Introduction
The retinoic acid (RA) nuclear receptor isotypes retinoic acid receptor (RAR)α, RARβ and RARγ have many overlapping as well as unique functions [
1‐
4]. The RARs belong to the steroid/thyroid hormone superfamily of ligand-dependent transcription factors [
5‐
8], bind both all-trans retinoic acid (ATRA) and its isomer, 9-cis RA, and form heterodimers with the retinoid × receptor isotypes (RXRs α-γ) [
9]. ATRA functions as a pan-agonist of all three RAR isotypes thereby playing crucial roles in embryonic morphogenesis, cell differentiation and maintenance of adult epithelia [
10,
11]. These findings together with preclinical, epidemiological and clinical observations [
12] have prompted extensive inquiries into ATRA's potential use as an anti-tumor agent. Despite its demonstrated anti-tumor activity
in vitro and in a limited number of cancer models [
13‐
21] and the highly positive response observed in acute promyelocytic leukemia patients [
22‐
24], clinical trials using ATRA as a treatment for solid tumors have produced disappointing results overall [
25‐
29].
Although the RAR isotypes display overlapping functions as evidenced by their ability to modulate common target genes [
30,
31], Husman
et al. [
32] described evidence of antagonism between RAR isotypes. Specifically, RARγ1 inhibited functions of other RAR isotypes. In addition, different RAR isotypes can transcribe the same target gene with different efficiencies, with transcription further modulated by their phosphorylation status. Moreover, interactions between isotypes are dynamic and affected by both intracellular and extracellular environments such as changes in cell signaling induced by oncogenic stress and global kinase activity [
33].
Studies of the RAR isotypes and their roles in mammary development and breast cancer provide the first clues to the unique activities that certain RAR isotypes have and suggest that certain isotype-selective retinoids may have therapeutic potential against breast cancer. It was shown that specific activation of RARα induces the expression of RARβ, which is required for normal tissue differentiation [
10,
11]. Similarly, activation of RARα also induced the expression of the cellular retinol-binding protein-1 (CRBP1), a key retinol chaperone in the cellular metabolism of retinol to ATRA, and maintained the differentiated status of the mature epithelial phenotype [
34‐
37]. In contrast, the RARγ isotype had pro-tumorigenic activity in liver cancer models [
38], and its activation stimulated breast cancer cell proliferation [
39]. On this basis, we tested whether the unique functions of these RAR isotypes could be translated into an effective approach to anti-tumor therapy. To achieve this goal we selected the synthetic retinoid Am580, which is reported to be an RARα-selective transcriptional agonist [
40] that does not activate RARγ [
40]. Previously, we showed that in the MMTV-Wnt1 and MMTV-Neu transgenic mouse models of breast cancer, in which the oncogene expression is driven by the mouse mammary tumor virus promoter (MMTV), treatment with Am580 [
40] significantly prolonged tumor-free survival and impaired tumor growth [
39]. In contrast, treatment of MMTV-Neu transgenic mice with the RAR isotype pan-agonist ATRA, which also activates RARγ, promoted tumor growth [
41]. Most importantly, these results further demonstrated the reciprocal relationship between RARα and RARγ, whereby direct inhibition of RARγ activity either by a specific RARγ antagonist or by indirect inhibition by ligand-mediated RARα activation leading to down-regulation of RARγ, allowed expression of RARα and its target genes,
RARβ and
CRBP1. Blocking RARγ while simultaneously activating RARα, strongly impinged on oncogene-induced growth pathways to attenuate the transforming potential of both
Neu and
Wnt1 oncogenes [
39,
42,
43].
This newly discovered cross-regulation of RAR isotypes, coupled with the cancer-promoting role of RARγ and anti-cancer role of RARα, prompted us to investigate their roles in the MMTV-Myc mammary cancer mouse model, in which parous females develop mammary carcinomas with 100% incidence following a latency period of several months. This model is representative of about 30% of human breast cancer cases in which
c-Myc is amplified and/or over-expressed [
44]. The
c-myc gene is often over-expressed in tumors having mutations in the
BRCA1 gene [
44]. By forming a heterodimer with Max, c-Myc transactivates several proliferation-related genes and consequently prevents Max from forming a Mad/Max heterodimer [
45,
46] that represses transcription of cell-cycle/growth arrest genes such as
p21
waf1/CIP1
,
p27
kip1
and
gadd45, and the angiogenesis inhibitor thrombospondin-1 [
47‐
49], which are also RARα regulated genes.
Here, we examined whether c-Myc over-expression affected the expression of RAR isotypes and their target genes in normal mouse mammary gland epithelial cells. The goal was to determine whether RARγ expression was enhanced and whether this increase affected c-Myc-induced tumor growth. We also evaluated whether specific activation of RARα by Am580 had anti-tumor effects in c-Myc-induced tumorigenesis.
Overall, our results bring new insights to our understanding of the effect of the c-Myc oncogene on RAR isotype expression, c-Myc/RAR isotype reciprocal relationships, and the novel tumor-promoting role of RARγ. We propose that the characterization of RAR expression in breast cancer will identify patients that would benefit from RAR isotype-selective retinoid treatment.
Materials and methods
Antibodies
Antibody sources were as follows: anti-p27 and anti-E-cadherin (BD Transduction Labs, Hoboken, NJ, USA); anti-CYP26A1 and anti-CRBP1 (Santa Cruz Biotechnology, CA); anti-RARα, -RARβ and -RARγ (Abcam, Cambridge, MA, USA), anti-Akt, anti-pAkt anti-pErk, anti-Erk and anti-pRB (Cell Signaling, Beverly, MA, USA); anti-GAPDH (Calbiochem, Gibbstown, NJ, USA); anti-tubulin (Sigma Diagnostics, St. Louis, MO, USA).
Immunoblotting
Tumor samples were mechanically homogenized in radioimmunoprecipitation assay (RIPA) buffer (1% Triton X-100, 10 mM Tris, pH 8, 140 mM NaCl, and 0.1% SDS). Primary cultures derived from wild type FVB mice mammary gland epithelium were washed in PBS, pH 7.4, and lysed with RIPA buffer. Immunoblotting was performed following standard procedures as described previously [
39].
In vivostudies
Three-month-old uniparous MMTV-Myc female mice (NCI Frederick Mouse Repository, Frederick, MD, USA) (30 mice/group) were fed with 0.3 mg/kg/day of the RARa agonist Am580 (4-[(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthyl)carboxamido]benzoic acid), which was kindly provided by Dr K Shudo (Research Foundation Itsuu Laboratory, Tokyo, Japan). Am580 was mixed into their regular diet by the vendor (Purina 5053, Richmond, IN, USA). Food consumption was measured to calculate the amount of Am580 to be added to the diet to achieve the daily dose as described previously [
39]. Regular diet was used as the control. Because the objective was to study the effect of Am580 on tumor initiation and development, mice that developed tumors within the first month were removed from the study on the assumption that their tumors had developed before treatment began. Mice were palpated twice weekly and the onset of tumor development was recorded. Once palpable, the tumor sizes were measured weekly in two dimensions and volumes calculated using the equation Vol = Dxd
2/2 (where D = major diameter and d = minor diameter). Tumor-free survival was calculated from Kaplan-Meier curves, and statistical significance was determined using the log-rank test for survival and the
t-test for tumor growth. Metastasis dissemination was evaluated by dissecting the lungs from euthanized mice and inspecting the Bouin-fixed (Sigma, St. Louis, MO, USA) lung surface for lesions using a stereoscope (Nikon SMZ800 stereoscope X3 to X5). For xenograft experiments, 8-week-old syngenic FVB mice were used (NCI Frederick Mouse Repository). Cells or tumor fragments were inoculated into the mammary fat pad of the inguinal mammary glands (gland numbers 4 and 8) under soft anesthesia and analgesia in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines.
In Vivoprotocol approval
Protocols designed and used in the in vivo experiments were approved by the Mount Sinai School of Medicine (MSSM) IACUC and conducted following its guidelines.
Immunohistochemistry
Tumor samples were fixed in 10% buffered formalin for 24 h, transferred to 70% ethanol and kept at 4ºC until use. Sections were prepared from eight tumors per group, subjected to standard antigen retrieval and incubated with primary antibody overnight at 4ºC. Sections were processed using the VectaStain ABC Elite Kit (Vector Laboratories, Burlingame, CA, USA), signals were detected using the Metal Enhanced DAB Substrate Kit (Pierce Laboratories, Rockford, IL, USA) and sections counterstained with Harris Hematoxylin Solution (Sigma Diagnostics).
Cell lines, culture conditions and plasmid transfection
After euthanasia by CO
2 overdose, tumors from untreated female MMTV-Myc mice and mammary glands from wild-type-FVB female mice were removed by dissection and minced into fragments, which were then digested with 5 ml of 1.5 mg/ml collagenase (Sigma, St. Louis, MO, USA) in PBS containing 25 mg/ml BSA (Sigma), 100 mM Ca
2+ and 100 mM Mg
2+ per approximate 500 mg of tissue at 37ºC for 30 to 45 minutes with gentle agitation. Cells were maintained in DMEM-F12 (Cellgro, Manassas, VA, USA) containing 5% (MMTV-Myc cells) or 10% (wt-FVB cells) fetal bovine serum (FBS) and 4 μg/ml insulin (Sigma Diagnostics). Immortalized nontumorigenic human MCF-10A breast epithelial cells, which were a generous gift from Dr J Brugge (Department of Cell Biology, Harvard Medical School, Boston, MA, USA), were maintained as described by Debnath
et al. [
50]. Human breast cancer MCF-7 (ER
+) and MDA-MB-231 (ER
-) cells were obtained from the American Tissue Culture Collection (ATCC, Manassas, VA, USA). Both cell lines were propagated using ATCC protocols. FVB or Myc cells were transfected with either pSG5-empty, pSG5-
RARγ, pSG5-m
CRBP1 (kindly provided by Dr Chambon, IGBMC, Strasbourg, France), pcDNA3, or pcDNA3-h-
c-Myc (Addgene, Ricci
et al. [
51]) vectors using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) in Opti-MEM (Gibco, Carlsbad, CA, USA). At 24 h after transfection, cells were treated for 16 h with 1 μM ATRA (Sigma, St. Louis, MO, USA) in dimethyl sulfoxide (DMSO) (0.01% final concentration; Sigma Diagnostics).
siRNA transfection using Lipofectamine RNAiMAX
MCF-10A, MCF-7 and MDA-MB-231 cells were plated and grown to 30 to 40% confluence at 24 h. One hour prior to transfection the medium was replaced with 1X Opti-MEM reduced serum medium (Gibco, Carlsbad, CA, USA). Anti-RARγ siRNA and a scramble control siRNA sequence (Sigma-Aldrich; seq1 #SASI_Hs01_00012455, seq2 #SASI_Hs01_00012456, Scramble seq #SIC001) were transfected according to the Invitrogen protocol using Lipofectamine RNAiMax reagent (Invitrogen catalogue number 13778-075, Grand Island, NY, USA).
Cell lines and 3D cultures
Primary Myc cell suspensions obtained after a 45-minute collagenase digestion (1.5 mg/ml collagenase and 25 mg/ml BSA in PBS plus 100 mM Ca
2+ and 100 mM Mg
2+) of MMTV-Myc tumor fragments were grown in DMEM/F12 medium supplemented with 5% horse serum, 100 ng/ml cholera toxin, 5 mg/ml insulin, 0.5 mg/ml hydrocortisone, 1% Pen/Strep, 1% glutamine (Gibco), 1% non-essential amino acids (Invitrogen) and 20 ng/ml EGF (PeproTech, Rock Hill, NJ, USA). Cells were transfected with shRARγ(2 μg) in 35-mm dishes, following the manufacturer's instructions (Open Biosystems, Huntsville, AL, USA) and seeded (3 × 10
3/well) in quadruplicate onto Matrigel
® (BD Bioscience, San Jose, CA, USA) beds in 8-well culture slides (BD Bioscience, Bedford, MA, USA) to prepare three-dimensional cultures as described by Debnath
et al. [
50]. The media was changed every 48 h for 8 consecutive days. An additional shRARγ transfection was done at day 4 to maintain
RARγ knockdown. Colony morphology was determined by phase-contrast microscopy. RARs were silenced using shRNAs from Open Biosystems (Open Biosystems, Huntsville, AL, USA) shRARα (Oligo ID: V2MM_6881) , shRARβ (Oligo ID: V2HS_239292) or shRARγ (Oligo ID: V2MM_62330). Scrambled shRNA (Open Biosystems, Huntsville, AL, USA) was used as the control.
Proliferation assay
Primary Myc cells (2 × 104) were seeded in triplicate in 6-well culture dishes and allowed to attach overnight. They were then washed with culture medium and treated with the RARγ antagonist SR11253 (2-(4-carboxyphenyl)-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthyl)-1,3-dithiolane) at increasing doses (10, 50 and 250 nM) or with DMSO (0.001% final concentration) alone, and then detached with 0.05% trypsin (Gibco) and counted every 24 h for 4 days. Statistical significance was determined by t-test. Following the same protocol described above for Myc, MCF-10A, MCF-7 and MDA-MB-231 cells were treated with 1 μM ATRA (RAR pan-agonist), 200 nM Am580 (RARα agonist), 100 nM CD437 (the RARγ/β agonist 6-[3-(1-adamantyl)-4-hydroxyphenyl]-2-naphthalenecarboxylic acid, AHPN) (Sigma), 30 nM BMS961 (the RARγ agonist 3-fluoro-4-[[2-hydroxy-2-(5,5,8,8-tetramethyl-5,6,7,8-tetrahydro-2-naphthalenyl)acetyl]amino]benzoic acid) (Tocris, Bristol, UK) and SR11253 or DMSO (0.001% final concentration) alone as the control.
Real-time PCR
Total RNA from FVB mammary gland epithelial cells was isolated using the RNeasy mini kit (Qiagen, Valencia, CA, USA) and reverse-transcribed using the iScript cDNA synthesis kit (Bio-Rad, Philadelphia, PA, USA). cDNA was amplified by real-time PCR using an iQ5 Real-Time PCR detection systems kit (Bio-Rad) and SYBR green PCR master mix (Applied Biosystems, Carlsbad, CA, USA). Primer sequences were: RARγ1-sense: 5'-TGG GGC CTG GAT CTG GTT AC-3', RARγ1-antisense: 5'-TTC ACA GGA GCT GAC CCC AT; RARγ2-sense: 5'-GCC GGG TCG CGA TGT ACG AC and RARγ2-antisense: 5'- TTC ACA GGA GCT GAC CC CAT; RARβ2-sense: 5'-ATG GAG TTC GTG GAC TTT TCT GTG-3' and RARβ2-antisense: 5'-CTC GCA GGC ACT GAC GCC AT-3'; CRBP1-sense: 5'-ACG GGT ACT GGA AGA TGC TG-3' and CRBP1-antisense: 5'-CCA TCC TGC ACG ATC TCT TT-3'; Hmq-c-Myc-sense: 5'-AGC GAC TCT GAG GAG GAA CA-3' and Hmq-c-Myc-antisense: 5'-AGT GGG CTG TGA GGA GGT TT-3'; GAPDH-sense: 5'-CGT AGA CAA AAT GGT GAA GG-3' and GAPDH-antisense: 5'-GAC TCC ACG ACA TAC TCA GC-3'.
Chromatin immunoprecipitation assay
Cells from FVB mammary gland epithelium were isolated as described above and seeded in 140-mm plates pre-coated with Matrigel
® (BD Bioscience, San Jose, CA, USA). At 70% confluency, cells were transfected with pSG5 or pSG5-
RARγ (10 μg) using Lipofectamine 2000 in Opti-MEM, 24 h after transfection, cells were treated with 1 μM ATRA for 16 h, and then crosslinked using 1% formaldehyde. Cell nuclei were isolated using hypertonic buffer A (150 mM NaCl, 50 mM Tris-HCl, pH 8, 1% NP40, 1% sodium deoxycholate, 0.5% SDS, and 2 mM EDTA), centrifuged, lysed in SDS lysis buffer (50 mM Tris, pH 8,10 mM EDTA, 1% SDS), and then sonicated. Chromatin from the nuclear fraction obtained was pre-cleared with protein G agarose/sperm salmon (Millipore, Billerica, MA, USA) and irrelevant immunoglobulin G (IgG). RARγ was immunoprecipitated overnight at 4°C, and immunoprecipitates were incubated with protein G agarose/sperm salmon for 2 h at 4°C. Immunoprecipitates were washed and eluted with elution buffer (100 mM Na
2CO
3 and 1% SDS) from agarose beads and incubated in 5 M NaCl at 65°C overnight. DNA was isolated by digestion with proteinase K, treatment with RNAse A, and purification using the QIAquick DNA purification kit (Qiagen, Valencia, CA, USA). Primers for the detection of the RA response element (RARE) sequence of
CRBP1 were designed based on Smith
et al. [
52]:
CRBP1-sense 5'-CTT GCC TAC CCT GAT GGT GT-3' and
CRBP1-antisense: 5'-CCC TTC TCA CCT GCT ACC TG-3'. As a control, nonspecific primers were designed using 9000-bp upstream of the
CRBP1 promoter; nonspecific-sense: 5'-GCA AGA CTG CTT GCT CTC CT-3' and nonspecific-antisense: 5'-AAC ACA TCG TGG GTG GTC TT-3'.
Discussion
Our data show that c-Myc over-expression in mammary epithelial cells alters the RARα -β/RARγ balance through RARγ up-regulation. The relative increase in RARγ isotype expression relative to the α and β isotypes correlates with tumor progression and lack of expression of RARα target genes involved in cell-cycle arrest and differentiation. In contrast, a shift towards RARα activation by the selective RARα agonist Am580 induces a tumor-inhibiting response in MMTV-Myc mice. Of the Am580-treated mice, approximately 63% responded to Am580 treatment by showing an anti-tumor effect with decreased tumor growth rates and lower incidence and number of lung metastases, while the rest (37%) did not respond to Am580. Only the responders displayed reduced levels of RARγ protein in their tumors, accompanied by increased expression of RARα target genes.
Collectively, several specific results obtained here identify RARγ as a pro-oncogenic RAR isotype. First, antagonism of RARγ transactivation by pharmacologic levels of its antagonist SR11253 dose-dependently halted the rapid growth of cells that over-expressed the oncogene Myc. Second, RARγ knockdown in MMTV-Myc cells, enabled RARα-selective Am580 to more effectively inhibit cell growth and to increase levels of the RARα target gene,
CRBP1. Moreover,
CRBP1 expression was higher in MMTV-Myc tumors in mice that responded to Am580 treatment. This result is important because we previously reported that re-expression of CRBP1 was associated with impaired tumor progression [
42,
43].
CRBP1 expression also decreased after normal mammary epithelial cells were ectopically transfected with RARγ, whereas specific pharmacologic activation of RARγ increased CRBP1 expression. The fact that in normal mammary epithelia RARγ-mediated repression of CRBP1 expression occurred only in the absence of its agonist, or at low agonist concentration (due to the presence of retinol in the serum), suggests that additional changes such as those induced by such oncogenes as c-Myc are required to potentiate the gene repressive function of RARγ in the presence of ligand.
In addition to the role of RARγ in Myc-mouse mammary epithelial cells, we documented its pro-oncogenic role in human breast cancer cells lines. Knockdown of RARγ expression in immortalized epithelial MCF-10A (ER
-, non-tumorigenic), MCF-7 (ER
+, tumorigenic) and MDA-MB-231 (ER
-, PR
- and HER2
-,tumorigenic) breast cancer cell lines led to their reduced proliferation. With the exception of MDA-MB-231 cells, proliferation was further reduced by treatment with RARα agonist Am580. The lack of response by MDA-MB-231 cells may most likely be due to their low levels of RARα [
59] (also see Figure
3). This reduction in cellular proliferation by the addition of Am580 implies that RARα is the tumor suppressor isotype. This conclusion was further strengthened by the experiments (Figure
5) in which selective down-regulation of RARα or RARβ in MCF-10A cells promoted the appearance of a less differentiated, more proliferative phenotype, while reduction of
RARγ expression led to cell-cycle arrest and reduced cell survival (Figure
5).
The results from the experiments using a genetic approach to RAR isotype regulation were recapitulated using synthetic RAR isotype-specific retinoids in both human and the mouse cell lines. Individually, a RARα agonist and a RARγ antagonist reduced cell growth and their combination was even more effective. The RARγ agonist significantly induced the growth of MCF7 breast cancer cells. Interestingly, MDA-MB-231 cells, which did not respond to Am580, became sensitized to Am580 by co-treatment with RARγ antagonist SR11253. The combination of Am580 with CD437/AHPN in the Myc cells gave the strongest growth inhibition; we believe that this strong response is probably related to the RAR-independent pro-apoptotic effects described for the CD437 that interestingly include the inhibition of c-Myc expression [
63‐
65]. In this regard Paroni
et al. [
66] have shown that RARα was co-amplified in approximately one third of ERBB2
+ human breast cancers. In culture, treating such cells with the ERBB2 inhibitor lapatinib combined with ATRA synergistically inhibited growth, and induced cell differentiation and apoptosis. We previously showed that in transgenic mice bearing MMTV-Neu- and -Wnt1-driven tumors [
39], and report here that in mice bearing MMTV-Myc-driven tumors and in human breast cancer cell lines,
RARγ expression is counterproductive to the anti-cancer effects of ATRA. On this basis, we propose that substituting a RARα-selective or specific agonist for the RAR pan-agonist ATRA should improve the therapeutic response.
Our findings are unique in that they describe a specific role for c-Myc-mediated differential regulation of RAR isotypes in which predominance of RARγ facilitates the pro-oncogenic phenotype. Our results are in agreement with the proliferative role of RARγ in hematopoiesis [
3] and hepatocellular carcinoma [
38]. We find that in the context of
c-Myc expression the oncogenic function of RARγ is potentiated. Myc increases
RARγ expression, leading to changes in the stoichiometry between RAR isotypes. Knockdown of RARγ in MMTV-Myc-derived cancer cells impaired tumor growth in a xenograft model and induced markers of cell-cycle arrest, differentiation and acinar-like morphogenesis in three-dimensional cultures.
Functional specificity of RAR depends on the ability of these isotypes to bind and recruit other co-repressors or co-activators [
67]. However, most studies to date have focused on the presence of RARα or the lack of RARβ. Privalsky
et al. showed that RARγ has low affinity for classical co-repressors associated with nuclear receptors, such as N-CoR or SMRT [
68], suggesting a different mechanism of repression would be induced by RARγ. In a recent paper [
69], RARγ was shown to control both the association and dissociation of the chromatin repressor Suz12 from transcriptional complexes. Suz12 belongs to the Polycomb Repressive Complex 2 and its up-regulation has been associated with cancer and stem cell compartment maintenance [
70,
71]. It is possible that oncogenic stress induces an up-regulation of both RARγ and Suz12, promoting an aberrant association between the two proteins leading to abnormal RARγ function and gene expression outcome.
Polycomb Group (PcG) proteins are known to control the recruitment of DNA methyltransferases (DNMT) to target promoters [
72] and it is possible that over-expressed RARγ bound to the
CRBP1 promoter favors this process. This repressive effect of RARγ on
CRBP1 expression may be an early response to c-Myc oncogenic transformation that reduces
CRBP1 expression, which later becomes permanent, by promoter hypermethylation [
73,
74]. In such a scenario RARα may no longer be able to compete for
CRBP1 promoter occupancy so that cells became insensitive to Am580 treatment. Although, this is an attractive and testable scenario, we can exclude the possibility that the decrease in
CRBP1 expression caused by RARγ up-regulation is due to the lack of recruitment of transcription activators to
CRBP1 promoter, suggesting that RARγ is a less efficient transactivator instead of being a strict repressor. Further analysis is required to determine how a shift in the RAR isotypes induced by c-Myc affects co-activators, co-repressors and other regulatory components of the transcriptional complexes.
As a whole, our results suggest that RARγ has an opposite role in tumorigenesis than RARα or RARβ. Oncogenic stress, which deregulates important cell signaling pathways and alters the normal function of nuclear receptors such as RARs, can induce RARγ expression and its pro-survival function inducing RARα-β/RARγ imbalance and aberrant repression of RARα-target genes. Pharmacologic activation of RARα, or inhibition of RARγ activity, reduces cancer cell growth in vitro, thereby suggesting that a specific RARα agonist would be a more effective method of cancer treatment than an RAR pan-agonist such us ATRA.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AB generated, assembled and analyzed the data on gene expression, protein levels and ChIP analysis. SPB generated and genotyped the transgenic mice required for the in vivo experiments, and collected and analyzed the in vivo data. YL conducted the in vivo treatments, collected the spontaneous tumor development data and performed the immunohistochemical analysis of the tumors. AG participated in the collection and analysis of the in vitro an in vivo data. AMJ conducted and assembled the in vitro proliferation assays. MID supplied RARγ agonist and antagonist, technical assistance in the compound used and critical editing of the manuscript. EFF conceived and designed the study, collected and assembled the data, performed data analysis and wrote the manuscript. All the authors participated in the edition of the manuscript and have read and approved the final manuscript.