agonistandantagonistofretinoicacidreceptorscausesimilar...

Agonist and Antagonist of Retinoic Acid Receptors Cause Similar

Changes in Gene Expression and Induce Senescence-like

Growth Arrest in MCF-7 Breast Carcinoma Cells

Yuhong Chen,1Milos Dokmanovic,

1Wilfred D. Stein,

1,2Robert J. Ardecky,

3and Igor B. Roninson

1

1Cancer Center, Ordway Research Institute, Albany, New York; 2Institute of Life Sciences, Hebrew University, Jerusalem, Israel; and3Ligand Pharmaceuticals, Inc., San Diego, California

Abstract

Biological effects of retinoids are mediated via retinoic acid(RA) receptors (RAR) and retinoid X receptors (RXR). Thebest-characterized mechanism of retinoid action is stimula-tion of transcription from promoters containing RA responseelements (RARE). Retinoids induce senescence-like growtharrest in MCF-7 breast carcinoma cells; this effect isassociated with the induction of several growth-inhibitorygenes. We have now found that these genes are induced byRAR-specific but not by RXR-specific ligands. Genome-scalemicroarray analysis of gene expression was used to comparethe effects of two pan-RAR ligands, one of which is a strongagonist of RARE-dependent transcription, whereas the otherinduces such transcription only weakly and antagonizes theinducing effect of RAR agonists. Both RAR ligands, however,produced very similar effects on gene expression in MCF-7cells, suggesting that RARE-dependent transcription is only aminor component of retinoid-induced changes in geneexpression. The effects of RAR ligands on gene expressionparallel changes associated with damage-induced senescence,and both ligands induced G1 arrest and the senescentphenotype in MCF-7 cells. The RAR ligands up-regulatedmany tumor-suppressive genes and down-regulated multiplegenes with oncogenic activities. Genes that are stronglyinduced by RAR ligands encode secreted bioactive proteins,including several tumor-suppressing factors. In agreementwith these observations, retinoid-treated MCF-7 cells inhibitedthe growth of retinoid-insensitive MDA-MB-231 breastcarcinoma cells in coculture. These results indicate thatRARE-independent transcriptional effects of RAR ligands leadto senescence-like growth arrest and paracrine growth-inhibitory activity in MCF-7 breast carcinoma cells. (CancerRes 2006; 66(17): 8749-61)

Introduction

Retinoids, natural and synthetic derivatives of vitamin A,regulate growth, differentiation, and survival of different types ofnormal and tumor cells. Retinoids are used in the treatment ofpromyelocytic leukemia and in chemoprevention of severalcancers, including breast carcinoma. The antitumor effect of

retinoids is most often attributed to the induction of differentia-tion, but these compounds were also shown to stop the growth oftumor cells by inducing apoptosis or accelerated senescence (1, 2).In particular, treatment of two human breast carcinoma cell lineswith all-trans retinoic acid (RA) or fenretinide, in vitro or in vivo ,induces a senescence-like phenotype characterized by increasedcell size and expression of senescence-associated h-galactosidase(SA-h-gal; refs. 3, 4). This phenotype, as investigated in MCF-7 cells,is associated with irreversible growth arrest and up-regulation ofseveral intracellular and secreted proteins with known growth-inhibitory activities. These include intracellular growth-inhibitoryproteins, such as UBD (also known as FAT10) and putative tumorsuppressor EPLIN, as well as secreted growth-inhibitory factors,including insulin-like growth factor-binding protein 3 (IGFBP3)and an extracellular matrix component, TGFBI also known ashIG-h3 (ref. 4).Induction of gene expression by retinoids is mediated at the level

of transcription through binding to dimeric transcription factorsformed by RA receptors (RAR) and retinoid X receptors (RXR). Thebest-known mechanism of action of these receptors involves theirbinding to RA response elements (RARE) in the promoters ofretinoid-responsive genes. Nevertheless, retinoid receptors alsoaffect transcription through RARE-independent mechanisms, suchas repression of transcription factor activator protein (AP-1; Jun/Fos; ref. 5), or by modulating the interaction of Sp1 and GC-richDNA via ternary complex formation (6). Remarkably, a survey ofBalmer and Blomhoff (7) concluded that only a minority of all thepublished retinoid-inducible genes are induced through the RARE-dependent mechanism. In the case of retinoid-treated MCF-7 cells,only 1 of 13 genes found to be strongly up-regulated at the onset ofsenescence-like growth arrest contained a putative RARE sequencein its promoter, whereas the other genes had no identifiable RAREsites and showed a slow kinetics of retinoid response, requiring upto 3 days for maximal induction (4). Such genes may be inducedeither by an entirely RARE-independent mechanism or as asecondary consequence of some early RARE-dependent changesin gene expression.In the present study, we have used pan-RAR– and pan-RXR–

specific agonists and antagonists to investigate the roles ofretinoid receptors in the induction of growth-inhibitory genes inMCF-7 cells. We have found that these genes are induced by theagonist of RAR (but nor RXR) and, surprisingly, by an RAR ligandthat was developed as an antagonist of RARE-dependenttranscriptional activation. Microarray analysis of gene expressionshowed that the agonist and the antagonist of RARE-dependenttranscription produced very similar effects on global geneexpression in MCF-7 cells. In agreement with these effects, bothRAR agonist and RAR antagonist induced senescence-like growtharrest in MCF-7 cells. We have also identified numerous genes with

Note: Current address for Milos Dokmanovic: Memorial Sloan-Kettering CancerCenter, Cell Biology Program, Sloan-Kettering Institute for Cancer Research, New York,NY 10021.Requests for reprints: Igor B. Roninson, Cancer Center, Ordway Research

Institute, 150 New Scotland Avenue, Albany, NY 12208. Phone: 518-641-6471; Fax: 518-641-6305; E-mail: [email protected].

I2006 American Association for Cancer Research.doi:10.1158/0008-5472.CAN-06-0581

www.aacrjournals.org 8749 Cancer Res 2006; 66: (17). September 1, 2006

Research Article

Research. on June 5, 2018. © 2006 American Association for Cancercancerres.aacrjournals.org Downloaded from

http://cancerres.aacrjournals.org/

oncogenic or tumor-suppressive properties that are affected byRAR ligands and shown that the up-regulation of secreted tumor-suppressive factors in retinoid-treated MCF-7 cells is associatedwith paracrine growth-inhibitory effect on retinoid-insensitivebreast carcinoma cells. These results indicate that retinoids inhibitMCF-7 cell growth primarily through RARE-independent effects oncellular gene expression.

Materials and Methods

Cellular assays. MCF-7 cells were obtained from American Type Culture

Collection (Manassas, VA) and cultured in DMEM with 10% fetal bovine

serum. Cells were plated at 105 per P100 ( for assays requiring up to 6 days ofculture) or at 104 per P60 ( for longer assays), in the presence of different

concentrations of all-trans-RA ( from Sigma, St. Louis, MO), pan-RAR

agonist LGD1550, pan-RXR agonist LGD1268, pan-RAR antagonist

LG100815 or pan-RXR antagonist LG101208 (Ligand Pharmaceuticals,Inc., San Diego, CA), or DMSO carrier. Following treatment, the number

of attached cells was measured using Coulter counter, and staining for

SA-h-gal was carried out as described (8). For cell cycle analysis, cells werestained with propidium iodide by standard procedures. Cellular DNA

content was determined by flow cytometry using BD LSRII fluorescence-

activated cell sorting, and the percentages of cells in G1, S, or G2-M were

determined using ModFit software.For coculture assays, MDA-MB-231 cells that were modified to express

green fluorescent protein (GFP) from lentiviral vector pLL3.7 (9) were plated

either alone (at 2 � 105 per P100) or mixed with MCF-7 (at 105 cells from

each cell line). In some assays, MCF-7 cells pretreated with 100 nmol/L RAfor 8 days were collected by trypsinization before mixing with MDA-MB-

231. After culture in the presence or in the absence of 100 nmol/L RA, total

cell number was counted and the percentage of MDA-MB-231 GFP cells was

determined by flow cytometry.Gene expression analysis. MCF-7 cells were plated at 5 � 105 per P100;

exposed to different drugs or carrier on the next day; and cells were

collected after 24, 48, or 72 hours treatment. Total cellular RNA was isolatedusing the RNeasy kit (Qiagen, Valencia, CA). Gene expression levels of

IGFBP3, EPLIN, UBD, TGFBI, and TRIM31 were quantitated by reverse

transcription-PCR (RT-PCR) as described (4) and by quantitative real-time

RT-PCR (QPCR), using an ABI 7900HT real-time PCR instrument. In QPCR,serial cDNA dilutions were used for primer validation and the comparative

CTmethod for relative quantitation of gene expression (Applied Biosystems,

Foster City, CA) was used to determine expression levels for target genes.

h-Actin was used as a normalization standard. Primer sequences will beprovided upon request. For gene expression profiling, RNA samples were

provided to the Microarray Core Facility at the Genomics Institute of the

NYSDOH Wadsworth Center, which carried out biotinylated targetpreparation (using 2 Ag RNA per assay) and hybridization with AffymetrixU133 Plus 2.0 microarrays. Data analysis was carried out using GeneSpring

software (Agilent, Palo Alto, CA). Gene function analysis was carried out

using Pathway Assist (Ariadne Genomics, Rockville, MD) and PubMed.RARE-dependent transcription was analyzed using a plasmid construct

that expresses firefly luciferase from a RARE-containing artificial promoter

DR5 (Stratagene, La Jolla, CA). Cells were plated at 3 � 105 per P60 24 hours

before transient transfection. DR5 reporter plasmid (4 Ag) was mixed withthe SV40-driven Renilla luciferase control plasmid (0.04 Ag) and transfectedusing LipofectAMINE Plus (Invitrogen, Carlsbad, CA) as described by the

manufacturer. Three hours after transfection, cells were rinsed thrice withPBS, trypsinized, and replated in a 12-well plate to the density of 5 � 104 per

well. Retinoid agonists or antagonists were added 48 hours later, and the

luciferase assay was done after another 24 hours.

Results

RAR agonists and an antagonist induce the expression ofgrowth-inhibitory genes. To determine which classes of retinoidreceptors mediate the induction of growth-inhibitory genes in

retinoid-treated MCF-7 cells, we have analyzed the effects of pan-RAR– and pan-RXR–specific agonists and antagonists on these cells.The tested compounds include LGD1550, a pan-RAR agonist thatshows greater selectivity for RAR than natural retinoids (10);LGD1268 (a pan-RXR agonist); LG100815 (a pan-RAR antagonist);and LG101208 (a pan-RXR antagonist). We have also used all-trans-RA, a natural RAR agonist that was originally used to define retinoid-induced senescence and changes in gene expression in MCF-7 cells(3, 4). The structures of these compounds are shown in Fig. 1.In the first set of experiments, we asked how these compounds

affect the expression of several growth-inhibitory genes previouslyfound to be strongly up-regulated under the conditions of retinoid-induced senescence of MCF-7 cells. The tested genes includedEPLIN-b, UBD (also known as FAT10), IGFBP3 , and TGFBI (alsoknown as bIG-h3), as well as TRIM31 , a gene strongly induced byRA that has a putative RARE element in its promoter (4). RNA wasextracted from MCF-7 cells that were either untreated or treatedwith individual compounds or their combinations for 3 days, theperiod previously shown to be required for the maximal effect ofRA or fenretinide (4). Gene expression was analyzed both bysemiquantitative RT-PCR (not shown) and by QPCR; both assaysproduced essentially the same results. The outcome of arepresentative set of QPCR assays is shown in Table 1.In agreement with the previous study, all five of the tested genes

were induced by 100 nmol/L RA, the concentration found to besufficient for maximal induction of gene expression (4), withTRIM31 showing greater fold induction than the other four genes.The pan-RAR agonist LGD1550 induced all five genes as strongly asRA; a 100 nmol/L concentration of this agonist was also sufficientfor the maximal effect. In contrast, the pan-RXR agonist LGD1268showed detectable induction of only two of the five genes (TGFBIand TRIM31), and the extent of induction was much lower thanobserved for RA or the pan-RAR agonist. The pan-RXR antagonistLG101208 showed no effect on the expression of any of the fivegenes (Table 1). When combined with 100 nmol/L RA, 10 Amol/L ofthe pan-RXR antagonist inhibited the induction of only one ofthese genes (IGBP3) by RA (Table 1). These findings suggest thatretinoid-inducible expression of growth-inhibitory genes is acti-vated primarily through RAR rather than RXR.Surprising results were obtained, however, with the pan-RAR

antagonist LG100815. This compound acts as a competitiveinhibitor of RAR agonists, which binds to RAR but does notefficiently activate RARE-dependent transcription (11).4 To verifythe antagonistic effect of LG100815 on RARE-dependent transcrip-tion, we analyzed the effects of this compound and of RAR agonistsRA and LGD1550 on the expression of firefly luciferase reportertranscribed from a RARE-containing artificial promoter, DR5.Figure 2 shows the results of DR5-luciferase transient transfectionassays, carried out in the presence of LGD1550, RA, and LG100815,alone or in pairwise combinations. One hundred nanomoles perliter concentrations of RA or LGD1550 agonists activated theRARE-containing promoterf50-fold, whereas 10 Amol/L LG100815antagonist (the concentration used in the literature for maximaleffect) produced an order of magnitude weaker (4.2-fold) induction.On the other hand, the addition of LG100815 to RAR agonists RA orLGD1550 diminished the induction of transcription by the lattercompounds 2.5 to 3 times (Fig. 2). These results confirm thatLG100815 is relatively inefficient in stimulating RARE-dependent

4 W. Lamph, personal communication.

Cancer Research

Cancer Res 2006; 66: (17). September 1, 2006 8750 www.aacrjournals.org



transcription and that it antagonizes the effect of RAR agonists onsuch transcription.In contrast to the results of the promoter assays, QPCR analysis

showed that LG100815 antagonist induced RNA expression of allthe tested endogenous genes to a level comparable with RAR

agonists (Table 1). The fold induction of these genes by 1 or10 Amol/L LG100815 was only slightly lower than the highestinduction by either RA or LGD1550 agonist, with the only largedifference (3- to 4-fold) observed for TRIM31. When 10 Amol/LLG100815 was combined with RA, it decreased the induction of

Figure 1. Structures of retinoid receptorligands used in the present study. Structurerepresentations in two-dimensional andthree-dimensional formats were generatedusing ChemDraw Pro version 10.0 andChem3D Ultra version 10.0, respectively.

Table 1. QPCR analysis of the effects of retinoid receptor ligands on gene expression

Compounds Fold induction

IGFBP3 EPLIN UBD TGFBI TRIM31

RA (100 nmol/L) 2.5 F 1.36 3.1 F 1.38 8.5 F 1.62 6.5 F 1.27 34.4 F 1.12LG1550 (100 nmol/L) 4.2 F 1.34 4.2 F 1.21 9.5 F 1.02 5.1 F 1.13 42.7 F 1.04

LG1550 (1 Amol/L) 3.7 F 1.17 3.2 F 1.1 7.7 F 1.15 3.5 F 1.11 30.3 F 1.04

LG1550 (10 Amol/L) 2.8 F 1.29 4.9 F 1.55 7.3 F 1.16 5.6 F 1.02 45.0 F 1.04

LGD100815 (1 Amol/L) 2.1 F 1.3 3.8 F 1.5 4.6 F 1.14 4.3 F 1.2 7.7 F 1.62LGD100815 (10 Amol/L) 2.0 F 1.28 3.2 F 1.61 2.8 F 1.24 4.4 F 1.09 11.7 F 1.4

RA (100 nmol/L) + LGD100815 (1 Amol/L) 2.4 F 1.35 1.7 F 1.3 8.0 F 1.22 6.4 F 1.38 30.3 F 1.12

RA (100 nmol/L) + LGD100815 (10 Amol/L) 1.9 F 1.31 0.9 F 1.63 2.2 F 1.06 3.7 F 1.28 10.0 F 1.56

LG1268 (100 nmol/L) 1.2 F 1.26 1.2 F 1.13 1.0 F 1.16 1.9 F 1.72 1.3 F 1.38LG1268 (1 Amol/L) 0.9 F 1.23 1.0 F 1.14 0.8 F 1.31 2.2 F 1.57 1.4 F 1.56

LG1268 (10 Amol/L) 1.2 F 1.21 1.4 F 1.4 1.0 F 1.25 2.4 F 1.12 3.8 F 1.21

LGD101208 (10 Amol/L) 1.0 F 1.24 0.8 F 1.53 1.0 F 1.33 0.9 F 1.47 1.0 F 1.3RA (100 nmol/L) + LGD101208 (10 Amol/L) 1.1 F 1.46 2.9 F 1.25 7.8 F 1.13 5.7 F 1.23 28.6 F 1.73

RAR Ligand Effects on Gene Expression and Cell Growth




gene expression by RA to the levels that were similar to thoseobserved with the antagonist alone, as expected for a competitiveeffect (Table 1). Hence, LG100815, although acting as a weakinducer and a competitive antagonist of RARE-dependent tran-scription, mimics the effects of RAR agonists in stimulating theexpression of the tested growth-inhibitory genes. This surprisingobservation prompted us to compare the effects of the RAR agonistand the RAR antagonist on the expression of other genes and on

the phenotype of MCF-7 cells. In most of the subsequent studies,we used 100 nmol/L LGD1550 agonist and 10 Amol/L LG100815antagonist, the concentrations that produced the maximal effectsin the above-described assays.RAR agonist and antagonist produce similar effects on

genome-scale gene expression. To analyze the effects of RARligands on the expression of essentially all the cellular genes, wetreated MCF-7 cells with RAR agonist LGD1550 (100 nmol/L) orRAR antagonist LG100815 (10 Amol/L) for 24, 48, or 72 hours, andanalyzed RNA from the untreated or treated cells by hybridizationwith Affymetrix U133 Plus 2.0 oligonucleotide arrays, containing56,000 probe sets representing 48,500 human transcripts. Thetime course of changes in gene expression observed in themicroarrays (Fig. 3A) was in agreement with the results of RT-PCR assays (not shown) and with previous studies on RA- orfenretinide-treated MCF-7 cells (4), with most of the responsivegenes showing the response on day 1 with subsequent increasesup to day 3. All 13 genes shown by RT-PCR to be induced by RAor fenretinide (4) also showed induction by the pan-RAR agonistand antagonist (Fig. 3B).Strikingly, the effects of the RAR agonist and the RAR antagonist

were exceedingly similar. Seventy-four percent of the genesshowing z1.5-fold induction and 77% of the genes showing z1.5-fold inhibition by the antagonist were also induced or inhibited,respectively, at least 1.3-fold by the agonist, and vice versa (77% and69%, respectively). Figure 4A plots (on the log scale) the maximalchanges in gene expression (at any time point) produced by the

Figure 2. Effects of RAR agonists (RA and LGD1550, 100 nmol/L each) andRAR antagonist LG100815 (10 Amol/L) on luciferase expression from DR5RARE-containing promoter in MCF-7 cells. The assays were carried out asdescribed in Materials and Methods, in triplicate.

Figure 3. Microarray analysis of changesin gene expression in MCF-7 cells treatedwith RAR agonist LGD1550 or RARantagonist LG100815, plotted usingGeneSpring software. X axis, differenttime points of treatment with RAR ligands(0 point correspond to cells cultured for3 days with DMSO carrier). Y axis,changes in gene expression on log scale.The groups of genes shown in (A ), (G ),and (H ) represent GO categories, with theexclusion of genes showing raw signalintensity <10 in MCF-7 cells. The selectionof the other gene groups is describedin the text.

Cancer Research




agonist versus those produced by the antagonist for 11,729 probesets that showed >1.3-fold changes in gene expression aftertreatment with either the agonist or the antagonist. The effectsof the agonist and antagonist on gene expression show highlysignificant correlation. The regression through all the data pointshas an R2 value of 0.6955 with n = 11,729 (Student’s t test yieldsP < 0.0001). The regression line has a slope of 1.027 F 0.006,indicating that the RAR agonist and the RAR antagonist have thesame effect on the majority of the responsive genes. The similarityof the overall effects of the agonist and the antagonist on geneexpression agrees with the results of QPCR analysis of selectedgenes (Table 1), but contrasts with an order of magnitude weakereffect of the antagonist on RARE-dependent transcription (Fig. 1).On the other hand, 316 probe sets showing the strongest

(>5-fold) induction or inhibition by RAR ligands were significantlymore responsive to the agonist than to the antagonist (at P <0.0001), with the regression slope increasing to 1.297 F 0.038 (thistranslates tof2-fold stronger average effect of the agonist relativeto the antagonist; Fig. 4B). We considered whether preferentialinduction of the most responsive genes by the agonist couldindicate the presence of RARE sequences in the correspondingpromoters. Indeed, the gene showing the strongest induction bythe agonist (Fig. 4B) encodes RA-metabolizing enzyme CYP26A1(induced 220-fold by the agonist and 23.5-fold by the antagonist),which was reported to contain two synergistically acting RARE

sequences in its promoter (12). Another cytochrome P450 involvedin RA metabolism, CYP26B1, showed the most selective response tothe RAR agonist among all the genes (180-fold induction by theagonist, compared with only 2.4-fold induction by the antagonist).We have also looked at the effects of the agonist and the antagoniston the expression of 40 human genes, identified by Balmer andBlomhoff (13) as containing canonical and evolutionarily conservedRARE sequences in their promoters. Only 7 of these 40 genes wereinduced z1.5-fold in MCF-7 cells by the RAR agonist and just threegenes were induced by the antagonist. RARE-containing genesinduced by both ligands showed 3- to 5-fold stronger response tothe agonist than to the antagonist (Fig. 3C). The most responsivegene in this group is HOXA1 , induced 95-fold by the agonist butonly 18-fold by the antagonist. However, when we examinedpromoter sequences of 10 randomly chosen genes that showpreferential induction by the agonist for the presence of putativeRARE sequences (this analysis was carried out using MatInspectorprogram as previously described; ref. 4), only 1 of 10 promoters wasfound to contain putative RARE sequences. We also examined 28randomly selected genes from the group identified by Balmer andBlomhoff (7) as indirectly inducible by retinoids (‘‘category 0’’) andfound that these genes showed a similar response to the agonistand the antagonist (Fig. 3D and data not shown). Altogether,these results indicate that the majority of genes induced by theRAR ligands are induced through indirect, RARE-independentmechanisms.RAR agonist and antagonist induce senescence-associated

changes in gene expression and cellular phenotype. MCF-7cells treated with 100 nmol/L RA eventually (after z4 days)undergo cytostatic growth arrest, which is accompanied by the lossof clonogenic potential upon removal of the drug and thedevelopment of the senescent phenotype (3, 4). We comparedchanges in gene expression induced in MCF-7 cells by LGD1550and LG100815 with the results obtained in a well-characterizedsystem of drug-induced senescence of tumor cells. In that system,HCT116 colon carcinoma cells were transiently exposed todoxorubicin, the surviving cells were separated into proliferatingand senescent populations after drug treatment, and differentiallyexpressed genes were identified using a cDNA microarray (14). Wehave recently repeated this analysis using Affymetrix U133 Plus 2.0array.5 We now asked how treatment of MCF-7 cells with RARligands affects the expression of genes found in the latter analysisto be strongly (>5-fold) induced or inhibited in senescent HCT116cells. As shown in Fig. 3E and F, most of these genes changed theirexpression in the same direction in MCF-7 cells treated with RARagonist or antagonist. Specifically, 88% of 231 genes down-regulated in senescent HCT116 cells and expressed in MCF-7 cellswere down-regulated by the RAR agonist, and only 3% were up-regulated. Among 353 genes up-regulated in senescent HCT116cells and expressed in MCF-7, 53% were up-regulated by theagonist and 10% were down-regulated. Similar results wereobtained with the RAR antagonist. This analysis indicatesprofound similarities between the effects of retinoids anddoxorubicin-induced senescence on gene expression. On the otherhand, there were also notable differences in gene expressionbetween RAR ligand-treated MCF-7 cells and HCT116 cellsundergoing doxorubicin-induced senescence. These differencesmay be attributed to a large extent to the fact that only the latter

Figure 4. Comparison of changes in gene expression produced by RAR agonistand antagonist. The maximal changes in gene expression for 11,729 probe setsrepresenting genes that show >1.3-fold (A) or >5-fold (B) effect by either theagonist or the antagonist (dots ) are plotted on a log scale. Trend lines representpower regression. Gene names for selected probe sets strongly affected byeither ligand are shown in (B).

5 Y. Chen and I.B. Roninson, in preparation.





but not the former up-regulated cyclin-dependent kinase (CDK)inhibitor p21Waf1 (CDKN1A), a key regulator of gene expression insenescent cells (see Discussion).The results of gene expression studies suggested that RAR

agonist and antagonist (but not RXR ligands) are likely to inducesenescence-like growth arrest in MCF-7 cells. To test this, wetreated MCF-7 cells with RAR and RXR agonists and antagonists for7 days and analyzed the effects of the treatment on the cell number(Fig. 5A), cell cycle distribution (Fig. 5B), and the expression of the

SA-h-gal marker of senescence (Fig. 5C, D ; ref. 8). The RXR agonistLGD1268 did not inhibit MCF-7 cell growth (Fig. 5A) and did notinduce the senescent phenotype (Fig. 5D); in fact, LGD1268treatment produced a moderate but reproducible increase in cellgrowth (Fig. 5A). The RXR antagonist LG101208 had no effecton either the cell growth (Fig. 5A) or the senescent phenotype(Fig. 5D). In contrast, the RAR agonist LGD1550 and the RARantagonist LG100815 inhibited cell growth to the extent similar tothat of RA (Fig. 5A). Cell cycle analysis showed that RA, LGD1550,and LG100815 all increased the G1 and decreased the S fraction,indicating cell cycle arrest in G1 (Fig. 5B), in agreement withprevious observations on RA-treated MCF-7 cells (15). LGD1550and LG100815 increased SA-h-gal activity (Fig. 5C), as has alsobeen shown for RA (3), and this effect of all three RAR ligands wasquantitatively similar (Fig. 5D). Hence, RAR agonists and theantagonist induce senescence-like growth arrest in MCF-7 cellswith a similar efficiency.Effects of RAR ligands on tumor suppressor and oncogene

expression and paracrine growth-inhibitory activity of reti-noid-treated MCF-7 cells. The most prominent Gene Ontologycategories of genes that were largely inhibited by RAR ligands aregenes involved in DNA replication (Fig. 3G) or mitosis (Fig. 3H).Strong down-regulation of such genes has been associated with cellcycle arrest induced by chemotherapeutic drugs. In contrast,inhibition of these genes by RAR ligands was moderate (f2-fold),with none of the genes showing z5-fold inhibition. On the otherhand, we noticed that many of the genes that show the greatestfold inhibition by RAR ligands are known or putative oncogenes.(Here, we define genes as putative oncogenes or tumor suppressorsif such genes have been reported in the literature to play afunctional role in tumor cell growth, survival, tumorigenesis, ormetastasis, as determined by targeted inhibition, gene over-expression or protein addition studies.) A total of 26 oncogeneswere found to be inhibited by RAR ligands, with the strongestinhibition found for VAV3, SPDEF (also known as PDEF), AMIGO2,MYB, RET, C4.4A, and MAFB (Table 2). On the other hand, we alsoidentified a smaller number (10) of genes with reported tumor-suppressive activities that were inhibited by RAR ligands (Table 2),most notably caveolin proteins CAV1 and CAV2, as well as cellularRA-binding protein 2 (CRABP2), a retinoid-binding proteinreported to sensitize MCF-7 cells to growth inhibition by RA (16).We have also identified 22 known or putative oncogenes and 34tumor suppressors as up-regulated by RAR ligands (Table 3). Theproducts of these genes include both secreted factors (see below)and cell-associated proteins. The most highly induced cell-associated growth inhibitors were the previously identifiedretinoid-inducible genes UBD, EPLIN (4), and SOX9 (17), followed

Figure 5. Effects of retinoid agonists and antagonists on MCF-7 cell growth andthe senescent phenotype. A, cell number after 7 days of culture with the additionof DMSO (control), 100 nmol/L RA, 100 nmol/L RAR agonist LGD1550,100 nmol/L RXR agonist LGD1268, 10 Amol/L RAR antagonist LG100815, and10 Amol/L RXR antagonist LG101208. Experiments were done in triplicate,and the results are expressed relative to the average of the control. B, changesin cell cycle distribution in untreated cells or in cells treated with 100 nmol/LRA, 100 nmol/L LGD1550 RAR agonist, 10 Amol/L LG100815 RAR antagonist,or DMSO carrier (untreated). For cells treated with RAR ligands, 0 pointrepresents cells cultured for 2 days with DMSO carrier. The analysis was carriedout as described in Materials and Methods. C, staining of cells that weretreated for 8 days with DMSO carrier (control), 100 nmol/L LGD1550, or10 Amol/L LG100815 for senescence marker SA-h-gal. Photographed at �100magnification. D, percentages of SA-h-gal+ cells after 8 days of treatmentwith the indicated compounds (in triplicate), at the same concentrations as in (A ).

Cancer Research




Table 2. Oncogenes and tumor suppressors inhibited by RAR ligands

Gene name Affymetrix

probe no.

Genbank Description Basal raw

signal

Max fold inhibition

by agonist

Max fold inhibition

by antagonist

Oncogenes

Cell -associatedVAV3 218807_at NM_006113 Vav 3 oncogene 1,181 14.5 2.7

SPDEF 220192_x_at NM_012391 SAM pointed domain containing

ets transcription factor

522 14.2 3.0

AMIGO2 222108_at AC004010 Adhesion molecule with

immunoglobulin-like domain 2

1,051 11.7 4.1

MYB 204798_at NM_005375 V-myb myeloblastosis viral

oncogene homologue

716 9.9 2.2

RET 211421_s_at M31213 Ret proto-oncogene 350 8.5 1.8

C4.4A 204952_at NM_014400 GPI-anchored metastasis-

associated protein

homologue

190 8.1 3.2

MAFB 218559_s_at NM_005461 V-maf musculoaponeurotic

fibrosarcoma oncogene

homologue B

774 7.5 2.1

FGFR3 204379_s_at NM_000142 Fibroblast growth factorreceptor 3

173 5.0 2.4

GREB1 205862_at NM_014668 GREB1 protein 285 4.8 2.8

MSX2 210319_x_at D89377 Msh homeo box homologue 2 395 4.0 3.2PDLIM2 219165_at NM_021630 PDZ and LIM domain 2 (mystique) 136 4.0 2.2

ENTPD5 205757_at NM_001249 Ectonucleoside triphosphate

diphosphohydrolase 5

51 3.4 3.0

MALAT1 224559_at AF001540 Metastasis-associated lungadenocarcinoma transcript 1

200 2.5 3.4

PBX3 204082_at NM_006195 Pre-B-cell leukemia

transcription factor 3

311 3.4 2.5

CCNA2 203418_at NM_001237 Cyclin A2 415 3.3 2.3G6PD 202275_at NM_000402 Glucose-6-phosphate

dehydrogenase

1,380 2.2 3.2

HMMR 207165_at NM_012485 Hyaluronan-mediatedmotility receptor (RHAMM)

540 2.9 2.1

CCND1 208711_s_at BC000076 Cyclin D1 695 2.1 2.9

ECT2 219787_s_at NM_018098 Epithelial cell transforming

sequence 2 oncogene

545 2.8 1.6

PTTG1 203554_x_at NM_004219 Pituitary tumor-transforming 1 1,236 2.7 2.2

MYBL1 213906_at AW592266 V-myb myeloblastosis viral

oncogene homologue-like 1

387 2.6 2.0

IRS1 204686_at NM_005544 Insulin receptor substrate 1 2,117 2.6 1.6PIK3R1 212240_s_at AI679268 Phosphoinositide-3-kinase,

regulatory subunit 1 (p85a)1,206 2.0 2.1

SecretedCXCL12 209687_at U19495 Chemokine (C-X-C motif) ligand 12

(stromal cell-derived factor 1)

596 5.7 1.6

TFF3 204623_at NM_003226 Trefoil factor 3 (intestinal) 387 4.0 3.0BMP7 209591_s_at M60316 Bone morphogenetic protein 7

(osteogenic protein 1)1,008 1.8 2.0

Tumor suppressors (cell associated)

CAV1 212097_at AU147399 Caveolin 1, caveolae protein, 22 kDa 516 21.8 4.3

GPR30 210640_s_at U63917 G protein-coupled receptor 30 163 7.0 2.6CRABP2 202575_at NM_001878 Cellular RA binding protein 2 4354 6.1 3.5

CAV2 203324_s_at NM_001233 Caveolin 2 643 5.0 2.8

PHLDA2 209803_s_at AF001294 Pleckstrin homology-like domain,

family A, member 2

351 4.4 2.7

BARD1 205345_at NM_000465 BRCA1-associated RING domain 1 543 4.1 2.3

NAT1 214440_at NM_000662 N-acetyltransferase 1

(arylamine N-acetyltransferase)

1484 4.1 2.2

BLM 205733_at NM_000057 Bloom syndrome 122 3.3 2.0

SFN 33322_i_at X57348 Stratifin 2,066 3.1 1.9TOB1 228834_at BF240286 Transducer of ERBB2, 1 3,287 2.9 2.1





Table 3. Oncogenes and tumor suppressors induced by RAR ligands


probe no.


signal

Max fold induction

by agonist

Max fold induction

by antagonist

Oncogenes

Cell-associatedHOXA1 214639_s_at S79910 Homeobox A1 8 95.3 18.3

RGC32 218723_s_at NM_014059 Response gene to complement 32 9 49.6 34.0

ATF3 202672_s_at NM_001674 Activating transcriptionfactor 3

24 6.9 3.3

CXCR4 217028_at AJ224869 Homo sapiens CXCR4 gene

encoding receptor CXCR4.

213 5.5 2.5

ELF3 210827_s_at U73844 E74-like factor 3 (ets domaintranscription factor,

epithelial-specific)

318 4.9 2.9

SIX1 228347_at N79004 Sine oculis homeobox

homologue 1

179 4.6 2.3

MITF 226066_at AL117653 Microphthalmia-associated

transcription factor

27 4.1 3.2

FGFR2 203638_s_at NM_022969 Fibroblast growth factor

receptor 2

57 3.3 4.1

BHLHB2 201170_s_at NM_003670 Basic helix-loop-helix domain

containing, class B, 2

95 3.5 2.1

MDM2 217373_x_at AJ276888 Mdm2, transformed 3T3 celldouble minute 2, p53

binding protein

39 3.1 1.9

LYN 202625_at AI356412 V-yes-1 Yamaguchi sarcoma viral

related oncogene homologue

77 2.8 3.0

JUN 201464_x_at BG491844 V-jun sarcoma virus 17

oncogene homologue

190 2.6 2.6

S100P 204351_at NM_005980 S100 calcium binding protein P 523 2.3 2.1

JAG1 209099_x_at U73936 Jagged 1 111 2.2 1.9Secreted

IL-8 202859_x_at NM_000584 Interleukin-8 10 14.4 8.2

PLA2G2A 203649_s_at NM_000300 Phospholipase A2, group IIA 30 7.4 1.9EDN1 222802_at J05008 Endothelin-1 22 4.1 1.7

CLU 208792_s_at M25915 Clusterin (complement lysis inhibitor,

SP-40,40, sulfated glycoprotein 2,

testosterone-repressed prostatemessage 2, apolipoprotein J)

723 3.4 2.7

PDGFC 218718_at NM_016205 Platelet-derived growth factor C 11 3.4 2.6

FGF13 205110_s_at NM_004114 Fibroblast growth factor 13 186 2.9 2.2

TGFA 205016_at NM_003236 Transforming growth factor, a 61 2.7 1.5PSAP 200866_s_at M32221 Prosaposin 674 1.9 1.8

Tumor suppressors

Cell-associatedUBD 205890_s_at NM_006398 Ubiquitin D 36 35.9 17.5

SOX9 202935_s_at AI382146 SRY (sex-determining region Y)-box 9 50 26.0 10.8

EPLIN 217892_s_at NM_016357 Epithelial protein lost in neoplasm h 128 18.8 7.4

CEACAM1 206576_s_at NM_001712 Carcinoembryonic antigen-relatedcell adhesion molecule 1

15 16.2 2.8

PPARG 208510_s_at NM_015869 Peroxisome proliferative

activated receptor, g64 15.9 4.0

MARCKS 201669_s_at NM_002356 Myristoylated alanine-rich proteinkinase C substrate

64 9.3 4.6

BTG2 201236_s_at NM_006763 BTG family, member 2 87 8.3 6.5

NKX3-1 209706_at AF247704 NK3 transcription factor-related, locus 1 35 7.5 2.9

IRF1 202531_at NM_002198 IFN regulatory factor 1 43 5.3 2.4SOD2 215223_s_at W46388 Superoxide dismutase 2, mitochondrial 327 4.9 4.8

NBL1 37005_at D28124 Neuroblastoma, suppression of

tumorigenicity 1

61 4.0 3.0

(Continued on the following page)

Cancer Research




by CEACAM1, PPARG, MARCKS, BTG2, and NKX3-1. The moststrongly induced oncogenes were the RARE-regulated transcriptionfactor HOXA1 and RGC32, a positive regulator of the cell cycle. Thetime course of the induction or inhibition of cell-associatedoncogenes and tumor suppressors listed in Table 2 is shown in Fig.3I and J, respectively.Many genes up-regulated by RAR ligands encode secreted

proteins. Some of these proteins have cancer-relevant activities,stimulating or inhibiting cell growth, survival, invasion, orangiogenesis. The induced secreted proteins with tumor-promoting(oncogenic) or tumor-suppressing activities are listed in Table 3,whereas the inhibited secreted proteins are listed in Table 2. Figure3K and L show the effects of RAR ligands on the expression of thecorresponding genes. The genes for secreted proteins showing thestrongest induction include four tumor-suppressing factors(TGFBI, IGFBP3, FBLN5, and GDF15) and two tumor-promotingproteins [interleukin-8 (IL-8) and PLAG2A].Because retinoid-treated MCF-7 cells up-regulate genes for

secreted factors with different activities, we carried out a

functional test to determine whether such cells produce primarilypromitogenic or antimitogenic paracrine effects. In this assay, wemixed MCF-7 cells 1:1 with MDA-MB-231 breast carcinoma cells(insensitive to retinoids). The latter cells had been transducedwith GFP, allowing us to distinguish and quantitate the two celllines by flow cytometry. The cocultures were treated for 5 dayswith 100 nmol/L RA (used because of limited availability ofLGD1550 and LG100815). RA or MCF-7 cells alone had no effecton the growth of MDA-MB-231 cells. In contrast, retinoidtreatment in coculture with MCF-7 decreased the number ofMDA-MB-231 cells by f25% (Fig. 6A). In another type ofexperiment, shown in Fig. 6B , MCF-7 cells were pretreated for8 days with RA to allow for complete growth arrest anddevelopment of the senescent phenotype. The treated MCF-7cells were collected by trypsinization and cocultured for 3 dayswith MDA-MB-231, in the presence or in the absence of RA.Coculture with RA-pretreated MCF-7 cells was sufficient toinhibit MDA-MB-231 cell growth by f30%, compared withcoculture with untreated MCF-7 cells (Fig. 6B). The addition of

Table 3. Oncogenes and tumor suppressors induced by RAR ligands (Cont’d)


probe no.


signal

Max fold induction

by agonist

Max fold induction

by antagonist

GADD45G 204121_at NM_006705 Growth arrest and DNA-damage-

inducible, g37 3.8 2.7

PHLDA1 217996_at AA576961 Pleckstrin homology-like domain,

family A, member 1

359 3.7 2.7

HIPK2 225368_at BF218115 Homeodomain interactingprotein kinase 2

114 3.7 2.8

CDKN2B 236313_at AW444761 Cyclin-dependent kinase inhibitor 2B

(p15, inhibits CDK4)

14 3.2 3.1

BATF 205965_at NM_006399 Basic leucine zipper transcriptionfactor, ATF-like

213 2.7 2.9

VHL 1559227_s_at BF972755 Von Hippel-Lindau tumor

suppressor

25 2.0 2.7

EI24 216396_s_at AF131850 Etoposide induced 2.4 mRNA 147 2.2 2.6KLF6 208961_s_at AB017493 Kruppel-like factor 6 25 2.2 2.5

BTG1 200921_s_at NM_001731 B-cell translocation gene 1,

antiproliferative

509 2.5 1.8

DDIT3 209383_at BC003637 DNA-damage-inducible transcript 3 131 2.5 2.2FOXO3A 224891_at AV725666 Forkhead box O3A 1172 2.3 1.8

GSN 200696_s_at NM_000177 Gelsolin 435 1.7 1.9

PDCD4 212593_s_at N92498 Programmed cell death 4(neoplastic transformation inhibitor)

1433 1.7 1.6

Secreted

TGFBI 201506_at NM_000358 Transforming growth factor,

h-induced, 68 kDa127 53.0 12.8

IGFBP3 210095_s_at M31159 Insulin-like growth factor

binding protein 3

15 32.2 12.3

FBLN5 203088_at NM_006329 Fibulin 5 7 10.1 19.0

GDF15 221577_x_at AF003934 Growth differentiation factor 15 378 7.2 3.4TGM2 211003_x_at BC003551 Transglutaminase 2 (C polypeptide,

protein-glutamine-g-glutamyltransferase)

17 5.8 2.9

SULF1 212354_at BE500977 Sulfatase 1 126 2.9 2.9

TGFB2 228121_at AU145950 Transforming growth factor, h2 20 2.0 2.7

IGFBP6 203851_at NM_002178 Insulin-like growth factor

binding protein 6

81 2.6 2.1

PRSS8 202525_at NM_002773 Protease, serine, 8 (prostasin) 186 2.4 1.9

PRSS11 201185_at NM_002775 Protease, serine, 11 (IGF binding) 50 1.9 1.5





RA to the coculture had no significant effect on this growthinhibition (Fig. 6B), indicating that MDA-MB-231 cell growth wasinhibited not by the retinoid but by factors secreted by retinoid-treated MCF-7 cells.

Discussion

In the present study, we have used pan-RAR– or pan-RXR–specific agonists and antagonists to investigate the mechanism ofchanges in gene expression associated with senescence-like growtharrest, which is induced by retinoids in MCF-7 breast carcinomacells. Biological effects of retinoids are commonly attributed to theeffects of ligand-bound retinoid receptors on transcription fromRARE-containing promoters. However, the majority of genesinduced by retinoids in different systems (7), including MCF-7cells treated with RA or fenretinide (4), do not contain RAREelements in their promoters. Nevertheless, RARE-mediated induc-tion of transcription could still be the primary response toretinoids, triggering a chain of events leading to indirect changes inthe expression of other genes. The results of the present studysuggest, however, that RARE-dependent transcription plays only aminor role in the effects of retinoids on gene expression in MCF-7cells, and that changes in gene expression responsible forsenescence-like growth arrest are due primarily to RARE-indepen-dent transcriptional effects.These conclusions are based on the following arguments. (a) The

limited role of RARE promoter sequences in determining transcrip-tional effects of retinoids is indicated by the findings that only asmall fraction of the genes induced by RAR ligands contain RAREelements in their promoter, and that the majority of RARE-containing genes examined are not affected by the RAR ligands. (b)RARE-dependent transcription is also unlikely to be the initial effectresponsible for subsequent global changes in gene expression, asindicated by the comparison of the effects of RAR agonist LGD1550and RAR antagonist LG100815. The RAR agonist was an order ofmagnitude more efficient than the antagonist in stimulating aRARE-containing promoter (at concentrations producing themaximal effects), and it induced the responsive genes known tocontain functional RARE sequences 3 to 10 times stronger than theantagonist. One would expect therefore that genes that are inducedas a consequence of RARE-dependent early events should also showpreferential response to the agonist, even if such genes do not

contain RARE elements. Indeed, we have identified a set of genes,primarily among the strongest responders, which are preferentiallyinduced or inhibited by the agonist relative to the antagonist. Suchgenes, however, were a small minority (e.g., 1,616 genes were up-regulated >2-fold by either ligand, but only 261 of these genesshowed z2-fold stronger response to the agonist than to the anta-gonist), and the average effects of the agonist and the antagonist onthe bulk of the responsive genes were essentially the same (Fig. 4A).(c) Nevertheless, one could still argue that RARE-dependent trans-cription, which is preferentially induced by the agonist, could be thekey determinant of the effect of retinoids on the growth of MCF-7cells, whereas RARE-independent changes in global gene expressioncould be epiphenomena irrelevant to the biological response.However, the RAR antagonist LG100815 was just as efficient as theRAR agonists (RA and LGD1550) in inducing cell growth inhibition,G1 arrest, and the senescent phenotype (Fig. 5).A number of RARE-independent mechanisms of regulation of

gene expression by retinoids have been described in the literature.Not all of these mechanisms are transcriptional; for example,retinoids were suggested to affect RNA stability (18). We haveanalyzed mRNA stability of IGFBP3, TGFBI, UBD, and EPLINduring 24-hour treatment with actinomycin D and found thesemessages to be as stable as h-actin mRNA, with no detectableeffect of RA on their stability (19). Hence, the effect of retinoidson these genes is more likely to be exerted at the level oftranscription. In the case of IGFBP3, the induction of transcrip-tion by RA in MCF-7 cells has been shown by nuclear run-onassays (20). Some examples of RARE-independent transcriptionregulatory mechanisms include the binding of RAR/RXR–basedtranscription factor complexes to cis-acting sequences distinctfrom RARE (21) or interactions between retinoid receptors andother transcription factors, such as Sp1 (6, 22) or AP-1 (5, 23).Furthermore, retinoids can regulate the activity of protein kinaseC (PKC; ref. 24), and RA-activated PKC was shown to bind to RARand stimulate its transcriptional activity (25). The effects of RARligands on gene expression, observed in the present study, mostlikely reflect many different RARE-independent transcriptionaleffects of retinoids rather than any single mechanism.In light of our findings, it would seem more appropriate to

describe compounds such as LG100815 not as RAR antagonistsbut rather as a novel type of RAR modulators that are poorinducers of RARE-dependent transcription and that act primarily

Figure 6. Effects of coculture with RA-treated MCF-7cells on MDA-MB-231 cell growth. A, MDA-MB-231 cells(GFP-expressing) were plated either alone or in 1:1 mixturewith MCF-7 cells, in the presence of 100 nmol/L RA orDMSO carrier. Columns, mean MDA-MB-231 cell numberafter 5 days of culture relative to cell number in the absenceof MCF-7 or RA, calculated from three independentexperiments; bars, SD. B, MDA-MB-231 cells (GFP-expressing) were plated as 1:1 mixtures with MCF-7 cellsthat were either untreated or treated for 8 days with100 nmol/L, and grown in the presence of DMSO carrier or100 nmol/L RA. Columns, mean MDA-MB-231 cell numberrelative to MDA-MB-231 cell number in coculture withuntreated MCF-7 without RA, calculated from threeindependent experiments; bars, SD.

Cancer Research




by stimulating RARE-independent transcriptional effects. Not allthe RAR antagonists seem to belong to this class. For example, aRAR-a–specific antagonist LG100629 was shown to block essen-tially completely the induction of IGFBP3 and TGFB2 by RA inhuman bronchial epithelial cells (26), in contrast to our findingswith pan-RAR modulator LG100815. On the other hand, severalRAR antagonists, similarly to LG100815, were found to inhibittumor cell growth (27–30). It would be of interest to determine ifthese compounds act as LG100815-type RAR modulators.Genome-scale microarray analysis of the effects of RAR ligands

has confirmed and expanded the results of our previous analysis ofthe effects of RA in MCF-7 cells, carried out using a smallermicroarray (4). The principal conclusion of the prior study was thatretinoid-induced arrest of MCF-7 cells was associated withconcerted induction of several growth-inhibitory genes encodingboth cell-associated (UBD and EPLIN) and secreted proteins(TGFBI and IGFBP3). The present analysis showed that RARligands LGD1550 and LG100815 also strongly induced these genes,and it revealed many additional tumor-suppressive genes inducedby retinoids (e.g., SOX9, CEACAM1, MARCKS, GDF15, FBLN5,BTG2, NKX3-1, NBL1, and IRF1). Importantly, we discovered in thepresent study that RAR ligands not only induce tumor suppressorsbut also inhibit multiple oncogenes or proto-oncogenes expressedin MCF-7 cells, with the strongest effects observed for VAV3,SPDEF, AMIGO2, MYB, RET, and C4.4A. We have investigatedthe effects of overexpressing several retinoid-inducible growth-inhibitory genes, including UBD (4), EPLIN , and TGFBI , as well asIGFBP3 protein product (19), on MCF-7 cells, and found that eachof these genes produced only weak growth inhibition and did notinduce the senescent phenotype. In light of these findings, wehypothesize that retinoids induce senescence-like growth arrest inMCF-7 cells through a cumulative effect of multiple changes in theexpression of different growth-regulatory genes.Many of the genes found in the present study to be affected to a

similar extent by RAR agonist and antagonist (i.e., RARE-independent genes) have been previously reported to be retinoidresponsive in MCF-7 and other breast carcinoma cell lines. Ofspecial interest, these include tumor suppressor genes proposed tomediate the antiproliferative effects of retinoids, such as IGFBP3,induced by RA in MCF-7 and Hs578T lines (31), as well as SOX9 andPDCD4, induced by RAR-specific but not by RXR-specific agonistsin MCF-7 and T-47D cells (17, 32). Other genes identified by ourmicroarray analysis have been implicated specifically in thephenotype of breast carcinoma cells, including MCF-7. Amongthe genes that we found to be inhibited by retinoids, ets trans-cription factor SPDEF (PDEF) was shown to be a key mediator ofmotility and invasion in MCF-7 and other breast carcinoma celllines (33), PDLIM2 (Mystique) is a negative regulator of anchorage-independent growth and migration in MCF-7 cells (34), and GREB1was reported to mediate the proliferative response of MCF-7 cellsto estrogen (35). Among the genes found here to be induced byRAR modulators, an isoform of adhesion molecule CEACAM1inhibits cell growth and restores normal morphology of MCF-7cells (36), and IRF1 was shown to mediate p53-independent tumorsuppression and apoptosis in MCF-7 and other breast carcinomacell lines (37). Growth inhibition and apoptosis in retinoid-treatedMCF-7 cells have been previously associated with the drasticallyoverexpressed IGFBP3 (38) and with the retinoid-binding proteinCRABP2 (39), which, paradoxically, we find here to be stronglydown-regulated rather than induced by RAR ligands. Down-regulation of CRABP2 gene expression upon retinoid treatment

could potentially represent a negative regulatory mechanism thatlimits retinoid-induced apoptosis. Another surprising observationis the inhibition of caveolin proteins CAV1 and CAV2. Caveolinswere shown to act as tumor suppressors in the mammary glandbut as oncogenes in other tissue contexts, such as the prostate (40).CAV1 expression in MCF-7 cells was found to inhibit anchorage-independent growth but at the same time protect cells fromapoptosis (41). Altogether, despite some changes to the contrary,the overall effect of RAR modulators on the expression ofoncogenes and tumor suppressors in MCF-7 cells seems to beindicative of the reversal of the neoplastic phenotype.Several retinoid-induced tumor suppressors (BTG2, BTG1,

GDF15, EPLIN, and CEACAM1) are also induced by DNA-damagingdrugs (such as doxorubicin) and remain constitutively up-regulatedin HCT116 colon carcinoma cells that become permanently growtharrested and develop the senescent phenotype after exposure todoxorubicin (14). Aside from these genes, we have found a strikingoverlap between the large groups of genes that are induced orinhibited in doxorubicin-induced senescence of HCT116 cells andin retinoid-induced senescence of MCF-7 cells (Fig. 3K and L). Onthe other hand, an important difference between retinoid-treatedMCF-7 cells and most of the characterized systems of DNAdamage–induced senescence is the lack of induction of thedamage-responsive CDK inhibitor p21Waf1 (CDKN1A), which wasfound here and in previous studies (15) to be moderately down-regulated in MCF-7 cells after retinoid treatment. In fact, the onlyCDK inhibitor that we found to be up-regulated in retinoid-treatedMCF-7 cells was p15Ink4b (CDKN2B), which was recently identifiedas a marker of oncogene-induced senescence (42). In contrast, p21is drastically induced in MCF-7 cells after doxorubicin treatment,concordantly with the development of the senescent phenotype(43). p21 induction leads to rapid inhibition of genes involved inmitosis or DNA replication (44), and p21 knockout prevents theshutdown of such genes in doxorubicin-treated cells (14). Althoughgenes involved in mitosis and DNA replication are clearly inhibitedin RAR ligand–treated MCF-7 cells (Fig. 2F and G), their inhibitionis not as drastic as in DNA-damaged cells (14), possibly due to thelow levels of p21. The lack of p21 induction is also likely to accountfor our observations that a number of genes that are induced inresponse to p21 and implicated in tumor promotion or thedevelopment of age-related diseases, such as LGALS3, APP, or SHC1(44), were not induced in RAR modulator–treated MCF-7 cells.Induction of senescence-like permanent growth arrest could be

one of the most desirable treatment responses in tumor cells,because it occurs at low drug concentrations that produce littlesystemic toxicity, and because senescent cells not only fail to growbut also secrete proteins with paracrine growth-inhibitory effects.On the other hand, senescence induced by treatment with p21-inducing DNA-damaging agents is also associated with overpro-duction of proteins with the opposite, tumor-promoting activities,which stimulate the growth or survival of the neighboringnonsenescent cells (45). In our previous study on retinoid-inducedsenescence of MCF-7 cells (4), we observed overproduction ofsecreted growth-inhibitory proteins (TGFBI and IGFBP3). We havenow found that MCF-7 cells treated with RAR ligands up-regulatenot only additional tumor-suppressing proteins (GDF15 andFBLN5) but also tumor-promoting factors (IL-8 and PLAG2A),although tumor-suppressing proteins show overall greater induc-tion, at least at the RNA level (Fig. 3K and L). The results of ourcoculture studies with retinoid-sensitive and retinoid-insensitivebreast carcinoma cell lines (Fig. 6) show that retinoid-treated





MCF-7 cells indeed secrete tumor-suppressing factors that producea moderate but reproducible decrease in the growth of neighboringretinoid-insensitive MDA-MB-231 cells. Exploitation of this para-crine tumor-suppressive effect of retinoid-treated tumor cells offersa new insight into the clinical applications of retinoids.Despite the shown efficacy of retinoids in PML treatment

and in cancer chemoprevention trials, retinoid treatment iscomplicated by systemic toxic responses, such as intracranialhypertension and headaches, dyspnea, hypertriglyceridemia, hyper-calcemia, and hyperleukocytosis (46). An experimental approachproposed to avoid these side effects is the use of RXR-selectiveagonists that show lower toxicity than the conventional RARagonists (47). However, RXR agonists may not be effective against allthe types of transformed mammary epithelial cells, as indicatedin the present study by the observation that LGD1268 did not inhibitbut rather stimulated MCF-7 cell growth (Fig. 5A). As an alternativeapproach, RAR antagonists have been developed to prevent or treatretinoid toxicity (48). RAR antagonists were reported to be very welltolerated even at very high doses (49) and in some cases producedin vivo tumor-suppressive effects on their own (49, 50). The resultsof the present study suggest an explanation for these observations.

LG100815-type RAR modulators are inefficient in inducing themajor effect of conventional retinoids (RARE-dependent trans-activation), which provides the likeliest explanation for theirnontoxicity. On the other hand, LG100815 mimics the effects ofconventional RAR agonists in inducing senescence-like growtharrest of tumor cells and up-regulating the expression of genesfor tumor-suppressing secreted factors. This novel type of RARmodulators may be viewed therefore as a prototype of a potentiallyinteresting new class of drugs that should provide a hightherapeutic ratio for cancer treatment or chemoprevention.

Acknowledgments

Received 2/14/2006; revised 5/22/2006; accepted 6/28/2006.Grant support: NIH grants R01 CA62099 and R01 AG17921 (I.B. Roninson).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 accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.We thank Dr. William Lamph (Ligand Pharmaceuticals, Inc., San Diego, CA), for

providing RAR and RXR ligands used in this study and for helpful advice; Dr. ChangLim for assistance with flow cytometry; Drs. Eugenia Broude, Errin Lagow, andCharitha Madiraju for helpful discussions; and the late Dr. Robin Pietropaolo and theMicroarray Core Facility at the Genomics Institute of the New York State Departmentof Health Wadsworth Center, who carried out microarray hybridization.

References

1. Altucci L, Gronemeyer H. The promise of retinoids tofight against cancer. Nat Rev Cancer 2001;1:181–93.2. Roninson IB, Dokmanovic M. Induction of senescence-associated growth inhibitors in the tumor-suppressivefunction of retinoids. J Cell Biochem 2003;88:83–94.3. Chang BD, Broude EV, Dokmanovic M, et al. Asenescence-like phenotype distinguishes tumor cellsthat undergo terminal proliferation arrest after exposureto anticancer agents. Cancer Res 1999;59:3761–7.4. Dokmanovic M, Chang BD, Fang J, Roninson IB.Retinoid-induced growth arrest of breast carcinomacells involves co-induction of multiple growth-inhibitorygenes. Cancer Biol Ther 2002;1:24–7.5. Pfahl M. Nuclear receptor/AP-1 interaction. EndocrRev 1993;14:651–8.6. Husmann M, Dragneva Y, Romahn E, Jehnichen P.Nuclear receptors modulate the interaction of Sp1 andGC-rich DNA via ternary complex formation. Biochem J2000;352 Pt 3:763–72.7. Balmer JE, Blomhoff R. Gene expression regulation byretinoic acid. J Lipid Res 2002;43:1773–808.8. Dimri GP, Lee X, Basile G, et al. A biomarker thatidentifies senescent human cells in culture and in agingskin in vivo . Proc Natl Acad Sci U S A 1995;92:9363–7.9. Rubinson DA, Dillon CP, Kwiatkowski AV, et al. Alentivirus-based system to functionally silence genes inprimary mammalian cells, stem cells and transgenicmice by RNA interference. Nat Genet 2003;33:401–6.10. Zhang L, Nadzan AM, Heyman RA, et al. Discovery ofnovel retinoic acid receptor agonists having potentantiproliferative activity in cervical cancer cells. J MedChem 1996;39:2659–63.11. Lee HY, Sueoka N, Hong WK, et al. All-trans -retinoicacid inhibits Jun N-terminal kinase by increasing dual-specificity phosphatase activity. Mol Cell Biol 1999;19:1973–80.12. Loudig O, Maclean GA, Dore NL, Luu L, Petkovich M.Transcriptional co-operativity between distant retinoicacid response elements in regulation of Cyp26A1inducibility. Biochem J 2005;392:241–8.13. Balmer JE, Blomhoff R. A robust characterization ofretinoic acid response elements based on a comparisonof sites in three species. J Steroid Biochem Mol Biol2005;96:347–54.14. Chang BD, Swift ME, Shen M, et al. Moleculardeterminants of terminal growth arrest induced intumor cells by a chemotherapeutic drug. Proc Natl AcadSci U S A 2002;99:389–94.

15. Zhu WY, Jones CS, Kiss A, et al. Retinoic acidinhibition of cell cycle progression in MCF-7 humanbreast cancer cells. Exp Cell Res 1997;234:293–9.16. Budhu AS, Noy N. Direct channeling of retinoic acidbetween cellular retinoic acid-binding protein II andretinoic acid receptor sensitizes mammary carcinomacells to retinoic acid-induced growth arrest. Mol CellBiol 2002;22:2632–41.17. Afonja O, Raaka BM, Huang A, et al. RAR agonistsstimulate SOX9 gene expression in breast cancer celllines: evidence for a role in retinoid-mediated growthinhibition. Oncogene 2002;21:7850–60.18. von Marschall Z, Riecken EO, Rosewicz S. Inductionof matrix metalloprotease-1 gene expression by retinoicacid in the human pancreatic tumour cell line Dan-G. BrJ Cancer 1999;80:935–9.19. Dokmanovic M. Growth inhibitory targets ofretinoids. Ph.D. thesis, University of Illinois atChicago; 2003.20. Shang Y, Baumrucker CR, Green MH. Signal relay byretinoic acid receptors a and h in the retinoic acid-induced expression of insulin-like growth factor-bindingprotein-3 in breast cancer cells. J Biol Chem 1999;274:18005–10.21. Wang J, Yen A. A novel retinoic acid-responsiveelement regulates retinoic acid-induced BLR1 expres-sion. Mol Cell Biol 2004;24:2423–43.22. Shimada J, Suzuki Y, Kim SJ, et al. Transactivation viaRAR/RXR-Sp1 interaction: characterization of bindingbetween Sp1 and GC box motif. Mol Endocrinol 2001;15:1677–92.23. Kamei Y, Xu L, Heinzel T, et al. A CBP integratorcomplex mediates transcriptional activation and AP-1inhibition by nuclear receptors. Cell 1996;85:403–14.24. Imam A, Hoyos B, Swenson C, et al. Retinoids asligands and coactivators of protein kinase Ca. FASEB J2001;15:28–30.25. Kambhampati S, Li Y, Verma A, et al. Activation ofprotein kinase Cy by all-trans -retinoic acid. J Biol Chem2003;278:32544–51.26. Han GR, Dohi DF, Lee HY, et al. All-trans -retinoicacid increases transforming growth factor-h2 andinsulin-like growth factor binding protein-3 expressionthrough a retinoic acid receptor-a-dependent signalingpathway. J Biol Chem 1997;272:13711–6.27. Toma S, Isnardi L, Raffo P, et al. RARa antagonist Ro41-5253 inhibits proliferation and induces apoptosis inbreast-cancer cell lines. Int J Cancer 1998;78:86–94.28. Yang L, Munoz-Medellin D, Kim HT, et al. Retinoicacid receptor antagonist BMS453 inhibits the growth of

normal and malignant breast cells without activatingRAR-dependent gene expression. Breast Cancer ResTreat 1999;56:277–91.29. Hammond LA, Van Krinks CH, Durham J, et al.Antagonists of retinoic acid receptors (RARs) are potentgrowth inhibitors of prostate carcinoma cells. Br JCancer 2001;85:453–62.30. Keedwell RG, ZhaoY, Hammond LA, et al. Anantagonist of retinoic acid receptors more effectivelyinhibits growth of human prostate cancer cells thannormal prostate epithelium. Br J Cancer 2004;91:580–8.31. Colston KW, Perks CM, Xie SP, Holly JM. Growthinhibition of both MCF-7 and Hs578T human breastcancer cell lines by vitamin D analogues is associatedwith increased expression of insulin-like growthfactor binding protein-3. J Mol Endocrinol 1998;20:157–62.32. Afonja O, Juste D, Das S, Matsuhashi S, Samuels HH.Induction of PDCD4 tumor suppressor gene expressionby RAR agonists, antiestrogen and HER-2/neu antago-nist in breast cancer cells. Evidence for a role inapoptosis. Oncogene 2004;23:8135–45.33. Gunawardane RN, Sgroi DC, Wrobel CN, et al. Novelrole for PDEF in epithelial cell migration and invasion.Cancer Res 2005;65:11572–80.34. Loughran G, Healy NC, Kiely PA, et al. Mystique is anew insulin-like growth factor-I-regulated PDZ-LIMdomain protein that promotes cell attachment andmigration and suppresses Anchorage-independentgrowth. Mol Biol Cell 2005;16:1811–22.35. Rae JM, Johnson MD, Scheys JO, et al. GREB 1 is acritical regulator of hormone dependent breast cancergrowth. Breast Cancer Res Treat 2005;92:141–9.36. Kirshner J, Chen CJ, Liu P, Huang J, Shively JE.CEACAM1-4S, a cell-cell adhesion molecule, mediatesapoptosis and reverts mammary carcinoma cells to anormal morphogenic phenotype in a 3D culture. ProcNatl Acad Sci U S A 2003;100:521–6.37. Bouker KB, Skaar TC, Riggins RB, et al. Interferonregulatory factor-1 (IRF-1) exhibits tumor suppressoractivities in breast cancer associated with caspaseactivation and induction of apoptosis. Carcinogenesis2005;26:1527–35.38. Gucev ZS, Oh Y, Kelley KM, Rosenfeld RG. Insulin-likegrowth factor binding protein 3 mediates retinoic acid-and transforming growth factor h2-induced growthinhibition in human breast cancer cells. Cancer Res1996;56:1545–50.39. Donato LJ, Noy N. Suppression of mammarycarcinoma growth by retinoic acid: proapoptotic genes

Cancer Research




are targets for retinoic acid receptor and cellularretinoic acid-binding protein II signaling. Cancer Res2005;65:8193–9.40. Williams TM, Lisanti MP. Caveolin-1 in oncogenictransformation, cancer, and metastasis. Am J PhysiolCell Physiol 2005;288:C494–506.41. Ravid D, Maor S, Werner H, Liscovitch M. Caveolin-1inhibits cell detachment-induced p53 activation andanoikis by upregulation of insulin-like growth factor-Ireceptors and signaling. Oncogene 2005;24:1338–47.42. Collado M, Gil J, Efeyan A, et al. Tumour biology:senescence in premalignant tumours. Nature 2005;436:642.

43. Elmore LW, Di X, Dumur C, Holt SE, Gewirtz DA.Evasion of a single-step, chemotherapy-induced senes-cence in breast cancer cells: implications for treatmentresponse. Clin Cancer Res 2005;11:2637–43.44. Chang BD, Watanabe K, Broude EV, et al. Effects ofp21Waf1/Cip1/Sdi1 on cellular gene expression: implica-tions for carcinogenesis, senescence, and age-relateddiseases. Proc Natl Acad Sci U S A 2000;97:4291–6.45. Roninson IB. Tumor cell senescence in cancertreatment. Cancer Res 2003;63:2705–15.46. Warrell RP. Differentiation agents 2001;:489–94.47. Wu K, Zhang Y, Xu XC, et al. The retinoid X receptor-selective retinoid, LGD1069, prevents the development

of estrogen receptor-negative mammary tumors intransgenic mice. Cancer Res 2002;62:6376–80.48. Standeven AM, Johnson AT, Escobar M, ChandraratnaRA. Specific antagonist of retinoid toxicity in mice.Toxicol Appl Pharmacol 1996;138:169–75.49. Toma S, Emionite L, Scaramuccia A, Ravera G,Scarabelli L. Retinoids and human breast cancer:in vivo effects of an antagonist for RAR-a. Cancer Lett2005;219:27–31.50. Fanjul AN, Piedrafita FJ, Al Shamma H, Pfahl M.Apoptosis induction and potent antiestrogen receptor-negative breast cancer activity in vivo by a retinoidantagonist. Cancer Res 1998;58:4607–10.





2006;66:8749-8761. Cancer Res Yuhong Chen, Milos Dokmanovic, Wilfred D. Stein, et al. CellsSenescence-like Growth Arrest in MCF-7 Breast CarcinomaSimilar Changes in Gene Expression and Induce Agonist and Antagonist of Retinoic Acid Receptors Cause

Updated version

http://cancerres.aacrjournals.org/content/66/17/8749

Access the most recent version of this article at:

Cited articles

http://cancerres.aacrjournals.org/content/66/17/8749.full#ref-list-1

This article cites 46 articles, 22 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/66/17/8749.full#related-urls

This article has been cited by 7 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://cancerres.aacrjournals.org/content/66/17/8749To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/content/66/17/8749.full#ref-list-1

http://cancerres.aacrjournals.org/content/66/17/8749.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]



agonistandantagonistofretinoicacidreceptorscausesimilar...

Documents