Peroxisome Proliferator Activated Receptor-α/Hypoxia Inducible Factor-1α Interplay Sustains Carbonic Anhydrase IX and Apoliprotein E Expression in Breast Cancer Stem Cells

Aims Cancer stem cell biology is tightly connected to the regulation of the pro-inflammatory cytokine network. The concept of cancer stem cells “inflammatory addiction” leads to envisage the potential role of anti-inflammatory molecules as new anti-cancer targets. Here we report on the relationship between nuclear receptors activity and the modulation of the pro-inflammatory phenotype in breast cancer stem cells. Methods Breast cancer stem cells were expanded as mammospheres from normal and tumor human breast tissues and from tumorigenic (MCF7) and non tumorigenic (MCF10) human breast cell lines. Mammospheres were exposed to the supernatant of breast tumor and normal mammary gland tissue fibroblasts. Results In mammospheres exposed to the breast tumor fibroblasts supernatant, autocrine tumor necrosis factor-α signalling engenders the functional interplay between peroxisome proliferator activated receptor-α and hypoxia inducible factor-1α (PPARα/HIF1α). The two proteins promote mammospheres formation and enhance each other expression via miRNA130b/miRNA17-5p-dependent mechanism which is antagonized by PPARγ. Further, the PPARα/HIF1α interplay regulates the expression of the pro-inflammatory cytokine interleukin-6, the hypoxia survival factor carbonic anhydrase IX and the plasma lipid carrier apolipoprotein E. Conclusion Our data demonstrate the importance of exploring the role of nuclear receptors (PPARα/PPARγ) in the regulation of pro-inflammatory pathways, with the aim to thwart breast cancer stem cells functioning.


Introduction
Breast cancer is a heterogeneous set of diseases that constitute the leading cause of cancer among women in western countries [1,2]. In the recent past, a minor sub-population of tumor cells endued with the characteristics of stem cells (named cancer stem cells, CSCs) has been identified [3][4][5]. It is currently proposed that CSCs provide the cellular substrate for metastatic spreading and relapse and constitute the ultimate targets for innovating cancer therapy [4][5][6]. CSCs can be studied in vitro by expanding multicellular spheroids (mammospheres, MS) from breast cancer surgical specimens and cell lines [7][8][9]. The pro-inflammatory cytokine network is of key importance in breast CSCs biology [9,10]. In particular, the pro-inflammatory nuclear factor-kB (NF-kB) pathway, as well the NF-kB-regulated cytokines tumor necrosis factor a (TNFa) and interleukin 6 (IL6) trigger MS survival and self-renewal [9][10][11][12][13]. In haematopoietic and prostate CSCs, such a pro-inflammatory phenotype has been associated with a kind of ''inflammatory addiction'', which makes CSCs likely targets of anti-inflammatory drugs, that may act as potential enhancers of cancer therapy [14][15][16].
The stromal cell is the cornerstone of the stem cell niche [17,18]. Stroma-derived inflammatory mediators, such as prostaglandins and IL6 promote MS growth and survival [9,19]. Similarly to its normal counterpart [20], the CSCs niche is characterized by low oxygen tension (hypoxia) which promotes stem cell survival [21]. Hypoxia inducible factor1a (HIF1a) affects a variety of malignant features, such as hypoxic cancer cell survival, via the regulation of a large number of genes, including carbonic anhydrase IX (CAIX) [22,23].
We recently reported that peroxisome proliferator activated receptor-a (PPARa) modulates the expression of stem cell genes (e.g. Jagged1) and apolipoprotein E (ApoE) in breast CSCs [24]; ApoE is a lipoprotein over-expressed in MS [7]. PPARa belongs to the PPAR nuclear receptor family, enlisting also PPARb and PPARc among its members. PPARa plays a key role in lipid metabolism and it is activated by fatty acids, leukotriene and synthetic fibrates [25]. PPARc binds natural molecules, such as prostaglandin J2, polyunsaturated fatty acids, and synthetic compounds, such as Pioglitazone (PGZ) [26]. Ligands of PPARc reduce the viability of cancer cell lines and breast CSCs [24,[27][28][29]. Noteworthy, PPARc plays a significant inhibitory role on the inflammatory process [30][31], while PPARa exerts pro-inflammatory activity [32]. Interestingly, the expression of PPARa increases, while that of PPARc decreases in neural stem cells exposed to hypoxia [33]. In this study, MS from normal (N-MS) and tumor (T-MS) tissues, as well as from tumorigenic MCF7 (MCF7-MS) and non tumorigenic (MCF10-MS) human breast cell lines, were exposed to the supernatant of normal mammary gland and breast tumor associated fibroblasts. We undertook this approach to elucidate the regulation of PPARa and PPARc in the context of breast CSCs inflammatory pathway activation.

Enhanced Autocrine TNFa Loop in Breast Cancer Tissue Derived MS Exposed to the Supernatant of Tumor Associated Fibroblasts
We recently reported the increase of NF-kB activity in breast tumor MS (T-MS), compared to their normal counterparts (N-MS) [24]. The major trigger of NF-kB pathway is TNFa, a potent inducer of MS formation [12,13]. Here we found higher expression of tumor necrosis factor receptor-1 (TNFR1), and of the NF-kB targets TNFa and IL6 in tumorigenic MCF7-MS and T-MS, compared to non tumorigenic MCF10-MS and N-MS, respectively ( Figure 1A, B, C). We also observed that exogenous TNFa elicited MS formation in MCF7 to a higher extent than in MCF10 cells, a phenomenon that was mimicked by the administration of the supernatant of tumor associated fibroblasts (TAF), but not of normal mammary gland fibroblasts (NAF, Figure 1D). Both TAF and NAF secreted very low levels of TNFa, whereas TAF secreted higher levels of transforming growth factor-b1 (TGFb) compared to NAF ( Figure 1E). TGFb is a potent MS growth factor and synergizes with TNFa to induce stem cell features in breast cancer cells [34][35]. With respect to this issue, we observed that TGFb was able to induce TNFa expression in MS ( Figure 1F). Moreover, the TAF supernatant induced TNFa expression in T-MS and MCF7-MS to a higher extent than in N-MS and MCF10-MS, respectively ( Figure 1G). Finally, the increase in T-MS formation following the TAF supernatant administration was halted by TNFa blocking antibody administration ( Figure 1H). These data show that autocrine TNFa signalling in breast CSCs is enhanced by TAF secretion of TGFb.
Opposing Roles of miRNA130b and miRNA17-5p on the PPARa/HIF1a Interplay To mechanistically elucidate the PPARa/HIF1a interplay, we examined microRNA130b (miR130b) expression. This microRNA was chosen basing on two considerations: i. its capability to increase HIF1a expression via the down-regulation of the HIF1a mRNA translation inhibitory protein DDX6 [36]; ii, the presence of its binding site at PPARc mRNA 39UTR [37]. We observed that the miR130b up-regulation in TAF supernatant-exposed MCF7-MS ( Figure 4A) was paralleled by the reduction of DDX6 expression in MS ( Figure 4B). Accordingly, the administration of miR130b antagonist (a-miR130b) in MCF7-MS reduced the activity of HRELuc ( Figure 4C) and induced DDX6 expression ( Figure 4D). We then observed the up-regulation of miR130b expression in MCF7-and MCF10-MS compared to adherent cells ( Figure 4E). In keeping with the results above reported, miR130b expression was down-regulated by siHIF1 and siPPARa, and it was induced by PPARa over-expression in MCF7-MS ( Figure 4F). Finally, the transfection of pre-miR130b increased PPARa expression in MCF7-MS ( Figure 4G).

ApoE Over-expression in T-MS is Under the Control of the PPARa/HIF1a Interplay
PPARa is involved in a wide variety of cellular functions, including lipid homeostasis [39]. Moreover, PPRE consensus sequence occurs at the promoters of lipid transporters, such as apolipoproteins [39]. Interestingly, ApoE is over-expressed in MS [7,24]. Here, we were able to quantify the over-expression of ApoE in T-MS compared to N-MS, as well as in MCF7-MS compared to MCF10-MS ( Figure 7A). We also found ApoE overexpression in MCF7-MS and MCF10-MS in response to exogenous TNFa and to the TAF supernatant administration ( Figure 7B). Then, we demonstrated that PPARa over-expression induced the mRNA expression ( Figure 7C) and protein (+81%, p,0.01, Figure S9A) of ApoE in MCF7-MS. As expected, the phenomenon was mimicked by the administration of WY (+127%, p,0.005, in MCF10-MS and +135%, p,0.005, in MCF7-MS, Figure S9B). Furthermore, HIF1 vector elicited (+44%, p,0.05, Figure S9C), as well as siHIF1 reduced the expression of ApoE in MCF7-MS ( Figure 7D). These data show that ApoE overexpression in breast cancer stem cells is under the control of the PPARa/HIF1a interplay.

Discussion
Our investigation started with the observation that human models for breast CSCs (T-MS and MCF7-MS) display increased TNFR1 compared to their normal/non tumorigenic counterparts (N-MS and MCF10-MS). Breast CSCs also exhibit an increased growth response to the TAF supernatant. Though TAF secrete low amounts of TNFa, we demonstrate that TAF elicit TNFa expression in CSCs by secreting TGFb, thus setting up an autocrine TNFa loop that enhances MS growth. TNFR1 signalling is a major inducer of inflammatory response via NF-kB activation [40][41][42][43]. The expression of the two NF-kB targets TNFa and IL6 is higher in CSCs, and both these cytokines have been previously shown to elicit MS formation [12][13]44,45]. Collectively, the findings here reported, together with previous data concerning the increase of NF-kB activity in T-MS [24], contribute to the tenet that CSCs are endowed with proinflammatory phenotype [14,16,18].
We then pinpoint that the core of such TAF-promoted pathway involves the PPARa/HIF1a interplay. In particular, the two proteins induce each other expression and trigger T-MS growth to a higher extent than their normal counterpart. In this regard, we show the involvement of miR130b. Over-expression of this microRNA has been previously reported in liver cancer and pluripotent stem cells [46,47]. Moreover, miR130b is induced by hypoxia and increases HIF1a protein expression by facilitating HIF1a mRNA translation, via the down-regulation of DDX6 expression [36]. Here, we report that miR130b up-regulates  HIF1a activity (as well as down-regulates DDX6 expression), and that the up-regulation of the PPARa/HIF1a interplay is paralleled by the down-regulation of PPARc, a miR130b target [37]. In regard to this issue, we provide evidence that the PPARc agonist PGZ, which up-regulates PPARc expression and activity [24], hinders the PPARa/HIF1a interplay in breast CSCs. We propose that this phenomenon is mediated by miR17-5p which targets both PPARa and HIF1a mRNA 39UTRs [48]. With respect to this issue, we show that miR17-5p expression is up-regulated by PGZ and that miR17-5p knock-down increases PPARa expres- Thus, miR17-5p may mediate the PPARa/HIF1a interplay switch-off throughout the induction of PPARc over-expression. Interestingly, miR17-5p has been previously found to be repressed by hypoxia [49]. Nevertheless, miR17-5p has been reported as a pro or anti oncogenic miR depending upon the genetic and environmental context [50].
We identified two acknowledged regulators of breast CSCs, namely IL6 and SLUG as targets of the PPARa/HIF1a interplay. The former has been characterized as crucial mediator of breast CSCs growth capacity in vitro [10]. The latter was recently demonstrated to play a pivotal role in normal and tumor mammary gland self renewal in human and mice [5,12]. We however focussed our attention on two additional targets of the PPARa/HIF1a interplay. The former target is the hypoxia inducible gene CAIX, which we found over-expressed in T-MS, in compliance with its original definition as tumor antigen [51]. We previously reported that CAIX expression is crucial for MS hypoxia survival [9,22]. Other investigations showed that CAIX sustains breast cancer survival and invasive behaviour [52]. CAIX is an HIF1a target and contributes to cancer aggressiveness in various biological contexts, such as the basal-like breast tumor subtype [12,53]. The data here presented lead to hypothesize that CAIX over-expression in CSCs may be the consequence of cues that pertain to the cancer stem cell niche. These data encourage to pursue the ongoing research on CAIX inhibitory molecules as anti cancer agents [54].
The latter target controlled by the PPARa/HIF1a interplay is ApoE, a major component of circulating lipoproteins [55].
Here we found the over-expression of ApoE in T-MS, recalling a similar finding in prostate CSCs [14]. We then demonstrate that ApoE knock-down reduces MS formation and the expression of CAIX, IL6 and SLUG [5,12]. These data agree on the role of ApoE in breast cancer aggressiveness. In fact, ApoE plays a crucial role in human pathology, as the e4 allele represents a frailty variant that predisposes to various agerelated diseases [56]. ApoE knock-out mice disclose increased mammary tumor incidence, likely in relationship with their hyperlipidemic state [57]. Intriguingly, ApoE physically interacts with HCCR-1, an onco-protein that promotes breast cancer [58]. However, the relationship between breast cancer and ApoE in humans is still under debate, and ApoE may impact disease susceptibility or response to therapy [59,60]. Interestingly, whereas ApoE is likely to impact cardiovascular diseases due to an alteration of the circulating lipidic profile, its role in cancer seems to be independent of this association [61].
In conclusion, we show that the PPARa/HIF1a interplay, triggered in breast CSCs by the tumor associated fibroblast secreted TGFb, engenders the expression of two acknowledged breast CSCs regulatory genes (IL6 and SLUG), as well as upregulates two less characterized regulators of breast CSCs, namely CAIX and ApoE (Figure 9). This molecular machinery is counter-acted by PPARc expression. Our data lead to envisage the possibility to harness nuclear receptor regulation of pro-inflammatory pathways to negatively interfere with CSCs survival.

Cell Cultures and Generation of MCF7 and MCF10 MS
MCF7 were grown in RPMI 1640 medium supplemented with 10% FBS, penicillin-streptomycin and glutamine. MCF10 [62]  were grown in DMEM medium with 20% FBS, supplemented with 10 mg/ml insulin, 10 mM hydrocortisone and 10 mg/ml EGF. Hypoxia (1% pO 2 ) was generated in an Vivo 2 300 hypoxic workstation (Ruskinn Technologies, Ireland). MCF7-MS and MCF10-MS were generated by plating 2500 cells into 3-cm 2 low-attachment wells (Corning, NY, USA) in mammary epithelial growth medium (MEGM), supplemented with B27, 10 ng/ml epidermal growth factor (EGF), 10 ng/ml basic fibroblast growth factor (bFGF), 10 mg/ml insulin, 10 26 M hydrocortisone (Voden Medical, Rome, Italy). Primary MS formation usually occurs after 48 to 72 h. To examine the effects of chemicals on MS formation and MS gene expression, MCF7-MS and MCF10-MS were exposed to each molecule and assessed after 24 h to 72 h. MS with an apparent diameters $50 mm were scored and photographed using a inverted microscope (Olympus CKX41, digital cameras Olympus C-5060, Japan). To examine the impact of each specific expression vector, siRNA or pre/antago-miR on MS formation and MS gene expression, adherent MCF7 or MCF10 (10 5 cells in a 3-cm 2 well) were transfected with 1 mg/well of each siRNA or pre/antago-miR using Lipofectamine 2000 (Invitrogen, USA). After 6 h of incubation, cells were re-suspended, seeded in 24-well ultra-low attachment plates at a density of 2500 cells per well and assessed after 48 h or 72 h. The Effect of each specific treatment was determined by at least n = 3 independent experiments. MDA-MB231 cells were grown in RPMI 1640 medium supplemented with 10% FBS, penicillin-streptomycin and glutamine.

Generation of MS from Normal and Breast Carcinoma Human Tissues
Twenty-two fresh surgical specimens, obtained from patients with ductal breast carcinoma who underwent quadrantectomy or mastectomy, were collected for the study (Table S1). Normal and tumor samples were processed as previously described [9,12]. Briefly, tissues were placed in sterile Epicult (Voden Medical), minced with sterile scalpels, and incubated for 6-12 h in the presence of 1000 U Collagenase/Hyaluronidase enzyme mix (Voden Medical). Samples were centrifuged at 806g, and the pellet was digested by Dispase and DNAse (Voden Medical), and then pelleted at 4506g. Pellets were resuspended, filtered through a 40-mM nylon mesh (Voden Medical), and plated into 3-cm 2 -well low attachment plates (Corning, NY, USA), filled with 3 ml MEGM, supplemented with B27, 10 ng/ml EGF, 10 ng/ml bFGF, 10 mg/ml insulin, 10 26 M hydrocortisone (Voden Medical). Primary MS started forming after 4-6 days and were processed at day 14. Self renewal of MS was tested by assessing the capacity of primary MS to generate secondary MS after trypsin disaggregation. Transfection in primary T-MS was performed by mixing 1 mg of each expression vector or siRNA with in vitro JET-PEI reagent (Poly-plus transfection, USA).

Isolation of Fibroblasts from Normal and Tumour Breast Tissues and Collection of the Fibroblasts Supernatant
Fibroblasts were collected by centrifuging the Collagenase/ Hyaluronidase digested tissue lysates used for MS generation at 5006g for 5 min (see above). The fibroblasts containing pellet was re-suspended and cultured in DMEM medium with 20% fetal bovine serum (FBS, Euroclone, Milan, Italy), penicillin-streptomycin and glutamine (Sigma), in 6-well plates. When the fibroblasts reached confluence, medium was discarded and replaced with fresh medium, containing DMEM+FBS 0.5% (1.5 mL/well), for 24 h. Supernatants were then collected, centrifuged for 5 min at 10 5 6g to remove debris and conserved at 280uC. For the experimental setting supernatants were diluted in MS medium (MEGM) at the final concentration of 10%, and already formed MS were exposed for 24-72 h.

RNA Extraction, Real-time Reverse Transcription Quantitative PCR (qPCR) and Reverse Transcription PCR (RT-PCR)
Total RNA was extracted from cultured cells, MS and fibroblasts using TRIzol (Life Technologies, Rockville, MD, USA) reagent following the customer's instructions. Real-time Reverse Transcription quantitative PCR (qPCR) analysis was performed by TaqMan approach in a Gene Amp 7000 Sequence Detection System (Life Technologies, Rockville, MD, USA), as previously described [24]. Each sample was analyzed in replicates (n = 3). Sets of primers and fluorogenic probes specific for the target genes (Table S2) were purchased from Applied Biosystems; qPCR conditions are: pre-denaturation step at 95uC for 2 min; 28 cycles of denaturation at 95uC for 1 min, annealing at the appropriate temperature for 1 min, extension at 72uC for 1 min; final extension at 72uC for 7 min. Human beta-glucuronidase was used as an endogenous control for mRNA level. 6 URNP was used as an endogenous control for miRNA level. The relative amount of each target mRNA or miRNA was calculated as: N target 2 2 (DCt sample2DCt calibrator) , where DCt values of the sample and calibrator were determined by subtracting the Ct value of the endogenous control gene from the Ct value of each target gene. RT-PCR analysis was performed using the Master RT plus PCR system kit according to the instruction of the supplier (Life Technologies). Actin was used as an internal control. RT-PCR was performed for 31 cycles (1 minute/annealing) for each primer, except for CAIX that was performed for 34 cycles. Primer sequence and PCR parameters are reported in Table S3.

Luciferase Assay
SLUGLuc, containing the 2800/+10 bp SLUG promoter sequence in the pGL3 basal vector, was kindly provided by Dr. Togo Ikuta (Saitama Cancer Centre, Saitama, Japan). ERaLuc plasmid, which contains 3 copies of estrogen response element (ERE), was kindly provided by Dr. Rakesh Kumar (Department of Molecular and Cellular Oncology, MD Anderson Cancer Center, Houston, Texas). CAIXLuc, containing the 2179/+34 bp promoter sequence of CAIX, was provided by J Pastorek (Slovak academy of science, Bratislava). Hypoxia responding element (HRE-Luc), containing 3 copies of HIF1 consensus was kindly provided by Dr. Giovanni Melillo (Tumor hypoxia laboratory, National Cancer Institute, Frederick, MD, USA). IL6Luc, containing the 22161 to 241 bp IL6 promoter sequence, was kindly provided by Dr. WL Farrar (NCI-Frederick Cancer Research and Development Center, Frederick, MD, USA). PPRELuc, containing 7 copies of PPARs consensus, was kindly provided by Professor Ronald Evans (Salk Institute, La Jolla, CA). NF-kBLuc was previously described [12]. Each of the above plasmids (1 mg) were co-transfected with a thymidine kinase promoter driven Renilla luciferase (400 ng) plasmid as a reference control (Promega, USA). MS transfection was performed with JET-PEI reagent (Poly-plus transfection) (3 mL for 1 mg plasmid) and Luciferase activity was assayed after 48 h using the Dual-LuciferaseH Reporter Assay System (Promega), according to the manufacturer's instructions. Luciferase activity was normalized over Renilla activity and all reported experiments were performed in triplicates.

Elisa Test
Determination of TNFa and TGFb level in TAF and NAF supernatant were evaluated by ELISA (S.I.C., Rome, Italy). Briefly, cells were seeded in a 6-well plate at the density of 3610 5 cells per well and collected in serum-free medium for 24 h. The harvested medium was centrifuged at 5006g for 5 min (4uC) to remove floating cells and the supernatants were collected and assayed following the customer's instructions.

Statistical and Bioinformatic Analysis
Statistical significance was assessed by ANOVA followed by Bonferroni's multiple comparison test or two-tail Student's t-test, as appropriate, using PRISM 5.1 (Graphpad Software, La Jolla, CA, USA). The level for accepted statistical significance is p,0.05. mRNA 39-UTR were analyzed for miRNA binding site by the online software Targetscan (www.Targetscan.com).