Oncostatin-M Promotes Phenotypic Changes Associated with Mesenchymal and Stem Cell-Like Differentiation in Breast Cancer

ORIGINAL ARTICLE Oncostatin-M promotes phenotypic changes associated with mesenchymal and stem cell-like differentiation in breast cancer

NR West1,2, JI Murray1 and PH Watson1,2,3

Cancer stem cell (CSC) biology and the epithelial-to-mesenchymal transition (EMT) are thought to be mechanistically linked and may be key components of cancer development and progression. However, stimuli that induce EMT and CSC-like features (‘stemness’) are poorly defined. We and others have shown that the inflammatory cytokine oncostatin-M (OSM) mediates phenotypic changes in breast cancer that are consistent with EMT and dedifferentiation, including enhanced migration and loss of hormone receptors. In this study, we have expanded on these prior observations to determine whether OSM is a cell-extrinsic driver of EMT and/or stemness. OSM stimulation of the luminal breast cancer cell lines MCF7 and T47D induced EMT features including loss of membranous E-cadherin and induction of snail and slug expression. OSM treatment markedly enhanced the formation of mammospheres (up to 20-fold, Po0.001), which displayed high expression of the pluripotency factor SOX2. The proportion of cells with a CD44highCD24 À /low phenotype was similarly increased by OSM (Po0.001). OSM-induced mammosphere formation and CD44highCD24 À /low induction was dependent on PI3K signalling. In silico analysis of human breast tumours (from a publicly available data set, n ¼ 322) confirmed that co-expression of a PI3K transcriptional signature, but not MAPK or STAT3 signatures, was necessary to detect an association between OSMR and poor prognosis. Assessment of a second in silico data set (n ¼ 241 breast tumours) confirmed a significant relationship between OSMR, markers of EMT and CSCs, and chemotherapy resistance. Direct analysis of mRNA expression by RT–PCR in a third cohort (n ¼ 72 breast tumours) demonstrated that high expression of OSM is associated positively with indicators of EMT (SNAI1, Po0.001) and stemness (SOX2, Po0.05). Our data suggest for the first time that OSM may promote a clinically relevant EMT/CSC-like phenotype in human breast cancer via a PI3K-dependent mechanism.

Oncogene (2014) 33, 1485–1494; doi:10.1038/onc.2013.105; published online 15 April 2013 Keywords: oncostatin-M; breast cancer; cancer stem cell; cytokine; epithelial-to-mesenchymal transition; inﬂammation

INTRODUCTION signalling is associated with poor prognosis and aggressive An improved understanding of the processes that underlie breast subtypes of breast cancer (basal-like and claudin-low) that 10–14 cancer progression and treatment resistance are urgently needed. frequently contain abundant leukocytes and cytokines, as 9,11 To this end, considerable attention has been devoted to the well as high expression of genes associated with EMT and CSCs. phenomenon of epithelial-to-mesenchymal transition (EMT), dur- OSM stimulation occurs through heterodimeric receptors ing which cells lose epithelial polarity and gain mesenchymal incorporating gp130 and either OSM receptor-b (OSMR) or the morphology and motility. EMT is postulated to occur in cancer and leukaemia inhibitory factor receptor (LIFR). These signal via janus may underpin the processes of migration and invasion that are kinases (JAKs) to activate the STAT3 (signal transducer and central to cellular dissemination during metastasis.1 Associated activator of transcription 3), PI3K (phosphatidylinositol-3-kinase) 15 with EMT is the cancer stem cell (CSC) hypothesis, which and MAPK (mitogen-activated protein kinase) cascades. postulates that tumours are generated and maintained by a Given that OSM is predominantly produced by leukocytes such 16 small population of self-renewing stem-like cells.1,2 In this model, as T cells, macrophages and neutrophils, the above observations CSCs are considered the primary agents of disease recurrence due imply that OSM may be an extrinsic factor in the tumour to their superior apoptotic resistance and capacity for tumour microenvironment that promotes EMT and/or CSC-like features generation. Importantly, mechanisms that are thought to regulate in breast cancer cells. To further explore this concept, we have EMT may jointly inﬂuence CSC-like features, and increasing examined the effect of OSM on EMT and CSC features in vitro and evidence suggests that the tumour microenvironment can the association of OSM signalling with markers of these inﬂuence the emergence of both EMT and CSC phenotypes in phenomena in vivo in human breast cancer. neoplastic cells.1–4 We and others have shown that oncostatin-M (OSM), a member of the interleukin-6 (IL-6) cytokine family, can potently induce RESULTS migration and invasiveness in human breast cancer.5–8 In addition, OSM induces an EMT-like process we have shown that OSM suppresses oestrogen receptor-a (ER) Increased migration is an EMT hallmark that OSM stimulates in expression,9 which is a key feature of luminal epithelial breast cancer cells.5–9 To expand on this observation, we differentiation in breast cancer cells.10 In human tumours, OSM examined the expression of several EMT-related genes in MCF7

1Deeley Research Centre, British Columbia Cancer Agency, Victoria, British Columbia, Canada; 2Department of Biochemistry and Microbiology, University of Victoria, Victoria, British Columbia, Canada and 3Department of Pathology and Laboratory Medicine, University of British Columbia, Vancouver, British Columbia, Canada. Correspondence: Dr PH Watson, Deeley Research Centre, 2410 Lee Avenue, 3rd Floor Research, Victoria, British Columbia, Canada V8R 6V5. E-mail: [email protected] Received 6 October 2012; revised 23 January 2013; accepted 7 February 2013; published online 15 April 2013 OSM promotes EMT and stem cell features in breast cancer NR West et al 1486 and T47D cells after 2 days of OSM stimulation. These cell lines Figure S2). Intriguingly, OSM treatment caused E-cadherin to were chosen because they are established models of well- shift from a membranous to an intracellular localization, suggest- differentiated, luminal breast cancer. OSM treatment induced ing that OSM can regulate cell adhesion by controlling E-cadherin expression of the transcription factors snail and slug (SNAI1 and at the post-translational level (Figure 1c). In both cell lines, SNAI2) in both cell lines (Figure 1a). However, OSM did not induce perturbation of E-cadherin was accompanied by a similar twist (TWIST1) or vimentin (VIM), while fibronectin (FN1), FOXC1, redistribution of b-catenin, such that many cells displayed and ZEB1 were variably induced between the two cell lines. high levels of b-catenin in the cytoplasm and/or nucleus Intriguingly, although MCF7 cells suppressed E-cadherin (CDH1) (Figure 1d; Supplementary Figure S2). Notably, cells with high and increased N-cadherin (CDH2) expression, this was not levels of non-membranous b-catenin typically lacked intercellular observed in T47D cells, despite the impact of OSM on E-cadherin adhesion and displayed apparent migratory morphology protein localization (see below). The effect of OSM on snail and (Figure 1d). E-cadherin expression was confirmed by western blot assays We have recently reported that OSM suppresses ER expression (Figure 1b). E-cadherin suppression in MCF7 cells occurred rapidly in breast cancer cells.9 As snail can suppress ER,18 we blocked following OSM stimulation (within 3 hours), and protein levels OSM-induced snail expression in MCF7 cells using siRNA remained diminished with ongoing treatment (Supplementary (Supplementary Figure S3a), but observed no change in OSM- Figure S1a). E-cadherin levels recovered within 24 h following mediated ER suppression. However, snail knockdown did prevent withdrawal of OSM, however, indicating that OSM signalling does induction of S100A7, an OSM target gene that promotes migration not induce permanent E-cadherin suppression (Supplementary in MCF7 cells.5 Figure S1b). As ER can augment E-cadherin expression,19 we tested the Loss of E-cadherin is considered an essential hallmark of EMT,17 effects of oestrogen stimulation on MCF7 cells. Oestrogen and its interaction with b-catenin is an important component of treatment did indeed promote E-cadherin expression, but this adherens junctions. Under control conditions, immunofluore- was abrogated in cells concurrently treated with OSM scence microscopy demonstrated membranous colocalization (Supplementary Figure S3b). Thus, loss of ER expression in MCF7 of E-cadherin and b-catenin in both cell lines (Supplementary cells may contribute to OSM-induced E-cadherin repression.

Figure 1. Induction of EMT-associated features by OSM. (a) EMT marker mRNA expression in OSM-stimulated MCF7 and T47D cells. Bars represent mean (±s.d.) mRNA expression from triplicate RT–PCR assays relative to untreated cells. (b) Western blot detection of E-cadherin and snail following OSM stimulation. Immunoﬂuorescent detection of E-cadherin (c) and b-catenin (d) in MCF7 and T47D cells, demonstrating intracellular redistribution of both proteins in OSM-stimulated cells. Scale bars equal 150 mm.

Oncogene (2014) 1485 – 1494 & 2014 Macmillan Publishers Limited OSM promotes EMT and stem cell features in breast cancer NR West et al 1487 OSM induces expression of pluripotency genes and expression of SNAI1, SNAI2 and SOX2 (Figure 4). S100A7 induction mammosphere formation was significantly but incompletely inhibited by OSMR knockdown, Given the mechanistic link between EMT and CSC phenotypes,1 suggesting that S100A7 may require relatively low levels of OSMR we examined the expression of three transcription factors that are for its induction. Knockdown of OSMR also abrogated the ability of characteristic of pluripotent cells:20 POU5F1 (OCT4), NANOG and OSM to induce a CD44highCD24 À /low phenotype (Supplementary SOX2 (Figure 2a). While neither cell line displayed increased Figure S5). POU5F1 expression following OSM treatment, both increased expression of SOX2; however, this was reproducible and significant OSM-induced stemness requires PI3K signalling only for MCF7 cells (P 0.05 in 3/3 replicate experiments). NANOG o Signal transduction from the OSM receptor primarily involves was consistently and significantly upregulated in T47D cells. JAK-dependent activation of the STAT3, MAPK and PI3K pathways. The ability to form mammospheres is established as an in vitro 21–23 As expected, treatment with the pan-JAK inhibitor pyridone-6 indicator of tumourigenic mammary CSCs. OSM treatment completely abrogated OSM-induced mammosphere formation, as clearly enhanced mammosphere formation by MCF7 and T47D did treatment with the PI3K inhibitor LY294002 (Figures 5a and b). cells (Figures 2b and c). The average baseline proportion of PI3K inhibition similarly prevented OSM from inducing a mammosphere-forming MCF7 cells was 0.57%, while OSM high /low CD44 CD24 À phenotype (Po0.001; Figure 5c). In contrast, treatment increased this value to 2.93%. Similarly, T47D cells OSM significantly enhanced mammosphere formation in the displayed an average baseline proportion of 0.34%, compared presence of U0126 and S3I201, which respectively inhibit MEK1/2 with an OSM-stimulated proportion of 7.27%. OSM-induced fold and STAT3 (Figures 5a and b). As b-catenin can exert transcrip- changes in mammosphere-forming capacity ranged from tional effects that promote EMT and stemness, and because 4.17–6.38 in MCF7 cells and 20.3–23.1 in T47D cells (P 0.001; o membranous b-catenin is lost following OSM stimulation Figure 2c). OSM-stimulated mammospheres were also significantly (Figure 1d), we asked whether b-catenin was necessary for OSM- larger than those grown under control conditions (Po0.0001 for induced mammosphere formation. However, treatment with both cell lines; Figure 2d). XAV939, which specifically blocks b-catenin signalling, had no We next assessed gene expression in control and OSM-induced measurable impact on mammosphere growth (data not shown). mammospheres relative to conventionally cultured cells (that is, Thus, PI3K signalling appears to be the major element required for adherent growth with 5% serum). OSMR and S100A7 are OSM-induced stemness. transcriptional targets of OSM signalling and were highly To determine if the PI3K pathway has a clinically relevant expressed in OSM-stimulated mammospheres (Figure 2e). As association with OSM signalling, we examined data from a publicly expected, SNAI1 was highly expressed in mammospheres, but available cohort from the University of North Carolina that showed no consistent changes with respect to OSM treatment. included over 300 breast tumours with whole-transcriptome gene Similar results were obtained with POU5F1 and NANOG (data not expression data (UNC cohort11). We have previously shown that shown). SOX2, however, was expressed at consistently high levels high OSMR expression (43rd quartile) is associated with poor in OSM-treated mammospheres relative to control cultures clinical outcome in this data set.9 When cases were categorized (Figure 2e). Given its recent identification as an essential factor 24 according to their expression of OSMR and a gene signature of for mammosphere formation and its known role as a driver of 27 25 PI3K activity, those with high levels of each had a dismal 5-year induced pluripotency, SOX2 induction may contribute to the recurrence-free survival (RFS) rate of only 12.4% (Figure 5d). In mammosphere-promoting effect of OSM. contrast, cases with high OSMR levels but low expression of the PI3K signature had a significantly better 5-year DFS rate of 63.4% high À /low OSM signalling promotes a CD44 CD24 cell surface (HR ¼ 4.48, 95% CI 2.04–9.85; P ¼ 0.0002). Notably, expression of phenotype gene signatures that are associated with Ras/MAPK28 and STAT329 The primary immunophenotypic assay for identifying mammary activity had no impact on the prognostic relevance of OSMR CSCs is detection of high CD44 and low CD24 (CD44highCD24 À /low) expression (Figure 5d). expression by flow cytometry.2,26 As the optimal staining thresholds for identifying tumourigenic CD44highCD24 À /low cells 2 The OSM pathway is associated with EMT/CSC features in human are unclear, we drew gates on control samples to include a tumours CD44highCD24 À /low population that comprised the same To determine if OSM signalling is associated with markers of EMT proportion of cells observed to form mammospheres under and CSCs in vivo, we examined a cohort of 72 human invasive control conditions. OSM-treated cells displayed increased CD44 breast carcinomas using RT–PCR to measure expression of SNAI1, expression such that the proportion of gated CD44highCD24 À /low POU5F1 and SOX2 relative to OSM and OSMR. Data for these genes cells increased B6-fold (P 0.001) and 22-fold (P 0.001) in MCF7 o o were median-normalized and cases were separated on the basis of and T47D cells, respectively (Figures 3a and b). These values are high (upper quartile) and low expression of SNAI1, POU5F1 or very consistent with the OSM-induced fold-changes in mammo- SOX2. Although high expression of these genes tended to sphere formation noted above (Figure 2). The ratio of CD44 to associate with high levels of OSMR, this was not statistically CD24 expression correspondingly increased by two- to three-fold significant (data not shown). However, cases high in SOX2 or SNAI1 following OSM stimulation in both cell lines (data not shown). were significantly enriched for expression of OSM (Figure 6a). As Assessment of OSMR levels by flow cytometry revealed a fold OSM-induction of S100A7 is dependent on snail and may mediate increase of 42.5 in both cell lines following OSM treatment some functional aspects of EMT, we also examined S100A7 (P 0.001; Supplementary Figure S4a). Notably, OSMR expression o expression and found that it was significantly associated with OSM correlated with CD44 such that OSM induced a considerable (Figure 6a). increase in CD44highOSMRhigh cells (Supplementary Figure S4b). Both the UNC data set and our in-house cohort are comprised The OSM-induced enrichment of CD44highCD24 À /low cells was of surgical tumour specimens, which contain significant popula- also more pronounced among OSMRhigh cells (data not shown). tions of stromal cells. To focus on gene expression profiles of tumour epithelia and reduce the impact of contributions from OSM effects are mediated by OSMR inflammatory cells, we used a second publicly available data set We previously reported that OSM suppresses ER via OSMR rather that included microarray gene expression data derived from fine than LIFR.9 Knockdown of OSMR expression in MCF7 cells using needle aspirates of 278 breast carcinomas that were subsequently siRNA confirmed that OSMR is required for OSM-induced treated with neoadjuvant chemotherapy (MAQC cohort30).

& 2014 Macmillan Publishers Limited Oncogene (2014) 1485 – 1494 OSM promotes EMT and stem cell features in breast cancer NR West et al 1488

Figure 2. Induction of CSC-associated features by OSM. (a) RT–PCR analysis of pluripotency factor mRNA in OSM-stimulated cells. Data are expressed relative to untreated cultures. (b) Representative images of 10-day mammosphere cultures with or without OSM. Scale bars equal 100 mm. (c) Mammosphere formation rates in a representative assay including four replicate cultures. (d) Size distribution of mammospheres 470 mm in diameter. (e) mRNA expression in control or OSM-stimulated mammospheres relative to pooled samples from three independent standard cultures. Bars in all panels represent means±s.d., with signiﬁcance determined by Student’s t-tests. *P ¼ 0.01–0.05; **P ¼ 0.001–0.01; ***Po0.001.

Oncogene (2014) 1485 – 1494 & 2014 Macmillan Publishers Limited OSM promotes EMT and stem cell features in breast cancer NR West et al 1489

Figure 3. Flow cytometry analysis of MCF7 and T47D cells. (a) CD24 and CD44 expression. Quadrants denote positivity thresholds based on negative staining controls. Elliptical gates denote CD44highCD24 À /low cells. (b) Fold changes (relative to control cultures) in the proportion of cells with a CD44highCD24 À /low phenotype following 2 days of OSM stimulation. Bars represent means (±s.d.) of at least three independent assays, with signiﬁcance established by one-sample t-test (versus a hypothetical mean of 1). ***Po0.001.

Figure 4. Blockade of OSM signalling via OSMR knockdown. MCF7 cells were transfected with GFP- or OSMR-speciﬁc siRNA for 2 days before OSM treatment. Bars represent means (±s.d.) from triplicate RT–PCR assays, expressed relative to untreated controls (compared using the Student’s t-test). *P ¼ 0.01–0.05; **P ¼ 0.001–0.01; ***Po0.001.

We found that OSMR expression correlated significantly with 9 of 10 predictor of pCR in this data set (Fisher’s exact test, Po0.0001) and EMT genes including, respectively, strong (Po0.0001) positive and high OSMR expression (upper quartile) was associated with inverse relationships with CDH2 and CDH1 (Table 1). Significant ER À tumours (Po0.05). OSMR was highly expressed in tumours relationships were also observed with each CSC-associated factor from ER À patients who did not achieve a pCR (P ¼ 0.03, Mann– we assessed, including the ratio of CD44 to CD24, members of the Whitney U-test). To determine the optimal OSMR cutpoint for NOTCH family, the transmembrane drug transporter ABCG2, and the predicting pCR, we performed receiver–operator characteristic pluripotency genes SOX2, POU5F1 and NANOG (Table 1). Although analysis, which led us to define OSMR-high status as the upper ALDH1A1 has been associated with CSCs, its expression correlated 20% of cases. OSMR-high patients had a pCR rate of 3/23 (13%), with neither OSMR nor other CSC markers. while 47% (43/91) of OSMR-low patients achieved pCR As CSCs may contribute to therapeutic resistance in breast (OR ¼ 0.167, 95% CI: 0.046–0.603; P ¼ 0.004). We performed a cancer, we examined the relationship between OSMR expression similar analysis of genes that encode receptors for other and rates of pathological complete response (pCR) in the inflammatory cytokines and IL-6 family members (using the upper ER À subset of the MAQC cohort. We focused on ER À cases 20th percentile to define ‘high’ status), including LIFR, IL6R (IL-6 (n ¼ 114) because the vast majority of pCRs following neoadjuvant receptor a-chain), IL1R1 (interleukin-1 (IL-1) receptor-1) and chemotherapy are achieved by ER À patients,31–34 and OSMR is TNFRSF1A (tumour necrosis factor (TNF) receptor). None were associated with ER-negativity. Indeed, ER À status was a strong significantly associated with pCR frequency, nor were any

& 2014 Macmillan Publishers Limited Oncogene (2014) 1485 – 1494 OSM promotes EMT and stem cell features in breast cancer NR West et al 1490

Figure 5. OSM induces stemness via PI3K signalling. Mammosphere formation in (a) MCF7 cells and (b) T47D cells treated with combinations of OSM and signal transduction inhibitors. Relative sphere formation values were calculated as the ratios of sphere formation in OSM- stimulated cultures to those of control cultures (normalized to 1) in each treatment condition. Bars represent means (±s.d.) of four replicate cultures, compared using Student’s t-test. (c, left) CD24 and CD44 expression in T47D cells following 2 days of OSM treatment with or without PI3K inhibition. Gating is equivalent to that in Figure 4. (c, right) Fold changes (relative to control cultures) in the proportion of cells with a CD44highCD24 À /low phenotype following 2 days of OSM stimulation alone (black bars) or in the presence of PI3K inhibition (white bars). Bars represent means (±s.d.) of three independent experiments, compared using Student’s t-test. (d) 5-year recurrence-free survival among patients categorized according to OSMR expression (high, 43rd quartile) and gene expression signatures indicative of STAT3, Ras or PI3K signalling (high, 4median). Hazard ratios and log-rank P-values are indicated for all comparisons involving the OSMR-high/signature-high groups. **P ¼ 0.001–0.01; ***Po0.001.

standard clinical parameters (Figure 6b). Most of the genes listed that OSM signalling is associated with EMT and CSC gene in Table 1 that correlated with OSMR in the whole data set also did expression patterns in vivo, and that OSMR may be associated so in the ER À subset (Supplementary Table S1 and Figure 6c). with resistance to chemotherapy in ER À breast cancer. Notably, when we excluded Her2 þ and PR þ cases to focus on the clinically challenging triple-negative subset of the MAQC cohort (that is, ER À PR À Her2 À ), pCR occurred in 0 of 17 OSMR- DISCUSSION high cases and 25 of 54 (46%) OSMR-low cases (OR ¼ 0.033, 95% CI Expression of a mesenchymal or CSC phenotype by malignant 0.002–0.578; P ¼ 0.0003). Again, EMT and CSC genes were cells is increasingly thought to promote metastasis and ther- enriched in triple-negative, OSMR-high cases (data not shown). apeutic resistance. We have now shown using models of luminal Genes that correlated strongly with OSMR in triple-negative breast cancer that OSM can induce features consistent with an tumours included the CD44/CD24 ratio (rs ¼ 0.4431, P ¼ 0.0001), EMT/CSC phenotype. Furthermore, the OSM pathway is associated SOX2 (rs ¼ 0.4380, P ¼ 0.0001) and CDH1 (rs ¼À0.4317, with these features in human breast tumours, which may explain P ¼ 0.0002). Collectively, our gene expression analyses suggest the relationship between OSMR expression and poor prognosis.

Oncogene (2014) 1485 – 1494 & 2014 Macmillan Publishers Limited OSM promotes EMT and stem cell features in breast cancer NR West et al 1491 Table 1. Correlation of OSMR with EMT and CSC-related genes in the MAQC cohort (n ¼ 278)

Category Gene Spearman rrs, 95% CI P-value

EMT factor ZEB2 0.3383 0.2264–0.4413 o0.0001 EMT factor CDH2 0.3292 0.2167–0.4330 o0.0001 EMT factor FOXC1 0.322 0.2091–0.4265 o0.0001 EMT factor ZEB1 0.2697 0.1536–0.3784 o0.0001 EMT factor SNAI2 0.2487 0.1316–0.3590 o0.0001 EMT factor CDH1 À 0.3213 À 0.4258– o0.0001 À 0.2083 EMT factor FOXC2 0.1821 0.06240–0.2967 0.0023 EMT factor VIM 0.1781 0.05821–0.2929 0.0029 EMT factor SNAI1 0.1283 0.0073–0.2456 0.0325 EMT factor TWIST1 À 0.0165 À 0.1373–0.1048 0.7842 CSC factor ABCG2 0.4564 0.3549–0.5473 o0.0001 CSC factor EGFR 0.4156 0.3101–0.5110 o0.0001 CSC factor NOTCH4 0.395 0.2877–0.4926 o0.0001 CSC factor SOX2 0.3875 0.2795–0.4858 o0.0001 CSC factor CD133 0.289 0.1740–0.3962 o0.0001 CSC factor CD44/ 0.2625 0.1460–0.3718 o0.0001 CD24a CSC factor NOTCH1 0.2279 0.1098–0.3396 0.0001 CSC factor NANOG 0.2168 0.0983–0.3293 0.0003 CSC factor POU5F1 0.1713 0.0513–0.2865 0.0042 aCD44/CD24 denotes the ratio of CD44 to CD24 expression.

the phenotypic similarity of metastases to primary tumours, possibly as a result of the so-called mesenchymal-to-epithelial transition.1,39 The variability of EMT marker expression in vivo was recently demonstrated in a quantitative immunofluorescence analysis of 4200 invasive breast carcinomas.40 In contrast with conventional models of EMT, no clear correlation between loss of E-cadherin and gain of N-cadherin was observed, nor was there evidence for significant relationships between N-cadherin, vimentin and fibronectin. However, snail and slug were significantly correlated, without any apparent relationship to other EMT markers. This bears some resemblance to our experimental data, which revealed consistent upregulation of snail and slug following OSM Figure 6. Association of OSM signalling with EMT/CSC features and treatment, but few consistent trends for other EMT markers. chemotherapy response in vivo.(a) RT–PCR assessment of 72 Notably, the maintenance of E-cadherin expression following OSM invasive breast carcinomas, showing OSM expression relative to stimulation in T47D cells (which instead employed E-cadherin SNAI1, POU5F1, and SOX2 levels. Box-and-whisker plots are compared internalization) demonstrates that E-cadherin repression is not a using the Mann–Whitney U-test. (b) Association of OSMR, other universal effect of the snail and slug transcription factors. cytokine receptors, and clinical parameters with response to Two of the most common approaches for identifying putative neoadjuvant chemotherapy in the ER À subset of the MAQC cohort. mammary CSCs, both of which are validated methods for Odds ratios reflect the probability of pathological complete identifying cells with enhanced tumourigenic capacity, are the response versus residual disease, calculated using Fisher’s exact clonal propagation of mammospheres21–23 and detection of a test. (c) Average expression values of EMT and stem cell-associated high À /low 2,26 genes (see Table 1) according to OSMR status in the ER À subset of CD44 CD24 cell surface phenotype. In our hands, OSM the MAQC cohort. Distributions were compared using the Mann– promoted both features at similar magnitudes in MCF7 and T47D Whitney U-test. *P ¼ 0.01–0.05; **P ¼ 0.001–0.01; ***Po0.001. cells, providing clear evidence that OSM can augment CSC features in breast cancer. It should be noted that prior studies have demonstrated a cytostatic effect of OSM on breast cancer In addition to the commonly observed pro-migratory effect of cells, including those used in this study.6,41–43 We have similarly OSM,5–9 we found that OSM could induce several, but not all of observed reduced proliferation of MCF7 and T47D cells when the stereotypical EMT-associated features.35 These include loss of treated with OSM during conventional adherent culture; however, epithelial morphology, suppression of epithelial markers such as our mammosphere experiments demonstrate that OSM may exert hormone receptors and E-cadherin,9 and gain of the EMT-driving significantly different effects in a three-dimensional culture transcription factors snail and slug. However, the variation system. It is possible that OSM does not promote rapid between MCF7 and T47D cells and the unexpected suppressive proliferation of mammosphere cells per se, but rather enhances effect of OSM on vimentin expression suggest that OSM may their survival, resulting in the outgrowth of a greater number of promote carcinoma dedifferentiation, rather than complete clones that recover more rapidly under selective conditions and transdifferentiation to a mesenchymal state.3,36 Several lines of begin proliferating earlier in the assay. evidence suggest that EMT is incomplete during cancer In both mammospheres and conventional cultures, OSM progression, including the expression of epithelial markers by reproducibly induced expression of SOX2, an essential factor for bone marrow-resident and circulating tumour cells,37,38 as well as the generation of pluripotent stem cells from fibroblasts.25 SOX2 is

& 2014 Macmillan Publishers Limited Oncogene (2014) 1485 – 1494 OSM promotes EMT and stem cell features in breast cancer NR West et al 1492 required for efficient mammosphere and xenograft formation by To firmly establish an oncogenic role for OSM in breast cancer, MCF7 cells and may be preferentially expressed in basal-like two additional sets of studies are required. The first is to examine tumours, which exhibit poor differentiation and poor the effects of OSM on tumourigenesis in preclinical animal models, prognosis.24,44 OSM also induced S100A7 expression in and the second is to confirm the relationship between OSM mammospheres, which is known to promote anoikis resistance signalling and prognosis in clinical cohorts. The former should and tumourigenesis in ER À breast cancer and may be required focus on establishing whether OSM can induce tumourigenicity for OSM-induced migration.5,45,46 Finally, OSMR mRNA expression through the generation of self-renewing stem-like cells. It would was heightened in OSM-induced mammospheres and OSMR also be interesting to determine if OSM confers resistance to protein was co-expressed with CD44, suggesting that high OSMR chemotherapy and radiotherapy in vivo. Human clinical analyses levels may be associated with cytokine-induced stemness. will require the validation of antibodies for reliable immunohis- Notably, OSMR and S100A7 were also highly expressed in tochemical detection of OSMR in archived tissue specimens. If the mammospheres in the absence of OSM. As the expression of apparent importance of OSM signalling is verified in these each is driven by STAT3,5,47 this could reflect heightened STAT3 settings, OSM and OSMR could serve as novel therapeutic targets activity in stem-like cells, possibly via autocrine IL-6 stimulation. for breast cancer. Notably, we found that PI3K signalling was essential for OSM to induce both mammosphere formation and a CD44highCD24 À /low phenotype. Furthermore, we found that high OSMR expression is MATERIALS AND METHODS associated with poor prognosis only if a transcriptional signature Cell culture and reagents of PI3K activation is co-expressed. In contrast, the STAT3 and Human breast carcinoma cell lines MCF7 and T47D (obtained originally MAPK pathways appear to be dispensable for OSM-induced from ATCC) were cultured in DMEM with 5% FBS under standard stemness in vitro, and did not influence the prognostic impact of conditions. Human OSM (Peprotech, Rocky Hill, NJ, USA) was used at OSMR in survival analyses. Intriguingly, STAT3 is crucial for 100 ng/ml. Cells were stimulated for 2 days before analysis. Experiments 9 OSM-induced migration,8 while MAPK signalling mediates OSM- involving siRNA were performed as described. Inhibitors to MEK1/2 induced loss of hormone receptor expression.9 A very recent (U0126, 10 mM; Cell Signaling, Danvers, MA, USA), JAKs (JAK inhibitor paper confirmed that STAT3 is important for OSM-induced EMT- I/pyridine-6, 10 mM), PI3K (LY294002, 10 mM), STAT3 (S3I201, 100 mM), and Wnt/b-catenin (XAV939, 3 mM; Calbiochem, San Diego, CA, USA) were like changes in MCF7 cells and identified STAT3-regulated added to cultures 30 min before cytokine stimulation. components including let-7 and miR-200 family members that may control EMT.48 OSM thus appears to rely on multiple signal transduction pathways to exert its major effects. It should be Clinical cohorts noted that STAT3 and MAPK inhibition suppressed baseline Three cohorts of human breast cancer were utilized. The first included 72 invasive breast carcinomas that were obtained from the Manitoba Breast mammosphere formation in our experiments, in agreement with 9 prior studies that indicate a role for these pathways in CSC Tumour Bank (MBTB) and described previously. The second cohort (‘UNC’) 22,49,50 is publicly available and was acquired from the University of North Carolina maintenance. Nevertheless, OSM was capable of inducing Microarray Database.11 Gene expression data were downloaded using the mammosphere formation with or without functional STAT3 or default quality filtering criteria of the UNC database (for example, genes MAPK signalling. Previous studies have supported the notion that had a log2 Lowess normalized ratio of 42 s.d. from the mean in at least PI3K contributes crucially to mammary CSC maintenance. For one array and an absolute value of 42 in at least two arrays). For example, knockdown of PTEN (phosphatase and tensin assessment of PI3K activity in the UNC cohort, we extracted all available homologue), a critical negative regulator of the PI3K-Akt data for the top 100 genes that correlated positively with PI3K activity, as 27 pathway, causes expansion of tumourigenic CSCs among MCF7 reported in the study by Creighton et al. These data were averaged 51 within individual cases to yield mean PI3K activity scores. We similarly cells and other breast cancer cell lines. Similarly, PTEN was 29 28 shown to be a key regulator of the survival, proliferation and examined gene expression indices of STAT3 and Ras-MAPK activity. 50 Genes used for each signature are listed in Supplementary Table S2. The tumourigenicity of MCF7 side population cells. third cohort (‘MAQC’) is publicly accessible and contains gene expression A potentially important aspect of OSM signalling is the rapidity data derived from 278 fine needle aspirate biopsies of invasive breast of its action. Many studies of EMT/CSC modulators depend on carcinoma (Gene Expression Omnibus accession number GSE2019430). constitutive expression strategies, raising questions of physiologi- Patients were treated after biopsy with neoadjuvant anthracycline-taxane- cal relevance. This has occurred for genes such as twist and snail, based chemotherapy. In both microarray data sets, median-normalized the latter being a particularly pertinent example because it data were log2-transformed before analysis. normally inactivates its own promoter.17 In some studies of EMT/CSC-related cytokines such as transforming growth factor-b Western blots (TGF-b) and TNF, cells were exposed to these factors for lengthy Immunoblotting was performed as described.9 Primary antibodies periods of time (ranging from 12 to 60 days) before EMT/CSC detected ERa (1:1000; Santa Cruz Biotechnology, Santa Cruz, CA, USA), 52–54 phenotypes were assessed. In contrast, OSM-stimulated GAPDH (1:3000; Stem Cell Technologies, Vancouver, BC, Canada), snail breast cancer cells exhibit dedifferentiation and a readily (1:500; Cell Signaling), and E-cadherin (1:1000; Abcam, Cambridge, MA, detectable EMT/CSC-like phenotype within two days. The ability USA). Secondary antibodies were HRP-conjugated (1:3000; Santa Cruz to effect change on a short time scale may be important in vivo, Biotechnology). given the dynamic nature of the tumour microenvironment. OSM now joins a growing list of immune-derived cytokines that Real-time quantitative polymerase chain reaction (RT–PCR) apparently modulate EMT and stemness in breast cancer cells, 9 52–54 54,55 56–58 59 RT–PCR was performed as described. Primer sequences for each target are including TGF-b, TNF, IL-6 and interleukin-8. Such listed in Supplementary Table S3. studies illustrate the ability of immunological processes to profoundly influence the differentiation state of neoplastic Immunofluorescence microscopy mammary cells. To date, these cytokines have generally been examined on an individual basis. Human breast tumours, Cells were grown on sterile glass coverslips and fixed with 3.7% 13 formaldehyde, followed by permeabilization with PBS containing 0.2% however, are known to contain a diverse repertoire of cytokines, Triton X-100 (PBST), blocking with 1% bovine serum albumin in PBST, and and studies of single cytokines in isolation are unlikely to overnight staining with antibodies against E-cadherin (1:200, Abcam) and/ reproduce in vivo conditions. Going forward, it will be important or b-catenin (1:200, Santa Cruz Biotechnology). Secondary antibodies were to explore the combinatorial effects of multiple cytokines during Alexa Fluor 488 or 555-conjugated (1:1000, Invitrogen). Cells were tumourigenesis. mounted onto glass slides using Prolong Gold antifade reagent containing

Oncogene (2014) 1485 – 1494 & 2014 Macmillan Publishers Limited OSM promotes EMT and stem cell features in breast cancer NR West et al 1493 DAPI (Invitrogen). Images were captured using a BX53 light/fluorescence 6 Holzer RG, Ryan RE, Tommack M, Schlekeway E, Jorcyk CL. Oncostatin M microscope (Olympus Canada, Richmond Hill, ON, Canada) with a Nuance stimulates the detachment of a reservoir of invasive mammary carcinoma cells: multispectral camera (CRi, Hopkinton, MA, USA). role of cyclooxygenase-2. Clin Exp Metastasis 2004; 21: 167–176. 7 Jorcyk CL, Holzer RG, Ryan RE. Oncostatin M induces cell detachment and enhances the metastatic capacity of T-47D human breast carcinoma cells. Mammosphere culture Cytokine 2006; 33: 323–336. Cells were cultured with or without OSM and/or signal transduction 8 Zhang F, Li C, Halfter H, Liu J. Delineating an oncostatin M-activated STAT3 sig- inhibitors for 2 days prior to trypsinization and filtration through 40 mm naling pathway that coordinates the expression of genes involved in cell cycle nylon strainers to remove clumps (BD Biosciences, Mississauga, ON, USA). regulation and extracellular matrix deposition of MCF-7 cells. Oncogene 2003; 22: 2500 viable cells were plated per well into 96-well ultra-low adhesion 894–905. plates (Corning, Lowell, MA, USA) containing 150 ml of mammosphere 9 West NR, Murphy LC, Watson PH. Oncostatin-M suppresses estrogen receptor- culture medium, composed as follows: serum-free DMEM with 1% alpha expression and is associated with poor outcome in human breast cancer. methylcellulose (R&D Systems, Minneapolis, MN, USA), 20 ng/ml EGF Endocr Relat Cancer 2012; 19: 181–195. (Sigma-Aldrich, Oakville, ON, Canada), 20 ng/ml bFGF (Peprotech), and 1 Â 10 Rakha EA, El-Sayed ME, Reis-Filho JS, Ellis IO. Expression profiling technology: its insulin-transferrin-selenite (Sigma). Addition of methylcellulose is crucial to contribution to our understanding of breast cancer. Histopathology 2008; 52: prevent cells from diffusing through the medium and forming clumps that 67–81. can be mistaken for clonal colonies. Colonies with a diameter greater than 11 Prat A, Parker JS, Karginova O, Fan C, Livasy C, Herschkowitz JI et al. Phenotypic 70 mm were recorded after 10 days. Generally, cells pretreated with OSM and molecular characterization of the claudin-low intrinsic subtype of breast received continued stimulation at a reduced dose (50 ng/ml) throughout cancer. Breast Cancer Res 2010; 12: R68. the assay. For assays involving inhibitors, cells were treated with drugs 12 Fulford LG, Easton DF, Reis-Filho JS, Sofronis A, Gillett CE, Lakhani SR et al. Specific and/or OSM for only the 2 days prior to plating. morphological features predictive for the basal phenotype in grade 3 invasive ductal carcinoma of breast. Histopathology 2006; 49: 22–34. 13 Chavey C, Bibeau F, Gourgou-Bourgade S, Burlinchon S, Boissiere F, Laune D et al. Flow cytometry Oestrogen receptor negative breast cancers exhibit high cytokine content. Breast Cells were harvested using 1 mM EDTA in PBS, fixed briefly with 3.7% Cancer Res 2007; 9: R15. formaldehyde to prevent antigen internalization, and stained with 14 Teschendorff AE, Journee M, Absil PA, Sepulchre R, Caldas C. Elucidating the CD44-FITC and CD24-PE monoclonal antibodies at the manufacturer’s altered transcriptional programs in breast cancer using independent component recommended concentration (BD Biosciences) in FACS buffer (3% FBS in analysis. PLoS Comput Biol 2007; 3:e161. PBS). Trypsin was not used for harvesting to avoid cleavage of surface 15 Heinrich PC, Behrmann I, Haan S, Hermanns HM, Muller-Newen G, Schaper F. 60 proteins and potential loss of signal. OSMR is induced following OSM Principles of interleukin (IL)-6-type cytokine signalling and its regulation. Biochem 61 stimulation, but much of this is present as an intracellular pool. Cells were J 2003; 374(Pt 1): 1–20. thus fixed and stained intracellularly with an OSMR-APC antibody at the 16 Tanaka M, Miyajima A. Oncostatin M, a multifunctional cytokine. Rev Physiol recommended concentration (eBioscience, San Diego, CA, USA). Cells were Biochem Pharmacol 2003; 149:39–52. analysed using a BD FACSCalibur flow cytometer and FlowJo 7.6.5 (Tree 17 de Herreros AG, Peiro S, Nassour M, Savagner P. Snail family regulation and Star, Ashland, OR, USA). epithelial mesenchymal transitions in breast cancer progression. J Mammary Gland Biol Neoplasia 2010; 15: 135–147. 18 Dhasarathy A, Kajita M, Wade PA. The transcription factor snail mediates epithelial Statistical analysis to mesenchymal transitions by repression of estrogen receptor-alpha. Mol Experiments were repeated at least thrice. Individual statistical tests are Endocrinol 2007; 21: 2907–2918. specified in figure and table legends, with significance established at 19 Ye Y, Xiao Y, Wang W, Yearsley K, Gao JX, Shetuni B et al. ERalpha signaling o0.05; all tests were performed using Prism 5.0 (GraphPad, La Jolla, CA, through slug regulates E-cadherin and EMT. Oncogene 2010; 29: 1451–1462. USA). In all figures, asterisks denote significance levels as follows: 20 Plath K, Lowry WE. Progress in understanding reprogramming to the induced *P ¼ 0.01–0.05, **P ¼ 0.01–0.001, ***Po0.001. pluripotent state. Nat Rev Genet 2011; 12: 253–265. 21 Dontu G, Abdallah WM, Foley JM, Jackson KW, Clarke MF, Kawamura MJ et al. In vitro propagation and transcriptional profiling of human mammary stem/pro- CONFLICT OF INTEREST genitor cells. Genes Dev 2003; 17: 1253–1270. 22 Yu F, Yao H, Zhu P, Zhang X, Pan Q, Gong C et al. let-7 regulates self renewal and The authors declare no conflict of interest. tumorigenicity of breast cancer cells. Cell 2007; 131: 1109–1123. 23 Ponti D, Costa A, Zaffaroni N, Pratesi G, Petrangolini G, Coradini D et al. Isolation and in vitro propagation of tumorigenic breast cancer cells with stem/progenitor ACKNOWLEDGEMENTS cell properties. Cancer Res 2005; 65: 5506–5511. We are grateful to members of the Deeley Research Centre for helpful discussion. 24 Leis O, Eguiara A, Lopez-Arribillaga E, Alberdi MJ, Hernandez-Garcia S, Elorriaga K We thank the Manitoba Breast Tumour Bank (a member of the Canadian Tumour et al. Sox2 expression in breast tumours and activation in breast cancer stem cells. Repository Network) for providing breast cancer tissue specimens and the authors of Oncogene 2011; 31: 1354–1365. the UNC and MAQC cohorts for making their data freely available. Financial support 25 Takahashi K, Yamanaka S. Induction of pluripotent stem cells from mouse was from the Canadian Breast Cancer Foundation (CBCF), BC/Yukon chapter and the embryonic and adult fibroblast cultures by defined factors. Cell 2006; 126: 663–676. BC Cancer Foundation. NRW was supported by a US DOD Breast Cancer Research 26 Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. Prospective Program Predoctoral award (W81XWH-08-1-0781). JIM is supported by a CBCF BC/ identification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 2003; 100: Yukon Region postdoctoral fellowship. 3983–3988. 27 Creighton CJ, Fu X, Hennessy BT, Casa AJ, Zhang Y, Gonzalez-Angulo AM et al. Proteomic and transcriptomic profiling reveals a link between the PI3K pathway and lower estrogen-receptor (ER) levels and activity in ER þ breast cancer. Breast REFERENCES Cancer Res 2010; 12: R40. 1 May CD, Sphyris N, Evans KW, Werden SJ, Guo W, Mani SA. Epithelial-mesench- 28 Loboda A, Nebozhyn M, Klinghoffer R, Frazier J, Chastain M, Arthur W et al. A gene ymal transition and cancer stem cells: a dangerously dynamic duo in breast expression signature of RAS pathway dependence predicts response to PI3K and cancer progression. Breast Cancer Res 2011; 13:202. RAS pathway inhibitors and expands the population of RAS pathway activated 2 Lawson JC, Blatch GL, Edkins AL. Cancer stem cells in breast cancer and metas- tumors. BMC Med Genomics 2010; 3:26. tasis. Breast Cancer Res Treat 2009; 118: 241–254. 29 Alvarez JV, Febbo PG, Ramaswamy S, Loda M, Richardson A, Frank DA. Identifi- 3 Klymkowsky MW, Savagner P. Epithelial-mesenchymal transition: a cancer cation of a genetic signature of activated signal transducer and activator of researcher’s conceptual friend and foe. Am J Pathol 2009; 174: 1588–1593. transcription 3 in human tumors. Cancer Res 2005; 65: 5054–5062. 4 Korkaya H, Liu S, Wicha MS. Breast cancer stem cells, cytokine networks, and the 30 Popovici V, Chen W, Gallas BG, Hatzis C, Shi W, Samuelson FW et al. Effect of tumor microenvironment. J Clin Invest 2011; 121: 3804–3809. training-sample size and classification difficulty on the accuracy of genomic 5 West NR, Watson PH. S100A7 (psoriasin) is induced by the proinflammatory predictors. Breast Cancer Res 2010; 12:R5. cytokines oncostatin-M and interleukin-6 in human breast cancer. Oncogene 2010; 31 Colleoni M, Bagnardi V, Rotmensz N, Gelber RD, Viale G, Pruneri G et al. Increasing 29: 2083–2092. steroid hormone receptors expression defines breast cancer subtypes non

& 2014 Macmillan Publishers Limited Oncogene (2014) 1485 – 1494 OSM promotes EMT and stem cell features in breast cancer NR West et al 1494 responsive to preoperative chemotherapy. Breast Cancer Res Treat 2009; 116: STAT5 in mammary epithelial cells define the roles of these STATs in mammary 359–369. development. Mol Endocrinol 2006; 20: 675–685. 32 Colleoni M, Viale G, Zahrieh D, Pruneri G, Gentilini O, Veronesi P et al. 48 Guo L, Chen C, Shi M, Wang F, Chen X, Diao D et al. Stat3-coordinated Chemotherapy is more effective in patients with breast cancer not expressing Lin-28–let-7–HMGA2 and miR-200–ZEB1 circuits initiate and maintain steroid hormone receptors: a study of preoperative treatment. Clin Cancer Res oncostatin M-driven epithelial–mesenchymal transition. Oncogene 2013; 32: 2004; 10: 6622–6628. 5272–5282. 33 Guarneri V, Broglio K, Kau SW, Cristofanilli M, Buzdar AU, Valero V et al. Prognostic 49 Dangi-Garimella S, Yun J, Eves EM, Newman M, Erkeland SJ, Hammond SM et al. value of pathologic complete response after primary chemotherapy in relation to Raf kinase inhibitory protein suppresses a metastasis signalling cascade involving hormone receptor status and other factors. J Clin Oncol 2006; 24: 1037–1044. LIN28 and let-7. EMBO J 2009; 28: 347–358. 34 Ring AE, Smith IE, Ashley S, Fulford LG, Lakhani SR. Oestrogen receptor status, 50 Zhou J, Wulfkuhle J, Zhang H, Gu P, Yang Y, Deng J et al. Activation of the PTEN/ pathological complete response and prognosis in patients receiving neoadjuvant mTOR/STAT3 pathway in breast cancer stem-like cells is required for viability and chemotherapy for early breast cancer. Br J Cancer 2004; 91: 2012–2017. maintenance. Proc Natl Acad Sci USA 2007; 104: 16158–16163. 35 Zeisberg M, Neilson EG. Biomarkers for epithelial-mesenchymal transitions. J Clin 51 Korkaya H, Paulson A, Charafe-Jauffret E, Ginestier C, Brown M, Dutcher J et al. Invest 2009; 119: 1429–1437. Regulation of mammary stem/progenitor cells by PTEN/Akt/beta-catenin signal- 36 Christiansen JJ, Rajasekaran AK. Reassessing epithelial to mesenchymal transition ing. PLoS Biol 2009; 7: e1000121. as a prerequisite for carcinoma invasion and metastasis. Cancer Res 2006; 66: 52 Mani SA, Guo W, Liao MJ, Eaton EN, Ayyanan A, Zhou AY et al. The epithelial- 8319–8326. mesenchymal transition generates cells with properties of stem cells. Cell 2008; 37 Mego M, Mani SA, Cristofanilli M. Molecular mechanisms of metastasis in breast 133: 704–715. cancer—clinical applications. Nat Rev Clin Oncol 2010; 7: 693–701. 53 Morel AP, Lievre M, Thomas C, Hinkal G, Ansieau S, Puisieux A. Generation of 38 Balic M, Lin H, Young L, Hawes D, Giuliano A, McNamara G et al. Most early breast cancer stem cells through epithelial-mesenchymal transition. PLoS One disseminated cancer cells detected in bone marrow of breast cancer patients 2008; 3: e2888. have a putative breast cancer stem cell phenotype. Clin Cancer Res 2006; 12: 54 Asiedu MK, Ingle JN, Behrens MD, Radisky DC, Knutson KL. TGFbeta/TNF(alpha)- 5615–5621. mediated epithelial-mesenchymal transition generates breast cancer stem cells 39 Polyak K, Weinberg RA. Transitions between epithelial and mesenchymal states: with a claudin-low phenotype. Cancer Res 2011; 71: 4707–4719. acquisition of malignant and stem cell traits. Nat Rev Cancer 2009; 9: 265–273. 55 Bhat-Nakshatri P, Campbell RA, Patel NM, Newton TR, King AJ, Marshall MS et al. 40 Dubois-Marshall S, Thomas JS, Faratian D, Harrison DJ, Katz E. Two possible Tumour necrosis factor and PI3-kinase control oestrogen receptor alpha protein mechanisms of epithelial to mesenchymal transition in invasive ductal breast level and its transrepression function. Br J Cancer 2004; 90: 853–859. cancer. Clin Exp Metastasis 2011; 28: 811–818. 56 Iliopoulos D, Hirsch HA, Wang G, Struhl K. Inducible formation of breast cancer 41 Liu J, Spence MJ, Wallace PM, Forcier K, Hellstrom I, Vestal RE. Oncostatin stem cells and their dynamic equilibrium with non-stem cancer cells via IL6 M-specific receptor mediates inhibition of breast cancer cell growth and down- secretion. Proc Natl Acad Sci USA 2011; 108: 1397–1402. regulation of the c-myc proto-oncogene. Cell Growth Differ 1997; 8: 667–676. 57 Sansone P, Storci G, Tavolari S, Guarnieri T, Giovannini C, Taffurelli M et al. IL-6 42 Grant SL, Hammacher A, Douglas AM, Goss GA, Mansfield RK, Heath JK et al. An triggers malignant features in mammospheres from human ductal breast carci- unexpected biochemical and functional interaction between gp130 and the EGF noma and normal mammary gland. J Clin Invest 2007; 117: 3988–4002. receptor family in breast cancer cells. Oncogene 2002; 21: 460–474. 58 Sullivan NJ, Sasser AK, Axel AE, Vesuna F, Raman V, Ramirez N et al. Interleukin-6 43 Douglas AM, Goss GA, Sutherland RL, Hilton DJ, Berndt MC, Nicola NA et al. induces an epithelial-mesenchymal transition phenotype in human breast cancer Expression and function of members of the cytokine receptor superfamily on cells. Oncogene 2009; 28: 2940–2947. breast cancer cells. Oncogene 1997; 14: 661–669. 59 Ginestier C, Liu S, Diebel ME, Korkaya H, Luo M, Brown M et al. CXCR1 blockade 44 Rodriguez-Pinilla SM, Sarrio D, Moreno-Bueno G, Rodriguez-Gil Y, Martinez MA, selectively targets human breast cancer stem cells in vitro and in xenografts. J Clin Hernandez L et al. Sox2: a possible driver of the basal-like phenotype in sporadic Invest 2010; 120: 485–497. breast cancer. Mod Pathol 2007; 20: 474–481. 60 Hardt O, Wild S, Oerlecke I, Hofmann K, Luo S, Wiencek Y et al. Highly sensitive 45 Emberley ED, Murphy LC, Watson PH. S100A7 and the progression of breast profiling of CD44( þ )/CD24(-) breast cancer stem cells by combining global mRNA cancer. Breast Cancer Res 2004; 6: 153–159. amplification and next generation sequencing: Evidence for a hyperactive PI3K 46 Emberley ED, Niu Y, Curtis L, Troup S, Mandal SK, Myers JN et al. The S100A7-c-Jun pathway. Cancer Lett 2012; 325: 165–174. activation domain binding protein 1 pathway enhances prosurvival pathways in 61 Blanchard F, Wang Y, Kinzie E, Duplomb L, Godard A, Baumann H. Oncostatin M breast cancer. Cancer Res 2005; 65: 5696–5702. regulates the synthesis and turnover of gp130, leukemia inhibitory factor receptor 47 Clarkson RW, Boland MP, Kritikou EA, Lee JM, Freeman TC, Tiffen PG et al. The alpha, and oncostatin M receptor beta by distinct mechanisms. J Biol Chem 2001; genes induced by signal transducer and activators of transcription (STAT)3 and 276: 47038–47045.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

Oncogene (2014) 1485 – 1494 & 2014 Macmillan Publishers Limited