Stromal oncostatin M cytokine promotes breast cancer progression by reprogramming the tumor microenvironment

The tumor microenvironment (TME) is reprogrammed by cancer cells and participates in all stages of tumor progression. The contribution of stromal cells to the reprogramming of the TME is not well understood. Here, we provide evidence of the role of the cytokine oncostatin M (OSM) as central node for multicellular interactions between immune and nonimmune stromal cells and the epithelial cancer cell compartment. OSM receptor (OSMR) deletion in a multistage breast cancer model halted tumor progression. We ascribed causality to the stromal function of the OSM axis by demonstrating reduced tumor burden of syngeneic tumors implanted in mice lacking OSMR. Single-cell and bioinformatic analysis of murine and human breast tumors revealed that OSM expression was restricted to myeloid cells, whereas OSMR was detected predominantly in fibroblasts and, to a lower extent, cancer cells. Myeloid-derived OSM reprogrammed fibroblasts to a more contractile and tumorigenic phenotype and elicited the secretion of VEGF and proinflammatory chemokines CXCL1 and CXCL16, leading to increased myeloid cell recruitment. Collectively, our data support the notion that the stromal OSM/OSMR axis reprograms the immune and nonimmune microenvironment and plays a key role in breast cancer progression.


Introduction
The tumor microenvironment (TME), composed of different cell types (e.g., fibroblasts, adipocytes, endothelial and infiltrating immune cells), harbors complex cell interactions that are often manipulated and hijacked by tumor cells in every step of cancer progression (1). Tumor cells corrupt the local microenvironment and promote the recruitment of primarily immune-suppressor cells from circulation (1). In addition, growth factors released by cancer and stromal cells (including serum growth factors) have an important role in tumor proliferation and malignant progression (2). However, the contribution of stromal cells to the reprogramming of the TME is poorly understood. Cancer-associated fibroblasts (CAFs) are a key cell population in the tumor stroma. Apart from their very well-known functions in matrix deposition and extracellular matrix remodeling, CAFs have recently been shown to interact with the immune system by responding to and secreting chemokines and cytokines (3). They differ from non-CAFs in multiple aspects and have distinctive properties, including a particular The tumor microenvironment (TME) is reprogrammed by cancer cells and participates in all stages of tumor progression. The contribution of stromal cells to the reprogramming of the TME is not well understood. Here, we provide evidence of the role of the cytokine oncostatin M (OSM) as central node for multicellular interactions between immune and nonimmune stromal cells and the epithelial cancer cell compartment. OSM receptor (OSMR) deletion in a multistage breast cancer model halted tumor progression. We ascribed causality to the stromal function of the OSM axis by demonstrating reduced tumor burden of syngeneic tumors implanted in mice lacking OSMR. Single-cell and bioinformatic analysis of murine and human breast tumors revealed that OSM expression was restricted to myeloid cells, whereas OSMR was detected predominantly in fibroblasts and, to a lower extent, cancer cells. Myeloid-derived OSM reprogrammed fibroblasts to a more contractile and tumorigenic phenotype and elicited the secretion of VEGF and proinflammatory chemokines CXCL1 and CXCL16, leading to increased myeloid cell recruitment. Collectively, our data support the notion that the stromal OSM/OSMR axis reprograms the immune and nonimmune microenvironment and plays a key role in breast cancer progression.
Stromal oncostatin M cytokine promotes breast cancer progression by reprogramming the tumor microenvironment increased levels of apoptosis but a similar degree of proliferation (Supplemental Figure 1, F and G). Finally, OSMR deficiency produced a remarkable reduction in the percentage of animals with lung metastasis (Figure 1, H and I).
These results show that OSM signaling is causally associated with tumor aggressiveness but, surprisingly, by using syngeneic cancer models, we found that this association requires, at least in part, the presence of the OSM/OSMR axis in the tumor stroma. We injected TS1 cells, derived from a MMTV-PyMT tumor (18), orthotopically into the mammary gland of syngeneic Osmr-KO and wild-type (WT) control mice (Figure 2A). This model allows the assessment of the contribution of stromal OSM signaling to cancer progression, as OSMR is only depleted in the TME, while TS1 cancer cells express OSMR that can be activated by host-derived OSM (Supplemental Figure 2, A-C). Depletion of OSMR in the TME resulted in delayed tumor onset and tumor growth (Figure 2, B-E, and Supplemental Figure 2, B and C), confirming that stromal OSMR signaling contributes to cancer progression. Conversely, OSMR depletion in cancer cells by the CRISPR/Cas9 technique did not show any effect on tumor onset and tumor growth in WT animals (Supplemental Figure 2, D-G).
Analyses of published gene expression profiles of breast cancer demonstrated that both OSM and OSMR are increased in human breast cancer stroma, compared with the cancer epithelial compartment and healthy stroma (Figure 2, F and G). A similar pattern of OSM/OSMR expression was observed in other cancer types, including colorectal and ovarian cancers (Supplemental Figure 3A). We also observed that increased OSM mRNA levels associated with decreased disease-free survival in the Metabric (19) and Wang (20) breast cancer data sets (Supplemental Figure 3B). Analysis of The Cancer Genome Atlas (TCGA) data by Kaplan-Meier Plotter (21) showed that high OSM levels were significantly associated with worse overall survival in other cancer types (Supplemental Figure 3C).
The OSM/OSMR signaling module exhibits a distinct microenvironment-restricted expression. As we found an unexpected contribution of the stromal OSM/OSMR axis to breast cancer progression, we performed single-cell RNA sequencing (scRNA-seq) analysis of mammary tumors from the MMTV-PyMT model to decipher which cells were responsible for producing OSM and expressing OSMR in the breast cancer context ( Figure 3A). Our data indicate that Osm is almost exclusively expressed by the myeloid cell population, while Osmr is mainly expressed by fibroblasts and some of the cancer epithelial clusters (Figure 3, A-C). The OSM/OSMR signaling module exhibits a distinct microenvironment-restricted expression and it differs from the one observed for other cytokine-receptor pairs of the same family such as IL-6/IL-6R and LIF/LIFR (Figure 3, B and C), supporting the notion that OSM exerts distinct and unique functions from other members of the family (22). Il6st (GP130) encodes the common receptor subunit for OSM, IL-6, LIF, and other cytokines of the family and is ubiquitously expressed (Figure 3, B and C), while the expression of the other receptor subunits is more restricted and tightly regulated. Reverse transcription quantitative PCR (RT-qPCR) analysis of FACS-isolated breast TS1 orthotopic tumors (18) confirmed expression of Osm in the myeloid population and expression of Osmr in fibroblasts and in cancer cells (Supplemental Figure 4, cytokine and chemokine secretory profile (4). However, CAFs are very heterogeneous and different subsets of functional fibroblasts have been proposed, some with predominant secretory functions and some with a prominent matrix remodeling phenotype (3).
Here, we discovered that the cytokine oncostatin M (OSM) acts as a central regulator of the crosstalk between immune cells, CAFs, and cancer cells, and that these immune-stromal-cancer cell interactions favor breast cancer progression and metastasis. OSM belongs to the IL-6 family, which is considered one of the most important cytokine families in the process of tumorigenesis and metastasis (5). IL-6 and OSM are acute-phase mediators of inflammation mainly produced by activated leukocytes. They can activate both epithelial and stromal cells to produce a wide array of additional inflammatory mediators (6). Nevertheless, the role of OSM in the TME remains unclear. OSM was first described as an antitumoral cytokine, owing to its antiproliferative effect in melanoma and other cancer cells (7). However, in recent years, OSM has been associated with tumor progression, as it induces epithelial-mesenchymal transition (EMT), cancer stem cell-like features, cell migration, and metastasis in animal and cellular models (8,9). High OSM receptor (OSMR) expression in clinical samples of glioblastoma, breast cancer, and cervical cancer correlates with decreased survival in those patients (10)(11)(12). However, the information regarding the role of OSM signaling in the TME is scarce, restricted to reports describing increased mRNA expression in this compartment (13,14) and a role for OSM in macrophage M2 polarization (15).
In this study, we used samples from human primary breast tumors, transgenic and orthotopic mouse models of breast cancer, genetically modified mice lacking OSM signaling, single-cell analysis, and in vitro cultures to demonstrate that OSM is a central node for multicellular interactions within the breast TME.

Results
The stromal OSM/OSMR axis promotes breast cancer progression. First, we studied the contribution of OSM signaling in the genetic mouse model MMTV-PyMT, which is widely used to study breast cancer progression in a fully competent TME and immune system (16). We crossed Osmr-deficient mice (Osmr-KO) with MMTV-PyMT as illustrated by the experimental scheme in Figure 1A.
Osmr-KO mice are viable but show mild defects in acute inflammation, liver regeneration, thymic hypoplasia, and net metabolism of bone and fat (17), suggesting that OSMR deficiency is not completely compensated. MMTV-PyMT Osmr-KO females showed a significant delay in tumor onset, tumor growth, and a reduced tumor burden at 14 weeks of age ( Figure 1, B-D, and Supplemental Figure 1, A-D; supplemental material available online with this article; https://doi.org/10.1172/JCI148667DS1). Importantly, OSMR deletion also reduced the malignancy of the tumors, assessed by histopathological analysis, as it reduced the percentage of mice with malignant carcinomas and increased the proportion of mice with premalignant adenomas/mammary intraepithelial neoplasia (MIN) or no tumors ( Figure 1E and Supplemental Figure 1E; P value = 0.007 for χ 2 test comparing malignant lesions versus premalignant lesions or no lesions). Interestingly, when compared with their controls, tumors in Osmr-KO mice showed decreased levels of the extracellular matrix protein fibronectin, predominantly produced by CAFs (4) (Figure 1, F and G), and immune cell lines (Supplemental Figure 5B). To demonstrate the relevance of our previous findings in human cancer clinical data, we used the TIMER (25) and xCell (26) web resources to analyze the association between OSM and OSMR expression and TME composition in 2 different clinical breast cancer data sets (21,27). TIMER analysis showed that OSMR mRNA expression significantly correlates with fibroblast enrichment in human breast cancer ( Figure 4B), while OSM mRNA levels show the most significant associations with myeloid macrophage and neutrophil populations ( Figure 4C). This analysis also showed that OSM and OSMR mRNA expression inversely correlated with tumor cell purity. The OSMR and OSM associations with fibroblast and myeloid cell infiltration, respectively, were validated by xCell in a different clinical data set ( Figure 4D). A similar pattern of OSM/OSMR expression was observed in FACS-isolated colorectal tumors (Supplemental A-C). Similar results were obtained when analyzing FACS-isolated populations of MMTV-PyMT tumors (Supplemental Figure  4D). An identical pattern of OSM/OSMR expression is maintained in the human setting, as demonstrated by RT-qPCR quantification in a large panel of human cell lines and analysis of RNA-seq data from the Human Protein Atlas (ref. 23, Figure 4A, and Supplemental Figure 5, A and B). OSM mRNA expression was restricted to undifferentiated and macrophage-like differentiated HL-60 cells (ref. 24 and Supplemental Figure 5A) and lymphoid and myeloid cell lines (Supplemental Figure 5B). Conversely, OSMR was only detected by RT-qPCR in breast cancer cells and fibroblasts, showing significantly higher expression in fibroblasts compared with epithelial cells ( Figure 4A and Supplemental Figure 5A). Analysis of a battery of human cell lines (23) confirmed expression of OSMR only in epithelial, endothelial, and fibroblast cell lines and not in Two independent experiments were performed, and the results were combined in B, D, and E. (F and G) OSM and OSMR mRNA expression in paired cancer epithelial versus cancer stroma (F, GSE10797) and normal stroma versus cancer stroma breast cancer samples (G, GSE9014). Data were downloaded from NCBI GEO data sets. P values were calculated using the Mantel-Cox test (B), 2-way ANOVA with post hoc Sidak's multiple-comparison test (C), or unpaired, 2-tailed Student's t test (D-G). ***P < 0.001, ****P < 0.0001 for experiment 1 and # P < 0.05, ## P < 0.01, #### P < 0.001 for experiment 2.
tility (4). Importantly, OSM treatment enhanced the capacity of CAFs (CAF-173 and CAF-318) to contract collagen matrices and, interestingly, this effect was not observed in noncancerous skin and breast fibroblasts (HS27 and RMF-31, respectively) ( Figure 5, A and B). The effect of OSM in CAF contractility was blocked by pretreatment with the highly potent inhibitor of Rho-associated kinase (ROCK), Y-27632, and could not be reproduced by LIF, a member of the OSM cytokine family (Supplemental Figure 6A). In agreement, OSM activated molecular markers of actomyosin contractility (MLC2 and FAK) in CAF-173, but not in normal RMF-31 fibroblasts, and the effect was mediated by ROCK (Supplemental Figure 6, B-F). To further investigate the role of OSM in potentiating CAF activation, we selected RMF-31 to be used as a model Figure 5C). Altogether, our data reveal that OSM and OSMR are stroma-expressed molecules, and point to paracrine OSM/OSMR signaling in cancer, as ligand and receptor are expressed by different cell types in the TME.
OSM activates CAFs, promoting tumor progression. As we previously observed that fibroblasts were the cell population with higher levels of OSMR within the tumor (Figures 3 and 4 and Supplemental Figures 4 and 5), we performed complementary in vitro and in vivo experiments to assess the effect of OSMR activation in mammary CAFs and normal fibroblasts derived from human breast tumors and reduction mammoplasty surgeries, respectively (28). The ability to remodel the extracellular matrix is a hallmark of CAFs that depends on the activation of actomyosin contrac-  Figure 5E). Of interest, OSMR was similarly expressed in normal fibroblasts and CAFs ( Figure 4A) and the pathway was functional in both cell types, as suggested by OSM induction of OSMR expression in both cell lines (Supplemental Figure 7A), a classical hallmark of OSMR activation (29). Gene set enrichment analysis (GSEA) of transcriptomic data of CAF-173 treated with OSM or vehicle showed that OSM induced signatures related to fibroblast activation and JAK/STAT3 signaling, in agreement with increased STAT3 phosphorylation by OSM (Fig-ure 5F, Supplemental Figure 7, B and C, and Supplemental Table  1). A transcriptional signature composed of the top differentially expressed genes by OSM in CAF-173 was enriched in the NCBI's Gene Expression Omnibus breast cancer stroma data set (GEO GSE9014) compared with normal stroma (Supplemental Figure  7D). Importantly, the top 4 genes induced by OSM in CAF-173 (SERPINB4, THBS1, RARRES1, and TNC; Supplemental Figure  7E) are associated with decreased overall survival in breast cancer patients ( Figure 5G). In addition, THBS1, RARRES1, and TNC levels correlate with OSMR expression in breast cancer clinical samples (Supplemental Figure 7F). These results indicate that OSM induces in CAFs the expression of promalignant genes, including fibroblast activation markers and genes associated with JAK/ STAT3 signaling. Of interest, OSM induced changes in the transcriptome of CAF-173 that were different from the ones observed  Table 1), suggesting that OSM activates unique signaling pathways in CAFs.
Moreover, the changes induced by OSM in CAF-173 contributed to breast cancer malignancy, as conditioned media (CM) from OSM-treated CAF-173 stimulated cancer cell migration in vitro (Supplemental Figure 9A). To test whether the OSM-induced changes in CAFs contributed to breast cancer progression in vivo, we pretreated CAF-173 with OSM or vehicle for 4 days in vitro and orthotopically coinjected them with MDA-MB-231 breast cancer cells into athymic nude-Foxn1 nu mice, as described in the experimental timeline in Figure 6A. Activation of fibroblasts by OSM promoted tumor growth ( Figure 6, B-D) and exhibited a trend to increase lung colonization ( Figure 6E), as assessed by qPCR analysis of human Alu DNA sequences in the lungs (30). The presence of metastasis in the lung of these mice was confirmed by vimentin staining of cancer cells (Supplemental Figure 9B). Conversely, OSMR downregulation by shRNA in CAF-173 delayed tumor by recruiting leukocytes to the tumor site. Analysis of a panel of 31 chemokines by antibody array showed that OSM induced expression of important chemoattractants ( Figure 7C and Supplemental Figures 11C and 12). Some of these factors were exclusive to CAFs (mainly CXCL10 and CXCL12), others to cancer cells (mainly CXCL7 and CCL20), and some factors, such as CCL2, were common to both cell types. Vascular endothelial growth factor (VEGF) can also modulate tumor immunity by inducing macrophage and myeloid-derived suppressor cell (MDSC) recruitment (31), and we previously showed that it is an OSMR target (29). As seen in  Figure 10E). Moreover, the tumors with OSMR silencing in CAFs showed reduced levels of GFP (Supplemental Figure 10E), suggesting reduced levels of CAFs in this experimental group, probably due to impaired CAF proliferation upon OSMR reduction, in line with the increased size of CAF spheres observed after OSMR activation ( Figure 5, C and D). Together, our data demonstrate that OSM/OSMR signaling activates CAFs and that this contributes to cancer progression.
OSM signaling induces chemokine secretion and myeloid recruitment. In an attempt to understand how OSMR activation in the stroma was inducing malignancy, we probed our transcriptomic data of CAFs (CAF-173) treated with OSM. Microarray data indicated that pathways and signatures related to leukocyte chemotaxis and inflammatory response were significantly enriched by OSM (Figure 7, A and B). Interestingly, transcriptomic analysis of breast cancer cells (MDA-MB-231) activated by OSM showed enrichment of similar pathways (Supplemental Figure 11, A and B). These data suggested that, upon OSMR activation by OSM, both CAFs and cancer cells could be involved in shaping the TME  Figure 13E). We did not observe this effect with CM from OSMactivated CAFs or with OSM itself. Therefore, we have discovered an unprecedented positive feed-forward loop between cancer cells, CAFs, and myeloid cells in which (a) tumor-infiltrating myeloid cells secrete OSM; (b) CAFs become activated, promoting further myeloid cell recruitment; and (c) OSM-induced secretome by cancer cells promotes sustained OSM production by myeloid cells.
Analysis of OSM protein levels in 141 samples of early breast cancer samples confirmed the association between OSM expression and increased inflammation in a clinical setting (Figure 9, A and B). Inflammation was assessed by a pathologist as infiltration of inflammatory cells from all lymphoid and myeloid subtypes. Immunohistochemistry (IHC) confirmed that the tumor inflammatory infiltrate was composed, at least, of T cells (CD3 + ), macrophages (CD68 + ), and neutrophils (CD15 + ) (Supplemental Figure  14). We observed that OSM was mainly expressed by myeloid-like cells as determined by their larger size and more irregular shape ( Figure 9A). Lymphoid cells, characterized by being smaller and KO and control tumors. We observed that these 2 populations were reduced in MMTV-PyMT Osmr-KO tumors compared with Osmr-WT tumors ( Figure 8C). The decreased number of macrophages in MMTV-PyMT Osmr-KO tumors was confirmed by FACS analysis of CD45 + CD11b + GR1 med F4/80 + macrophages (Supplemental Figure 13, A and B). Of interest, there was no difference in the percentage of M2-like protumoral (CD206 + ) and M1-like antitumoral (CD80 + ) macrophages in those tumors (Supplemental Figure 13C), suggesting that OSMR affects macrophage recruitment without altering their polarization. We did not observe a reduction in CD45 + CD11b + GR1 hi neutrophils in Osmr-KO tumors by FACS analysis (Supplemental Figure 13, A and B), suggesting that Ly6G staining in MMTV-PyMT Osmr-KO tumors may be marking neutrophils but also other myeloid-derived cells such as MDSCs (36,37). Marker analysis showed that most of the tumor-infiltrating neutrophils in MMTV-PyMT Osmr-WT and -KO tumors exhibited a protumoral and immunosuppressive phenotype, as assessed by CXCR4 and CCR5 positivity (refs. 38, 39, and Supplemental Figure  13D). Interestingly, tumor-bearing MMTV-PyMT Osmr-KO mice compared with control mice showed reduced serum VEGF and CXCL16 levels and exhibited a trend toward a decrease in CXCL1 ( Figure 8D), all factors being involved in myeloid cell recruitment (31,40,41). Our findings are clinically relevant, as VEGF, CXCL1, and CXCL16 mRNA expression is associated with OSM and/or OSMR levels in breast cancer patients and with decreased overall survival (Figure 8, E and F). In summary, these results show that    Figures 4 and 5). While the ligand OSM was only expressed by the myeloid cell populations, we found that the receptor OSMR was mainly expressed by fibroblasts, cancer cells, and endothelial cells. Whether there is one myeloid cell population mainly responsible for OSM production in breast cancer, or whether OSM is secreted by different immune cell types (including neutrophils, macrophages, or even circulating monocytes) remains to be determined. Of interest, a recent report identified OSM as one of the key signaling mediators of neutrophil-cancer cell interactions in prometastatic clusters of neutrophils and circulating tumor cells (44).
The cell population showing the most significant association with OSMR expression in human breast cancer samples is the CAF compartment, and our data point to an important role for this cell type in transducing OSM signaling within the TME. While the protumoral effect of OSMR activation in cancer cells has been extensively described (10,12,45), little is known about the effects of OSM in the tumor stroma and our results shed light on the effects of OSM signaling in CAFs. It has been previously reported that OSM and its related cytokine LIF stimulate actomyosin contractility and matrix remodeling by oral squamous cell carcinoma-derived CAFs (46,47). However, we did not observe any effect of LIF on collagen contraction assays in our experimental setting, maybe owing to the particularities of the different protocols used. In addition to an effect on CAF contractility, we observed an increase in CAF proliferation and a proinflammatory phenotype in OSMactivated CAFs. Moreover, the secretome of OSM-activated CAFs promoted cancer cell migration. To our knowledge, this is the first report describing that OSM induces the activation of a proinflammatory transcriptional program in CAFs. In line with our results, round and by having a round nucleus with little cytoplasm, showed very low or negative OSM expression ( Figure 9A). Importantly, high OSM protein levels were associated with decreased overall survival in this data set (P = 0.029; Figure 9C).

Discussion
Cytokines are important players in inflammation, a process associated with tumor progression (42). Even cancers not directly associated with persistent infections or chronic inflammation, such as breast cancer, exhibit tumor-elicited inflammation, which has important consequences in tumor promotion, progression, and metastasis (5,43), facilitating the acquisition of cancer hallmarks (2). Understanding how inflammatory signals orchestrate promalignant effects in the different cell compartments within the TME is key to designing new therapeutic strategies to target tumor-promoting inflammation. The oncogenic activity of inflammatory signaling factors such as IL-6 and OSM has been classically attributed to cell-intrinsic mechanisms within the cancer cell. However, our results reveal a key aspect of OSM/OSMR signaling that is instrumental for breast cancer progression beyond the epithelial compartment. Genetic and molecular analyses reveal that the tumor stroma responds to altered OSM production and signaling to influence breast cancer biology. Loss of OSMR in the nontumoral tissue hampers tumor aggressiveness, thus demonstrating that tumor cell-extrinsic OSM signaling is a pivotal factor in breast cancer progression. Our study identifies the proinflammatory cytokine OSM as a crucial mediator of the crosstalk between different cell types within the tumor by activating an intriguing protumoral "ménageà-trois" between myeloid cells, CAFs, and cancer cells.
Our scRNA-seq and FACS analyses revealed that OSM and OSMR have a unique expression pattern in breast tumors, com- calculated by adding the volume or the weight of all the tumors from the same animal. For whole-mount analysis of early lesions, abdominal mammary glands from 9-week-old MMTV-PyMT Osmr-KO and control female mice were spread out on a glass slide, fixed overnight in Carnoy's solution, stained with Carmine Alum, and cleared in ethanol and xylene. Pictures were taken with a Nikon D5000 at 60 mm focal length. For the generation of syngeneic orthotopic tumors, 300,000 viable murine control or Osmr-KO TS1 cells (derived from a MMTV-PyMT tumor in FVB/NJ mice; ref. 18) in growth factor-reduced (GFR) matrigel (1:1 ratio, Corning), were injected into the fourth right mammary fat pad of anesthetized (with 4% isoflurane) 6-to 8-week-old female FVB/NJ Osmr-KO or control mice. For the orthotopic coinjections of human MDA-MB-231 breast cancer cells and CAF-173 CAFs, cells were injected into the fourth right mammary fat pad of anesthetized (with 4% isoflurane) 6-week-old female Athymic Nude-Foxn1 nu immunocompromised mice (Envigo). In OSM activation experiments, CAF-173 were treated with 10 ng/mL OSM for 4 days, prior to coinjection with MDA-MB-231 (500,000 cells per cell line) in GFR matrigel (1:1 ratio). For OSMR knockdown experiments, 100,000 MDA-MB-231-hOSM cells and 500,000 shOSMR-infected CAF-173 were coinjected in GFR matrigel (1:1 ratio). In all mouse experiments, animals were monitored 3 times a week and tumor growth was measured using a caliper. Animals were culled once tumors reached the maximum allowed size. After animal culling, lungs were visually inspected for macroscopic metastases, and mammary glands and lungs were fixed in neutral buffered formalin solution (Sigma-Aldrich). Microscopic metastases were determined by H&E staining of formalin-fixed, paraffin-embedded sections. Tumors were divided in portions for (a) preparation of tissue sections for H&E and IHC (fixed in formalin) and (b) protein and RNA extraction (snap frozen).
Gene expression analyses of clinical data sets and bioinformatics analyses. Disease-free survival of patients based on OSM mRNA expression was calculated using data from the publicly available METABRIC (19) and Wang (20) data sets with the CANCERTOOL interface (59). Kaplan-Meier curves showing overall survival of patients from various cancer types according to the expression of different genes were obtained from the Kaplan-Meier Plotter website (21). Expression values were stratified by median. RNA-seq data from 64 cell lines was retrieved from the Human Protein Atlas (23). RNA consensus-normalized expression values were plotted for OSM and OSMR transcripts using GraphPad software. Associations between OSMR and OSM mRNA expression and infiltration of different cell types from the TME were analyzed by using xCell (26) on 1,809 breast cancer samples from Kaplan-Meier Plotter website (21) and TIMER2.0, which incorporates 1,100 breast cancer samples from TCGA (25). TIMER2.0 was also used to analyze gene expression correlations, after purity adjustment. All correlations were calculated with Spearman's rank correlation coefficient. scRNA-seq. Drop-seq data set (65) raw data for MMTV-PyMT (WT) tumors were obtained from Valdes-Mora et al. (66). This sub-a similar proinflammatory program was described to be activated by OSM in intestinal stromal cells in inflammatory bowel disease (48). Importantly, our results reveal that OSMR activation in CAFs promotes the recruitment of OSM-producing myeloid cells to the tumor through OSM-induced secretion of chemokines, thereby inducing a feed-forward loop. It has been shown that blocking myeloid recruitment to the premetastatic niche with anti-Ly6G antibodies inhibits metastasis (49), and impairing recruitment of tumor-associated macrophages reduces tumor incidence and metastasis (50) in MMTV-PyMT mice. Thus, decreased numbers of Ly6G + and F4/80 + myeloid cells may explain, at least in part, the strong antitumoral and antimetastatic effect of OSMR depletion in the MMTV-PyMT cancer model. Altogether, our data show that OSMR activation in CAFs could be promoting tumor progression by different and complementary mechanisms, including increased matrix contractility and proliferation, activation of an inflammatory response, secretion of chemokines, and promotion of myeloid cell recruitment and cancer cell migration.
In summary, our results demonstrate that OSM orchestrates an intriguing protumoral crosstalk between myeloid cells, CAFs, and cancer cells that has important consequences in tumor progression. Therapies aimed at modulating inflammatory responses in the TME have been of great interest in recent years (51). Interestingly, targeting IL-6 is problematic and anti-IL-6 drugs have not yielded significant results against solid tumors in clinical trials (52,53). Our results strongly support the notion that therapeutic targeting of OSM signaling is a valid alternative to blocking tumor-promoting inflammation in cancer that is worth exploring. The OSM/OSMR signaling module exhibits a unique microenvironment-restricted expression pattern, distinct from the rest of the members of the family, supporting the idea that targeting OSM/ OSMR will potentially avoid the toxic effects of anti-IL-6 drugs. OSM-OSMR interactions could be blocked by antibody-based inhibition, a strategy that has had a major impact on cancer (54), which makes them a promising candidate for therapeutic targeting. Interestingly, humanized anti-OSM antibodies have proven to be safe and well tolerated (55) and are now in phase II clinical trials for the treatment of inflammatory diseases, such as systemic sclerosis and Crohn's disease. Together, our findings further strengthen the case for the preclinical investigation of OSM/OSMR-blocking antibodies as a targeted anticancer therapy.

Methods
Mouse studies. Generation of the congenic strain MMTV-PyMT Osmr-KO was accomplished by mating MMTV-PyMT mice [FVB/N-Tg (MMTV-PyVT)634Mul/J, The Jackson Laboratory], which spontaneously develop mammary tumors and lung metastases (56), with Osmr-KO mice (B6.129S-Osmr tm1Mtan , Riken BRC; refs. 57, 58). To transfer the transgenic Osmr-KO line (with a C57BL/6J background) to the genetic background of the tumor-prone animals (FVB/NJ), the Osmr-KO mice were previously backcrossed with FVB/NJ mice (Charles River) for 9 generations. Osmr-WT, -HET (heterozygous), and -KO animals used for experiments were female littermates. Tumor onset was monitored by palpation and tumors were measured once a week using a caliper, and volume was calculated as (4π/3) × (width/2) 2 × (length/2). Animals were culled at 14 weeks of age, once tumors in the control group reached the maximum allowed size. Tumor burden was seeded in triplicate or quadruplicate in 24-well plates. After polymerization, collagen gels were detached, and they were treated with OSM, LIF (both at 10 ng/mL), or vehicle. Pictures were taken 48 hours later, and the area of collagen disks was analyzed using Fiji ImageJ software.
Cell migration assays. MDA-MB-231 and CAF-173 cells were treated with a single dose of OSM (10 ng/mL) or vehicle (PBS) for 72 hours in serum-reduced media (2% FBS) for CM generation. For breast cancer cell migration experiments, MDA-MB-231 cells were treated with the corresponding CM (diluted 1:2) for 72 hours and subjected to migration assays for 48 hours by seeding 25,000 cells at the top of 24-well Transwell inserts (8 μm pore, Corning). FBS was used as chemoattractant. Chambers were fixed in 10% formalin (20 minutes) and stained with crystal violet solution (20 minutes). For the quantification of migrated cells, crystal violet was solubilized with 600 μL of 1% SDS (30 minutes) and absorbance was measured at 570 nm. For monocyte migration experiments, 750,000 HL-60-derived monocytes were seeded at the top of the Transwell inserts with 600 μL of the corresponding CM in the lower chamber. Cells were allowed to migrate for 3 hours and the number of migratory cells in the lower compartment was counted using a hemocytometer.
3D fibroblast cell cultures. Fibroblast spheres were formed by seeding 8,000 cells/well in 96-well ultra-low attachment Corning plates (Costar). Cells were treated with 30 ng/mL OSM or PBS for 3 (for transcriptomic microarray analysis) or 4 days (for RT-qPCR, Western blot analysis, and quantification of sphere area). Pictures were taken using an EVOS FL Cell Imaging System (Thermo Fisher Scientific) and the area of the spheres was analyzed using Fiji ImageJ software. Spheres were collected for RNA and protein analyses.
Flow cytometry. Freshly obtained tumors and mammary glands from 14-week-old MMTV-PyMT Osmr-KO, -HET, and -WT mice were mechanically disrupted in 7 mL of digestion medium (collagenase type 1, Merck) and incubated for 1 hour at 37°C. The single-cell suspension was filtered through a 70 μm cell strainer (Falcon) and treated with ACK lysis buffer (Invitrogen) for 3 minutes at room temperature. Then, cells were stained with fluorochrome-labeled antibodies described in Supplemental Table 2 and with DAPI (1:5,000, Thermo Fisher Scientific) in FACS buffer (eBioscience). Flow cytometry analysis was performed with a BD FACSymphony flow cytometer and data were analyzed using FlowJo (BD Biosciences). The gating strategy is shown in Supplemental Figure 13B. FACS isolation of TS1 GFP + cells in CRISPR/Cas9 D10A nickase experiments was performed with a BD FACSJazz (2B74YG) cell sorter. For FACS experiments of TS1derived tumors, TS1 cells were injected orthotopically in FVB mice as described above, and 15 days after injection, freshly obtained TS1 tumors were dissociated into single-cell suspensions and stained with the antibodies described in Supplemental Table 2. Flow sorting was performed with a BD FACSAria II cell sorter. Gating strategy for experiments is shown in Supplemental Western blotting. Cells and tumors were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (cOmplete ULTRA Tablets, Mini, EASYpack Protease Inhibitor Cocktail; and PhosSTOP, both from Roche). Total lysates were quantified by BCA (Pierce BCA set was subsequently analyzed using Seurat v3.2 (67). Briefly, a total of 9,636 sequenced cells from 8 MMTV-PyMT tumors passed the QC filter, with less than 5% mitochondrial-to-nuclear gene content (65), and fewer than 8,000 molecules/cell, as they potentially represented cell doublets. Downstream analysis was performed according to Butler et al. (67), using 30 principal components to build a shared nearest neighbor (SNN) graph calculating k-nearest neighbor (Jaccard Index) for each cell, subsequent cluster calling, and UMAP dimensional reduction projection (68).
Collagen gel contraction assays. To assess the collagen remodeling capacity (71), fibroblasts were treated for 4 days with recombinant hOSM (R&D Systems) or human LIF (Millipore) at 10 ng/mL or vehicle (PBS) before being embedded (250,000 cells per matrix) in collagen matrix (2 mg/mL rat tail collagen type I, Corning, in DMEM + 10% FBS) in the presence or absence of Y-27632 (10 μM, Tocris), and minutes with streptavidin-peroxidase complex diluted 1:400 in TBS (Invitrogen). The chromogen was 3,3′-diaminobenzidine (Vector Laboratories). Nuclei were counterstained with Harris's hematoxylin for 1 minute. Pictures were obtained using the Nikon Eclipse 80i microscope with a Nikon DS-5M camera incorporated. The number of positive cells and total cells per area was counted manually in 5 to 15 different areas of samples from 5 to 7 mice per experimental group, using Fiji ImageJ software. For immunofluorescence analysis, paraffinembedded sections from murine lungs or cells fixed on a coverslip were permeabilized with 0.1% or 0.2% Triton X-100 and 0.01% SDS after performing antigen retrieval with citrate buffer for 15 minutes. The slides were then blocked with 3% BSA/PBS containing 3% normal goat serum and 0.1% Tween 20 for 30 minutes to 3 hours and stained overnight with the corresponding primary antibodies (Supplemental Table 2) followed by secondary antibody incubation (1 hour at room temperature). F-actin was stained with Phalloidin CruzFluor 633 Conjugate (Santa Cruz Biotechnology). Nuclei were counterstained with DAPI (Thermo Fisher Scientific). Finally, sections were mounted with Mowiol (Thermo Fisher Scientific).
Cytokine and chemokine analysis. Cytokine and chemokine levels were analyzed in CM from CAF-173 treated with OSM (30 ng/mL) or vehicle for 72 hours (n = 4), and from MDA-MB-231-hOSM and corresponding control cells (n = 6). A panel of 31 human chemokines was analyzed by Human Chemokine Array Kit (Proteome Profiler Array, R&D Systems), and VEGF levels were quantified by Human VEGF Quantikine ELISA Kit (R&D Systems) following the manufacturer's instructions. Mouse VEGF, CXCL1, and CXCL16 levels in plasma from 14-week-old MMTV-PyMT Osmr-KO, -HET, and -WT mice were analyzed by mouse Premixed Multi-Analyte Kit (Magnetic Luminex Assay, R&D Systems) following the manufacturer's instructions. Detection was carried out with the MAGPIX detector and data analysis was performed using xPOTENT software, both from R&D Systems.
Tissue microarrays. Formalin-fixed, paraffin-embedded blocks of 141 tumor tissues from cases surgically resected at the University Hospital Basel between 1991 and 2013, and included in tissue microarrays (TMAs), were used for analysis of OSM protein expression in human samples. Complete histopathological information (Supplemental Table 6), date and cause of death, as well as date of local and/or distant relapse were available for all the patients. TMAs were generated by punching a 1-mm spot of each sample. Tissue sections were subjected to a heat-induced antigen retrieval step prior to exposure to primary antibodies (Supplemental Table 2). Immunodetection was performed using the Roche Ventana BenchMark ULTRA IHC staining system, with DAB as the chromogen. Cases were reviewed by 2 independent pathologists and OSM staining was evaluated by the semiquantitative method of H-score (or "histo" score), used to assess the extent of immunoreactivity in tumor samples (78). Inflammation was semiquantitatively assessed by a pathologist as high or low tumor infiltration of immune cells according to their morphology.
Data availability. RNA-seq raw data were obtained from Valdes-Mora et al. (66) and are available in the GEO repository (GSE158677). The mRNA data sets generated during the current study are available in the GEO repository (GSE195787). Source data on uncropped Western blots are provided in the supplemental material. The gene list for the fibroblast activation signature used in Figure 5 was derived from Sahai et al. (4) and is shown in Supplemental Table 4. The gene list for the CAF-173 OSM signature used in Supplemental Figure 7 includes Protein Assay Kit, Thermo Fisher Scientific), resolved by SDS-PAGE, and transferred to nitrocellulose membranes. After blocking with 5% (wt/vol) nonfat dry milk in TBS-Tween, membranes were incubated with the corresponding antibodies (Supplemental Table 2) overnight at 4°C. Secondary antibodies (Supplemental Table 2) were chosen according to the species of origin of the primary antibodies and detected by an enhanced chemiluminescence system (Bio-Rad). Densitometric analysis of the relative expression of the protein of interest versus the corresponding control was performed with Fiji ImageJ software. Complete unedited blots can be found in the supplemental material.
DNA and RNA extraction, RT-qPCR, and transcriptomic analysis. Lung genomic DNA was extracted from frozen lungs using the QIAmp DNA mini kit (Qiagen) for qPCR analysis. RNA was obtained from snap-frozen animal tissue or cell pellets and extracted using TRIzol reagent (Invitrogen) or Recover all Total Nucleic Acid Isolation kit (Invitrogen), for RT-qPCR and microarray analysis, respectively. cDNA was obtained with the Maxima First-Strand cDNA synthesis kit (Thermo Fisher Scientific) with DNAse treatment incorporated. qPCR was performed using Power SYBR Green PCR master mix (Applied Biosystems) and oligonucleotide sequences are described in Supplemental Table 3 (all purchased from Condalab). Expression levels of genes were determined using the ΔΔCt method (73) and normalized against 3 housekeeping genes optimized for each reaction (74). Human Alu sequences (30) were normalized against the 18S housekeeping gene using primers capable of recognizing both human and mouse DNA. Microarray analysis was performed using the Human Clariom S assay (Thermo Fisher Scientific). RNA quality was evaluated using the 2100 Bioanalyzer (Agilent) and microarray chips were processed on the Affymetrix GeneChip Fluidics Station 450 and Scanner 3000 7G (Affymetrix) according to standard protocols (n = 3 per experimental condition). Data were analyzed using the Transcriptome Analysis Console 4.0 (TAC). Genes with FDR less than 0.1 and absolute fold change greater than 2 were considered significantly modulated. GO analysis was performed using Panther (75). GSEA was performed as previously described (76). FDR less than 0.25 or 0.05 was regarded as statistically significant, depending on the type of permutations performed. We compiled the GSEA signatures used in Figures 5 and 7 and Supplemental Figures 7 and 11 from the Molecular Signatures Database (MsigDB) by the Broad Institute or they were manually curated from the literature. The gene list for each signature is publicly available at http://software.broadinstitute.org/gsea/msigdb/search.jsp, in Pein et al. (77), or in Supplemental Tables 4 and 5.
Histopathology, IHC, and immunofluorescence analyses. Histological analysis of murine tumors and lung metastasis was performed in H&E-stained sections. Immunohistochemical staining was performed in formalin-fixed, paraffin-embedded sections using Novolink Polymer Detection Systems (Leica). Antigen retrieval was performed using boiling 10 mM citrate buffer, pH 6.0, for 15 minutes. Endogenous peroxidase activity was inactivated by incubation with 3% hydrogen peroxide in methanol (15 minutes at room temperature). Tissue sections were incubated in a humidified chamber (overnight, 4°C) using the antibodies described in Supplemental Table 2 diluted in Tris-buffered saline (TBS). For negative controls, primary antibodies were replaced by nonimmune serum. After 3 rinses in TBS (5 minutes each), samples were incubated with the corresponding secondary antibody (Supplemental Table 2). After a 30-minute incubation, tissue sections were washed in TBS (5 minutes, 3 times) and immediately incubated for 30 J Clin Invest. 2022;132 (7) the 233 genes differentially upregulated in CAF-173 upon OSM treatment and can be found in Supplemental Table 5. All other data files supporting the findings of this study are available from the corresponding author upon reasonable request.
Statistics. Statistical analyses were performed using GraphPad Prism or SPSS software (IBM). For Gaussians distributions, the Student's t test (paired or unpaired) was used to compare differences between 2 groups. Welch's correction was applied when variances were significantly different. One-or 2-way ANOVA with post hoc Tukey's, Dunnett's, or Sidak's multiple-comparison tests were used to determine differences between more than 2 independent groups. For non-Gaussian distributions, the Mann-Whitney test was performed. The χ 2 test was used to determine differences between expected frequencies. For Kaplan-Meier analyses, the log-rank (Mantel-Cox) test was used. P values less than 0.05 were considered statistically significant. Unless otherwise stated, results are expressed as mean ± standard error of the mean (SEM).
Study approval. All patients whose samples were included in the TMAs have given written informed consent for their archival tissue to be used for scientific research in accordance with the Declaration of Helsinki, and the TMA construction was approved by the responsible local Ethical Committee EKBB (Ethikkommission beider Basel), number 361/12. All procedures involving animals were performed with the approval of the Biodonostia Animal Experimentation Committee and Gipuzkoa Regional Government, according to European official regulations.

Author contributions
AMA, AA, PA, JR, and IOQ performed all the cellular and molecular experiments. AMA, AA, JILV, and MMC performed the animal experiments. JMF performed immunohistochemistry and analyzed mouse histopathology. JILV, SM, and LP analyzed mouse immunostaining. FVM and DGO obtained and analyzed the scRNA-seq data. AMA, LJ, LEM, and AP performed FACS isolation and FACS analyses. FC and NF provided RNA from FACS-isolated tumors. AT and SEC collected and analyzed patient data, generated the TMAs, performed immunohistochemistry, and interpreted and statistically analyzed the resulting data. RR and MMC interpreted and analyzed the TMA staining. AMA, AA, PA, NMM, FC, AC, and MMC performed bioinformatic analysis. PFN and PB generated the fibroblast cell lines. PB, CV, and AA performed and analyzed immunofluorescence staining. NC, IAL, and AU contributed to experimental design. CMI and CHL contributed to experimental design and helped with supervision of the project. AMA, AA, and MMC designed and supervised the study, analyzed the data, and wrote the manuscript. Both AMA and AA contributed equally and have the right to list their name first in their CV and should be considered co-first authors. The order of co-first authors was determined by seniority in the labo-