Oncostatin M (OSM) Cytostasis of Breast Tumor Cells: Characterization of an OSM Receptor β–Specific Kernel

The interleukin-6 cytokine oncostatin M (OSM) induces potent growth-inhibitory and morphogenic responses in several different tumor cell types, highlighting the importance of OSM signaling mechanisms as targets for therapeutic intervention. The specific molecular pathways involved are not well understood, as OSM can signal through two separate heterodimeric receptor complexes, glycoprotein 130 (gp130)/leukemia inhibitory factor receptor (LIFR) α and gp130/OSM receptor β (OSMRβ). In this investigation, we used a LIFR antagonist to help resolve signaling responses and identify patterns of gene expression elicited by the different receptor complexes. OSM-induced biological effects on breast tumor–derived cell lines were specifically mediated through the gp130/OSMRβ complex. Each cytokine tested exhibited differential signaling capability and manifested both shared and unique patterns of gene activation, emphasizing compositional differences in activator protein-1 transcription factor activity and expression. In particular, OSM strongly activated the c-Jun NH₂-terminal kinase (JNK) serine/threonine kinase and downstream components, including activating transcription factor (ATF)/cyclic AMP-responsive element binding protein family member, ATF3. JNK/stress-activated protein kinase kinase inhibition abrogated cell morphogenesis induced by OSM, indicating an important role for this pathway in OSM specificity. These findings identify a core signaling/transcriptional mechanism specific to the OSMRβ in breast tumor cells. (Cancer Res 2006; 66(22): 10891-901)

Receptor-mediated signal transduction pathways are a major target for anticancer drug development (1). In principle, suppression of tumor cell growth can be achieved by inhibiting pathways that activate cell multiplication or by activating pathways that inhibit cell multiplication. Molecular definition of the pathways that negatively regulate cell multiplication in tumor cells would therefore open up new prospects for the manipulation of tumor cell growth. The characterization of negative growth-regulatory signal transduction pathways has been greatly aided by the identification of ligands that suppress tumor cell growth. Here, we analyze the molecular mechanisms of signal transduction mediated by the cytostatic cytokine oncostatin M (OSM), which result in the suppression of tumor cell proliferation.

OSM is a multifunctional, 28-kDa glycoprotein that was originally cloned based on its ability to inhibit the growth of melanoma cells (2, 3). OSM is produced by activated T lymphocytes and monocytes (2, 4) and is a member of the glycoprotein 130 (gp130) cytokine family, which exert their actions via a shared signal-transducing receptor, gp130 (5). Members of this family include interleukin (IL)-6, human herpesvirus 8 IL-6, IL-11, leukemia inhibitory factor (LIF), ciliary neurotrophic factor, cardiotrophin-like cytokine, cardiotrophin-1, and neuropoietin (6, 7). Different members of the gp130 family signal either by the formation of homodimeric gp130 complexes (facilitated in some cases by nonsignaling coreceptors) or via heterodimeric receptor complexes containing gp130 and a second signaling receptor.

Human OSM is unusual in that it can form two types of heterodimeric signaling complexes. The type I OSM receptor complex consists of the gp130 receptor and the LIF receptor α (gp130/LIFRα; ref. 8). This configuration is also used by LIF. As a consequence, human OSM and LIF can exhibit similarities in biological activity. The type II OSM receptor complex comprises gp130 and the OSM receptor β (gp130/OSMRβ), and this is activated by OSM only (9). The type II receptor complex may activate signaling pathways distinct from the OSM type I complex (10, 11), thus conferring OSM-specific signaling functions.

There is a particular interest in dissecting the specific signaling functions of human OSM, as it has been identified as a potent suppressor of tumor cell multiplication in a variety of contexts. Targets for human OSM-mediated growth inhibition include melanoma (3, 4), glioblastoma (12), lung carcinomas (13), ovarian carcinomas (14), and breast tumors (15, 16). In addition, human OSM induces differentiation of several tumor cell types (12, 17). These two effects highlight the potential application of human OSM as a therapeutic agent. However, a key outstanding issue is the role played by the different OSM receptor complexes in mediating these biological outcomes.

The generic features of gp130 cytokine signaling are well defined (6). The formation of a high-affinity transmembrane signaling complex results in activation of the receptor-associated Janus-activated kinase (JAK) kinase family, leading to phosphorylation of gp130, recruitment of the adaptor protein Shc, and activation of the mitogen-activated protein kinase (MAPK) cascade. The second target of receptor-activated JAK kinases is the signal transducers and activators of transcription (STAT) family of transcription factors, which, on JAK-mediated phosphorylation, migrate to the nucleus and activate the transcription of downstream gene targets. However, this generic pathway cannot readily account for the distinctive properties of OSM in suppressing tumor cell growth, suggesting that either dynamic features of gp130 cytokine signaling dictate cell responses or additional receptor-activated growth-inhibitory pathways remain to be discovered.

In this study, we exploit the properties of a specific type I receptor antagonist LIF-05 (18, 19) to identify the receptor complexes involved in mediating the growth-inhibitory functions of human OSM on a panel of breast tumor cell lines. We find that growth inhibition and morphologic transformation of these cells is an exclusive feature of the type II receptor complex. Using microarray approaches, we identify both shared and specific molecular signatures of type I and type II OSM signaling. Further analysis of the specific type II cytostatic signaling pathway reveals that the type II complex is a potent activator of the JNK/stress-activated protein kinase (SAPK) pathway, which leads to activation of a specific set of target genes that have been implicated in mediating cytostasis in other cell systems. Thus, we conclude that human OSM suppresses breast tumor cell proliferation and induces morphologic transformation, in part, by activating a specific stress/senescence program of gene expression mediated by the JNK/SAPK pathway.

Cells and reagents. T47D, MCF-7, and MDA-MB-231 human breast cancer cells were obtained from the American Type Culture Collection (Manassas, VA). Cells were cultured in DMEM (Life Technologies Ltd., United Kingdom) supplemented with 10% FCS, 1 mmol/L glutamine, and 1 mmol/L streptomycin/1 mmol/L penicillin. Chemical inhibitors SU6656, RG13022, SU5614, PD153035, PD98059, SP600125, LY294002, and AG490 were purchased from Merck Biosciences Ltd., United Kingdom and used in a final concentration of DMSO of <0.1%.

Cytokines. The human OSM recombinant expression plasmid was prepared by K.R. Hudson (University of Birmingham, United Kingdom). The expression vector pGEX-3C-OSM was modified from an original clone as described previously (20). The design of pGEX-2T-hLIF and pGEX-2T-hLIF05 plasmids (LIF wild-type and LIF antagonist) has also been described previously (18, 21). OSM, LIF, and LIF-05 were expressed as glutathione S-transferase fusion proteins in Escherichia coli strain JM109. Expression, purification, and cleavage of proteins were carried out as described (21, 22). Protein concentrations were determined using the Coomassie blue protein assay (Perbio Science UK Ltd., United Kingdom). Human recombinant IL-6 was purchased from Calbiochem.

Cell growth/morphology. Cells were seeded in 24-well plates (growth/inhibitor studies) or onto coverslips in six-well Costar (Cambridge, MA) cluster plates (morphology studies) at an initial density of between 0.5 × 10⁴ and 1 × 10⁴ cells/mL. Fresh medium and cytokines/inhibitor were added 24 hours after initial seeding and replenished daily. After 7 days of cytokine/inhibitor exposure, cells were washed with PBS and trypsinized and viable cell numbers were counted by trypan blue exclusion. For morphology studies, coverslips supporting T47D cells were transferred to chamber slides containing normal medium and cell morphology of live cells was assessed by light microscopy. MDA-MB-231 cells, which are less visible in live culture, were fixed in 4% paraformaldehyde for 20 minutes, permeabilized in TBS containing 0.1% Triton X-100, and stained with May-Grunwald-Giemsa (Sigma-Aldrich, United Kingdom) for viewing by light microscopy. Images for both cell types were captured using OpenLab (Improvision Ltd., United Kingdom).

Western blot. Following serum starvation for 24 hours, T47D cells were stimulated with increasing concentrations of cytokine for varying times at 37°C. Cells were washed in PBS and solubilized in either Triton X-100 lysis buffer (50 mmol/L Tris-HCl, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 50 mmol/L NaF, 1% Triton X-100) or radioimmunoprecipitation assay (RIPA) buffer (50 mmol/L Tris-HCl, 150 mmol/L NaCl, 5 mmol/L EDTA, 50 mmol/L NaF, 0.1% SDS, 1% Triton X-100, 0.5% sodium deoxycholate) plus 1 mmol/L sodium orthovanadate (Na₃VO₄) and one complete protease inhibitor tablet (Roche Diagnostics GmbH, United Kingdom) per 10 mL lysis buffer. Even amounts of whole-cell lysate were loaded onto 4% to 20% Tris-glycine gels (Invitrogen Life Technologies Ltd., United Kingdom) and electrophoresed at 100 V for 1 hour. For Western blotting, proteins were transferred onto polyvinylidene difluoride (PVDF; Millipore, Billerica, MA) using a standard protocol. Membranes were blocked overnight in TBS buffer [20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 0.1% Tween 20] plus 5% bovine serum albumin and subjected to immunodetection using specific antibodies: anti-phosphorylated STAT1 (Tyr⁷⁰¹), STAT3 (Tyr⁷⁰⁵), p38 (Thr¹⁸⁰/Tyr¹⁸²), extracellular signal-regulated kinase 1/2 (ERK1/2; Thr²⁰²/Tyr²⁰⁴), JNK/SAPK (Thr¹⁸³/Tyr¹⁸⁵), MAPK kinase kinase (MKK) 4 (Thr²⁶¹), MKK7 (Ser²⁷¹/Thr²⁷⁵), activating transcription factor (ATF) 2 (Thr⁷¹), c-Jun (Ser⁶³), anti-STAT1, STAT3, c-Jun, c-Myc, IFN regulatory factor 1 [IRF1; New England Biolabs (UK) Ltd., United Kingdom], anti-ATF3, JunB, ETS-2 (Santa Cruz Biotechnology, Santa Cruz, CA), anti-phosphorylated STAT5A/B (Tyr⁶⁹⁴/Tyr⁶⁹⁹; Millipore, Billerica, MA), and anti-α-tubulin (Sigma). Anti-KLF10/TIEG antibody was a kind gift of Malayannan Subramaniam (Mayo Clinic College of Medicine, Rochester, MN). Membranes were then probed with secondary anti-rabbit or anti-mouse horseradish peroxidase–conjugated antibodies (Amersham Biosciences, United Kingdom), and blots were developed using SuperSignal West Pico Enhanced Chemiluminescence (Pierce).

cDNA microarray hybridization and scanning. T47D breast tumor cells were stimulated with IL-6 (100 ng/mL) or OSM (100 ng/mL) ±30-minute preincubation with LIF-05 (500 ng/mL) for 30/60 minutes or with LIF (100 ng/mL) or LIF-05 (500 ng/mL) for 60 minutes. MCF-7 and MDA-MB-231 breast tumor cells were stimulated with OSM (100 ng/mL) for 60 minutes. Control cDNA was hybridized against cDNA from OSM/LIF-05-, OSM-, IL-6-, LIF-, or LIF-05-treated cells. Total RNA was isolated by Trizol extraction according to the manufacturer's instructions (Life Technologies), and equal amounts (20 μg) of RNA were used as starting material for cDNA synthesis. This was done overnight at 42°C using SuperScript II RNase H Reverse Transcriptase (Invitrogen). Random nonamers (3 mg/mL), oligo(dT) (1.0 μg/μL), and Cy3-dCTP or Cy5-dCTP were obtained from Amersham Biosciences, United Kingdom. The RNA template was removed by alkaline hydrolysis, and unincorporated nucleotides were removed with QIAquick PCR purification spin columns as per manufacturer's instructions (Qiagen Ltd., United Kingdom). Equal amounts (130 pmol) of Cy3- and Cy5-labeled probe from control and stimulated samples were mixed together in the presence of human Cot-1 DNA (Life Technologies) and concentrated using YM30 filter (Millipore). Prewarmed (37°C) hybridization buffer (Invitrogen), formamide (BDH Chemicals, United Kingdom), and pd-dA40-60 (Amersham Biosciences, United Kingdom) were mixed with the probe. Genome-wide microarray chips were spotted at the Cancer Research UK DNA microarray facility (Institute of Cancer Research, Sutton, Surrey, United Kingdom) and carried human DNA representing 6,528 pairs of duplicate DNA spots.¹

The mixed probe was heated at 92°C for 3 minutes and cooled to room temperature for 10 minutes before application to the array. This was covered by a LifterSlip (Erie Scientific Co., Portsmouth, NH). Hybridization was done overnight at 42°C in a hybridization chamber (Fisher Scientific Co., United Kingdom). The array was washed five times (2 minutes per wash) in prewarmed (42°C) 2× SSC buffer (Sigma) containing 0.2% SDS and once each (2 minutes) in 1× SSC and 0.5× SSC buffer, respectively. The array was dried quickly with nitrogen gas, and signals were detected with an Axon slide scanner 4000B and GenePix Pro 3 software (Molecular Devices, United Kingdom) at a resolution of 10 μm at 100% laser power and photomultiplier tube voltage of 550 to 650. Data were analyzed with GeneSpring data software package (Agilent Technologies, United Kingdom). Briefly, each data set was transformed into log base 2 expression ratios, normalized using “Lowess” intensity-dependent normalization, and subjected to minimal expression restriction (intensity minimum of 400 for control/raw data). For each pathway, two hybridizations were done (dye swap) to give mean values of n = 2. Only genes overexpressed or underexpressed (2-fold cutoff) in both hybridizations were deemed to be significant. All GenePix ‘gpr’ data files and associated image files have been deposited at the Gene Expression Omnibus repository² [expression platform accession number GPL996 (23); sample accession numbers GSM102927-928, GSM104792-793, GSM104751-770, GSM104788-781; series entry GSE4661].

The LIF antagonist LIF-05 discriminates between type I and type II OSM actions on T47D and MDA-MB-231 breast tumor cells. Effects of gp130 cytokines on breast tumor cell proliferation have been reported previously (15, 24). Reverse transcription-PCR analysis confirmed expression of gp130, LIFRα, OSMRβ, and IL-6 receptor (IL-6R) mRNA by T47D, MCF-7, and MDA-MB-231 cells (data not shown). Here, we assessed the antiproliferative effect of OSM on estrogen receptor–positive T47D and estrogen receptor–negative MDA-MB-231 breast tumor cells. We examined breast tumor cell growth by measuring cell numbers after 7 days of exposure to OSM, LIF, and IL-6 (Fig. 1A). Cytokine concentrations of between 10 and 1,000 ng/mL OSM and IL-6 exhibited potent, dose-dependent growth inhibition of the T47D and MDA-MB-231 cell lines. OSM was more potent than IL-6 in mediating growth inhibition in these experiments. In contrast, only the highest nonphysiologic concentration (1,000 ng/mL) of LIF suppressed T47D and MDA-MB-231 cell growth (Fig. 1A), indicating that gp130/LIFRα heterodimer formation does not mediate sustained growth suppression. These findings confirm the significant cytostatic functions of human OSM on human breast tumor cells.

The specific LIFR antagonist, LIF-05, has previously been shown to block ligand-stimulated tyrosine phosphorylation of both gp130 and the LIFR and suppress biological responses mediated by the gp130/LIFRα heterodimeric complex (19). LIF-05, therefore, represents a useful tool to discriminate between responses mediated by the type I and type II OSM receptor complexes. T47D and MDA-MB-231 cells were incubated for 7 days in the presence of OSM (200 ng/mL) or OSM plus increasing concentrations of LIF-05 (10, 100, and 1,000 ng/mL). OSM, on its own, inhibited proliferation of T47D and MDA-MB-231 cells by 41% and 30%, respectively (Fig. 1B). Addition of LIF-05 up to 1,000 ng/mL did not block OSM growth inhibition, confirming that this effect is mediated through the type II gp130/OSMRβ complex and not the type I gp130/LIFRα complex. In fact, LIF-05 enhances the cytostatic activity of OSM, suggesting that type I and type II receptor signaling may exhibit antagonistic functions.

In the course of these experiments, we noted, as previously reported by others (15, 25), a striking effect of OSM on breast tumor cell morphology (Fig. 1C). T47D cells, in the absence of any added cytokine, are typically epithelioid in shape growing tightly in flat, monolayer colonies (Fig. 1C,, a, top). MDA-MB-231 cells grow as a monolayer of dissociated epithelial-like cells due to loss of E-cadherin expression and intercellular adhesion (Fig. 1C,, a, bottom). Exposure to OSM dramatically altered morphology, with T47D cells becoming spherical or fusiform in appearance, exhibiting decreased intercellular contact and forming aggregated colonies (Fig. 1C,, b, top). MDA-MB-231 cells became elongated, exhibiting a more fibroblast-like phenotype and developed long pseudopodia (Fig. 1C,, b, bottom). OSM also altered MCF-7 cell morphology (data not shown). These cells became very heterogeneous in size and shape, developed numerous cytoplasmic processes, and also exhibited decreased intercellular contact. T47D cells exposed to LIF (Fig. 1C,, c, top) or IL-6 (Fig. 1C,, d, top) remained epithelioid in shape and still grew in flat monolayer colonies, but there were a higher number of cells showing rounded or fusiform morphologies than occurred in control colonies. Cell borders also appeared more defined, indicating a small effect on cell-cell adhesion. Similarly, LIF and IL-6 had a less dramatic effect than OSM on MDA-MB-231 cell morphology, which remained predominantly epithelial-like (Fig. 1C,, c and d, bottom), although cells developed higher numbers of small, cytoplasmic extensions compared with untreated control cells. This suggests that LIF and IL-6 have a small effect on breast tumor cell morphology but OSM seems the prominent cytokine of this family to stimulate gross structural changes in both cell lines. Incubation with OSM in the presence of LIF-05 did not block OSM-induced changes in cell morphology of either T47D or MDA-MB-231 cells (Fig. 1C , e, top and bottom). LIF-05 alone did not stimulate any change in cell shape or structure compared with untreated controls (data not shown). Thus, the characteristic changes OSM induces in breast tumor cell morphology are also mediated via the type II gp130/OSMRβ receptor complex.

OSM, LIF, and IL-6 exhibit differential STAT and MAPK activation. Activation of both the JAK/STAT and MAPK pathways by gp130 cytokines has been well defined as a mechanism of signal transduction (6). In this study, we used our panel of cytokines to determine the repertoire of STATs and MAPKs activated by each receptor complex in T47D cells. We studied STAT1, STAT3, and STAT5 only, as other STATs are not implicated in gp130 family cytokine signaling in breast tumor cells. OSM, LIF, and IL-6 stimulated phosphorylation of STAT3 (Fig. 2A). In contrast, only OSM and IL-6 were able to induce appreciable STAT1 and STAT5 phosphorylation. This suggests that gp130/LIFRα heterodimerization does not initiate efficient activation of STAT1 or STAT5. In addition, we tested for activation of MAPKs, ERK1/2, p38, and JNK/SAPK. OSM, LIF, and IL-6 activated p42/44 ERK1/2 with LIF being more potent than OSM or IL-6 (Fig. 2A). None could induce p38 phosphorylation (data not shown). OSM and IL-6 mediated p46 JNK/SAPK phosphorylation, but the former cytokine was 10-fold more potent (Fig. 2A). In the MCF-7 breast tumor cell line, only OSM stimulation resulted in JNK/SAPK phosphorylation (data not shown).

Cells were then stimulated with OSM or LIF in the presence of LIF-05 to identify signals propagated by the type II gp130/OSMRβ complex. As the concentration of LIF-05 increased (10-1,000 ng/mL), LIF-induced STAT3 phosphorylation decreased but was unaffected in OSM-stimulated cells (Fig. 2B). Thus, human OSM has the potential to mediate STAT3 phosphorylation through both gp130/LIFRα and gp130/OSMRβ complex formation. LIF-05 had no effect on OSM-mediated STAT1 (Fig. 2B) or STAT5 phosphorylation (data not shown). This reveals that OSM activates both STATs primarily via the gp130/OSMRβ heterocomplex because LIF cannot stimulate effective phosphorylation of STAT1 or STAT5. LIF-05 also had no effect on OSM-induced JNK/SAPK phosphorylation, again defining a specific requirement for the gp130/OSMRβ heterocomplex in this pathway. Conversely, stimulation with OSM or LIF in the presence of LIF-05 decreased the level of ERK1/2 phosphorylation (Fig. 2B), suggesting that gp130/LIFRα dimerization significantly contributes to OSM-induced ERK1/2 activation. OSM, LIF, and IL-6 activation of STAT and MAPK pathways by each of the representative receptor complexes are summarized (Fig. 2C). The type II gp130/OSMRβ complex specifically activates signaling via the JNK/SAPK and STAT1/STAT5 pathways, whereas both type I and type II complexes strongly activate the canonical STAT3 and MAPK/ERK1/2 signaling pathways.

OSM and IL-6 regulate ‘public’ and ‘private’ subsets of early-response genes in T47D breast tumor cells. The previous experiments show that human OSM activates different signaling pathways depending on the identity of the receptor complex used. These signals result in the activation of gene expression, which leads to the biological response. In this case, we were interested to define those genes that are specifically regulated by the gp130/OSMRβ type II complex as mediators of the cytostatic and morphologic effects of human OSM on breast cancer cells. We characterized the repertoire of genes regulated by signaling through different OSM receptor complexes using a custom genome-wide cDNA microarray containing ∼6,500 genes selected for their roles in cell proliferation, checkpoint control, and survival.¹

T47D breast tumor cells were either preincubated with LIF-05 for 30 minutes before stimulation with OSM or stimulated with OSM, IL-6, LIF, or LIF-05 alone (Fig. 3A,, pathways A to E). Sets of genes up-regulated as a result of stimulation of the gp130/OSMRβ heterocomplex (pathway A) or arising via the combined activities of the gp130/OSMRβ and gp130/LIFRα receptor complexes (pathway B) were compared (Fig. 3B and C). After 30 and 60 minutes, 10 and 25 genes, respectively, were significantly (>2-fold) up-regulated in the presence or absence of the LIFR antagonist LIF-05, indicating OSMRβ-specific gene induction. LIF-05 stimulation (pathway E) had no effect on gene expression. LIF stimulation (pathway D) also had no effect on gene expression. This mirrors the lower signaling capability of LIF and lack of growth suppression mediated by 100 ng/mL of this cytokine. Genes up-regulated as a result of IL-6 activation of the gp130/gp130 homodimer (pathway C) were compared with the OSM-stimulated gene set of pathway B (Fig. 3B and C). Collectively, these define a set of 12 ‘shared’ OSM- and IL-6-regulated genes (7 and 11 common genes up-regulated after 30 and 60 minutes, respectively) and a set of 15 ‘specific’ gp130/OSMRβ-regulated genes and 5 ‘specific’ gp130/gp130-regulated genes (Fig. 3B and C). Of the shared genes up-regulated, five were early-response transcription factors (junb, atf3, klf10/tieg, irf1, and fra2). OSM specifically up-regulated a further four transcription factors (c-fos, c-myc, c-jun, and ets-2).

To identify an OSM-induced ‘gene signature’ common to multiple breast tumor–derived cell lines, microarray experiments were done on MCF-7 and MDA-MB-231 cells in addition to the T47D cell line (Fig. 4A). OSM mediated up-regulation of six genes (klf10/tieg, c-jun, junb, irf1, rgs16, and efna1) in all three cell lines and a further 15 genes (c-fos, egr1, atf3, c-myc, ets-2, socs3, fosl2/fra2, cdk5r1, ccl2, ccl7, ccl11, ctgf, icam1, ier3, and pggt1b) in at least two of the three cell lines.

Distilling this data, we arrive at a common core set of OSM/IL-6-regulated transcription factors (junb, atf3, klf10/tieg, and irf1) and a specific ‘OSM-only’ set of transcription factors (c-jun, c-myc, and ets-2) whose expression may be tightly associated with the cytostatic and morphologic responses. We next validated these genes by examining the induction of protein expression in response to OSM and IL-6 signaling. These results (Fig. 4B) showed that, at the protein level, ATF3 and JunB exhibited robust induction in response to OSM signaling, c-Jun and c-Myc were induced at lower levels, whereas IRF1, ETS-2, and KLF10/TIEG expression were unaffected. In summary, fold changes for ‘shared’ genes (Fig. 3B) and protein levels (Fig. 4B) were generally higher for OSM (especially JunB and ATF3) compared with IL-6 and potentially reflect the greater potency of biological action exhibited by OSM.

Long-term regulation of gene expression by OSMRβ. We extended the microarray study to longer time points (3 hours and 1, 3, and 6 days) to coincide with the onset and establishment of stasis with the objective of discerning the signature of gene expression associated with the cytostatic response to activation of gp130/OSMRβ. The results are detailed in Supplementary Table S1.² Prominent in this data set is the up-regulation of genes associated with inhibition of cell multiplication and growth arrest, such as RBP1 (26) and GADD45 (27), and down-regulation of genes associated with DNA synthesis (histones) and G₂-M cell cycle progression, such as CDC20 (28) and CENPF/mitosin (29).

These findings are consistent with an OSM-dependent growth arrest mechanism that has multiple axes acting at both the G₁-S and G₂-M checkpoints. In addition to genes implicated in cell cycle control, this data set also displays significant changes in genes involved in a wide variety of metabolic processes and immune regulatory functions (Supplementary Table S1). These results are consistent with the proposal that OSM-induced arrest of breast tumor cell multiplication and associated `morphologic differences resemble the induction of a partial terminal differentiation program of pleiotropic changes (17).

OSM specifically activates the JNK/SAPK pathway in T47D breast tumor cells. As OSM seemed to be a potent activator of JNK/SAPK serine/threonine kinase, we assessed activation of downstream effectors ATF2 and c-Jun. JNK/SAPK was originally associated with stress-induced and inflammatory responses but has known roles in other biological outcomes, such as growth and differentiation. JNK/SAPK is activated by stress signals (UV, γ-radiation, and osmotic shock) as well as inflammatory cytokines (tumor necrosis factor-α, IL-1β, and IFN-γ) becoming phosphorylated on threonine and tyrosine residues in response to upstream MAPK kinases MKK4 and MKK7 (30). Once activated, JNK/SAPK translocates to the nucleus where it can phosphorylate members of the activator protein-1 family of transcription factors (c-Jun, JunB, JunD, and ATF2). We used antibodies to phosphorylated MKK4, MKK7, c-Jun, and ATF2 to investigate the ability of OSM and IL-6 to activate JNK/SAPK pathway components. OSM stimulation resulted in phosphorylation of c-Jun and ATF2 in a dose-dependent manner, whereas IL-6 activated ATF2 and, to a much lesser extent, c-Jun (Fig. 5A). LIF stimulation did not phosphorylate JNK/SAPK or downstream components. In contrast to MKK7 (data not shown), T47D cells displayed a low basal level of MKK4 phosphorylation. This increased after OSM stimulation with maximum levels at 10 to 15 minutes coinciding with optimum JNK/SAPK phosphorylation (Fig. 5B), implicating this threonine/tyrosine kinase to be upstream of JNK/SAPK. OSM stimulation for 2 hours also caused a dose-dependent up-regulation of ATF3 (Fig. 5C), a known target of c-Jun and ATF2 (31).

We used a specific ATP-competitive inhibitor of JNK/SAPK serine/threonine kinase, SP600125, to confirm that phosphorylation of c-Jun and ATF2 was mediated by OSM-induced JNK/SAPK activity. At a concentration of 10 μmol/L SP600125, phosphorylation of both transcription factors was blocked on stimulation with OSM (Fig. 5D), confirming a requirement for JNK/SAPK kinase activity. Phosphorylation of JNK/SAPK itself was also partially blocked. These results show that OSM, acting through the gp130/OSMRβ, is a potent activator of the JNK/SAPK pathway. JNK/SAPK inhibition also reduced OSM-mediated ATF3 and c-Jun expression (Fig. 5E). However, there was minimal effect on c-Myc expression and no effect on JunB expression. This suggests that gp130/OSMRβ signaling via the JNK/SAPK pathway results in regulation of at least two key early-response genes, ATF3 and c-Jun.

Inhibition of JNK/SAPK abrogates OSM-induced morphologic changes of T47D breast tumor cells. We assessed the effect of JNK/SAPK inhibition on OSM-induced growth inhibition and morphologic changes of T47D cells (Fig. 6A and B). Ligand stimulation of IL-6R oligomerization and subsequent phosphorylation of tyrosine residues in the receptor cytoplasmic regions have the potential to attract several effector molecules, which link these receptors to downstream signaling pathway. We assessed the effect of a panel of chemical inhibitors of tyrosine kinase activity on OSM-mediated growth inhibition. T47D cells, exposed to OSM, were incubated for 7 days in the presence or absence of specific kinase inhibitors. Viable cell numbers were counted by trypan blue exclusion. Effects of tyrosine kinase inhibition are shown for T47D cells (Fig. 6A). Inhibition of platelet-derived growth factor receptor/vascular endothelial growth factor receptor (PDGFR/VEGFR; SU5614), JNK/SAPK (SP600125), phosphatidylinositol 3-kinase (PI3K; LY294002), and the JAK family of kinases (AG490) partially inhibited cell growth in the absence of OSM, indicating a requirement for these molecules in effective growth and survival. However, none of the tyrosine kinase inhibitors blocked the cytostatic effects of OSM on T47D cells (Fig. 6A) or MCF-7 cells (data not shown). This indicates that these effectors and associated downstream targets, such as ERK1/2 and STATs, cannot, on their own, mediate OSM-induced cytostatic activity, which may require combinatorial input from multiple signaling pathways.

For morphology studies, cells were grown for 7 days in the presence or absence of OSM (50 ng/mL) and SP600125 (100 nmol/L, 10 μmol/L). In normal growth medium, cells displayed flat, monolayer, epithelial-like cell morphology (Fig. 6B,, a), whereas OSM (50 ng/mL) induced the characteristic alterations in morphology (Fig. 6B,, b) as described above. Cells exposed to OSM in the presence of 100 nmol/L SP600125 underwent morphologic changes, but these did not seem as pronounced as with OSM alone (Fig. 6B,, c). In the presence of a higher concentration of SP600125 (10 μmol/L), at which phosphorylation of c-Jun and ATF2 is blocked, T47D cells retained their epithelial-like shape (Fig. 6B,, d). Exposure to SP600125 alone (100 nmol/L or 10 μmol/L) had little effect (Fig. 6B , e and f). These findings suggest that JNK/SAPK pathway activation is integral in OSM-induced changes on T47D breast tumor cell morphology. However, because SP600125 alone did not abrogate OSM growth inhibition, this indicates that integration of additional signaling pathways (ERK1/2 or STATs) at the level of gene transcription is a necessary requirement for OSM cytostasis.

The aim of this study was to delineate the molecular pathways involved in the cellular responses of breast tumor cells to the cytostatic cytokine OSM. Studies using the receptor-specific antagonist LIF-05 revealed that the characteristic functions of OSM are executed by signaling through the gp130/OSMRβ complex and not by the gp130/LIFRα complex, which has a broader ligand specificity (5). Using this platform, we revealed that the gp130/OSMRβ and gp130/LIFRα dimers exhibit shared and distinct signaling pathways. The widely used STAT3 and ERK1/2 systems are common to both receptors, whereas a defining feature of signaling via the gp130/OSMRβ complex activates the JNK/SAPK pathway. We also noted that gp130/OSMRβ formation shares features with gp130 homodimer signaling, but not gp130/LIFRα signaling, in the activation of STAT5 and STAT1.

Two conclusions emerge from these experiments. The defining morphologic features stemming from gp130/OSMRβ complex formation involve the JNK/SAPK signaling pathway. Second, the gp130 cytokine family as a whole should be considered as a single multifunctional entity whose signaling outputs are qualitatively and quantitatively dictated by the repertoire of receptors expressed by the responding cell type. These considerations highlight the use of receptor-specific agonists and antagonists in eliciting receptor-specific cell responses in both experimental (19) and therapeutic settings (22, 32).

A second OSMRβ-dependent cytokine, IL-31, has recently been reported (33), which signals via heterodimerization with the gp130-related receptor GPL (34, 35). Characterization of the signaling pathways elicited by the OSMRβ/GPL heterodimer has revealed activation of the STAT1, STAT3, and STAT5 pathways but not the ERK1/2 or JNK/SAPK pathways, suggesting that OSMRβ signaling may be further influenced by the identity of the coreceptor participating in the activated signaling complex.

The execution of OSM-dependent biological responses is dictated by the transcriptional response to gp130/OSMRβ activation. Here, this is defined by microarray methods and validated at the level of protein expression for selected targets. In agreement with similar studies (36, 37), a time-dependent profile of transcriptional remodeling is observed where rapid induction, and subsequent disappearance, of an “early-response” set of genes is later followed by the induction and suppression of a broader set of genes in a process that extends in our study for 6 days. By comparing the early-response profile in several OSM-responsive breast tumor cell lines and using a variety of stimuli, including targeted activation of OSMRβ, we define a ‘kernel’ transcriptional profile of early responses to activation of the gp130/OSMRβ heterodimer common to the breast tumor cell lines under investigation. Further analysis of the kernel by validation of protein expression revealed four OSMRβ-dependent early genes whose expression was common to multiple breast tumor cell lines: atf3, c-jun, junb, and c-myc.

Members of the Jun family of transcription factors are well documented to respond to activation of both the ERK1/2 and JNK/SAPK pathways (38, 39) and are defined as having growth-promoting and growth-suppressing functions, respectively, with JunB having the property of suppressing the growth-promoting properties of c-Jun (40). ATF3 induction is a well-documented transcriptional response to a wide variety of genotoxic and nongenotoxic stress stimuli [including cytokine-mediated activation of the JNK/SAPK pathway (36)], which mediates cytostatic responses in tumor cell lines in part through stabilization of the p53 response at the protein level (41) and possibly via activation or suppression of downstream ATF3 gene targets (42). However, in other biological contexts, ATF3 has been reported to exert a multiplication enhancing effect (43). Hai et al. (44) have argued that the divergent cell responses to ATF3 activation reflect the context in which ATF3 functions are executed. In this respect, the results presented here support this notion in that transcriptional responses to a ‘pure’ OSMRβ signal differ quantitatively in cells that undergo a similar biological response (cytostasis). Recall that signaling via OSMRβ involves activation of shared or core pathways, such as STAT3 and ERK1/2, alongside private pathways, such as JNK/SAPK. This implies that, as described above, seeking a single target gene for mediation of OSMRβ-mediated cytostasis may be a false errand if the response depends on the integrated totality of transcriptional changes. The early gene transcriptional responses described here are, however, entirely consistent with the ability of signaling via OSMRβ to elicit robust and prolonged activation of the JNK/SAPK pathway, which leads to activation of a small kernel of early gene responses that, in turn, are modified by coinduction of response modifiers in a cell type–dependent context leading to morphologic changes and cytostasis.

The long-term transcriptional responses to OSMRβ activation are more diverse and numerous than early changes, and at time points when the cytostatic effect of OSM signaling is firmly established, we observe pleiotropic changes in the gene expression profile, including alterations in key components of G₁ and G₂ checkpoint controls, immune response regulators, and diverse metabolic enzymes. These findings support the proposals of Grant et al. (17, 45) that the effects of OSM on breast cancer cells involve significant long-term changes in cell phenotype, which resemble a differentiation program. To conclude, these results identify a core signaling/transcriptional mechanism specific to human OSM, which feed into the downstream biological effects induced on breast tumor cells. In addition, these findings provide the first OSMRβ-specific gene expression profile, at both early and late time points, which provide a framework from which to study OSM functions in other pathologic and nonpathologic contexts.

Grant support: Cancer Research UK.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Celine Jones and Anthony Jones for skilled technical support and advice, Ian Giddings (Institute of Cancer Research) for supplying the microarrays, and Suzie Grant and Glenn Begley for discussions and sharing of data.