RANK Induces Epithelial–Mesenchymal Transition and Stemness in Human Mammary Epithelial Cells and Promotes Tumorigenesis and Metastasis Free

Receptor activator of NF-κB (RANK) ligand (RANKL) and its receptor RANK, members of the TNF ligand and receptor super family, respectively, are key regulators of bone remodeling and metastasis (1), and mammary gland development (2). Impaired mammary gland development of the RANK and RANKL null mice is due to defective proliferation and increased apoptosis of mammary epithelium (2). Conversely, overexpression of RANK or RANKL in the mammary gland leads to increased proliferation of the mammary epithelia (3, 4). It has been postulated that paracrine signaling through RANK/RANKL is responsible for the expansion of mammary stem cells observed during pregnancy and luteal cycles (5, 6).

We and others have recently shown that activation of RANK signaling promotes mammary tumorigenesis in mice (4, 7, 8). MMTV-RANK transgenic mice are prone to mammary tumors (4, 7). Reciprocally, pharmacologic inhibition of RANKL or genetic ablation of RANK attenuates mammary tumor development (7, 8).

Epithelial to mesenchymal transition (EMT) involves the loss of E-cadherin–mediated cell–cell adhesion and apical–basal polarity, concomitantly with the acquisition of a motile behavior contributing to invasion and metastasis (9). Induction of EMT in immortalized human mammary epithelial cells also results in the expression of stem cell markers (10), suggesting that these 2 processes may be functionally linked. Because rodent studies have shown that activation of RANK signaling promotes tumor initiation, progression, and metastasis involving mechanisms including increased proliferation, survival (7, 8), and enhanced regenerative capacity of cancer stem cells (8), then we analyzed the contribution of the RANK pathway to human breast stem cells and EMT using a panel of normal and tumoral human mammary epithelial cells. Our results highlight the relevance of the RANK pathway in human mammary stem cells and breast cancer.

All cell lines were purchased from the American Type Culture Collection (ATCC), except UACC3199, which was obtained from the Arizona Cancer Center (Tucson, AZ). ATCC provides molecular authentication in support of their collection through their genomics, immunology, and proteomic cores, as described, by using DNA barcoding and species identification, quantitative gene expression, and transcriptomic analyses (ATCC Bulletin, 2010). UACC3199 cells harbor a methylated BRCA1 promoter suppressing gene transcription (11). UACC3199 was authenticated by its ability to reexpress BRCA1 after DNA demethylation treatment with 5-aza-2′-deoxycytidine. All lines were expanded and frozen within 2 weeks of purchase and used for a maximum of 2 months after resuscitation of frozen aliquots. MCF10A and HMECs immortalized with telomerase were cultured as described in Debnath and colleagues (12). Other cell culture conditions are described in Supplementary Methods.

Lentiviral infection using pLV409-RANK, pLV417 control that contain a luciferase reporter, or pLenti6/V5-DEST-RANK, pLenti6/V5-DEST-tubGFP was done following the manufacturer's indications (Invitrogen) and Supplementary Methods.

To evaluate activation of RANK signaling, MDA-MB-436, HCC1937, and UACC3199 cells were starved (S) with 0% FBS for 48 hours, MCF10A cells were starved in 2% serum without EGF overnight and then stimulated with hRANKL (100 ng/mL; Amgen). Proteins were isolated and analyzed following standard procedures and antibodies (Supplementary Methods).

Acinar structures were cultured and stained as described (12) and Supplementary Methods. Confocal analysis was carried out using a Leica confocal microscopy system equipped with argon and HeNe lasers. Images were captured using LasAF software (Leica).

Cells were seeded at 50% of confluence in growth medium. After 24 hours 100 ng/mL hRANKL (Amgen) were added. Forty-eight hours later medium was removed, MCF10A cells were washed and stained as reported in Supplementary Methods, and analyzed with the fluorescence-activated cell sorting Canto BD Flow Cytometer.

FL-RANK or parental MCF10A, MDA-MB-436, or UACC 3199 (3 × 10⁶) cells mixed with Matrigel (50%) were injected into the fat pad or (10⁶) in the tail vein of severe combined immunodeficient (SCID)/Beige (Charles River) or athymic nude (Harlan). After 110 to 160 days, MCF10A-injected mice were injected with 150 μg/g luciferin substrate and mammary glands and lungs were scored for bioluminescence and cell growth using (IVIS; Xenogen). Outgrowths were scored in mammary glands that were efficiently cleared. Other mice were observed for palpable tumors or health deterioration once a week. Lungs were collected at indicated times, sectioned every 100 μm, and scored for metastasis. All mice were bred and maintained in a specific pathogen-free AAALAC International accredited facility with controlled light/dark cycle, temperature, and humidity. Cages, bedding, food, and water were all autoclaved. Experimental procedures were approved by the Bellvitge Biomedical Research Institute (IDIBELL) ethics committee and were in accordance with Spanish and European regulations.

Samples from breast cancer patients for mRNA analysis were collected from the University Hospital of Bellvitge (details in Supplementary Methods), using protocols approved by the IDIBELL ethics committee and according to Declaration of Helsinki. Samples were collected immediately after surgery and subsequently frozen or fixed for RNA extraction or immunohistochemistry (IHC).

Samples from consented breast cancer patients for immunohistochemical analysis were ethically collected and procured through various human biospecimen providers (Asterand, Ardais, Bio-Options, Cytomyx Origene, and Zoion). Samples were obtained from providers through the Amgen Tissue Bank and immunohistochemical analysis done at Amgen.

Anti-human RANK or vimentin IHC was carried out on sections from formalin-fixed, paraffin-embedded specimens using anti-human RANK monoclonal antibodies (N-1H8 and N-2B10; Amgen) as described in Gonzalez-Suarez and colleagues (7) or anti-human vimentin.

To evaluate whether RANK/RANKL mRNA levels discriminate metastastatic tumors, we used Multivariate ANOVA (MANOVA) considering N0 = lymph node negative (n = 25); Ni+ = isolated tumor cells in nodes; isolated groups of epithelial cells less than 0.2 mm in diameter, only visible by IHC (n = 7); Nmi = lymph node metastases more than 0.2 mm but less than 2 mm in diameter (n = 5); N1 = lymph node positive or distant metastasis (n = 26). Analyses was adjusted by tumor subtype (S): ER⁺PR⁺Her2⁻ (n = 39); ER⁺PR⁺Her2⁺ (n = 5); ER⁻PR⁻Her2⁺ (n = 8); ER⁻PR⁻Her2⁻ (n = 10). Logarithmic transformation was required to avoid data heteroscedasticity.

Plasmids, antibodies, and other methods are included in Supplementary Methods.

To evaluate the impact of RANK overexpression in human mammary epithelial cells, we obtained stable MCF10A cells, immortal but nontransformed cells (12), expressing high levels of the full-length hRANK receptor (FL-RANK) and a control vector (PARENTAL; Fig. 1A). FL-RANK expression levels were maintained during serial passages, whereas endogenous RANK in parental cells was low or undetectable by our methodology. RANKL expression was very low and similar in both genotypes (Fig. 1A, Supplementary Fig. S1A). Functionality of the protein was shown by phosphorylation of IkBα and the p65 subunit of NF-kB upon RANKL stimulation (Supplementary Fig. S1B). Increased basal levels of RANK dowstream targets, P-IkBα, P-p38, P-Erk, and P-Akt were observed in FL-RANK when compared with parental cells (Supplementary Fig. S1C), indicating that the pathway is constitutively active. Parental cells showed highly organized cell–cell adhesion with cobble stone–like appearance at confluence, whereas FL-RANK MCF10A cells had an elongated appearance and loss of cell–cell contacts with a spindle-like fibroblast morphology, suggesting that the cells may be undergoing EMT (Fig. 1B). In monolayer cultures, parental MCF10A cells show high levels of the epithelial protein E-cadherin, low levels of vimentin, and no expression of fibronectin (mesenchymal proteins). In contrast, undetectable E-cadherin and strong staining of vimentin and fibronectin was observed in FL-RANK MCF10A cells, supporting an EMT phenotype (Fig. 1C). Quantitative real-time reverse transcriptase (RT-PCR) analyses confirmed changes observed for fibronectin, vimentin, and E-cadherin (73- and 25-fold increase, 13-fold decrease, respectively) and increased N-cadherin expression in FL-RANK cells (Fig. 1D). We observed a clear increase in mRNA levels of transcription factors known as repressors of E-cadherin promoter activity (13, 14), including, Snail (23-fold), Twist (3-fold), Zeb1 (6-fold), Zeb2 (10-fold), and Slug (2-fold) in FL-RANK MCF10A (Fig. 1D). The increased expression in EMT-related genes was confirmed using independently transduced pools of RANK expressing MCF10A and also in a distinct, nontransformed breast cell line, HMECS immortalized with telomerase (Supplementary Fig. S2A–C). Treatment of FL-RANK MCF10A cells with hRANKL for 48 hours led to further increases in EMT genes such as Vimentin (1.4 fold) or Snail (7-fold), suggesting that RANKL stimulation further induces the EMT characteristics of the FL-RANK cells (Supplementary Fig. S2D). In summary, analyses of morphologic and molecular changes indicated that RANK overexpression causes EMT in human mammary epithelial cells.

It has been shown that mammary cells undergoing EMT exhibit stem cell markers and properties of stem cells (10). We therefore looked at markers that have been related to stem cells and cancer stem cells in the human breast such as CD44, CD24, EpCAM, CD10, CD49f, and CD133 (15–20). We found that most FL-RANK MCF10A cells (82.6 ± 2.9%) were CD44⁺CD24⁻ (Fig. 2A, Supplementary Fig. S3A), compared with only 3.1 ± 0.15% parental cells. In correlation with their mesenchymal phenotype, most FL-RANK cells (99.6%) were negative for the luminal marker EpCAM and 95.6% positive for the basal marker CD10, whereas parental cells were EpCAM⁺ (87.7%), CD10⁺ (64.4%); lower levels of CD49f, and a slight increase in CD133 expression was observed in FL-RANK cells (Fig. 2B and Supplementary Fig. S3A). Both phenotypes related with basal/stem-like cells (16, 17, 20, 21). RANKL stimulation for 48 hours further enhanced the changes induced by RANK, including the increased frequency of CD44⁺CD24⁻ (10% more) and CD10 cells, and the decreased EpCAM and CD49fhi (Supplementary Fig. S3B). FL-RANK cells also show significantly higher levels of mRNA expression of SOX2, NANOG, and OCT4 (fold change 3×, 3×, and 2×, respectively), transcription factors expressed in mammary stem cells and breast tumor cells (ref. 22; Fig. 2C). To evaluate the in vivo stem cell activity, we used the cleared fat pad transplant assay (23, 24). FL-RANK MCF10A cells, in contrast to parental, were detected in the fat pad 4 months after injection, as revealed by bioluminiscence analyses (Fig. 2D); moreover, they formed small outgrowths showing that, in correlation with their stem characteristics, FL-RANK MCF10A cells were able to repopulate the mammary epithelia in vivo (Fig. 2E and F; Supplementary Fig. S3C). In summary, these results showed that in human mammary cells RANK overexpression induces the expression of mammary stem cell markers and transcription factors and the ability to generate mammary outgrowths in vivo, a function of stem cells.

The main functional consequence of EMT is enhanced cell scattering and migration. Thus, in wound healing assays, FL-RANK MCF10A cells migrated faster than the parental cells, both in the presence or absence of EGF. RANKL significantly increased the motility of FL-RANK as compared with EGF, and the most motile cells were the FL-RANK cells treated with EGF plus RANKL (Fig. 3A).

We next asked whether RANK overexpression could induce transformation. MCF10A cells grown in Matrigel with EGF, recapitulate several features of breast epithelium in vivo, including the formation of acinus-like spheroids with a hollow lumen, apicobasal polarization, and growth arrest (12, 25), whereas transformed mammary epithelial cells show a multiacinar phenotype, filled lumen, and absence of proliferative arrest (25). In contrast to parental, FL-RANK MCF10A cells formed large and disrupted structures at 20 days of culture (Fig. 3B). RANKL treatment resulted in aberrantly elongated tubular like structures (day 2) that by day 20 looked like aggregates (Fig. 3B). Confocal cross-sections revealed in FL-RANK cultures abnormal structures often composed of cells filling the luminal space and occasionaly showing protrusions into the Matrigel, characteristic of invasive cells (Fig. 3C). In FL-RANK structures E-cadherin expression was hardly detected in the cell–cell contacts and decreased expression or mislocalization of the basal marker CD49f was often observed in FL-RANK cells indicating loss of polarization (Fig. 3C). Unlike parental cells, FL-RANK acini/aggregates were not growth arrested after 20 days of culture as shown by the presence of ki67 and activated caspase 3 (Supplementary Fig. S4A). Moreover, FL-RANK overexpressing cells cultured with RANKL formed acini with filled lumens in the absence of EGF, whereas parental cells were unable to grow in these conditions (Supplementary Fig. S4B). Consistent with their nontransformed phenotype, parental MCF10A cells failed to grow in soft agar whereas FL-RANK MCF10A cells formed numerous foci (Fig. 3D), indicating that RANK expression conferes the cells the ability to grow in the absence of anchorage, a characteristic of tumor or stem cells. We concluded that RANK overexpression in human mammary epithelial cells resulted in hallmarks of transformation, including, increased motility, inability to respond to growth arrest signals, impaired polarization, luminal filling, and growth in soft agar. Despite in vitro observations consistent with the RANK-dependent transformation of MCF10A cells, these cells although able to grow in vivo (Fig. 2D and E) were nontumorigenic when injected in the cleared fat pad of immunodeficient mice and did not form lung metastasis when injected in the tail vein (Supplementary Fig. S4C).

We aim to evaluate whether RANK can cooperate with other mutations relevant in breast cancer such as BRCA1 deficiency (26) during tumorigenesis and metastasis. We therefore overexpressed RANK in breast cancer cells defective for BRCA1, MDA-MB-436, HCC1937, and UACC3199 cells (27) and confirmed its functionality (Fig. 4A and Supplementary Fig. S5A and B). RANK expression in the parental lines was low or undetectable by our methods; RANKL expression was low and comparable between both phenotypes (Fig. 4A, Supplementary Fig. S5A). As in MCF10A cells, RANK overexpression increased the frequency of CD44⁺CD24⁻ cells in all 3 cell lines due to reduced CD24 levels as compared with parental cell lines (Fig. 4B, Fig. S5C and D); a further reduction in CD24 expression was observed after RANKL stimulation in most cells (Supplementary Fig. S5C). Lower frequency of EpCAM⁺ (13% reduction) and CD49fhi (17% reduction) cells in FL-RANK MDA-MB-436 and higher of CD10⁺ in UACC3199-FL-RANK, as compared with the correspondent parental lines was observed (Supplementary Fig. S5C and D). Consistent with their mesenchymal morphology, MDA-MB-436 cells express fibronectin, vimentin, N-cadherin, and the transcription factors, Snail, Twist, Zeb1 and Zeb2, and even higher levels upon RANK overexpression (Fig. 4C, Supplementary Fig. S5E). In FL-RANK HCC1937 and UACC3199, the expression levels of EMT mesenchymal markers or transcription factors, and the frequency of EpCAM⁺ was similar to the corresponding parental lines (Fig. 4C, Supplementary Fig. S5D and E). These results showed that RANK induces CD44⁺CD24⁻ phenotype in breast cancer cell lines concomitantly with or separately from EMT.

We next investigated the functional consequences of RANK overexpression in BRCA1-defective breast cancer cell lines. Monolayer cultures revealed a modest increase in growth of the 3 FL-RANK BRCA1-defective cell lines (Supplementary Fig. S6A). Protusions into the Matrigel, indicative of an invasive phenotype, were often observed in MDA-MB-436 FL-RANK acini but, not in parental, and even more prominent in the presence of RANKL (Supplementary Fig. S6B). Quantification of proliferation in acinar cultures revealed a small increase upon RANK overexpression in some cell lines. FL-RANK cells showed lower levels of apoptosis when cultured in the presence of RANKL (Supplementary Fig. S6C).

To investigate whether RANK expression in BRCA1-defective cells promotes tumorigenesis and metastasis in vivo, we used the MDA-MB-436 cells in which EMT was enhanced upon RANK overexpression and UACC3199 cells in which RANK increased the frequency of CD44⁺CD24⁻ cells but no EMT was observed. After injection in the mammary fat pad, FL-RANK MDA-MB-436 and UACC3199 cells gave rise to tumors with faster growth and/or a shorter latency than parental cells (Fig. 5A and B). We next determined whether RANK overexpression would enhance the metastatic properties of these cells. A total of 6.5 weeks after tail vein injection, frequency and size of lung metastasis were higher in mice inoculated with FL-RANK cells as compared with parental MDA-MB-436 cells (Fig. 5C). By 12 weeks metastatic lesions in mice injected with FL-RANK MDA-MB-436 cells colonized most of the lung area as compared with smaller metastatic foci formed by parental cells (Fig. 5D; Supplementary Fig. S6D). UACC3199 cells were poorly metastatic as compared with MDA-MB-436, in correlation with the absence of EMT in these cells. Micrometastases were observed in 50% mice injected with FL-RANK UACC3199 cells (up to 56 metastatic foci) and in 33% (n = 6) mice injected with the parental UACC3199 (up to 3 lesions per mouse; Supplementary Fig. S6D). In summary, RANK overexpression enhanced tumorigenesis and metastasis in BRCA1-deficient cells.

To investigate the potential clinical relevance of RANK pathway in human breast cancer, we analyzed RANK and RANKL mRNA expression levels in adenocarcinomas considering their pathologic characteristics. We found significantly higher levels of mRNA RANK in human tumors that lack expression of the hormonal receptors for estrogen and progesterone (ER and PR) as compared with (ER⁺PR⁺) tumors that usually bear a better prognosis, and in tumors with high proliferation index (more than 40% of ki67⁺ cells) and high pathologic grade (Fig. 6A). Similar epithelial and leucocyte content was found within all groups (Supplementary Fig. S7A). To confirm the mRNA results, we analyzed RANK protein expression by IHC in a second collection of breast carcinomas. Consistent with our previous study (7), RANK protein expression was found within the carcinoma element of some tumors and a subset of infiltrating macrophages but was not observed on stromal cells, fiboblasts, or lymphocytes (Fig. 6B). A higher incidence of RANK protein expression was observed in ER⁻/PR⁻ (50%) versus ER⁺/PR⁺ (18%) and in grade III (32%) versus grade I (13%) or grade II (18%; Fig. 6C).

Our results with cell lines indicated that activation of RANK pathway enhances migration and metastasis. In human breast adenocarcinomas using a multivariant ANOVA analyses, we found that RANK/RANKL mRNA expression levels were able to discriminate between nonmetastatic (N0) and metastatic tumors (N1; either to the lymph nodel or other sites; Fig. 6D; Supplementary Fig. S7B–E). These results indicated that high RANK expression levels correlate with tumor aggressiveness and that RANK/RANKL expression in the primary tumor may indicate its metastatic behavior.

The RANK/RANKL signaling pathway, although dispensable for the initial development of the mammary gland in mice (2), controls the expansion of the stem cell compartment in adults (5, 6). Similarly, it has been recently shown that RANK is expressed in several stem cells of the skin, and that RANK signaling activate the hair cycle and epidermal growth (28). The mechanism responsible for this RANK-induced expansion of stem cells remains unknown. We show that RANK overexpression in human MCF10A cells results in constitutive activation of the pathway shown by the increased levels of RANK downstream targets (P-Iκβα, P-p38, P-Akt, and P-Erk). It is well established that overexpression of any TNF receptor family member leads to ligand-independent receptor oligomerization and ligand-independent activation of signal transduction pathways (29–32). RANK overexpression generates a EpCAM-/CD10+/CD49f lo phenotype, markers used to identify populations enriched in mammary stem cells in the healthy breast (15–17) and provides the MCF10A with the ability to reconstitute the cleared fat pad of immunodeficient mice in contrast to parental MCF10A that are unable to grow in vivo (33). Moreover, RANK overexpression increases the frequency of CD44⁺CD24lo/⁻, a phenotype ascribed to human breast cancer stem cells, in MCF10A and in breast cancer cell lines with nonfunctional BRCA1. Plasticity of breast cancer cells with respect to CD24, CD44, EpCAM, and CD49f has been reported (18, 34, 35). Although we cannot rule out that RANK is promoting proliferation or survival of stem cells, the rapid acquisition of stem markers by the entire MCF10A population upon RANK overexpression suggests that activation of RANK signaling may induce a dedifferentiation process.

An association between RANK expression and EMT has been reported in prostate cancer (36). Here we show that RANK overexpression directly induces a strong EMT phenotype in MCF10A cells (and HMECs). EMT induced by twist/snail generates cells with cancer stem cell properties (10), and usually CD44⁺CD24⁻ cells have a mesenchymal phenotype as compared with the luminal phenotype of CD24⁺ cells, leading to the belief that CSC and EMT are interchangeable events. However, despite the gain in CD44⁺CD24⁻ phenotype induced by RANK in all BRCA1-deficient cell lines, an increase in the mesenchymal markers is observed only in MDA-MB-436 cells, similar to results reported with urokinase receptor signaling (37). Thus, RANK overexpression induces CD44⁺CD24⁻ phenotype in nontransformed mammary epithelial cells and in breast cancer cell lines that can be accompanied by EMT depending on the cell of origin.

We have previously shown that RANK overexpression promotes mammary tumorigenesis in mice (7), but the relevance of this pathway in human breast disease remained largerly unknown, except for a single-nucleotide polymorphism in the locus of the TNFRSF11A gene (RANK) that associates with breast cancer risk (38). Now we show that in nontransformed mammary epithelial cells, RANK expression enhances migration, disrupts acinar formation, and allows growth in soft agar, a classical transformation assay. Changes observed in FL-RANK MCF10A acini are consistent with those described in mouse MMTV-RANK MECs (4, 7) and may reflect an imbalance response between proliferation and apoptosis that together with the impaired polarization results in increased organoid size and luminal filling. Despite these changes associated with transformation, FL-RANK MCF10A cells do not form tumors or metastasis when injected in immunodeficient mice. Similar results have been observed with several oncogenes, such as ErbB2/neu, that are able to disrupt acinus formation and transform MCF10A or HMLE in vitro (39) but fail to convey tumor growth as xenografts (10, 40). It has been shown that RANK signaling induces migration and bone metastasis in human breast cancer lines (41). Here, we provide a putative mechanism by which RANK promotes tumorigenesis, migration, and metastasis. The enhanced motility and ability to seed tumors in other locations provided by the CD44⁺CD24⁻ phenotype in cells with nonfunctional BRCA1, together with the EMT phentoype when exisiting, indicates that metastasis may not be restricted to the bone. In fact, pharmacologic inhibition of RANKL or genetic ablation of RANK significantly reduced the incidence and multiplicity of lung metastasis in MMTV-neu mice (7, 42). These findings show that RANK overexpression in human breast cancer cells increases their aggressiveness and may result in poorer clinical outcome, as recently shown (43). In fact, we observe that RANK expression progressively increases with pathologic grade and significantly associates with a high proliferative index in clinical samples of breast cancer patients. Higher levels of RANK mRNA expression were found in (ER⁻ PR⁻) tumors, which were consistent with the higher incidence of RANK protein expression in these tumors. ER⁻PR⁻ tumors are more aggressive, show a higher incidence of metastasis and worse prognosis than luminal tumors, and contain a higher frequency of CD44⁺/CD24⁻ cells (21, 44). Moreover, RANK/RANKL mRNA expression levels allow discrimination between metastatic and nonmetastatic adenocarcinomas.

In summary, we show that RANK overexpression in human mammary epithelial cells generates a “stem” phenotype in nontransformed basal human mammary epithelial cells and human breast cancer cells that can be accompanied by EMT. Our results highlight the relevance of RANK signaling pathway in human mammary tumorigenesis, and human mammary stem cells in correlation with the results previously obtained in mice, and suggest that RANK promotes tumor initiation, progression, and metastasis in breast cancer.

M. Tometsko, D. Branstetter, and WC. Dougall are Amgen employees and stockholders. No potential conflicts of interest were disclosed.

Conception and design: W.C. Dougall, E. González-Suárez

Development of methodology: M. Palafox, I. Ferrer, P. Pellegrini, S. Hernandez-Ortega, D. Branstetter, W.C. Dougall, E. González-Suárez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Palafox, A. Urruticoechea, F. Climent, M.T. Soler, M. Tometsko, D. Branstetter, W.C. Dougall, E. González-Suárez

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Palafox, I. Ferrer, P. Pellegrini, M. Tometsko, D. Branstetter, W.C. Dougall, E. González-Suárez

Writing, review, and/or revision of the manuscript: M. Palafox, I. Ferrer, P. Pellegrini, A. Urruticoechea, P. Muñoz, F. Viñals, W.C. Dougall, E. González-Suárez

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Pellegrini, S. Vila

Study supervision: E. González-Suárez

The authors thank M.A. Pujana, C. Maxwell, and M. Esteller for reactives and helpful advice, K. Rohrbach, R. Soriano, and L.Y. Huang for expert technical support of the IHC procedures, the IDIBELL animal facility service, histology service, and UB-SCT for technical support; X. Sole and V. Navarro for statistic analyses; O. Casanovas, C. Muñoz Pinedo, and members of the laboratory for useful discussions.

This work was supported by grants to E. González-Suárez by the Spanish Ministry of Science and Innovation MICINN (SAF2008-01975), by Ramon y Cajal (EGS), by the Scientific Foundation of the AECC (Catalunya), by Concern Foundation, and by institutional funds provided by the Generalitat de Catalunya to the PEBC. M. Palafox is recipient of a FPU grant, P. Pellegrini is recipient of a FPI grant, both from the MICINN.

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