Cancer Epigenetics and Biology Program, Bellvitge Biomedical Research Institute, IDIBELL, Barcelona, Spain
Correspondence: González-Suárez Eva, Ph.D., Cancer Epigenetics and Biology Program, Bellvitge Biomedical Research Institute, IDIBELL, Gran Via 199, L'Hospitalet de Llobregat, Barcelona 08908, Spain. Telephone: +34–932607253; Fax: +34–932607219; e-mail: firstname.lastname@example.org
Author contributions. P.P.: collection and/or assembly of data, data analysis and interpretation, final approval of manuscript; A.C.: collection and/or assembly of data; data analysis and interpretation; final approval of manuscript; M.I.G. and W.C.D: provision of study material, revisions of manuscript, final approval of manuscript; P.M.: revisions of manuscript, final approval of manuscript; P.A.M.: data analysis and interpretation, revisions of manuscript, final approval of manuscript; E.G.-S.: conception and design, financial support, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript. PP and AC contributed equally to this work.
Receptor Activator of NF-kappa B (RANK) pathway controls mammary gland development in mice but its role in mammary stem cell fate remains undefined. We show that constitutive RANK signaling expands luminal and basal mammary compartments including mammary stem and luminal progenitor cell pools and interferes with the generation of CD61+ and Sca1+ luminal cells and Elf5 expression. Impaired mammary cell commitment upon RANK overexpression leads to the accumulation of progenitors including K14+K8+ bipotent cells and the formation of heterogeneous tumors containing hyperplastic basal, luminal, and progenitor cells. RANK expression increases in wild-type mammary epithelia with age and parity, and spontaneous preneoplastic lesions express RANK and accumulate K14+K8+ cells. In human breast tumors, high RANK expression levels are also associated with altered mammary differentiation. These results suggest that increased RANK signaling interferes with mammary cell commitment, contributing to breast carcinogenesis. Stem Cells2013;31:1954-1965
The mouse mammary epithelium consists of a branched ductal system that develops mainly during puberty and a lobuloalveolar compartment containing secretory epithelial cells that undergo functional differentiation and become milk-producing during pregnancy. Mammary epithelial cells can be organized in two main compartments, luminal and basal. Expression of hormone receptors (estrogen and progesterone receptors, ER and PR, respectively) is confined to the mature luminal cells that are characterized as being CD24high (hi) CD49flow (lo) Sca1+, whereas CD24hi CD49flo Sca1− cells are largely ER− progenitors  and contain the CD61+ luminal progenitors that establish the alveolar lineage during pregnancy [2, 3]. The basal compartment, identified by the expression of CD24lo CD49fhi Sca1− CD61+, is mainly composed of contractile myoepithelial cells that surround ducts and alveoli and contains the population of mammary stem cells (MaSCs) identified by their ability to reconstitute an entire functional mammary gland [4-6]. Several cytokeratins are lineage markers within the mouse mammary epithelium: K5 and K14 mark basal/myoepithelial cells and are strongly expressed in MaSC, whereas K8/K18 expression marks luminal cells . A hierarchical organization of mammary epithelia, in which MaSC give rise to differentiated cell types via a series of multipotent and lineage-restricted intermediates has been proposed .
MaSC and intermediate progenitors have been postulated as the cells of origin of tumors, as their long lifespan might allow them to accumulate enough genetic lesions to generate a tumor . The observed expansion of MaSC and other progenitor populations with each round of pregnancy and estrous cycles [10, 11] provides a potential mechanistic explanation for age and parity as risk factors for the development of mammary tumors .
RANK and its ligand, RANKL, are key regulators of mammary gland development [13, 14]. The RANK signaling pathway mediates the major proliferative response of mouse mammary epithelium to progesterone during morphogenesis [10, 11, 15].
Disrupted mammary gland development during pregnancy and impaired lactation is observed as a consequence of RANK loss or overexpression [13, 14]. Defective alveologenesis in RANK-null mice could be attributed to decreased proliferation and survival of mammary epithelial cells , whereas mouse mammary tumor virus (MMTV)-RANK mice show high mammary epithelial cell proliferation during pregnancy , pointing to a defect in differentiation. We have recently shown that activation of RANK signaling promotes mammary tumorigenesis in mice  and that RANK overexpression in human mammary cell lines induces stemness and promotes tumorigenesis and metastasis .
Here, we investigated the role of RANK signaling in MaSC fate and found that activation of RANK signaling expands basal and luminal compartments disrupting mammary lineage commitment, and results in the accumulation of intermediate progenitor cells, eventually leading to hyperplasia and tumorigenesis.
Materials and Methods
All research involving animals was performed at the IDIBELL animal facility and complied with protocols approved by the IDIBELL Committee on Animal Care. For cell proliferation analysis, 5-bromo-2′-deoxyuridine (BrdU, 30 mg/kg of mouse) was injected intraperitoneally 2 hours before killing.
Mammary Cell Isolation
Single cells were isolated from tumors and the mammary glands of virgin young (12–15 weeks), old (31–81 weeks), or pregnant mice, as described previously . Briefly, fresh tissues were mechanically cutted with McIlwain tissue chopper and enzymatically digested with appropriate medium (Dulbecco's modified Eagle's medium [DMEM] F-12, 0.3% Collagenase A, 2.5U/mL dispase, 20 mM HEPES, and antibiotics) 40 minutes at 37°C. Samples were washed with Leibowitz L15 medium+10% fetal bovine serum (FBS) between each step. Erythrocytes were eliminated by treating samples with hypotonic lysis buffer and fibroblasts were excluded by incubation with DMEM F-12 + 10% FBS 1 hour at 37°C in a tissue culture flask (the majority of fibroblasts attach to the tissue culture plastic while most of epithelial organoids do not). Single epithelial cells were isolated by treating with trypsin (PAA Laboratories, Velizy-Villacoublay, France, http://www.paa.com) 2 minutes at 37°C and a mix of dispase (Gibco life technologies, Invitrogen Saint Aubin, France, http://www.invitrogen.com) DNAse (Invitrogen Saint Aubin, France, http://www.invitrogen.com) 5 minutes 37°C. Cell aggregates were removed by filtering cell suspension with 40 µm filter and counted.
Single cells were labeled with antibodies against CD24-PE or CD24-FITC (M1/69 BD Pharmingen, San Diego, CA, http://www.bdbiosciences.com), CD29-FITC (HMβ1-1, BD Pharmingen), CD49f-APC (GoH3 R&D systems, Minneapolis, MN US, www.rndsystems.com), CD61-FITC or CD61-PE (2C9.G2 BD Pharmingen), Sca-1-PE or Sca1-APC (Ly-6A/E, BD Pharmingen), and CD49b-alexa 647 (HMα2 Biolegend, San Diego, CA, http://www.biolegend.com). Lymphocytes and endothelial cells were excluded in flow cytometry using CD45-PECy7 (30-F11 Biolegend) and CD31-PECy7 (390 Biolegend) antibodies, respectively. A population of 10,000 live cells was captured in all fluorescence-activated cell sorting (FACS) experiments. FACS analysis was performed using FACS Canto (Becton Dickinson, San Jose, CA) and Diva software package. Cell sorting was performed using FACS Aria cell sorter (BD Becton Dickinson, San Jose, CA).
Colony Forming Assays and Immunofluorescence
For colony-forming assays, sorted cells were plated as described previously  in growth medium that contains B27, 5% FBS, EGF (10 ng/mL), Hydrocortisone 0.5 µg/mL, Insulin 5 µg/mL, Cholera toxin 100 ng/mL, and penicillin/streptomycin and RANKL (1 µg/mL; Amgen Inc, Thousand Oaks, CA, USA, http://www.amgen.com) as indicated. ROCK inhibitor (10 µM Y-27632, Sigma Aldrich, Saint-Quentin Fallavier, http://www.sigmaaldrich.com/france) was added to basal cell cultures. After 15 days of culture, colonies were fixed with paraformaldehyde (PFA) 2% for keratins staining. K5 and K8 immunostaining on three-dimensional cultures was performed following standard procedures .
Cleared Fat Pad Transplantation
Primary mammary epithelial cells were freshly isolated from mouse mammary glands, sorted based on CD24 CD49f/Sca1 values and resuspended in Leibowitz L15 medium (Invitrogen) with 10% FBS and antibiotics, diluted 1:1 with Matrigel Matrix (BD Biosciences, San Diego, CA, http://www.bdbiosciences.com) for a final volume of 20–30 µL and CD24lo/CD49fhi cells were injected at limiting dilution in cleared mammary fat pads of 3–4 weeks old FVB mice. After 8 weeks, the transplanted fat pads were whole-mounted and carmine-stained. Outgrowths filling at least 25% of fat pat were scored. The proportion of stem/progenitor cells was determined using the Extreme Limiting Dilution Analysis (http://bioinf.wehi.edu.au/software/elda/).
Tissue Section Histology and Immunostaining
Mammary glands and tumors were fixed in 4% PFA or formalin and embedded in paraffin. For histological analysis, 3 µm sections were cut and stained with hematoxylin and eosin. Immunostaining was performed on 3 µm mammary gland/tumor sections. Antigen heat retrieval with citrate was used for all antibodies except for RANK (protease XXIV) and BrdU (protease XXIV and HCl 2 M) and primary antibodies were incubated overnight at 4°C. Antibodies used include RANK (AF692; R&D Systems), PR (A0098 Dako, Trappes, France, http://www.dako.fr), K5 (AF-138, Covance, Princeton, NJ, http://www.covance.com.), K14 (AF-64 Covance), K8 (TROMA, dshl, Developmental Studies Hybridoma Bank, Iowa City, Iowa), and SMA (1A4; A2547, Sigma), BrdU (G3G4, Kaufman laboratory; University of Illinois). The antigen–antibody complexes were detected with streptavidin horseradish peroxidase (Vector Laboratories, Burlingame, CA, USA, email@example.com) for RANK, K8 immunostainings. K5, K14, PR, and SMA were detected with EnVision technology (Dako, Trappes, France, http://www.dako.fr); BrdU was detected with labeled polymer-horse-radish peroxidase (DAKO). Peroxidase was finally revealed with 3,3-Diaminobenzidine DAB (DAKO). For immunofluorescence, opportune fluorochrome-conjugated secondary antibodies were added after primary incubation and then mounted with VECTASHIELD Mounting Media.
RNA Preparation and real time reverse transcription PCR
Total RNA of cells, mammary glands and tumors was prepared with Tripure Isolation Reagent (Roche, Basel, Switzerland, http://www.roche.ch/en/standorte/basel-hq.htm) in accordance with the manufacturer's instructions. After tissue dissociation, 20 ng/mL of messenger RNA (mRNA) or similar cell numbers were pretreated with DNase I (Ambion Invitrogen Saint Aubin, France, http://www.invitrogen.com). Single-stranded complementary DNA was produced by reverse transcription using 1 µg of RNA DNA-free in a 20-µL reaction (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). Quantitative polymerase chain reaction (PCR) was performed using the TaqMan probe-based system (Applied Biosystems) on the ABI 7900HT as per the manufacturer's instructions (Applied Biosystems; rank Mm00437135_m1, rankl Mm00441908_m1, wap Mm00839913_m1, k8 Mm00835759_m1, K14 Mm00516876_m1, elf5 Mm00468732_m1, slug Mm00441531_m1, sox9 Mm00448840_m1).
Gene Set Enrichment Analysis (GSEA) Analyses
The GSEA tool was run using default values for all parameters. Preprocessed and normalized The Cancer Genome Atlas (TCGA) data were downloaded from the corresponding repository (version July 3, 2012; http://tcga-data.nci.nih.gov/tcga/tcgaHome2.jsp). Samples were clustered based on the expression of the 50 genes/probes from PAM50 and assigned to the intrinsic subtypes basal-like, HER2-enriched, luminal A, luminal B, or normal-like, based on a comparison with the PAM50 centroids . The PAM50 confidence scores were >0.90 for all the basal-like and luminal A tumors except for two cases in each subtype. The GSEA leading edges were analyzed for enrichment in pathways annotated by the Kyoto Encyclopedia of Genes and Genomes (KEGG) using the Database for Annotation, Visualization and Integrated Discovery (DAVID) . Only pathways over-represented at a false discovery rate <5% between the leading edge and the rest of the corresponding gene set were considered.
Statistical analysis of the differences between two mouse cohorts or conditions was performed with a two-tailed Student's t-test. We used Microsoft Excell or GraphPad Prism software for calculations and expressed the results as the means ± SD. To calculate the significance of differences affecting several variables a multivariate analysis of variance (MANOVA) test has been carried out. The null hypothesis is that all the tumors are identical in all variables.
RANK Overexpression Results in Morphologically Aberrant Ducts that Contain Dual Positive K5/K8 Cells
As we have shown that RANK induces stemness in human mammary epithelial cell lines , we hypothesize that increased levels of RANK may alter the balance in mammary differentiated/progenitor cell populations. To evaluate this hypothesis, we used our previously generated MMTV-RANK mouse model  backcrossed to the FVB background, most commonly used in mammary studies (Supporting Information Fig. S1A). RANK overexpression in mammary epithelia led to constitutive activation of RANK pathway , resulting in mammary epithelial cell hyper-proliferation (Supporting Information Fig. S1D, S1E), impaired secretory alveologenesis and absence of Wap milk protein expression during pregnancy (Supporting Information Fig. S1B, S1C). These phenotypes led to a lactation failure that could not be rescued by multiple pregnancies in correlation with previous results . Increased epithelial growth, enhanced ductal side-branching, and appearance of precocious small alveoli were observed in virgin MMTV-RANK glands (Fig. 1A; S1B) together with increased expression of basal, K14, and luminal, K8, mRNA markers as compared with wild type (WT) (Fig. 1B). Double immunofluorescence (IF) analyses of K5 and K8 (Fig. 1C) revealed frequent alterations within the basal and luminal compartments in young virgin MMTV-RANK glands. First, MMTV-RANK ducts often contained disorganized myoepithelium, occasionally with K5 cells that accumulate in several layers, areas lacking K5 expression in the outer layer or K5 cells localized within the luminal area, surrounded by K8+ cells. Second, accumulation of multiple layers of K8+ cells was noted in most young MMTV-RANK glands resulting in minimal luminal space or filled ducts (Fig. 1C, 1D). By contrast, in WT ducts, basal/myoepithelial K5 cells were well aligned and normally organized and a single layer of K8+ cells was detected. K14 and K5 showed similar staining patterns in the mammary glands of young WT and MMTV-RANK virgin glands; thus, K14 and K8 IF revealed similar phenotypes than K5 and K8 IF (data not shown). Additionally, both K5/K8 and K14/K8 double positive cells, which have been proposed as bipotent progenitors blocked in differentiation [22, 23], were more common in MMTV-RANK ducts than in WT ducts (Fig. 1C, 1E). These observations demonstrate that enhanced activation of RANK signaling disrupts basal/luminal epithelial morphology resulting in aberrant ducts containing a hyperplastic luminal layer, a disorganized basal layer and expansion of dual positive K5/K8 (or K14/K8) cells, described as bipotent mammary progenitors.
RANK Overexpression Interferes with MaSC Fate in the Basal and Luminal Compartment
Given the disrupted mammary epithelia observed in MMTV-RANK glands, we next investigated whether RANK overexpression leads to alterations in mammary cell populations in virgin mice. Quantitative expression analysis (real time reverse transcriptase PCR [RT-qPCR]) of Rank in fluorescence-activated cell sorting (FACS)-sorted WT and MMTV-RANK mammary populations demonstrated that MMTV promoter directs Rank expression to basal (B: CD24lo CD49fhi) and luminal (L: CD24hi CD49flo) cells, both Sca1+ (LM: luminal mature) and Sca1− (LP: luminal progenitor enriched) fractions (Supporting Information Fig. S2A). FACS analyses of freshly isolated epithelial cells revealed a two fold increase in basal and 1.6-fold increase in luminal populations in MMTV-RANK as compared to WT mammary glands (Fig. 2A, 2B), in agreement with the abundant epithelia and increased expression of K14 and K8 markers observed (Fig. 1A, 1B). These findings further indicate that Rank overexpression under the MMTV promoter, induces an expansion of the mammary basal and luminal compartments. We next analyzed Sca1, CD61, and CD49b lineage markers expression to discriminate mammary subpopulations [1, 2, 24, 25]. The distribution of Sca1+ and Sca1− cells was altered in MMTV-RANK luminal cells (Fig. 2A, 2B): in five of seven MMTV-RANK mice analyzed, there was a significant lower Sca1+ cell population within the luminal fraction compared with WT controls (Supporting Information Fig. S2B). Consistent with the known association between Sca1 and hormone receptor expression , the frequency of PR+ cells (detected by IHC analyses) was reduced in three of seven virgin MMTV-RANK mammary glands analyzed, compared to control glands (Fig. 2C). Thus, RANK overexpression results in a significant expansion of basal (CD24lo CD49fhi Sca1−) and luminal progenitor enriched (CD24hi CD49flo Sca1− PR−) compartments (Supporting Information Fig. S2C). The CD61+ luminal cells are contained in this latter fraction and are considered to be precursors of mature secretory alveolar cells . A dramatic decline in the proportion of CD61+ cells was observed in the luminal fraction of MMTV-RANK compared with WT glands (Fig. 2A, 2B). The reduction in CD61 expression in MMTV-RANK luminal cells was corroborated by IF analysis of MMTV-RANK and WT virgin mammary glands (Fig. 2D). Moreover, MMTV-RANK glands contained more CD49bhi cells than WT glands (Fig. 2A, 2B). These findings demonstrate that RANK activation/overexpression disrupts the distribution of mammary populations in virgin glands.
To characterize the impaired alveolar differentiation that takes place in MMTV-RANK glands we analyzed the expression of lineage markers during pregnancy. The gestation G14.5 WT glands contained fewer CD61+ cells in the luminal fraction than G10.5 or virgin WT glands, consistent with the differentiation of CD61+ luminal cells into alveolar mature secreting cells (Supporting Information Fig. S2D). In MMTV-RANK glands, the proportion of CD61+ in the luminal compartment was low at all time points (virgin, G10.5 and G14.5), despite the accumulation of luminal cells observed during pregnancy (Supporting Information Fig. S2D). Similarly, low levels of CD61+ luminal cells were observed in aged MMTV-RANK females, either virgin or multiparous (Supporting Information Fig. S2E). This dramatic reduction in the CD61+ alveolar progenitor population in MMTV-RANK glands would explain the lack of alveologenesis observed during pregnancy in these mice even after multiple pregnancies. The percentage of CD49bhi cells within the luminal population at G10.5 and G14.5 is significantly higher in MMTV-RANK than in WT glands suggesting that CD49bhi luminal cells accumulate in MMTV-RANK glands during pregnancy (Supporting Information Fig. S2D). Altogether these results demonstrate that RANK overexpression disrupts MaSC fate leading to unbalanced mammary epithelial cell populations in virgin and pregnant glands.
Activation of RANK Pathway Promotes MaSC and Luminal Progenitor Cell Activity
Given that RANK overexpression results in an expansion of the basal compartment, we investigated the impact on the MaSC contained in this population. Mammary transplantation assays into the cleared fat pad of FVB mice were used to test the repopulating ability of FACS-isolated basal cells (CD24lo CD49fhi) from virgin WT and MMTV-RANK mice. A 2.5-fold increase in the frequency of mammary repopulating units was found in assays using MMTV-RANK basal cells compared with WT cells (estimate 1/171 vs. 1/428, respectively, p = 0.05) (Fig. 3A), indicating that MMTV-RANK glands contain more MaSC than WT glands. Next, we tested the ability of basal WT and MMTV-RANK cells to form colonies in the presence or absence of RANKL. WT and MMTV-RANK basal cells form dense colonies with different morphologies (Fig. 3C). These colonies contained K5+ and K8+ as well as K5+/K8+ cells after 15 days of culture (Fig. 3C). As freshly isolated basal cells do not express K8 (tested by RT-qPCR, data not shown), K8+ and K5+/K8+ cells must be derived from K5+ cells (Fig. 3C). K14 staining was similar to K5 in these colonies (data not shown). Frequency of colony formation was higher in MMTV-RANK basal cells as compared to WT and even higher upon exogenous RANKL stimulation (Fig. 3B). These results demonstrate that activation of RANK signaling results in an expansion of the MaSC/basal progenitor pool. To investigate the putative mechanism responsible for this expansion, we analyzed Slug and Sox9 expression in FACS-sorted basal cells, as coexpression of these genes has been shown to induce MaSCs in the mouse mammary gland . RT-qPCR analyses revealed higher expression of Slug and Sox9 mRNA in some MMTV-RANK basal cells pools as compared to WT (Fig. 3D).
Elf5 mRNA expression has been shown to be enriched in the luminal progenitor fraction relative to basal and luminal mature cells . In FACS-sorted luminal cells, a dramatic reduction in Elf5 mRNA expression is found in MMTV-RANK as compared to WT cells indicating that the luminal compartment is disrupted in MMTV-RANK glands (Fig. 3D). To investigate the functional consequences of these alterations, we analyzed the colony-forming ability of WT and MMTV-RANK luminal cells. Within the luminal population, Sca1−, CD49bhi, and CD61+ cells have all been shown to have enhanced clonogenic capacity in vitro as compared to Sca1+, CD49b−, and CD61− cells, respectively, and have therefore been identified as luminal progenitors [1, 2, 24]. Our data confirm that luminal Sca1− cells form more colonies than luminal Sca1+ cells in all comparisons (Fig. 3B). Upon RANKL stimulation, colony forming ability of MMTV-RANK Sca1− and Sca1+ luminal cells was significantly higher than the corresponding WT populations (Fig. 3B), and colonies showed a higher cellular density , (more solid colonies upon RANKL stimulation) (Fig. 3B, 3C). Colonies derived from WT and MMTV-RANK cells contained mostly K8+ and rarely K5+ cells, consistent with a luminal origin. Notably, many K8+ cells were also positive for K14, but not for K5 (Fig. 3C). Together, these results show that despite the decline in CD61+ luminal cells and disruption of the luminal compartment, MMTV-RANK luminal cells retain their clonogenic ability which is further enhanced by RANKL. These findings demonstrate that activation of RANK signaling disrupts MaSC fate in the basal and luminal compartments.
Spontaneous Preneoplastic Lesions and Heterogeneous Mammary Tumors in MMTV-RANK Mice
Our results show that RANK signaling disrupts mammary differentiation in virgin and parous glands, leading to the accumulation of MaSC, luminal and intermediate progenitors. K14+K8+ cells were found in the ducts from aged multiparous MMTV-RANK glands at frequency of 3.1%, a 2.3-fold increase when compared with the ducts of virgin young mice (Supporting Information Fig. S3B). We hypothesize that under continuous RANK stimulation, progenitors blocked in differentiation accumulate, acquire mutations, and might initiate tumorigenesis. After multiple pregnancies, 38% of elderly MMTV-RANK/FVB mice spontaneously developed high-grade invasive tumors with evidence of pulmonary metastasis (data not shown) and some mice showed multiple tumors (Fig. 4A–4C). Upon histological examination, preneoplastic lesions, including extensive hyperplasias and mammary intraepithelial neoplasias, were observed in most (90%) multiparous MMTV-RANK mammary glands and also in 25% of virgin aged MMTV-RANK mice (Fig. 4B; Supporting Information Fig. S3A). Age-matched multiparous WT mice exhibited only focal hyperplasias (25%) but no evidence of invasive tumors (Fig. 4A, 4B).
Invasive tumors (T1–T5) from MMTV-RANK mice showed a complex histological intra- and inter-tumor heterogeneity with different expression patterns of keratin and SMA staining despite all being RANK+ PR− (Fig. 4C; Supporting Information Figs. S3D, S4, S5). Consistent with the expansion of earlier progenitors observed in MMTV-RANK mice, each MMTV-RANK tumor was morphologically distinct, even in the case of tumors found in the same mouse (T2 and T3). MMTV-RANK tumors contained luminal (K8+) and/or basal (K5+, K14+, or SMA+) cells and importantly, cells coexpressing K14/K8, which were more frequent in the poorly differentiated tumors (T4 and T5). In contrast, K5+/K8+ cells were scarce in all tumors (Fig. 4C; Supporting Information Fig. S3D, S4, S5). FACS analyses revealed further heterogeneity between tumors (Fig. 4D; Supporting Information Fig. S3C, S3D). Higher expression levels of CD24 were found in tumors with high K8 expression, suggesting a luminal origin (T1 and T3), as opposed to K5+ tumors (T2). Expression of Sca1, CD61, and CD49b varied greatly between tumors and even within the same tumor (i.e., T5 contains two cell populations with different expression of CD49b and Sca1, which may be indicative of a polyclonal origin) (Fig. 4D; Supporting Information Fig. S3C, S3D). These results suggest that RANK overexpression disrupts mammary cell fate leading to extensively heterogeneous tumors that may originate from different populations of basal or luminal cells or alternatively from multipotent progenitors.
RANK Expression Increases with Age and Multiparity in WT Mice
Reproductive history and age have been linked to mammary tumorigenesis , and we found that increased levels of RANK expression promote spontaneous mammary tumors in mice. Rank mRNA expression, normalized to the epithelial content, was significantly higher in mammary glands from aged than young virgin WT (FVB) and was even higher in multiparous than in virgin glands (Fig. 5A). RANK protein is strongly expressed in spontaneous mammary secretory hyperplasias and ductal adenomas found occasionally in aged multiparous females compared with young virgin mammary tissues (Fig. 5B). Glands from aged multiparous WT mice showed a moderate increase in the frequency of basal and luminal cells as compared to younger WT and enhanced side branching, resembling the RANK-overexpressing glands (Supporting Information Fig. S6A, S6B). Sox9 and Elf5 expression levels increase with age and parities (Supporting Information Fig. S6C). While K14+K8+ cells were very rare in normal ducts of young virgin WT mice (Fig. 1C, 1E), their frequency increases in the ducts of elderly multiparous females and increases further in hyperplastic lesions (Fig. 5C, 5D). Our findings show that RANK expression increases with age and parity in the WT mammary gland, emulating some of the phenotypes observed in the MMTV-RANK glands.
RANK Expression Levels in Human Breast Adenocarcinomas Are Associated with Altered Mammary Differentiation
Previous work suggested a role for activated RANK signaling in human breast carcinogenesis; RANK expression was found in a subgroup of luminal and basal human mammary adenocarcinomas and was associated with tumor aggressiveness . Based on these observations and the results presented here, RANK expression in these tumors may be associated with alteration of cell differentiation programs. To assess this, hypothesis data from TCGA  were analyzed for the expression levels of the genes that characterize mammary differentiation hierarchy in mice and humans, including MaSC, luminal progenitors, and luminal mature cells . The primary human tumors represented in TCGA dataset were first classified into the intrinsic subtypes using the PAM50 predictor. Subsequently, the basal-like and luminal A tumors were categorized into tertiles according to the expression levels of RANK (Fig. 6). No differences in the expression of the genes encoding for ER and PR were identified between the RANK-based basal-like or luminal A subgroups. Next, the ranks of expression differences between high and low RANK-expressing tumors were analyzed using the Gene Set Enrichment Analysis (GSEA) tool. The results revealed an association between high RANK expression in human breast tumors and altered mammary differentiation for both basal-like and luminal A tumors: while the gene sets that characterize MaSC and luminal progenitors were positively correlated with high RANK expression, an opposite pattern was observed for the gene sets that characterize the luminal mature cells (Fig. 6; Supporting Information Table 1). Regarding the link with altered signaling pathways, over-representation analysis in the GSEA leading edges only found the extracellular matrix (ECM)-receptor interaction pathway as significantly associated with high RANK expression and the MaSC signature (overexpressed genes) in luminal A tumors (Fig. 6). Specific interactions between cells and the ECM are mediated by transmembrane molecules, mainly integrins or other cell-surface-associated components and lead to a direct or indirect control of multiple cellular activities. The rest of leading edges did not show differences relative to what was described previously for the specific expression signatures . Together, these results indicate that, akin to the observations made from mice studies, activated RANK signaling associates with a cell dedifferentiation profile in different subtypes of human breast cancer.
The fact that RANK overexpression and RANK loss impaired mammary gland development [13, 14] suggests that RANK signaling controls mammary epithelial cell fate (schematized in Fig. 7). Our data reveal profound effects of the activated RANK/RANKL pathway on the mouse mammary epithelial hierarchy, affecting mammary cell fate and subsequently enhancing tumor initiation and progression. It was postulated that RANK signaling was the paracrine mediator of the expansion in the MaSC-enriched compartment driven by progesterone in mice [10, 11]. As no specific markers allow the isolation of MaSC from the bulk of basal cells, the identification of MaSC relies on their ability to repopulate the mammary fat pad and their clonogenic potential. The increased repopulation frequency of basal MMTV-RANK cells and their enhanced ability to form colonies, supports that activation of RANK signaling expands the MaSC population. Increased expression of Sox9 and Slug, genes that have recently been shown to mediate MaSC function  is observed in basal MMTV-RANK cells. These results corroborate in vivo our findings in human mammary epithelial cell lines, where we recently demonstrated that RANK overexpression induces stemness .
Constitutive activation of RANK in the mouse mammary gland expands not only the basal and MaSC population, but also the luminal compartment, and has dramatic effects on the distribution of luminal subpopulations. RANK overexpression resulted in an expansion of the luminal populations CD24hi CD49flo Sca1− and CD24hi CD49flo CD49b+ but interfered with the generation of the luminal CD24hi CD49flo CD61+ that establish the alveolar lineage during pregnancy [2, 3]. MMTV-RANK luminal cells retained their clonogenic ability which has further enhanced by RANKL, despite the decrease in CD61+ luminal cells, suggesting that the colony-forming ability is likely compensated by other luminal progenitors. The significant decrease in Elf5 expression in luminal cells from virgin MMTV-RANK mice fits with the notion that Elf5 is required for luminal cell differentiation . Notably, it has been recently shown that Elf5 loss, akin RANK overexpression, induces MaSC activity and epithelial to mesenchymal transition (EMT) [17, 22, 30]. Our findings now indicate that the decline in the CD61+ alveolar progenitors contributes to the failure of alveologenesis and lactation in MMTV-RANK mice.
The hyperplastic epithelia, increased side branching and aberrant ducts observed in virgin MMTV-RANK glands is similar to that induced by progesterone treatment , suggesting that RANK expression is emulating aspects of acute progesterone stimuli. We hypothesize that RANK could contribute to the increased risk of breast cancer development associated with reproductive history and aging by accumulation of undifferentiated cells or progenitors. Several findings are consistent with this hypothesis: RANK expression increases in the WT mammary gland with age and parity; mammary ducts and hyperplastic lesions of aged multiparous WT females strongly express RANK and contain K14+/K8+ cells at higher frequencies than ducts from young females; RANK overexpression results in an expansion of K14+/K8+ cells in the virgin gland; spontaneous tumors found in aged multiparous MMTV-RANK mice contain K14+/K8+ cells, suggesting that these cells may contribute to tumorigenesis. In the aged human mammary gland similar changes, a reduction in myoepithelial cells and an increase in luminal cells expressing basal markers, resulting from an age-dependent expansion of defective multipotent progenitors, were reported  and RANKL was shown to induce the expression of basal cell markers and prevent milk production in human mammary acini . Accumulation of K14+/K8+ cells, identified as intermediate progenitors blocked in differentiation and mislocalization of K14+ cells in the luminal compartment have been reported in the ducts of mice deficient for Elf5 during pregnancy and in spontaneous tumors [22, 23]. We show that K14+K8+ cells are found in colonies derived from both basal and luminal cells, in contrast to K5+K8+ cells that are only found in basal colonies. K5+K8+ cells, in contrast to K14+K8+ cells were scarce in MMTV-RANK spontaneous. These results support that K5 and K14 mark different populations of cells , and situates K14+/K8+ cells at the edge basal/luminal differentiation.
Different breast tumor subtypes may originate from different classes of stem/progenitor cells . In fact, it has been directly demonstrated that the cell of origin of a tumor is one of the key determinants of the tumor's histological features . Inter- and intratumor heterogeneity observed in spontaneous MMTV-RANK tumors suggest that each tumor may originate from a different population of luminal and basal cells or from multipotent progenitors that later differentiate into distinct phenotypes. This tumor heterogeneity in MMTV-RANK mice contrasts with other MMTV-driven models such as MMTV-neu or MMTV-PyMT, which are highly clonal and homogenous [37-39]. Thus, RANK-induced changes in luminal, basal, and progenitor cells can cooperate with additional stochastic changes that occur in these cells to promote tumor initiation and progression.
We had previously shown that RANK is expressed in 50% of basal and 20% of luminal human breast adenocarcinomas and that RANK overexpression in human mammary cell lines induces stemness . This GSEA analyses supports the hypothesis that RANK expression in luminal A and basal human breast adenocarcinomas is associated with a poorly differentiated phenotype and may therefore result in a poorer prognosis. At least signaling cascades mediated by protein kinases are induced by RANK signaling during osteoclastogenesis and activation —inhibitor of NF-κB kinase (IKK), c-Jun N-terminal kinase (JNK), p38, extracellular signal-regulated kinase (ERK), and Src pathways . Future studies will reveal which of the downstream targets of RANK signal transduction in mammary epithelia mediate RANK-driven alterations in MaSC fate. In conclusion, this study, as represented in Figure 7, demonstrates that activated RANK signaling targets the basal and luminal compartments for expansion, impairing mammary cell fate and eventually leading to hyperplasia and tumorigenesis.
This study provides the first evidence that RANK is a master regulator of MaSC fate. Constitutive RANK expression interferes with basal and luminal cells commitment, resulting in the accumulation of MaSC, luminal and bipotent progenitors. Our data support the hypothesis that RANK driven expansion of mammary progenitors underlies RANK induced tumorigenesis and make an argument for a physiological upregulation of RANK in the mammary epithelia during aging and parity providing a mechanistic rationale for the increased risk of breast cancer with age.
We thank G. Boigues, S. Vila, and the IDIBELL animal facility for their assistance with mouse colonies; J. Climent and M.T. Soler for pathological assessment of mouse tumors; UB-SCT services for technical assistance; and O. Casanovas and L. Planelles for useful discussions. This work was supported by grants from MICINN (SAF2008-01975; SAF2011–22893), AECC (Catalunya), FMM, Concern Foundation, and a Ramon y Cajal contract to G-S, E and by a grant from the ISCIII FIS (09/02483; to P.A.M.). PP is recipient of a FPI grant from the MICINN. P.P. and A.C. contributed equally to this work.
Disclosure of Potential Conflicts of Interest
The author W.C.D. is an employee and shareholder of Amgen Inc. All other authors indicate no potential conflicts of interest.