receptor–ligand interaction statistical analysis (p-values < 0.05). Also see Supplementary Fig. 1. Source data are provided as a Source Data file.

Fig. 2. Macrophage–tumour cell interactions induce inflammatory fibroblasts in vitro.

a Heatmap showing relative levels of soluble signals quantified by LUMINEX, clustered with unsupervised hierarchical clustering (n = 3 per culture condition). b Heatmap of relative expression levels of soluble signals measured by RT-qPCR, clustered with unsupervised hierarchical clustering (n = 4 per culture condition). c Volcano plot showing differentially regulated genes (DEGs) of PSCs isolated from PCC–PSC (n = 4) and PCC–PSC–MØ (n = 4) co-cultures. 3 W: PCC–PSC–MØ co-culture. 2 W: PCC–PSC co-culture. FC fold change. d Gene set enrichment analysis (GSEA) of RNAseq data from (c). NES normalised enrichment score, FDR false discovery rate. e RT-qPCR assay of selected myofibroblastic and inflammatory genes of PSCs stimulated with conditioned medium from 2 or 3 W co-cultures using different isolations of PCCs (PCC, PCC4 and PCC11). Left: Changes in expression state visualised by principle component analysis (PCA) across experimental conditions (n = 4). Right: Representative mRNA expression of inflammatory (Il6, Il4ra and Cxcl1) and myofibroblastic (Acta2) genes. Results displayed as mean ± SEM, n = 4, one-way ANOVA Tukey test. f RT-qPCR of selected myofibroblastic and inflammatory genes in PSCs stimulated with conditioned medium from 1, 2 or 3 W co-cultures of PCC, PSC and bone marrow-derived MØs. Left: Changes in expression state visualised by PCA plot across experimental conditions (n = 4). Right: Representative mRNA expression of inflammatory (Il6, Il4ra and Cxcl1) and myofibroblastic (Acta2) genes. Results displayed as mean ± SEM, n = 4, one-way ANOVA Tukey test. g Mass cytometry analysis of PSCs in mono- (n = 3) and co-culture with PCC (n = 3) or PCC and MØs (n = 3). Self-organising map clustering (FlowSOM) of PSCs in mono- and co-cultures displayed as t-Distributed Stochastic Neighbour Embedding (tSNE). h Mass cytometry analysis of PSCs from 3 W PCC–PSC–MØ co-culture. Number of macrophages plated in 3 W co-culture was titrated down and PDPN, CD73 and Integrin α₅ expression levels shown as biaxial plots. Also see Supplementary Fig. 2. Source data are provided as a Source Data file.

Fig. 3. Heterocellular OSM–OSMR signalling induces inflammatory fibroblasts.

a Cell-type-specific mRNA expression of Osm and Osmr across different culture conditions in vitro. Results displayed as mean ± SEM, n = 4, one-way ANOVA Tukey test. b RT-qPCR assay of myofibroblastic and inflammatory genes. PSCs were treated with recombinant OSM, conditioned medium or vehicle control, and gene expression analysis was performed. Left: Changes in expression state visualised by PCA. iCAF score is a sum of mean z-values of 15 selected inflammatory genes. Right: Representative mRNA expression of inflammatory and myofibroblastic genes of PSCs treated as indicated. Results displayed as mean ± SEM, n = 3, one-way ANOVA Tukey test. c RT-qPCR analysis of myofibroblastic and inflammatory genes in PSCs following pharmacological inhibitors and CRISPR–Cas9-medidated Osmr knockout (Osmr-KO) in PSCs. For inhibitor perturbation, PSCs were stimulated with conditioned medium supplemented with vehicle control, IKKβ inhibitor (TPCA-1, 1 µM), JAK2 inhibitor (AZD1480, 2.5 µM), control IgG antibody (2.5 ng/mL) or neutralising OSM antibody (αOSM, 2.5 ng/mL). For Osmr knockout, scrambled (Scr) control and Osmr-KO fibroblasts were stimulated with PCC–PSC–MØ conditioned medium and gene expression was analysed by RT-qPCR. Left: Changes in expression state visualised by a PCA plot. iCAF score is a sum of mean z-values of 15 selected inflammatory genes. Right: Representative mRNA expression of inflammatory (Il6 and Il4ra) and myofibroblastic (Thbs1 and Col4a6) genes of PSCs treated as indicated. Results displayed as mean ± SEM, n = 3, one-way ANOVA Tukey test. Also see Supplementary Fig. 3. Source data are provided as a Source Data file.

Fig. 4. Tumour cell-secreted GM-CSF increases macrophage secretion of OSM.

a Quantification of OSM levels in conditioned medium from bone-marrow-derived MØs stimulated with conditioned medium from PCC lines. Results displayed as mean ± SEM, n = 3. CM conditioned medium. b Quantification of GM-CSF levels in conditioned medium of PCC cell lines. Results displayed as mean ± SEM, n = 3. c Quantification of OSM levels in conditioned medium of bone-marrow-derived MØs stimulated with either recombinant murine GM-CSF (rm-GM-CSF) or PCC conditioned medium in the presence of neutralising GM-CSF antibodies (αGM-CSF-415, 1 µg/mL and αGM-CSF-7331, 1 µg/mL) or control. Results displayed as mean ± SEM (n = 3), one-way ANOVA Tukey test. d RT-qPCR assay of selected genes in PSCs stimulated with conditioned medium from 2 W (PCC–PSC) or 3 W co-cultures (PCC–PSCs–bone marrow-derived MØs) where recombinant murine GM-CSF substitutes for PCCs in the 3 W culture. Results displayed as mean ± SEM (n = 4), one-way ANOVA Tukey test. e Scatter plot of OSM (x-axis) and CSF2 (y-axis) expression levels in PDA patients (TCGA PAAD dataset, n = 179, two-tailed Spearman correlation test). RSEM RNA-seq by expectation-maximisation. Also see Supplementary Fig. 4. Source data are provided as a Source Data file.

Fig. 5. OSM supports the formation of an immunosuppressive microenvironment.

a Experimental workflow of in vivo experiments. b Protein levels of indicated soluble signals in wildtype (n = 7) and Osm^−/− (n = 7) tumours, measured by ELISA. Results displayed as mean ± SEM, two-tailed student t test. c Immunohistochemical analysis of αSMA in wildtype and Osm^−/− tumour sections. Left: Representative αSMA staining. Right: Quantification of αSMA^pos staining of tumour sections of wildtype (n = 6) and Osm^−/− (n = 6) animals. Results displayed as mean ± SEM, two-tailed student t test. d RT-qPCR gene expression analysis of FACS-isolated CAFs from wildtype (n = 5) and Osm^−/− (n = 5) tumours displayed as a heatmap (left). Number of isolated CAFs shown as bar plot (right). Results displayed as mean ± SEM, two-tailed student t test. e Immunohistochemical analysis of CD8 in wildtype (n = 6) and Osm^−/− (n = 6) animals. Upper: Total CD8^pos staining. Lower: Ratio of peripheral/central CD8^pos staining using the mean CD8^pos stain from four representative peripheral regions and four representative central regions per tumour. Results displayed as mean ± SEM, two-tailed student t test. f Mass cytometry analysis quantifying abundance of antigen presenting cells (APCs) expressing co-stimulatory markers in wildtype (n = 5) and Osm^−/− (n = 5) tumours. Results displayed as mean ± SEM. Unpaired two-tail student t test. g Mass cytometry analysis of CD45⁺/CD3⁺ T cells expressing indicated markers in wildtype (n = 5) and Osm^−/− (n = 5) tumours. Results displayed as mean ± SEM, two-tailed student t test. h, i Mass cytometry analysis of memory T cells expressing indicated exhaustion/dysfunction markers in wildtype (n = 5) and Osm^−/− (n = 5) tumours. Results displayed as mean ± SEM, two-tailed student t test. Tcm central memory, Tem effector memory, Trm resident memory. j Spearman correlation analysis between infiltration of indicated immune cells (determined by CIBERSORT) and OSMR expression levels in PDA patients from TCGA PanCancer PAAD dataset (n = 179). 95% confidence interval as shaded area. Also see Supplementary Figs. S5 and S6. Source data are provided as a Source Data file.

Fig. 6. PCC–PSC–MØ interactions reshape tumour cell signalling.

a Experimental workflow outlining SILAC-based phosphoproteomics analysis of stroma-regulated tumour cell signalling. “Light” (L), “Medium” (M) and “Heavy” (H) labelled PCCs were stimulated (5 min) with PCC or PCC–PSC or PCC–PSC–MØ conditioned medium, respectively (n = 5). Pooled L/M/H peptides (in a 1:1:1 ratio) were fractionated, phospho-peptides enriched and analysed by LC–MS/MS. b PhosphoPath phospho-network analysis summarising PCC–PSC–MØ-regulated tumour cell signalling, relative to PCC–PSC-regulated signals (p < 0.05, Fisher exact test with Benjamini–Hochberg multiple testing correction). Arrows indicate kinase-substrate relationships and blunt-ended edges indicate protein–protein interactions. c Western blots of phosho-STAT3 levels in PCCs treated with conditioned medium from 1, 2 or 3 W co-cultures (n = 2). d Western blots of phosho-STAT3 levels in PCCs treated with conditioned medium from 2 W or 3 W co-cultures containing either Osmr-KO or Scr-PSCs (n = 4). e Single-sample gene set enrichment analysis (ssGSEA) of PDA patients from TCGA PanCancer PAAD dataset (n = 179). Patients were stratified into OSMR^high (n = 89) and OSMR^low (n = 90) groups. f Representative western blot of EMT markers in PCCs treated with conditioned medium from 1, 2, or 3 W co-cultures; E-cadherin, Vimentin, SNAIL, and SLUG (n = 3). Also see Supplementary Fig. 7. Source data are provided as a Source Data file.

Fig. 7. OSM drives tumour growth and metastasis in vivo.

a Left: Representative iRFP imaging of resected pancreatic tumours from orthotopically transplanted immune-competent wildtype (Wt) (n = 19) and Osm^−/− (n = 13) mice. Right: Quantification of iRFP signal intensity, tumour volume and tumour weight. Results displayed as mean ± SEM (Wt, n = 19 and Osm^−/− n = 13), two-tailed student t test. b Left: Representative Pan-Cytokeratin (PanCK) staining of wildtype and Osm^−/− tumour sections. Right: Quantification of total PanCK^pos staining, mesenchymal PanCK^pos staining, and epithelial PanCK^pos staining following morphology analysis, and the epithelial–mesenchymal PanCK^pos ratio in tumour sections from wildtype (n = 6) and Osm^−/− (n = 6) mice. A shift of PanCK^pos cells from a mixed mesenchymal and epithelial morphology in wildtype tumours to a more epithelial-dominant morphology in Osm^−/− animals. Results displayed as mean ± SEM, two-tailed student t test. c Quantification of total Ki67,^pos CD31^pos, and SLUG^pos immunohistochemical staining of tumour sections from wildtype (n = 6) and Osm^−/− (n = 6) mice. Results displayed as mean ± SEM, two-tailed student t test. d Upper: Representative iRFP in vivo imaging (top) and haematoxylin and eosin staining (bottom) of resected livers from wildtype and Osm^−/− mice. Lower: Quantification of iRFP^pos liver metastasis incidents in wildtype (n = 11/19) and Osm^−/− (n = 0/13) animals. e Model outlining heterocellular OSM–OSMR signalling between pancreatic cancer cells, fibroblasts and macrophages. Pancreatic cancer cell secretion of GM-CSF induces macrophage secretion of OSM, which reprograms fibroblasts to an inflammatory phenotype, in turn accentuating tumour growth and metastasis. Also see Supplementary Fig. 8. Source data are provided as a Source Data file.