Hypoxia‐induced secretion stimulates breast cancer stem cell regulatory signalling pathways

It is well known that tumour cells are dependent on communication with the tumour microenvironment. Previously, it has been shown that hypoxia (HX) induces pronounced, diverse and direct effects on cancer stem cell (CSC) qualities in different breast cancer subtypes. Here, we describe the mechanism by which HX‐induced secretion influences the spreading of CSCs. Conditioned media (CM) from estrogen receptor (ER)‐α‐positive hypoxic breast cancer cell cultures increased the fraction of CSCs compared to normal growth conditions, as determined using sets of CSC assays and model systems. In contrast, media from ERα‐negative hypoxic cell cultures instead decreased this key subpopulation of cancer cells. Further, there was a striking overrepresentation of JAK‐STAT‐associated cytokines in both the ERα‐positive and ERα‐negative linked hypoxic responses as determined by a protein screen of the CM. JAK‐STAT inhibitors and knockdown experiments further supported the hypothesis that this pathway is critical for the CSC‐activating and CSC‐inactivating effects induced by hypoxic secretion. We also observed that the interleukin‐6‐JAK2‐STAT3 axis was specifically central for the ERα‐negative hypoxic behaviour. Our results underline the importance of considering breast cancer subtypes in treatments targeting JAK‐STAT or HX‐associated processes and indicate that HX is not only a confined tumour biological event, but also influences key tumour properties in widespread normoxic microenvironments.


Introduction
Breast cancer represents a diverse group of malignancies derived from ductal epithelium. In recent years, the microenvironments of these cancers have been shown to play a critical role in tumour progression as well as therapeutic response (Hu et al., 2009;Tyan et al., 2011). Importantly, different subtypes of breast cancer demonstrate contrasting direct effects towards a hypoxic environment (Fuchs et al., 2004;Korkaya et al., 2011). Our group has previously shown that reduced oxygen levels have an opposing, direct estrogen receptor (ER)-a-dependent effect on breast cancer stem cell (BCSC) populations (Harrison et al., 2013). Given that many reports highlight secretion as a mean for tumour cells to communicate within the tumour microenvironment, we hypothesized that hypoxic regions secrete factors to nearby cells and thereby influence cellular plasticity and expansion or reduction of stem cell properties (Harrison et al., 2018;Hu et al., 2009;Tyan et al., 2011).
Furthermore, breast cancer consists of heterogeneous cell compositions explained by clonal variation within the tumour as well as an intrinsic balance between different cellular differentiation stages, including cancer stem cell (CSC) pools and more differentiated cells (Campbell and Polyak, 2007;Marusyk and Polyak, 2010). These subpopulations of tumour cells with specific qualities represent a major challenge for treatment resistance in general as well as for monitoring therapy efficacy. Regarding clonal variations, characterization through, for example, sequencing of distal parts of tumour has highlighted the complexity and has identified driver mutations leading to clonal expansion (Vogelstein et al., 2013). Nevertheless, within this study we focused on cellular differentiation aspects of heterogeneous cellular populations.
Previously, much research focus has been on trying to understand inflammatory responses within the tumours through tumour interactions with the extracellular matrix, mesenchymal stem cells, immune cells, cancerassociated fibroblasts and tumour-activated macrophages (Criscitiello et al., 2014;Gilkes et al., 2014;Ginestier et al., 2010Ginestier et al., , 2011Li et al., 2012;Lu et al., 2014). A crosstalk between CSCs and their microenvironment via certain cytokines has been demonstrated (Bak et al., 2016;Di Stefano et al., 2010;Sansone et al., 2007;Singh et al., 2013). In 2011, Liu and coworkers reported that BCSCs are regulated by cytokine-related signalling pathways and highlighted the influence IL6 and IL8 production to maintain CSC properties Sansone et al., 2007). Additionally, IL6 has been shown to function as a molecular switch between non-CSCs and CSCs in breast cancer, suggesting a dynamic system that is dependent on communication between the different cell types through secretion to maintain the CSC niche (Iliopoulos et al., 2011).

Hypoxic cell culture
Cells were incubated for 48 h in the SCI-tiveN hypoxic workstation (Ruskinn, Sanford, ME, USA) in 1% O 2 , 5% CO 2 and 94% N 2 in a humidified environment at 37°C. Cells were seeded, cultured and harvested within the workstation to maintain hypoxia (HX) at all times. Confirmation of hypoxic conditions was carried out using western blot analysis of HX-inducible factor (HIF)-1a protein expression.

Mammosphere culture
Mammosphere culture was carried out as previously described (Shaw et al., 2012). In brief, following treatment, cells were resuspended as single cells using a 25gauge needle. A total of 5000 cells per 2 mL media were seeded into Polyhema-coated 6-well culture plates in MEM (Gibco TM , Thermo Fischer Scientific; phenol red free) containing 1% penicillin/streptomycin, 200 lg human recombinant EGF (VWr) and 1% B27 supplement (Life Technologies, Thermo Fischer Scientific). Cells were maintained for 5 days in a humidified incubator at 37°C at an atmospheric pressure of 5% (v/v) CO 2 /air and analysed for mammosphere formation by microscope. Mammospheres ≥ 50 lm were counted.

Patient ex vivo explants
Breast cancer tissue was obtained after written informed consent through Sahlgrenska University Hospital, Gothenburg, Sweden. 500 lm thick sections from breast cancer tumours were prepared at Department of Pathology at Sahlgrenska University Hospital using a vibrating blade microtome LEICA VT1000S and cultured in duplicate on a presoaked gelatin sponge (Johnson and Johnson, New Research, New Brunswick, NJ, USA) in six-well plates containing DMEM-F12 (Lonza; 10% FBS/1% PenStrep) (Centenera et al., 2013;Dean et al., 2012) without covering the tissue. Tissues were cultured in either normoxic (21% O 2 ) or hypoxic (1% O 2 ) conditions at 37°C for 48 h and then formalin-fixed and paraffin-embedded or preserved in RNAlater (Invitrogen, San Diego, CA, USA). Conditioned culture media were taken from explants and centrifuged at 300 rcf for 5 min to remove cellular debris. The conditioned media (CM) were then used to either treat cell lines or analysed using western blot.

Human antibody array
Culture media from 48 h normoxic (21% O 2 ) or hypoxic (1% O 2 ) conditions at 37°C were collected from ERapositive (MCF7) and ERa-negative (MDA-MB 231) cell cultures and analysed for protein secretion using human antibody array (RayBiotech, Inc., Norcross, GA, USA). The antibody array was performed according to the protocol provided by the manufacturer.
2.8. Transient transfections, inhibitor treatment, and cytokine treatment MCF7 and T47D cells were transfected with 25 nM small interference RNA (siRNA) against the ESR1 gene (Origene, Herford, Germany) using the transfection reagent Viromer blue (Lipocalyx, Halle, Germany). In brief, cells were seeded in a six-well plate 1 day prior to transfection and incubated at 37°C and 5% CO 2 overnight. The transfection was carried out in normal complete growth media. Cells were then incubated for 48 h in either normoxic (21% O 2 ) or hypoxic (1% O 2 ) conditions. CM media were collected and spun down at 300 rcf for 3 min to remove cell debris and dead cells. Overexpression of ERa with VP16ERa vector (Addgene plasmid 27 , Watertown, MA, USA) in MDA-MB 231 cells was carried out using Lipofectamine 2000 (Invitrogen) according to manufacturer 0 s description. VP16 empty vector was used as a control. Cells were cultured in normoxic or hypoxic conditions for 48 h following transfection. CM were collected and spun down at 300 g for 3 min to remove cell debris and dead cells. JAK inhibitor I (#420099; Millipore, Burlington, MA, USA), JAK2 inhibitor II (#420132; Millipore) and STAT5 inhibitor (#573108; Millipore) were reconstituted in DMSO upon delivery, aliquoted and stored at À20°C. STAT3 inhibitor I (#573096; Millipore) was diluted in water, and STAT3 inhibitor II (#573097; Millipore) was diluted in ethanol upon delivery. Working stock concentrations were achieved by dilution in culture media. Cell lines were treated with indicated concentrations for 48 h at 37°C, 5% CO 2 and 21% O 2 . Recombinant IL6 (R&D system, Minneapolis, MN, USA) and IL12RB2 (R&D systems) were reconstituted in PBS upon delivery, aliquoted and stored at À20°C. Working stock concentrations were achieved by dilution in culture media, and cell lines were treated with indicated concentrations for 5 min (IL6) or 48 h (IL12RB2) in either normoxic (21% O 2 ) or hypoxic (1% CO 2 ) conditions.

Single-cell collection
For single-cell collection, MCF7 cells were washed with 19 PBS, pH 7.4 (Sigma-Aldrich), and enzymatically dissociated with 0.25% trypsin-EDTA (PAA). Cells were resuspended in 19 PBS, pH 7.4, supplemented with 2% BSA (Sigma-Aldrich) and kept at 4°C until sorting. Cell aggregates were removed by filtering through a 35-lm cell strainer (BD Biosciences). Individual cells were sorted using the Becton Dickinson FACS ARIA II into 96-well PCR plates (Life Technologies), containing 5 lL, 1 lgÁlL À1 BSA and 2.5% glycerol (Thermo Scientific, Waltham, MA, USA) in DNase/RNase-free water (Life Technologies). In total, 264 cells were sorted from MCF7 cells treated with HX CM from MCF7 cells, 251 cells treated with normoxia (NX) CM from MCF7 cells, 181 cells treated with HX CM from MDA-MB 231 cells and 272 cells treated with NX CM from MDA-MB cells.

Single-cell reverse transcription
Single-cell cDNA synthesis was performed using TATAA GrandScript cDNA Synthesis Kit (TATAA Biocenter, Gothenburg, Sweden). Briefly, 2 lL 59 TATAA GrandScript RT Reaction Mix, 0.5 lL TATAA GrandScript RT Enzyme and 2.5 lL water were added to each sample for a final volume of 10 lL. The following thermal program was used: 22°C for 5 min, 42°C for 30 min and 85°C for 5 min, performed on a T100 Thermal Cycler (Bio-Rad, Hercules, CA, USA). Samples were diluted with 30 lL DNase/RNasefree water (Life Technologies) and stored at À20°C.

Preamplification
Targeted preamplification was performed in 50 lL reactions, using iQ Supermix (Bio-Rad), 40 nM of each primer and 20 lL diluted cDNA. Identical primer pairs were applied for targeted preamplification and downstream quantitative PCR (qPCR; Table S1). The following thermal profile was applied on a T100 Thermal Cycler (Bio-Rad) or a PTC-200 (MJ Researc, Waltham, MA, USAh): 95°C for 3 min, followed by 20 cycles of amplification (95°C for 20 s, 60°C for 3 min and 72°C for 20 s). During the final extended (10 min) elongation step, the samples were immediately frozen on dry ice, slowly thawed on ice, diluted 1 : 20 in 10 mM Tris and 1 mM EDTA, pH 8.0 (Life Technologies) and stored at À20°C until analysis.

Single-cell qPCR
Single-cell qPCR was performed in 6 lL reactions using the 29 SYBR GrandMaster Mix (TATAA Biocenter), 400 nM of each primer and 3 lL diluted preamplified cDNA as template. qPCRs were carried out in white 384well plates (Bio-Rad) applying the following thermal protocol: 95°C for 1 min: 40 cycles of: 95°C for 5 s, 60°C for 20 s, 70°C for 20 s (CFX384 Touch Real-Time PCR Detection System; Bio-Rad). Cycles of quantification (Cqs) were determined by the second-derivative maximum method (CFX MANAGER SOFTWARE, Bio-Rad, version 3.0). All qPCR experiments were conducted in accordance with the MIQE guidelines (Bustin et al., 2009). Experimental details unique for single-cell analysis are described elsewhere (Stahlberg et al., 2012(Stahlberg et al., , 2013.

Ethics approval and consent to participate
Breast cancer tissue was obtained with written informed consent through Sahlgrenska University Hospital, and all patients gave written consent for the use of material for research purposes. The study was approved by the Regional Ethic Authority in Gothenburg (reference #515-12), and methodologies conformed to the standards set by the Declaration of Helsinki. Mice were housed in Experimental Biomedicine Animal Unit, University of Gothenburg, and the study was approved by the Animal Research Ethical Committee of Gothenburg, and proper animal experimentation guidelines were followed.

Statistical analysis
All statistical analyses were carried out using GRAPHPAD PRISM 7.0 Software (GraphPad, San Diego, CA, USA). For all tests, data were considered significant if *P < 0.05, **. P < 0.01 and *** P < 0.001. Two-tailed unpaired and paired Student 0 s t-test was used to calculate differences in mammosphere formation and gene expression. For cytokine analysis, cytokines were considered to be overexpressed if NX-HX > 1, unaltered if NX-HX = 0 and underexpressed if HX-NX > 1. Mammosphere formation data were presented with error bars showing standard deviation (SD) and single-cell gene expression with error bars showing standard error of means (SEM). Pearson 0 s chi-square test was used to calculate significance between percentage positive cells in three different groups; differentiation positive/pluripotency negative, double positive for differentiation and pluripotency and differentiation negative/pluripotency positive in the gene expression data from the single-cell analysis.

Diverse response of hypoxia-induced secretion depends on ERa status of breast cancer cells
To delineate the potential aspect of an extended hypoxic signalling via secretion, two ERa-positive cell lines, MCF7 and T47D, and two ERa-negative cell lines, MDA-MB 231 and MDA-MB 468, were treated with CM from cells that had been grown under hypoxic or normoxic conditions for 48 h (Fig. 1A). Interestingly, CM from hypoxic ERa-positive cells increased the mammosphere-forming capacity in the recipient cells regardless of their own ERa status. In contrast, cells treated with CM from ERa-negative hypoxic cells showed a decrease in mammosphere formation (Fig. 1B). In addition, to further validate the mammosphere-forming capacity, the morphology of cell colonies was analysed as it has previously been shown that holoclones, which are round or oval in shape with tightly packed cells, contain the CSC subpopulation (Liu et al., 2013). Results show that the ability to form holoclones showed similar results as the mammosphere data with a significant increased response to hypoxic CM from ERa-positive treatment (Fig. S1A,B. Compare NX CM vs. HX CM scramble (scr) control). Next, we used patient-derived material from ERa-positive and ERa-negative breast cancer patients to confirm our previous findings. Tumour biopsies were sectioned and incubated on a gelatine sponge in HX or NX for 48 h followed by collection of CM. Results showed that MCF7 cells treated with CM from patient-derived tissue section cultures had similar effects on mammosphere formation as when using CM from breast cancer cell lines (Fig. 1C) with a pronounced contrasting response to hypoxic CM from ERa-positive versus ERa-negative patientderived tissue cultures. Data from tumour formation ability in mice, as a measure of the CSC presence, were nevertheless less conclusive. There was only a significant reduction (P = 0.022) in tumour take using ERa-negative hypoxic patient-derived CM treatment.
Moreover, to determine whether HX-linked secretion is directly dependent on the expression of ERa, siRNA specifically targeting ESR1 was used. When using CM from ERa-knockdown cells, the mammosphere-forming capacity as well as holoclone formation in recipient cells was reversed, mimicking the ERa-negative behaviour and suggesting a direct link between ERa-and HX-dependent secretion (Figs 1D and S1). In addition, overexpressing ERa in the ERa-negative cell line MDA-MB 231 results in a significant change on the effect of hypoxic secretion. However, ERa overexpression did not fully revert the ERa-negative behaviour to an ERa-positive response (Fig. 1E).

Increased pluripotency signature of cells cultured in hypoxic conditioned media from ERapositive breast cancer cells
To define and characterize subgroups of cells after treatment with CM from various cultures, we used single-cell gene expression profiling applying qPCR. In order to delineate subsets of cellular differentiation stages, we analysed genes involved in pluripotency (POU5F1, NANOG and SOX2), epithelial-to-mesenchymal transition (EMT; VIM and SNAI1), proliferation (PCNA and CCNA2) and differentiation (KRT18, ESR1 and EPCAM) as earlier established (Akrap et al., 2016). The overall effect of ERa-positive CM treatment showed a significant increased expression of POU5F1 in cells treated with hypoxic CM compared to normoxic CM and a significant decrease in ESR1 expression (Fig. S2A). These observations support the hypothesis of either an expansion of the POU5F1-expressing subpopulation or a loss of differentiation. To define gradual expression changes, we used a self-organizing map algorithm (SOM). Results show that after ERa-negative versus ERa-positive hypoxic CM treatment, cells subdivide into three SOM groups when using the expression of genes involved in pluripotency, EMT, proliferation and differentiation ( Fig. 2A). In addition, we observed that there was an overrepresentation of cells treated with ERa-negative hypoxic CM in SOM group 1, whereas there was an overrepresentation of ERa-positive hypoxic CM-treated cells in SOM group 3 (Fig. 2B). The gene expression in each SOM group revealed that cells treated with ERa-negative hypoxic CM illustrate a more differentiated and proliferating cell population while treatment with ERa-positive hypoxic CM demonstrates decreased differentiation and an increase in pluripotency signature (Fig. 2C). Moreover, the overall effect on the ERa-positive MCF7 cells treated with CM from the ERa-negative MDA-MB 231 cells showed no significant differences on individual gene level (data not shown) or difference in clustering using distribution analysis (Fig. S2B). However, when we correlate the index group differentiation and pluripotency and further separate the population into differentiation positive/pluripotency negative, differentiation negative/pluripotency positive, double positive and double negative, we observed a significant change where the hypoxic CM treatment decreased the differentiation negative/pluripotency positive group (Fig. S2C), suggesting a loss of the most primitive, stem-like cells.
3.3. The JAK-STAT pathway is involved in regulating the contrasting secretome of ERapositive and ERa-negative breast cancer cells In order to define potential differences in cytokine secretion during hypoxic conditions from the various cell lines used, we screened the CM using a human   Results are expressed as relative mammosphere formation AE SD normalized to control (NX), and statistical significance was tested using unpaired ttest (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001. (C) Mammosphere-forming capacity was analysed in MCF7 cells treated for 48 h with CM from patient-derived material after 48-h culturing on dental sponges in NX and HX. Results are expressed as relative mammosphere formation AE SD, and statistical significance was tested using unpaired t-test (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001. (D) Western blot determination of ERa after siRNA knockdown of the ESR1 gene or scr control in T47D cells. Mammosphere formation assay was carried out on MCF7 receiving cells to verify the effect of CM from siESR1-knockdown cells. Results are expressed as relative mammosphere formation AE SD, and statistical significance was tested using unpaired t-test (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001. (E) Transient overexpression of ERa using a vp16 ERa plasmid and empty vector as control in MDA-MB 231 cells followed by 48-h normoxic and hypoxic incubation. Mammosphere assay in MCF7 receiving cells treated with CM from MDA-MB 231 with overexpression of ERa. Results are expressed as relative mammosphere formation AE SD, and statistical significance was tested using unpaired t-test (n = 3).). *P < 0.05, **P < 0.01 and ***P < 0.001.
biotin label-based cytokine antibody array with 507 human proteins. From this screen, 165 cytokines were identified as upregulated (NX-HX > 1) whilst 6 cytokines were reduced (HX-NX > 1) in hypoxic CM versus normoxic control media from MCF7 cells. Interestingly, CM from MDA-MB 231 cells showed an overall reduction of secreted factors after HX, with 174 reduced cytokines and only four increased cytokines (Fig. 3A, B). Further, the significantly altered cytokine hits were analysed using the KEGG pathway enrichment software to identify potentially enriched pathways involved in the regulation of CSCs. The fact that we observed that the mammosphere-forming capacity had contrasting effects dependent on ERa status of the secreting and hypoxic stimulated cells, we hypothesized that the hypoxic secretome from ERa-positive and ERa-negative cells is significantly different. Interestingly, the pathway analysis identified the JAK-STAT pathway in both the ERa-positive and ERanegative hypoxic CM, suggesting that this pathway could play a pivotal role in influencing the BCSC population (Figs 3C,D and S3A,B). To validate the role of JAK-STAT activation on the mammosphere-forming capacity, we used a panel of inhibitors targeting both JAK and STAT. Results showed that both JAK and STAT inhibitors reduce the HX-mediated increase in mammosphere formation by ERa-positive MCF7 CM (Fig. 3E), whereas only the STAT inhibitors among the JAK-STAT inhibitors tested for ERa-negative MDA-MB 231 CM treatment (Fig. 3F). Further, silencing JAK2 using siRNA in the recipient cells followed by CM treatment demonstrates an important role for this signalling pathway in mammosphereforming capacity (Fig. 3G).

Hypoxia-induced IL6-mediated STAT3 phosphorylation in breast cancer cells
Based on the cytokine array data, one of the most striking differences in cytokine secretion by the MDA-  MB 231 cells was IL6, which demonstrated a significant (P: 0.0104) reduction in hypoxic CM compared to normoxic CM (Fig. 4A). Further, we analysed the activation of downstream targets of IL6 and we could show that MDA-MB 231 cells treated with hypoxic CM from ERa-negative patient-derived explants demonstrate less phosphorylation of STAT3 (Fig. 4B), suggesting an IL6-mediated phosphorylation of STAT3. In addition, to verify the role of IL6 we treated MDA-MB 231 cells with hypoxic CM from MDA-MB 231 cells with addition of IL6. Results show that the HX-mediated decrease in mammosphere-forming capacity with ERa-negative CM treatment was rescued with the addition of IL6 (Fig. 4C) along with induction of pSTAT3 (Fig. 4D). Moreover, one of the most pronounced changes in ERa-positive hypoxic CM among the JAK-STAT-related secreted factors was an increase in IL12RB2 (P: 0.0201; Fig. 4E). Therefore, we tested whether treatment with this single peptide would mimic the hypoxic behaviour mediated by ERa-positive cells. In line with the hypothesis, we observed a significant increase in mammosphere formation in ERa-positive MCF7 recipient cells following IL12RB2 administration (Fig. 4F). IL12RB2 is known to be a direct binding partner to STAT4. Therefore, we investigated whether STAT4 was activated after IL12RB2 treatment. Results show that there was no activation of STAT4 after IL12RB2 treatment (data not shown) but instead an increase in phosphorylated STAT3 (Fig. 4G), suggesting that the subunit IL12RB2 of the interleukin (IL)-12 receptor is important for STAT3 activation.

Discussion
Microenvironmental factors such as HX strongly affect tumour cells, participating in driving tumour progression, and influence the response to treatment (Bustin et al., 2009;Dean et al., 2012). In this study, we examined whether HX influences subtypes of breast cancer.
Our results show that CM from hypoxic ERa-positive breast cancer cells increased the fraction of CSCs compared to normal growth conditions, whereas media from ERa-negative hypoxic cells instead decreased this subpopulation of cancer cells. Further, we could show that JAK-STAT-associated cytokines are involved in both the ERa-positive and ERa-negative linked hypoxic processes as determined by a cytokine screen of the CM. JAK-STAT inhibitors and knockdown experiments further supported that this pathway was critical for the contrasting CSC response in the ERa positive and ERa negative induced by hypoxic secretion. The JAK-STAT pathway was originally discovered in a study of interferon signalling, identifying how a growth factor leads to the activation of a transcription factor (Kiu and Nicholson, 2012). Cytokines, receptors, tyrosine kinases and protein inhibitors among other factors tightly regulate JAK-STAT activation. In contrast to normal cells, STAT1, 3, 4 and 5 are constitutively tyrosine phosphorylated in most malignancies, suggesting that this pathway play an important role in disease progression.
Based on the cytokine screen, the most striking differences in cytokine secretion from ERa-negative MDA-MB 231 cells exposed to hypoxic environment was seen in IL6, which showed a strong reduction in hypoxic CM compared to normoxic CM. This suggests that IL6 may be involved in the ERa-negative behaviours during hypoxic conditions, where we observed a decrease in stem cell activity. Elevated serum IL6 has been demonstrated to associate with clinical outcome in several cancers, including breast cancer and prostate cancer where IL6 has been shown to have pleiotropic properties (Fisher et al., 2014). IL6 was demonstrated to both stimulate and reduce tumour cell proliferation and growth, inhibit apoptosis and enhance invasiveness and metastasis. A constitutive activation of IL6-JAK2-STAT3 axis has previously been described in breast cancer, especially pronounced in basal subtypes (Iliopoulos et al., 2011). Here, we observed that the Results are expressed as relative mammosphere formation AE SD, and the significant differences in ratio were calculated between vehicle control (VC) treatment and each inhibitor using unpaired t-test (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001. (G) MCF7 cells treated with CM from siRNA knockdown of JAK2 or scr in T47D cells for 48 h. Results are expressed as relative mammosphere formation AE SD, and statistical significance was tested using unpaired t-test (n = 3) *P < 0.05, **P < 0.01 and ***P < 0.001. (100 ngÁmL À1 ) for 5 min (for IL6 treatment) and 1 h (for CM treatment) and analysed for mammosphere-forming capacity and by western blot for activation of STAT3. Results are expressed as relative mammosphere formation AE SD, and statistical significance was tested using unpaired t-test (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001. (E) Representation of cytokines involved in the JAK-STAT pathway that were altered between NX and HX CM from MCF7 cells. (F) MCF7 cells were treated with human recombinant IL12RB2 at various concentrations for 48 h followed by mammosphere-forming assay for 5 days.

CM
Results are expressed as relative mammosphere formation AE SD, and statistical significance was tested using unpaired t-test (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001. (G) Phosphorylated STAT3 was measured using western blot after 48 h treatment. Total STAT3 was used as a control.
IL6-JAK2-STAT3 axis was specifically central for the ERa-positive hypoxic behaviour. Our results show that the ERa-negative MDA-MB 231 cells treated with hypoxic CM from ERa-negative patient-derived explants revealed a decreased phosphorylation of STAT3. The fact that JAK-STAT is a downstream target of IL6 and that we observed a reduction of IL6 secretion in ERa-negative hypoxic CM suggests an IL6-mediated phosphorylation of STAT3. These results suggest that the reduction in activated STAT3 in hypoxic predisposed ERa-negative cells contributes to the decrease in stem cell activity. Moreover, the most pronounced change in ERa-positive CM among the JAK-STAT-related secreted factors was an increase in IL12RB2. IL12RB2 is a subunit of the IL-12 receptor, which together with IL12RB1 strongly bind IL-12. Previously, IL12RB2 has been reported as a direct binding partner to STAT4 as well as being required for activation of the STAT4 pathway (Yao et al., 1999). When examining this potential association with STAT4, there was nevertheless no activation of STAT4 after IL12RB2 treatment but instead a significant increase in phosphorylated STAT3. Collectively, the IL6 and IL12RB2 treatment data suggest that STAT3 signalling plays a role in both ERa-negative behaviour and ERa-positive behaviour in hypoxic conditions but in opposite directions.

Conclusion
Altogether, our data support a dynamic system that is dependent on communication between the different cell types through secretion to maintain the CSC niche. We show that BCSCs from ERa-negative and ERa-positive breast cancer are inversely influenced by HX-induced cytokine secretion, which highlights the importance of HX-induced secretion in spreading contrasting BCSC influencing signalling in subtypes of the disease. However, the HIF-a and HIF-a paralog associations to HX-induced CSC propagating secretion should nevertheless be explored in more detail. Our results also underline the importance of considering breast cancer subtypes in treatments targeting JAK-STAT or HX-associated processes and highlight that HX is not only a confined tumour biological event in cancer but will influence key tumour properties in widespread normoxic microenvironments.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. (a) MCF7 and (b) T47D were transfected with siRNA against ESR1 or scr control followed by 48-h incubation in normoxic (NX) and hypoxic (HX) conditions. Progesterone expression levels were used as a functional control for the siESR1 knockdown. A holoclone assay was carried out in MCF7 and MDA-MB 231 receiving cells treated with CM from siESR1 knockdown MCF7 or T47D cells. Results are expressed as relative holoclone formation AE SD and statistical significance was tested using unpaired t-test (n = 3). *P < 0.05, **P < 0.01 and ***P < 0.001 (c) Image of MCF7 and MDA-MB 231 holoclone. Scale bar represent 100 lm.