High expression of oncostatin M (OSM) receptor OSMR has recently emerged as a poor prognosis factor in several malignancies such as acute myeloblastic leukemia (AML), gliomas, pancreatic, gastric and kidney carcinomas. In this perspective, we perform a larger in silico survey of the prognostic value of high OSMR and OSM transcripts in 33 malignancies and discuss the mechanisms involved in the adverse effects of high OSM-mediated signaling in malignant cells and the tumor microenvironment, and potential treatments to target this signaling pathway.

OSM is an inflammatory cytokine of the interleukin-6 (IL-6) family. OSM binds to two different receptors made up of glycoprotein-130 (GP130, gene IL6ST in humans Il6st in mice) complexed with either transmembrane OSM receptor (OSMR, gene OSMR in humans or Osmr in mice) or with leukemia inhibitory factor receptor (LIFR) (1, 2). Both OSMR and LIFR have very long intracellular domains enabling the docking of many kinases and adaptors, eliciting similar but not identical signaling in response to OSM (3, 4). OSM is mostly produced by activated myeloid cells (granulocytes, monocytes and macrophages) as well as all cells of the osteoblast lineage (3, 5, 6), and astrocytes, neurons, microglia in the brain (7). In contrast, its receptor OSMR is not expressed by leukocytes or hematopoietic stem and progenitor cells (HSPC) but expressed by cells of the mesenchymal and endothelial lineages in the bone marrow (6) and other tissues. OSM is a regulator of both hematopoiesis and skeleton homeostasis mostly via the OSMR: GP130 receptor complex (6) and plays important roles in inflammatory responses in skeletal (8, 9) and cardiac (10) muscles, lung (11), liver (12, 13), intestine (14), adipose (15), skin (16, 17) and joints (18). Adding to the complexity of OSMR protein roles, OSMR protein complexed with the IL-31 receptor α chain IL31RA acts as a receptor for IL-31 (19). Recent literature now highlights important roles of both OSMR and OSM in the pathogenesis and response to treatment of several malignancies as detailed below.

In a recent paper, high throughput proteomic analysis on more than 550 newly diagnosed AML patients demonstrated that high plasma concentration of soluble OSMR protein was a strong independent predictor of poor survival and early mortality (20). These clinical data suggest an oncogenic role of OSMR-mediated signaling and are consistent with the anti-proliferative effect of OSMR gene inactivation in mice and humans. Indeed, Osmr ^-/- mice display mild anemia and thrombocytopenia (21–23) with decreased HSPC cycling in the bone marrow, increased HSPC chemotactic response, increased HSPC mobilization into the blood in response to G-CSF or CXCR4 antagonists, as well as decreased expression of genes associated with cell cycling, lipid metabolism, and erythropoiesis in hematopoietic stem cells (HSC) (5). Another recent report has shown that biallelic loss-of-function of the OSM gene in humans causes profound anemia, thrombocytopenia and neutropenia (24). Therefore, OSM and OSMR-mediated signaling contributes to increased HSPC cycling and retention within the bone marrow, enabling sufficient erythropoiesis and thrombopoiesis output. Interestingly, the hematopoietic effects of OSMR are mediated via the hematopoietic environment as OSMR mRNA is undetectable to very low in mouse and human HSPC (5) or in AML blasts (20). Reville et al. highlighted that high OSMR protein production and transcript expression was limited to mesenchymal stromal cells (MSCs) while undetectable in AML blasts in cultures of leukemic marrow aspirates (20). However, their culture system did not enable the survival and growth of bone marrow endothelial cells, which similar to MSCs, express high levels of OSMR transcripts and protein (5, 6) and are key functional regulatory elements of HSC niches and leukemia stem cell niches (25–30). Therefore, the source of sOSMR protein in the blood of AML patients could be endothelial cells as well as MSCs and the deregulation of these niche cell types in the bone marrow may be involved in the mechanisms of poorer outcome in patients with high sOSMR protein.

As the OSMR ligand OSM is expressed by myeloid cells (particularly neutrophils and macrophages) and osteoblasts in the bone marrow (3, 5), we plotted the expression of OSM transcripts in the 9,736 tumor and 8,587 normal tissue samples contained in The Cancer Genome Atlas (TCGA) database using Gene Expression Profiling and Interactive Analysis (GEPIA) website (31, 32) ( Figure 1A ). OSM transcripts were significantly higher in AML samples than normal bone marrow samples (“LAML” in Figure 1A ). Overall survival (OS) Kaplan-Meier plot from AML patients with 50% highest content and 50% lowest content in OSM transcripts is shown in Figure 1B . Although the p values of p=0.055 for differences were just above the significance threshold of p=0.05, AML patients with highest levels of OSM transcript had a trend to shorter OS with a 1.7 hazard ratio compared to AML patients with lowest OSM transcripts. Considering that high sOSMR protein is significantly associated with poorer prognosis in AML (20), confirmation of a significant association between poor prognosis and high OSM transcripts in a larger cohort of AML patients is warranted. In regard to the alternative OSMR ligand interleukin-31 (IL-31), IL31 transcripts were detected at least 2 orders of magnitude lower compared to OSM transcripts ( Figure 1C ) and there was no difference in OS between high and low IL31 expressing AML with the 50% cut-off ( Figure 1D ). Therefore, it is likely that the adverse effect of high OSM transcript and sOSMR protein in AML is mediated via its ligand OSM acting indirectly via bone marrow mesenchymal and endothelial cells which express both OSMR with its co-receptor GP130 (5, 22, 23) rather than IL-31-mediated signaling. One of the adverse effects of OSM-OSMR interaction may be mediated through its induction of E-selectin expression by endothelial cells (33) as E-selectin mediated signaling in AML stem cells increases their resistance to the cytotoxic effects of chemotherapy (30) (see model in Figure 1E ).

In respect to other hematological malignancies, OSM mRNA are reportedly high in Philadelphia-negative myeloproliferative neoplasm (Ph^- MPN) cells with a constitutively active JAK2 tyrosine kinase mutations such as JAK2 ^V617F (34), with increased OSM protein concentration in Ph^- MPN patient bone marrow cells and blood plasma (34). Furthermore cytokine production by primary fibroblasts from MPN patients is increased in response to OSM via OSMR expressed by fibroblasts (34). Therefore, sOSMR and its ligand OSM may also have prognostic value in Ph^- MPN. This hypothesis is consistent with the following findings: 1) Transduction of a JAK2 ^V617F mutant in myeloid cell lines dramatically increases OSM expression (34). 2) OSM promotes HSPC proliferation, erythropoiesis and thrombopoiesis in vivo (5, 21) indirectly via bone marrow niche cells that express OSMR (5, 6). 3) In a mouse model of lymphoproliferative neoplasm driven by activating mutation of the FLT3 tyrosine kinase domain (FLT3-TKD), co-transduction with OSM cDNA switched the lymphoproliferative FLT3-TKD neoplasm to a myeloproliferative neoplasm suggesting that OSM is a potent inducer of a pro-myeloid niche (35). 4) Transplantation of mouse HSCs overexpressing OSM resulted in fatal myelofibrosis (36), a common adverse evolution of Ph^- MPNs. Further along this latter point, it has recently emerged that activating JAK2^V617F mutation upregulates the expression of the “don’t eat me” antigen CD24 at the surface of neutrophils enabling the accumulation of old senescent neutrophils to escape efferocytosis and accumulate in the bone marrow promoting evolution to myelofibrosis (37). As neutrophils are a major source of OSM protein in the bone marrow (5, 6) and OSM overexpression triggers myelofibrosis (36), accumulation of OSM secreted by senescent JAK2^V617F neutrophils may be a driver of the evolution of polycythemia vera (the most frequent clinical manifestation of MPNs with JAK2^V617F driving mutation) to myelofibrosis which has worse clinical outcome.

In respect to lymphoid neoplasms, high sOSMR (19) and OSM (38) protein concentrations have also been reported in the serum of multiple myeloma patients. However, high plasma OSM had no significant prognostic value in newly diagnosed multiple myeloma patients (38) and the potential prognosis value of sOSMR concentrations in multiple myeloma has not been reported.

Endothelial cells (39, 40) and mesenchymal cells (41) are important players in cancer stem cell niches within solid tumors. As OSMR is expressed by these cells (5, 22, 23) as well as some mucosal and glandular epithelial cells (42), OSMR and OSM may also play important roles in several non-hematological malignancies by directly acting on tumor cells or indirectly via endothelial and mesenchymal cells such as fibroblasts in the tumor environment. In support of this, high OSMR transcript content is a poor prognosis factor in glioblastoma (43, 44), pancreatic adenocarcinoma (45), and gastric cancer (46). Analysis of OSMR transcript expression in all tumor types and corresponding healthy tissues in the TCGA database ( Figure 2A ) showed OSMR transcripts are significantly more expressed in tumors than the paired normal tissues in esophageal carcinoma (ESCA), glioblastoma multiform (GBM), kidney renal clear cell carcinoma (KIRC), pancreatic adenocarcinoma (PAAD), stomach adenocarcinoma (STAD) and thymoma (THYM). Lower grade glioma (LGG) had higher expression of OSMR when compared to healthy tissue, but this was not significant. As samples were generally higher in solid tumor datasets compared to AML, we used a more stringent quartile cut-off and plotted OS and disease-free survival (DFS) from the 25% highest OSMR transcript expressing tumors (High OSMR) and 25% lowest OSMR expressing tumors (Low OSMR). GBM, LGG, PAAD and STAD with highest OSMR transcripts had significantly worse OS, DFS and hazard ratios than the same tumors with lowest OSMR expression ( Figures 2B–E ). High OSMR transcripts in KIRC had no effect on OS but significantly poorer DFS ( Figure 2F ). Among tumor types with similar OSMR transcripts compared to paired normal tissue, kidney renal papillary cell carcinoma (KIRP) had significantly worse OS but no difference in DFS ( Figure 2G ). Unexpectedly, adrenocortical carcinoma (ACC), had significantly less OSMR transcripts compared to paired normal tissue ( Figure 2A ) but patients with high OSMR transcripts had significantly poorer DFS and equivalent OS ( Figure 2H ). We could not find any significant difference in OS and DFS between high and low OSMR expression in all other tumor types contained in the TCGA database (result not shown).

Reciprocally, OSM mRNA expression was significantly higher in GBM, PAAD and AML compared to corresponding healthy tissue ( Figure 1A ). OS was significantly worse in KIRC with high OSM transcripts but there was no significant difference in DFS ( Figure 2I ). There was no difference in OS or DFS between high and low OSM transcript levels in GBM or PAAD patients (result not shown). We also checked that these findings were similar with a median cut-off (50%) for OSMR and OSM transcripts ( Supplementary Figure 1 ). This was true except for KIRP and ACC which did not show significant difference for OS or DFS with a 50% cut-off.

The poorer prognosis of tumors with high OSMR transcripts is consistent with previously reported effects of OSM or OSMR gene knock-down on these tumor types. For instance, OSM protein has been reported to increase migration, invasiveness and mesenchymal phenotype of GBM cell lines (43), and OSMR protein can translocate to the mitochondrial matrix to protect GBM stem cells from irradiation (44). Likewise OSM has been found to enhance tumor growth and epithelial-to-mesenchymal transition of pancreatic ductal adenocarcinoma cell lines in vivo in immunodeficient mice via OSMR (45). Additionally, OSM produced by macrophages has been found to induce an inflammatory phenotype in PAAD-associated fibroblasts providing a pro-tumorigenic environment in PAAD (47). OSM also promotes the proliferation of gastric cancer cells in vitro and in vivo in Nude mice (46). Finally, OSM has been reported to increase migration, invasiveness and epithelial-to-mesenchymal transition of clear cell renal carcinoma cell lines in vitro (48). Importantly, deletion of the VHL gene specifically in kidney tubule cells in mice (approximately 70% of clear cell renal carcinoma have inactivating mutation of VHL) induces production of OSM which in turn pushes kidney endothelial cells into an endothelial-to-mesenchymal transition to support VHL-defective tubule cell transformation and promote macrophage polarization towards a tissue-supportive M2-like phenotype (49). This vicious loop between OSM-expressing VHL-defective tubule cells and activated endothelium expressing OSMR is functionally important as injection of neutralizing OSM antibodies into immuno-deficient mice transplanted with human clear cell renal carcinoma cells inhibited primary tumor vascularization, growth and metastasis (49).

Kaplan-Meier plots in Figures 1 , 2 and previous reports discussed above clearly indicate that the interaction of OSM with the OSMR: GP130 receptor complex plays a role in the pathobiology and response to treatment of a number of malignancies such as AML, MPNs, glioma and glioblastoma, pancreatic, gastric and renal carcinomas where high expression of OSMR or OSM transcripts has poor prognosis value. The effects of the OSM: OSMR interaction are multipronged as it can alter either the function of malignant stem cell niches as observed in malignant hematological stem cells that do not express OSMR such as AML or MPNs, or directly act on malignant cells that express OSMR with further support from tumor-associated fibroblasts and endothelial cells which also express OSMR.

As OSM protein is highly expressed by activated neutrophils, monocytes and macrophages, blood concentration of OSM protein may not be a reliable prognosis marker of cancer treatment outcome because OSM concentration is also elevated during infections, sepsis (50, 51), acute and chronic inflammation (14, 16, 52) or in response to therapeutic treatments with myelopoietic cytokines such as granulocyte colony-stimulating factor (5). On the other hand, OSMR is not expressed by leukocytes but by non-hematopoietic malignant cells, as well as endothelial and mesenchymal cells forming the tumor environment. As soluble sOSMR protein is produced by alternative splicing of the OSMR transcript in exon 8 (19), it is likely that sOSMR protein may be similarly increased in the plasma of patients with tumors expressing high levels of OSMR transcripts, although actual concentrations of plasma sOSMR were not measured in patient samples from the TCGA used for our in silico analysis.

Blood sOSMR concentration may represent an easily measurable prognosis marker not only in AML as recently reported (20) but also in the other malignancies discussed in this perspective, which warrants further clinical studies for its prognostic potential. In regards of possible treatments of patients with high levels of sOSMR protein or OSMR or OSM gene expression, the main axis of OSMR-mediated signaling is via tyrosine phosphorylation and activation of transcription factors STAT1 and STAT3 (and in some cell types STAT5) via tyrosine kinases JAK1 and JAK2 (4, 6). Therefore, adjunct treatments with small non-selective JAK1 and JAK2 tyrosine kinase inhibitors such as ruxolitinib, tofacitinib and others (53) may improve outcome in these patients. Therapeutics that specifically target OSMR or OSM such as the humanized neutralizing anti-OSM monoclonal antibody GSK2330811 may also be of interest to treat AML patients with high sOSMR protein or OSM transcripts as GSK2330811 was found to induce thrombocytopenia, anemia and mild neutropenia in patients with diffuse cutaneous systemic sclerosis (54). Effectiveness of such neutralizing OSM antibodies in other malignancies in which high OSMR expression leads to poorer outcome remains to be established. In conclusion, elevated OSMR-mediated signaling is emerging as a poor prognosis factor in a number of malignancies such as AML, GBM, LGG, STAD, KIRC and KIRP. Consequently, OSMR gene expression and sOSMR protein blood concentration should be systematically measured in patients with these diseases and clinical trials should be undertaken to test efficacy of adjunct therapies with JAK1/2 inhibitors or neutralizing OSM antibodies to current best treatments.