MOLECULAR AND CELLULAR BIOLOGY, Apr. 1993, p. 2258-22680270-7306/93/042258-11$02.00/0

Inhibition of Estrogen-Responsive Gene Activation by theRetinoid X Receptor ,B: Evidence for Multiple

Inhibitory PathwaysJAMES H. SEGARS,1 MICHAEL S. MARKS,' STEVEN HIRSCHFELD,' PAUL H. DRIGGERS,1

ERNEST MARTINEZ,2 JOSEPH F. GRIPPO,3 WALTER WAHLI,2 AND KEIKO OZATOl*

Laboratory ofMolecular Growth Regulation, National Institute of Child Health and Human Development,Bethesda, Maryland 208921; Institut de Biologie Animale, University of Lausanne, CH-1015 Lausanne,

Switzerland2; and Drug Metabolism, Hoffman-La Roche, Nutley, New Jersey 071103

Received 26 August 1992/Returned for modification 16 November 1992/Accepted 31 December 1992

The retinoid X receptor 13 (RXRI; H-2RIIBP) forms heterodimers with various nuclear hormone receptorsand binds multiple hormone response elements, including the estrogen response element (ERE). In this report,we show that endogenous RXR3 contributes to ERE binding activity in nuclear extracts of the human breastcancer cell line MCF-7. To define a possible regulatory role of RXRf3 regarding estrogen-responsivetranscription in breast cancer cells, RXRI3 and a reporter gene driven by the vitellogenin A2 ERE were

transfected into estrogen-treated MCF-7 cells. RXR13 inhibited ERE-driven reporter activity in a dose-dependent and element-specific fashion. This inhibition occurred in the absence of the RXR ligand 9-cis retinoicacid. The RXRB-induced inhibition was specific for estrogen receptor (ER)-mediated ERE activation becauseinhibition was observed in ER-negative MDA-MB-231 cells only following transfection of the estrogen-activatedER. No inhibition of the basal reporter activity was observed. The inhibition was not caused by simplecompetition of RXRI3 with the ER for ERE binding, since deletion mutants retaining DNA binding activity butlacking the N-terminal or C-terminal domain failed to inhibit reporter activity. In addition, cross-linkingstudies indicated the presence of an auxiliary nuclear factor present in MCF-7 cells that contributed to RXRI3binding of the ERE. Studies using known heterodimerization partners of RXR13 confirmed that RXRP/triiodothyronine receptor ai heterodimers avidly bind the ERE but revealed the existence of anothertriiodothyronine-independent pathway of ERE inhibition. These results indicate that estrogen-responsive genesmay be negatively regulated by RXRI3 through two distinct pathways.

Estrogen-mediated gene regulation plays a fundamentalrole in development of female sex organs and in reproductiveprocesses in vertebrates. In addition, estrogen has beenimplicated in control of gene expression unrelated to sexualdifferentiation and reproduction (12, 25, 33). Activation ofestrogen-responsive genes occurs via interaction of theestrogen receptor (ER) with 17-beta estradiol (E2) and sub-sequent receptor binding to estrogen-responsive elements(EREs) of target genes. The ER, a member of the nuclearreceptor superfamily, is composed of three distinct modulardomains with distinct functions (19, 20, 41, 43, 58, 75).Analysis of in Xenopus and chicken vitellogenin genes as

well as other estrogen-responsive genes have shown thatEREs contain a GGTCA motif, often in palindromic repeats(38, 55, 77). Despite the presence of the GGTCA motif inother hormone response elements, DNA sequence require-ments for functional ER binding have been found to berather stringent (42, 54, 59, 61, 65).

Reports suggest that control of estrogen-responsive generegulation is complex (for a review, see reference 32 andreferences therein). Several lines of evidence indicate thatgene regulation mediated by estrogen and EREs involves notonly the ER but also other transcription factors and DNA-binding proteins. Involvement of oncogene products such as

c-jun/c-fos (17, 26), NF-1 (11, 54), and H-ras (63) in E2/ERE-mediated gene regulation has been documented, althoughthese factors do not bind to EREs. In addition, it has been

* Corresponding author.

shown that some cells lacking the ER nevertheless contain a

nuclear factor(s) that can bind to EREs (21) and that some ofthese factors appear capable of affecting transcription fromERE-containing promoters (37). Further, other members ofthe nuclear hormone receptor family such as the thyroidhormone (triiodothyronine [T3]) receptor (T3R [28]) andCOUP-TF (1) have been shown to bind EREs in addition totheir cognate response elements. Despite their ability to bindto EREs, a functional role for these non-ER proteins inestrogen-regulated cells has remained elusive.We have previously reported isolation of a cDNA for the

retinoid X receptor 1B (mRXR,B; H-2RIIBP), a member of thenuclear hormone receptor superfamily which binds to theregion II enhancer element of major histocompatibility com-plex (MHC) class I genes and other elements containing theGGTCA motif (35). This motif is common to other hormoneresponse elements such as the thyroid response element andretinoic acid (RA) response element (3, 19, 54, 61, 76). Wehave shown that RXR1 binds a number of these hormoneresponse elements, including EREs (52). Two other mem-

bers of the RXR subfamily, RXRao and RXR-y, have subse-quently been cloned (49). All RXRs bind an isoform ofall-trans RA, 9-cis RA (36, 47), but can control transcriptionof target genes independently of RA receptors (RARs) (49,62). Further, we (51) and others (8, 40, 46, 80, 81) haveshown that RXRs form heterodimeric complexes with othernuclear hormone receptors, including the T3R, RAR, andvitamin D receptor (VDR). These studies have suggestedthat RXR heterodimers are more stable than the RXRhomodimer and bind to various hormone response elements

2258

Vol. 13, No. 4

INHIBITION OF E2-RESPONSIVE GENES BY RXRI 2259

with much greater affinities than do the homodimers of thecognate receptors. Despite the demonstration of ERE bind-ing by RXRO, the function of RXR,B with respect to ER-mediated gene activation has not yet been determined.

In this report, we have addressed the role of mRXR,B intranscriptional regulation of an estrogen-responsive gene inthe human breast cancer cell line MCF-7. This line of inquiryis supported by the observation that retinoids may functionas antiestrogens in breast cancer cells (4, 22). MCF-7 cellswere chosen for study since the cells permit a direct analysisof the action of mRXRI with respect to function of theendogenous ER (13, 15, 38, 68, 72). We show here that theendogenous RXR, expressed in MCF-7 cells contributes toERE binding activity, independently of the ER. Moreover,we show that transient overexpression of RXR,B results inspecific inhibition of ERE-driven reporter activity. Deletionanalysis of RXR, indicates that the mechanism of inhibitionis more complex than a direct interference of ER binding tothe target ERE by RXRI. Furthermore, inhibition of ERE-driven promoter activity by RXRI was observed to occurvia two pathways: (i) a ligand-independent pathway involv-ing RXRI and an unidentified auxiliary factor present inMCF-7 cells and (ii) a pathway involving RXRI in combina-tion with thyroid hormone and the T3R. These data thereforeshow that estrogen-responsive genes may be negativelyregulated by RXR,B through two distinct inhibitory path-ways.

MATERIALS AND METHODS

Cell culture. MCF-7 cells (passage 148) were obtainedfrom American Type Culture Collection (Rockville, Md.)and maintained at 37°C in improved minimal essential me-dium (IMEM; Biofluids, Rockville, Md.) supplemented with10% heat-inactivated fetal bovine serum (GIBCO) in anatmosphere of 5% CO2. Experiments were routinely con-ducted on cells (passages 149 to 170) maintained for aminimum of 24 h in phenol red-free IMEM plus 5% charcoal-stripped fetal bovine serum (13). MDA-MB-231 cells (9)were a gift of Kim Leslie (University of Colorado, Denver)and were cultured in conditions identical to those for MCF-7cells. Hormones were added to the media as indicated at thefollowing concentrations: E2, 10' M; T3, 10' M; all-transRA, 10-6 M; 9-cis RA, 10-6 M; and dexamethasone, i0-'M. Controls for hormone-treated cells were treated withidentical concentrations of vehicle alone which did notexceed 0.02% (vol/vol).

Gel retardation assays. Nuclear extracts were prepared bythe method of Dignam et al. (16), with the minor modificationof omitting the final dialysis step and freezing extracts inbuffer D. Five micrograms of nuclear proteins was mixedwith approximately 2 fmol of 32P-radiolabeled oligonucle-otide (A2 ERE sequence [5'-GGCTGATCAGGTCACTGTGACCTGACTT-3']) in a 20-,ul volume of binding bufferconsisting of 25 mM Tris-HCl (pH 7.6), 75 mM NaCl, 5 mMEDTA, 1 mM dithiothreitol, 1 mM MgCl2, and 2 ,ug ofpoly(dI-dC). poly(dI-dC) (Pharmacia Fine Chemicals) withor without 200-fold excess competitor DNA (H-2Ld, region Isequence [5'-GCTGGGGATTCCCCAT-3'] or AP-1 [5'-AATTAGTCAGCCATGGGG-3']). Unless otherwise stated,after a 30-min binding reaction at 4°C, the mixture waselectrophoresed at 140 V in a 4% polyacrylamide gel inlow-ionic-strength conditions of 0.4x TBE (52) buffer. Re-combinant ER produced in a baculovirus expression systemwas a generous gift of Myles Brown (Dana Farber CancerInstitute, Boston, Mass. [7]), COUP-TF was kindly donated

by Ming-Jer Tsai (Baylor College of Medicine, Houston,Tex.), and T3Ra was donated by Vera Nikodem (NationalInstitute of Diabetes and Digestive and Kidney Diseases[34]). The preparation of mRXRI is described elsewhere(52). Baculovirus-infected Sf9 cell extracts were added at aconcentration of 2 p,g of total proteins per lane unlessotherwise stated. Anti-ER monoclonal antibody H222 was agift of Geoffrey Greene (Ben May Institute, University ofChicago [31]). Anti-RXRO monoclonal antibody MOK 13.17has been described elsewhere (52). Monoclonal antibodyMOK 15.55, specific for RXRj, will be described elsewhere(53).

Methylation interference. An antibody-mediated methyla-tion interference assay using dimethyl sulfate-treated oligo-nucleotides was performed as described previously (52).Briefly, to determine G residues involved in binding ofhomodimeric RXR,, baculovirus-produced RXRI was firstcomplexed to anti-mouse immunoglobulin beads via a mono-clonal antibody (MOK 13.17) to RXR,. Methylated labeledprobes were then incubated with the bound RXR,B, the beadswere washed, and probe remaining in the precipitated com-plex was then extracted with phenol-chloroform and pro-cessed as described previously (52). To determine G resi-dues involved in T3Ra/RXRP heterodimer binding, 3.5 ,ug ofSf9 cell extracts containing each receptor was subjected togel shift, using methylated probe to visualize the het-erodimeric complex. The shifted probe was then eluted andprocessed as described previously (52).CAT assay. Twenty-four hours prior to transfection, 1 x

106 to 3 x 106 cells were plated in 100-mm2 plates containingphenol red-free IMEM and 5% charcoal-stripped serum.Medium was routinely changed 3 to 4 h prior to addition ofDNA. Unless otherwise stated, a total of 15 to 20 ,ug ofDNAwas added to plates by using the BES (N,N-bis[2-hydroxy-ethyl]-2-aminoethanesulfonic acid) buffer-based calciumphosphate precipitation method of transfection (62). Cellswere washed with phosphate-buffered saline approximately20 h later, and fresh medium containing hormones was addedas indicated. Following an additional 36- to 48-h incubation,cells were harvested and lysed by freeze-thaw in 100 mMKHP04 (pH 7.8) according to the method of deWet et al.(14). As indicated in figure legends, cell extracts werenormalized for protein concentration by using the Bio-Radassay system and/or for luciferase activity produced bycotransfected Rous sarcoma virus (RSV)-luciferase by usinga Monolight 2010 luminometer (Analytical LuminescenceLaboratory) and assayed as described previously (14).Chloramphenicol acetyltransferase (CAT) activity was as-sayed as described elsewhere (29). Quantitation of the per-centage of conversion to acetylated chloramphenicol wasperformed by using an AMBIS Radioanalytic Imaging sys-tem. Unless otherwise stated, plotted values represent themean of three experiments ± standard error of the mean.

Plasmid construction and DNA preparation. Constructionof the expression vector pRSV-H-2RIIBP and the RXRIdomain deletion mutants has been described elsewhere (62).NcoI digests of the respective deletion mutants were sub-cloned into pEXPRESS-O (24) in order to transcribe andtranslate the deletion mutants in vitro. Correct insertion ofthe mutants was confirmed by dideoxy sequencing. Thethymidine kinase (tk)-CAT reporter constructs A2 EREtkCAT, GRE (glucocorticord response element) tkCAT, A2ERE tkCAT, and basal tkCAT are described elsewhere (56).The T3Ra expression vector was a generous gift of V.Nikodem and is described elsewhere (64). The expressionvector for the estrogen receptor, HEO, and control vector,

VOL. 13, 1993

2260 SEGARS ET AL.

B PROBE A2 ERE

ANTIBODY -

COMPETITOR - ERE II NS

ER BAND > &.-A

RXRB BAND D" it

-:

- .ER C ,IRXRB CNTRL

D>*iN 0 w 1 4-

1 2 3 456

FREE PROBE

LANE 1 2 3 4 5 6 7 8 9 0 11

FIG. 1. Evidence that MCF-7 cells contain an endogenous ERE binding activity attributable to RXRf. (A) Western analysis. Nuclearextracts prepared from estrogen-treated MCF-7 cells or in vitro-translated mRXR13 were probed with a monoclonal antibody directed againstRXR,B (MOK 13.17; lanes 1 to 3). In vitro-translated murine ISCBP protein served as a negative control. The arrowhead indicates the RXR1band. Lanes 4 to 6, control antibody (MOK 15.42) directed against an unrelated antigen. (B) Gel retardation assay of ERE binding activityin MCF-7 nuclear extracts. Nuclear extracts prepared from estrogen-treated MCF-7 cells were incubated with an oligonucleotidecorresponding to the vitellogenin A2 ERE. Three reproducible bands are marked: open arrowhead, black arrowhead, and asterisk.Competition was performed with a 200-fold molar excess of A2 ERE, MHC class I region II (lane 2), or an unrelated oligonucleotide, MHCclass I region I (NS, lane 4). For supershift experiments, a monoclonal antibody reacting with the ER (aER; 1 p.l; lane 6) or RXR,B (MOK15.55; aRXRI3; lanes 8 and 9, 4 and 2 ,ul, respectively) or a control antibody (MOK 15.42; CNTRL; lanes 10 and 11, 4 and 2 pul, respectively)was added to the reaction mixture as culture supernatant for 15 min at 30°C prior to electrophoresis.

pKCR2 (6), were gifts of P. Chambon (30). pRSV-luciferasewas given by D. R. Helinski (14). Plasmids pRSV-Jm66 (62)and RSV-3-galactosidase (3-gal) (a generous gift of ChrisSax, National Eye Institute [4]) were used as controls.Western immunoblot analysis. Western blots were per-

formed as described previously, using the anti-RXR1 anti-body MOK 13.17 and a nonspecific isotype-matched anti-body, MOK 15.42 (52). Fifty micrograms of MCF-7 nuclearand cytoplasmic extracts was loaded per lane on 10.5%polyacrylamide gels for sodium dodecyl sulfate (SDS)-poly-acrylamide gel electrophoresis (PAGE) and subsequenttransfer.

Streptavidin-biotin-coupled DNA/cross-linking assay. Un-less otherwise indicated, the procedure was performed asdescribed elsewhere (51), using a 27-bp oligonucleotideidentical to the A2 ERE sequence used in mobility shiftassays but synthesized with a biotin group on the 5' nucle-otide of one strand (10). One to two microliters of radiola-beled, in vitro-translated protein produced from transcrip-tion of the respective pEXPRESS-RXR1 deletion constructswas incubated in 50 pul of buffer A (20 mM N-2-hydroxyeth-ylpiperazine-N'-2-ethanesulfonic acid [HEPES; pH 7.9], 50mM NaCl, 1 mM EDTA, 5% [wt/vol] glycerol, 0.05%[vol/vol] Triton X-100) containing 1.4 p,M nonspecific oligo-nucleotide (5'-CGAATCTAACGCTATrl'TlGGAATGCAAAT-3') with either (i) no added extracts, (ii) 15 p.g of nuclearproteins from MCF-7 cells treated for 48 h with 10-7 M E2,or (iii) 15 p.g of NTera-2 cytoplasmic extracts. After a 10-minpreincubation on ice, biotinylated oligonucleotide was addedto 0.1 p.M and samples were placed at 4°C for 1 h. Disuc-cinimidyl suberate (Pierce) was then added to reactionmixtures at 0.25 mM for 30 min.; disuccinimidyl suberatewas then quenched with 10 mM NH4Cl. Ten microliters of

streptavidin-agarose beads (Sigma) was added for a final 30-to 60-min incubation. Beads were then washed, and proteinswere resolved on an SDS-8% polyacrylamide gel as de-scribed. Parallel control experiments (not shown) were per-formed with a biotinylated oligonucleotide corresponding tothe murine H-2Ld region I element (5'-GCCCAGGGCTGGGGATTCCCCATCTCCT-3').

RESULTS

RXR3 is a nuclear protein expressed in MCF-7 cells. Theexpression of RXR,B was examined by Western blot analysisof nuclear and cytoplasmic extracts prepared from estrogen-treated MCF-7 cells. As seen in Fig. 1A, a monoclonalantibody (MOK 13.17) directed against RXR1 detected asingle 44-kDa band in nuclear extracts (lane 1, arrowhead).This band corresponded with the size of RXR1 detected inextracts of a number of mammalian cells (53). This bandcomigrated with the band produced by in vitro translation ofthe mRXR,B cDNA (lane 2). The same antibody exhibited noreactivity to an unrelated in vitro-translated protein, ICSBP(18). Analysis of MCF-7 cytoplasmic extracts confirmednuclear localization of RXR13 in these cells (not shown). Asexpected, the same proteins probed with an isotype-matchedcontrol antibody of unrelated specificity (MOK 15.42; lanes4 to 6) did not produce this band. These results indicate thatRXRI is expressed in MCF-7 cells as a 44-kDa nuclearprotein.Endogenous RXRfl constitutes an ERE binding activity in

MCF-7 cells. To determine whether the endogenous RXR1contributes to an ERE binding activity in MCF-7 cells, weperformed gel mobility shift experiments using oligonucle-otides corresponding to the ERE of the Xenopus vitellogenin

A anti-RXRBz

LL x mU n

ControlLSzN cc

CC CLLL X Cv a: U

-97 kD

--67 kD

-44 kD

MOL. CELL. BIOL.

INHIBITION OF E2-RESPONSIVE GENES BY RXRI 2261

Relative CAT Activity20 40 60 80 100

- Conltroltk -+ 1RSV-RXRCAT - -

A2ERE2 +tkCAT ; +

A2 -+ERE + +tCk +CAT + +

GRE + +t kCAT-

B HORMONE E2

LANE 1 2 3 4 5 6

0**

tmnRSV-RXRB - - - 2.5 7.5 15

RSV-CONTROL 2.5 7.5 15 - - -

%CONVERSION 34 28 25 36 4 0.9

CHormone

U,

Co.

Relative CAT activity25 so 75 100 125 150

. .. J

\I \

RSV B-galRSV-RXRB

FIG. 2. Inhibition of ERE tkCAT activity by RXR3 in MCF-7cells. (A) Design of tk-CAT reporter constructs. MCF-7 cells weretransfected with the indicated reporter in addition to 10 p.g ofexpression vector for either RXRI (RSV-RXRI) or control (RSV-,-gal). Mean relative CAT activity (± standard error of the mean)from three independent experiments controlled for luciferase activ-ity is shown. Data were normalized such that estrogen-stimulated(i.e., maximal) CAT activity of the A2 ERE tkCAT reporter con-struct was 100%. Cells were treated with E2 (10-' M) or dexameth-asone (Dex; 10- M) 24 h after transfection as indicated. Resultsusing the GRE tkCAT and A2 ERE tkCAT reporters were obtained

A2 gene as a probe and nuclear extracts from MCF-7 cellsthat were cultured in the presence or absence of E2. Asshown in lanes 1 to 4 of Fig. 1B, nuclear extracts preparedfrom estrogen-treated MCF-7 cells produced three majorbands, all ofwhich were removed by a 200-fold molar excessof unlabeled homologous ERE. In addition, oligonucleotidescorresponding to the RXR target sequence, murine H-2Ldgene (MHC class I) region II (Fig. 1B, lane 3), showed partialcompetition; these oligonucleotides removed the middleband (arrow) but not the uppermost and lowermost bands.An unrelated competitor (MHC class I, region I) did notaffect the mobility of any of these bands. These resultsindicate that ERE binding activities in MCF-7 cells consist ofmultiple factors. Since the murine H-2Ld gene region IIbinds RXR,B but not the ER (57), it is possible that the upperand lower bands correspond to an ER/ERE complex and themiddle band corresponds to an RXR3/ERE complex. Todetermine the identity of these bands, supershift experi-ments were performed with monoclonal antibodies specificfor the ER (H222 [31]) or RXRI (MOK 15.55 [51]) (Fig. 1B,lanes 5 to 11). Addition of anti-ER antibody shifted theposition of the uppermost band (open arrowhead to theposition marked by the arrow; lane 6) upward. This antibodydid not affect the position of the remaining bands. MCF-7extracts also produced a more rapidly migrating band (aster-isk in Fig. 1B) which may represent a proteolytic product ofthe ER as suggested previously (67). A matched controlantibody (lane 7) did not affect the migration of the upperband. In addition, the antibody directed against RXRI3altered the migration of the middle band but did not affect theuppermost band. Matched control antibodies did not affectthe position of any of the bands (lanes 10 and 11). Theseresults indicate that the uppermost band represents anER/ERE complex, while the middle band contains RXRI.Identical results were observed with nuclear extracts fromMCF-7 cells cultured in the absence of estrogen (not shown).These data led us to postulate that RXRI contributes to EREbinding activity in MCF-7 cell nuclear extracts and maytherefore play a role in estrogen-mediated gene regulation inMCF-7 cells. Consistent with prior data generated by usingrecombinant receptors (46, 51, 81), data in Fig. 1B argueagainst the existence of an RXR/ER heterodimer.RXR, inhibits ERE-tk promoter activity in MCF-7 cells. To

determine whether RXR3 influenced transcription of anestrogen-responsive gene in MCF-7 cells, we performedcotransfection experiments using a tk-CAT reporter drivenby the vitellogenin A2 ERE (38, 56) (Fig. 2A) and pRSV-H-2RIIBP, an mRXRO expression plasmid (62). It should benoted that pRSV-H-2RIIBP has been found to produce aplasmid-specific RXRI transcript in NTera-2 cells followingtransient transfection (62). As expected (69), MCF-7 cells

in the presence of estradiol (10-7 M) and were normalized such thatdexamethasone-stimulated reporter activity was 100%. (B) Dosedependence of RXR3 inhibition. Microgram amounts of controlplasmid pRSV-Jm66 or RSV-RXRI were added as indicated toMCF-7 cells transfected with 7.5 p.g of the A2 ERE tkCAT reporterin the presence of E2 (10-7 M). Amounts of DNA added to plateswere equalized by using pBluescript (Stratagene). (C) Evidence thatERE-tk reporter inhibition by RXR3 is ligand independent. MCF-7cells were transfected with A2 ERE tkCAT and RSV-,B-gal orRSV-RXRP as indicated. Data were normalized and controlled asdescribed above. In addition to estradiol (10-7 M), cells were treatedwith either T3 (10-7 M), all-trans RA (10-6 M), or 9-cis RA (10-6 M)as indicated.

CD130

a0F

A

Reporter Construct

-1054/54 tk

CAT

GGTCAnnnTGACC(2)

IGGTCA-n-TGACC

V IGGACAnnnTGTCC

VOL. 13, 1993

2262 SEGARS ET AL.

treated with estrogen showed an increase in CAT activitywhen transfected with the A2 ERE reporter in the presenceof a control pRSV vector not expressing RXR, (Fig. 2A).However, when RXR,B was cotransfected, CAT activityproduced by the A2 ERE reporter in E2-treated cells was70% lower than that obtained with the RSV control plasmid.Down-regulation was also seen when RXRP was transfectedinto cells not treated with estrogen. In these cells, the EREtkCAT activity was higher than that of the control construct,tkCAT. This finding suggests that in these cells the ERE isslightly activated in the absence of estrogen, possibly as aresult of remaining estrogenic activity in the serum. Mostimportantly, the RXR,B-dependent inhibition was specific forthe ERE, since the same tk promoter driven by either a GREor an altered ERE that is unable to respond to estrogen (A2ERE [56]) did not demonstrate down-regulation by cotrans-fected RXRI (Fig. 2A). Furthermore, the basal transcriptionactivity was not affected by RXR,B (tkCAT and A2 EREtkCAT constructs), indicating that RXRI is not inhibitinggeneral transcription. As shown in Fig. 2B, RXR, down-regulation of the ERE was dose dependent; the greatestsuppression observed was 3% of the control value, withhigher amounts of DNA than in Fig. 2A. Conversely, thecontrol RSV vector was without any effect at the dosestested. Other RSV vectors containing other unrelated insertsshowed no inhibition, further supporting the specificity ofthe inhibition (data not shown). Since RXR,B has been shownto transactivate a MHC class I promoter in an RA-dependentfashion (62) and since 9-cis RA has been shown to be a ligandfor RXR3 (36, 47), we tested the effect of all-trans RA and9-cis RA on RXRI inhibition of the A2 ERE-tk promoteractivity. As seen in Fig. 2C, there was no change in RXRIinhibition of promoter activity following addition of 9-cis RAat 10-6 M or all-trans RA at 10-6 M. It should be noted thata comparable concentration of all-trans RA led to transactivation of MHC class I promoter activity by RXRI (62).Together, these results indicate that (i) RXR3 inhibits estro-gen-dependent transcription of ERE-tk promoter activity inan element-specific fashion without requiring the addition of9-cis RA and (ii) there is no inhibition of basal transcriptionactivity.RXRO inhibits ER action. Because the strongest RXRP

inhibition was seen when cells were treated with estrogen,we sought to determine whether this down-regulation wasspecific for gene transcription attributable to the ER. Fur-ther, we sought to determine whether RXRP might posi-tively trans activate an ERE-tk promoter in the absence ofthe ER. To address these questions, we performed cotrans-fection experiments with an ER-negative breast tumor cellline, MDA-MB-231. These cells do not express a detectablelevel of the ER and thus do not respond to E2 (10). As seenin Fig. 3, the ERE-tk promoter activity was low in theabsence of the exogenous ER (lanes 1 and 2), and treatmentof these cells with E2 did not significantly change the level ofthe promoter activity (not shown). As expected, transfectionof HEO, an ER expression plasmid (30), resulted in an 8- to10-fold stimulation of the promoter activity only when cellswere treated with E2 (lanes 3 and 4). However, when RXRPwas cotransfected together with the ER, A2 ERE promoteractivity was decreased to 50% of that seen with the controlplasmid (compare lane 4 with lane 6). A similar down-regulation of ERE promoter activity was also observed whenthe same experiment was performed with HeLa cells (notshown). In agreement with the results in Fig. 2, cotransfec-tion of RXR, in the absence of ER did not significantly affectthe basal level of promoter activity, either in the presence or

120 7

100 D No hormone

* + Estrogen

so

) 60 T

40-

20-

0Control + +ER (HEO) - - + + + +

RSV-B-gal + - + +t . + + -:

RSV-RXRB - + + + __ + +

LANE I j2 3 5 6 7 8 9 10

FIG. 3. Evidence that RXRI inhibits ER-stimulated transcrip-tion of an ERE. Twelve micrograms of A2 ERE tkCAT reporter wastransfected in the ER-negative cell line MDA-MB-231 with either 5,ug of an ER expression vector (HEO) or 5 ,ug of a control (pKCR2)in addition to 10 ,ug of RSV-3-gal or RSV-RXR3 (pRSV-H-2RIIBP)as shown. Mean CAT activity (± standard deviation) of threeindependent experiments was controlled for luciferase expressionand normalized as described in the legend Fig. 2. Cells were treatedwith E2 (10-' M) as indicated.

in the absence of E2 (lanes 9 and 10). These results confirmthat the inhibition observed in MCF-7 cells was due to aninterference with ER-mediated gene activation caused byRXRP.Domain requirements for RXRf3-mediated ERE inhibition.

Similar to other nuclear hormone receptors, RXR,B is com-posed of a modular domain structure: the least conservedN-terminal domain, the highly conserved DNA bindingdomain, and the moderately conserved C-terminal domainthat is responsible for dimerization and ligand binding (49,51). In agreement with studies performed on other receptors(23), we have found that the C-terminal domain of mRXR, isessential for both homodimer and heterodimer formation(51). To assess domains involved in RXRP down-regulationof the ERE-tk promoter, we tested three RXRI deletionmutants deficient in each of the three modular domains butpossessing an intact nuclear localization domain. Figure 4Ashows a schematic representation of the deletion constructsand results of cotransfection assays performed in MCF-7cells. To exclude the trivial explanation that abnormaltranslation or protein transport of the deletion mutants couldinfluence the findings, a monoclonal antibody directedagainst a conserved region of RXRI was used to performimmunofluorescence studies of MCF-7 cells transfected withthe described deletion mutants. Those studies support theconclusion that the described deletion mutants are expressedand transported to the nucleus of transfected MCF-7 cells ina manner similar to that for intact RXRI (data not shown).As previously shown (Fig. 2A), the intact RXRI caused a70% reduction in the ERE-tk promoter activity. In contrast,deletion of the N-terminal domain resulted in a loss ofdown-regulation. The inability of the N-terminal deletion toinhibit ERE-tk promoter activity was not likely due to grossalteration of RXR,B tertiary structure because the samemutant has been shown to retain the ability to trans activate

MOL. CELL. BIOL.

INHIBITION OF E2-RESPONSIVE GENES BY RXRI 2263

DOMAIN DELETION CONSTRUCTS RELATIVE CAT FOLDACTIVITY DEC

DNAN-TerrinalIbindng c

N-del _

D-del

C-del

Amino acid t 79 147 254

B Native N-delwU >. wU >z u z

U. ULL E L.

2 X - a X

Control 100C-Terminal Native- RXRO 26±7

N-del 92±10

0.0

-2.7

-0.3

79±6 1-0.8

C-del 139+12

410

D-del

wj >.

Z 0)N-U.

00Q

01.4

C-del

z vr- r-UL U.

L0U- a

Mr kD

200-

>97-

67-

O44- __-_

29-

LANE 1 2 3 4 5 6 7 8 9 10 11 12

FIG. 4. (A) Deletion analysis of RXRI3 inhibition of ERE-depen-dent trans activation. Structures of RXRI3 domain deletion constructsare shown. Cotransfection experiments were performed in MCF-7cells as described for Fig. 2A but using 10 ,ug of the domain deletionconstructs instead of the native RXRI3. Relative CAT activity (mean+ standard error of the mean) of three experiments normalized asdescribed for Fig. 2A and controlled for protein is indicated for eachconstruct. Deletion constructs were designed to conserve the puta-tive nuclear localization signal. Values obtained with the estrogen-stimulated control represent a 3.7-fold stimulation over CAT activityin the absence of ligand and are shown as 100% relative CAT activity.Fold reduction relative to stimulated control is also shown. (B)Evidence that RXRI3 binding to the ERE involves an endogenousprotein present in MCF-7 nuclear extracts. 35S-labeled in vitro-

translated native RXRO or deletion construct was incubated alone orin combination with nuclear (MCF7 NE) or cytoplasmic (MCF7 CY)extracts prepared from estrogen-treated MCF-7 cells as indicated.After mixing with a biotinylated oligonucleotide corresponding to theA2 ERE, protein-DNA complexes were chemically cross-linked,precipitated with streptavidin-agarose, and resolved by SDS-PAGE(10% polyacrylamide gel). A retarded band (black arrowhead) wasobserved only when MCF-7 nuclear extracts were combined with35S-labeled native RXRI (lane 2) or N-terminally deleted RXR3 (lane5). No similar retarded band was observed in the presence of a

nonspecific biotinylated oligonucleotide (data not shown).

a MHC class I promoter (62). Similarly, deletion of the DNAbinding domain or the C-terminal domain resulted in a loss ofERE promoter inhibition. It should be noted that the mutantlacking the DNA binding domain was capable of het-erodimerizing with the T3R (53). Also, the identical C-termi-

nal deletion construct is capable of binding to the ERE invitro (Fig. 4B). These results suggest that in addition to theDNA binding domain, the N-terminal and C-terminal do-mains are involved in the repression of ERE-tk promoteractivity.ERE binding by RXRj3 involves an auxiliary protein present

in MCF-7 cells. Previous studies of RXR function support thepostulate that mRXR,B functions predominantly as a het-erodimer rather than a homodimer (or a monomer) in theintracellular environment (8, 40, 46, 51, 80, 81). If thispostulate is correct, the inhibition observed above may becaused by RXR, acting as a heterodimer. To test thispostulate, we used chemical cross-linking experiments toexamine MCF-7 nuclear extracts for the presence of proteinscapable of interacting with mRXR3. Figure 4B shows resultsof a streptavidin-biotin-coupled DNA/cross-linking assay inwhich in vitro-translated 5S-labeled mRXRP was mixedwith MCF-7 nuclear extracts in the presence of a biotinyl-ated ERE and subjected to chemical cross-linking, and thebound materials were then precipitated with streptavidin-agarose beads. When labeled native RXR,B was cross-linkedwith MCF-7 cell nuclear extracts (lane 2) and precipitatedwith the ERE, a high-molecular-size band (100 kDa; blackarrowhead) was seen in addition to the monomeric RXR (44kDa; white arrowhead). A high-molecular-size band was notobserved when MCF-7 cytoplasmic extracts were substi-tuted for the nuclear extracts (lanes 3 and 6). Similar tonative RXRI, a high-molecular-size band (lane 4) was ob-served when the labeled N-deletion mutant was combinedwith MCF-7 nuclear extracts. As expected, the C-deletionmutant lacking the dimerization domain failed to produce ahigh-molecular size band. The labeled DNA binding domainmutant failed to bind the ERE and thus could not beprecipitated in this assay (lanes 7 to 9). Control experimentsusing a biotinylated DNA sequence which fails to bindRXR, showed neither precipitation of monomeric RXR,Bnor the high-molecular-size band (not shown). These resultsindicate that MCF-7 nuclear extracts, but not cytoplasmicextracts, contain an auxiliary factor (likely a heterodimerpartner and not a homodimer; see below) that is capable ofassociating with RXR,B to bind the A2 ERE.RXR,3 heterodimerizes with T3Ra to bind the ERE. To look

for heterodimer partners contributing to ERE binding ofRXR,B, we tested the ability of T3Ra (34), COUP-TF (1, 78),and the ER (7) to dimerize with RXRP and bind to the ERE.It has been shown that T3R not only binds to the ERE (27)but also is capable of forming a heterodimer with RXRI (46,80) to bind various hormone response elements, includingthe "imperfect" vitellogenin Bi ERE (68), with strikinglyhigher affinities than do either RXR,B or T3Ra homodimers(51). Gel retardation assays were performed with labeled A2ERE oligonucleotides and extracts prepared from Sf9 cellsinfected with recombinant baculovirus encoding proteins forthe ER, mRXR,B, COUP-TF, or T3Ra. These extracts wereestimated to contain similar levels of receptor proteins (notshown). Addition of extracts containing either T3Ra (Fig.5A, lane 1) or mRXR,B (lane 9) alone to the binding reactionproduced a weak, barely detectable band. However, whenextracts containing both RXRP and T3Rax were combined, anew band of much greater intensity was seen (lane 2; blackarrowhead). The newly formed band was specifically com-peted for by the A2 ERE (lane 3) but not by a control AP-1element (lane 4) and was not produced using control extractsprepared from Sf9 cells infected with wild-type baculovirus(not shown). This new complex was most likely a RXR,/T3Ra heterodimer bound to the ERE, since much of this

A

VOL. 13, 1993

2264 SEGARS ET AL.

_ + + + _ _ + + + +

+ 4 + + + _ + _ _ + +

ER + + + +

COMPETITOR/ANTIBODY - - ERE API- 13.17 CNTRL

ER >

T3R/RXRB -

HETERODIMER

*

FREE PROBE

LANE 1 2 3 4 5 6

RXRB/T3R HETERODIN

CODING NONCODING

F B F B

< _

0 <~~~~< rs 2

V< <

UC' %v LOU&o

U

H_ _

A2 ERE U

band was supershifted in the presence of antibody specificfor RXR, (MOK 13.17; arrow, lane 10). A similar analysisusing COUP-TF did not generate a heterodimer band withRXRPI on the ERE (not shown). Addition of the ER alone tothe binding reaction produced a band which migrated moreslowly than the RXRF/T3Ra complex (open arrowhead; lane6). This band represents the ER/ERE complex since it was

- also competed for by excess unlabeled ERE and super-shifted with use of an antibody directed against the ER(H222) but not with use of a control or anti-RXR,3 antibody(not shown). When the three extracts containing the ER,RXRI, and T3Ra were mixed and added to the reaction,both the upper ER and the intermediate RXRP/T3Ra bandswere seen at comparable intensities. This gel retardationpattern resembled the pattern produced with MCF-7 nuclearextracts (Fig. 1B). These results suggest that under theseconditions, the RXR,F/T3Ra heterodimer binds to the A2ERE at an affinity comparable to that of the ER homodimer.This observation indicates that the ER and the RXRI3/T3Raheterodimer are competing for an overlapping binding site onthe A2 ERE. Consistent with this interpretation, methylationinterference experiments showed that the RXR3F/T3Ra het-erodimer (and the mRXRI homodimer) binding involves Gresidues within the A2 ERE which overlap residues shown

IER to be critical to binding by the ER (39) (Fig. 5C). Further, wenote that the ER and RXR,F/T3Ra heterodimer produced amethylation interference pattern slightly different from thatseen with the RXRP homodimer; the interference of thecentral G residues in the palindrome was consistentlyweaker with the ER or heterodimer than with the RXRPihomodimer.

RXRfI inhibits ERE-tk promoter activity by two pathways.The binding data presented above suggested that T3Rax mayserve as a partner for mRXR,B and thus contribute todown-regulation of the ERE-tk promoter. Such a het-erodimeric interaction could mediate the ERE inhibition byT3R described by Glass et al. (28). To further define the roleof T3Rac in the inhibitory effect, cotransfection experimentswere performed with MDA-MB-231 cells, using expressionplasmids for the ER, T3Rao, and mRXR,. MDA-MB-231cells were used for these experiments since they permittedadjustment of the amounts of RXRP and T3Ra vectorsrelative to the ER expression vector. Cells were cotrans-fected with HEO (ER) and either 0.2 or 2.0 ,ug of pRSV-H-2RIIBP or pRSV-T3Ra, either alone or in combination (Fig.6). Cotransfection of increasing amounts of RXR,B, as ex-pected, resulted in inhibition of the ERE promoter activity

C A2 ERE *s a.. RXRJi HOMODIMER5'-(-337)-AAGTCAGGTCACATGACETGATCA-(.313)-3

TTCAGTCCAGTJrCACTGGACTAGTa a o

A2 ERE * * o a RXRj5/T3Ro HETERODIMER5 -(-337)-AAGTCAGGTCAQrTGACtTGATCA-+-313)-3'

TTCAGTCCAGTPrCACTGGACTAGT

ERE . . ER HOMODIMER5'-TAGGTCACAGTGA CTA-3'ATCCAGTGTCACTGGAT

FIG. 5. Evidence that RXRI heterodimerization with T3Rct re-

sults in increased A2 ERE binding. (A) Gel retardation assay.32P-labeled A2 ERE oligonucleotide was mixed with extracts pre-pared from Sf9 cells infected with recombinant baculovirus express-ing RXR,B, T3Ra, or ER, alone or in combination, as indicated.Coincubation of T3Ra and RXR, generated a complex of mobilityintermediate between bands generated with either T3Ra or RXRI

alone (lane 2, black arrowhead). This new complex was competedfor with a 200-fold molar excess A2 ERE but not with a nonspecificoligonucleotide (AP-1). Coincubation of RXRP with the ER (lane 8)or COUP-TF (not shown) did not generate a new complex. Additionof 1 ,ul of antibody (MOK 13.17) directed against RXRI for 15 minat 30°C further shifted this band (black arrowhead) to the positionnoted by the arrow. One microliter of an antibody of unrelatedspecificity (MOK 14.44; CNTRL) had no such effect. (B) Methyla-tion interference. DNA isolated from free (F) and bound (B)protein/DNA complexes was resolved on a 12% denaturing poly-acrylamide gel. G residues whose methylation completely abolishesDNA binding appear as missing bands in the bound lanes and are

summarized as black dots in panel C. Partially interfering residuesare depicted as open circles. The published pattern of the ER on thesame sequence (ERE) is provided for comparison (39). Central Gresidues showing differential methylation between the receptors are

boxed.

A PROBE A2 ERE

RXRB

T3Rm

B RXRB HOMODIMER

CODING NONCODING

F B F B

I

(.r- M

0 1

H-i0

'C.

0

u

0

'C'c_Is

0

0;z0

..I

-0

O_ 0

0<:

F

u

0F-

- s

ku

A2 ERE

MOL. CELL. BIOL.

INHIBITION OF E2-RESPONSIVE GENES BY RXRI 2265

0

*

C.)

0

Cu=

175

150 -

125 -

100 -

75-

50 -

25 -

0

E2ER

RSV-B galRSV-RXRpRSV-T3R a

T3I

0.2ug2.0 ug5.0 ug

-i1 2 3 4 5 6 7 8 9 10 11+

+ V +

= -- =+ + + 1 + I +

_ - + + II+1 +1 +

FIG. 6. Evidence that inhibition of estrogen-responsive genetranscription by RXRP involves two pathways. Cotransfectionexperiments were performed with MDA-MB-231 cells, and datawere controlled and normalized as described for Fig. 3. All trans-fections were performed in the presence of E2 (10-' M) and 3.5 ,ugof ER (HEO) as indicated. Expression vectors were added forRXR, (RSV-RXR3), T3Ra (RSV-T3R), or control (RSV-P-fal) ateither 0.2, 2.0, or 5.0 ,ug as indicated. T3 was added at 10 M asindicated. Plates transfected with both T3Ra and RXR,B received 0.1or 1.0 ,ug (total of 0.2 and 2.0 pLg of expression vector, respectively)of each vector. pBluescript was used to equalize amounts of DNAadded to cells.

(blocks 2 and 3) compared with transfection of the RSV-p-gal control (block 1). Transfection of increasing amounts ofT3Rcx alone in the absence of thyroid hormone resulted ineither a slight stimulation or no reduction of ERE-tk pro-moter activity (block 4). Consistent with prior reports (28),addition of thyroid hormone to plates transfected with T3Raalone did result in down-regulation ofCAT activity (block 5).Cotransfection of RXR3 and T3Ra together caused a signif-icant reduction in promoter activity in the presence ofthyroid hormone. This down-regulation was dependent onthe amounts of pRSV-H-2RIIBP and pRSV-T3Rct added tothe cells. Similar T3-dependent inhibition of ERE-tk pro-moter activity was observed following cotransfection ofRXRP and T3RaL in MCF-7 cells (not shown). It is notewor-thy that in the absence of thyroid hormone, but in thepresence of T3Ra and RXR,, the reduction was similar tothat seen with RXR, alone. It should be noted that in theabsence of added ER, we did not observe stimulation ofreporter activity following addition of RXRI or T3Ra, eitheralone or in combination (blocks 9 to 11). Thus, there appearto be two distinct pathways leading to an inhibition ofERE-tk promoter activity: (i) a thyroid hormone-indepen-dent mechanism, seen with addition of mRXRj only, and (ii)a thyroid hormone-dependent mechanism, seen with T3Ratalone or in combination with RXR,B. These results suggestthat since the down-regulation of ERE-tk promoter activityseen with RXR,B alone did not require addition of T3 (Fig. 2

to 4), it is likely caused by the interaction of RXR, with anunidentified heterodimer partner and not by heterodimeriza-tion with T3Ra.

DISCUSSION

We have shown that cotransfection of RXR, leads tostrong element-specific inhibition of ERE-tk reporter activityin MCF-7 cells. This inhibition results from a block inER-mediated activation because the inhibition was not ob-served in MDA-MB-231 cells, which do not express the ER,but became detectable following cotransfection of the ER(Fig. 3). Further, MCF-7 cells were shown to express RXRPand an auxiliary protein which contributed to ERE bindingactivity. To determine whether a known heterodimer partnerof RXRf could contribute to the observed inhibition, westudied the role of T3Ra, which dimerizes with RXRP andbinds to the A2 ERE (Fig. 5). We found that transfection ofT3Ra alone or in combination with RXRP also inhibitedERE-tk reporter activity. In contrast to the inhibition causedby RXRI alone (which was observed in the absence of areceptor-specific ligand), T3Roa-mediated inhibition was de-pendent upon addition of thyroid hormone. These resultsindicate that estrogen-responsive genes are not only acti-vated by the ER, but are also repressed by RXRI. Becauseof the multiple partners with which RXRP is known todimerize, such negative regulation could involve multiplenuclear hormone receptors. These findings suggest that thenet ERE activity measured in MCF-7 cells represents abalance between a positive effect produced by the endoge-nous ER and a negative effect produced by the endogenousRXRI.Feavers et al. (21) have reported the existence of a non-ER

protein present in MCF-7 cells that can bind to EREs. Basedsolely on molecular size, this protein (70 kDa) is unlikely tobe RXRI (44 kDa; Fig. 1) but may represent an RXRheterodimer partner. Other members of the RXR gene family(i.e., RXRa and RXR-y [49]) may represent the proteindescribed by Feavers et al. (21), although an ERE-modulat-ing role for these RXR proteins has yet to be demonstrated.Involvement of a non-ER protein in transcriptional inhibitionof an ERE reporter in vitro has been reported by Kaling etal. (37); however, binding of that protein to the ERE has notbeen demonstrated. Similar to our findings, Sharif andPrivalsky (69) reported that v-ErbA, an oncogenic variant ofT3R, is capable of binding to the ERE and suppressingERE-driven CAT activity. These authors concluded that thesuppression was caused by interference with ER action byv-ErbA. The inhibition caused by RXR3 may operatethrough a similar mechanism. However, since v-ErbA is notnormally a constituent of mammalian cells, the significanceof v-ErbA repression in the physiologic regulation of estro-gen-responsive genes is less clear.ERE inhibition by RXR3 involves an auxiliary heterodimer

partner. The inhibition of ERE-tk promoter activity causedby RXRI and T3Ra, either alone or in combination, may beassumed to be mediated by a heterodimeric complex of thesereceptors. This assumption is supported by the demonstra-tion that RXR has a strong propensity to form heterodimericcomplexes with other nuclear hormone receptors (46, 51,80). Consistent with the involvement of RXR heterodimersin the inhibition of ERE-tk promoter activity, we show thatERE binding by RXROi involves an auxiliary factor presentin MCF-7 cell extracts (Fig. 4B). This finding resembles thedemonstration that T3Rs (and also VDRs and RARs) requirean accessory protein present in mammalian nuclei that

L

I

VOL. 13, 1993

2266 SEGARS ET AL.

dimerizes with them (i.e., RXRs) to result in optimal targetDNA binding and trans activation of target genes (27, 40,44-46). The poor binding of the RXRPi homodimer to the A2ERE (Fig. 5A) further supports the role of a heterodimerpartner. On the basis of these results, we favor the interpre-tation that the inhibition observed when RXRI alone istransfected (Fig. 2 and 4A) is caused by interaction with anunidentified heterodimeric partner rather than an RXRRhomodimer.The inhibition associated with transfection of T3Ra in the

presence of thyroid hormone (Fig. 6) is likely caused byT3Rao acting as a heterodimer rather than as a homodimer (ormonomer). Interestingly, the interaction of the RXRF/T3Raheterodimer with an ERE appears quite specific, since theclosely related RXRI3/T3Ra heterodimer has been reportedto bind poorly, if at all, to the ERE (81). In view of thetendency of T3Ra to function as a heterodimer, and theinability of the RXRC/T3Ra heterodimer to bind the ERE,the T3-induced T3Rao-mediated inhibition of ERE trans acti-vation reported by Glass et al. (28) may be produced byRXR,B/T3Ra heterodimers. Similarly, we observed T3Ra-dependent inhibition to be T3 dependent. However, since theRXR,-induced inhibition of ERE-tk promoter activity inMCF-7 cells was not dependent upon T3 treatment, it may beconcluded that the auxiliary endogenous RXR, heterodimerpartner (mentioned above) is not T3Ra.The mechanism of inhibition. Our finding that the RXRP/

T3Rax heterodimer contacts the ERE through a the same setof G residues as does the ER suggests that ER displacementfrom the ERE is required for the observed inhibition. Arequirement for DNA binding is further supported by thefinding that deletion of either the DNA binding domain or theC-terminal dimerization domain abolished the inhibition.This finding contrasts with reports describing inhibition ofERE-driven reporter activity mediated by the glucocorticoidreceptor (74) or c-foslc-jun (17), which did not require EREbinding. Even though binding of RXROi to the ERE is aprerequisite, mere binding to the ERE was insufficient tocause the observed inhibition. This is evident from thefinding that deletion of the N-terminal domain of RXR,Babolished the inhibitory activity. Of note, we show thatdeletion of the N-terminal domain did not prevent EREbinding in vitro (Fig. 4B). Furthermore, the deletion mutantused in this work was capable of trans activating an MHCclass I promoter as efficiently as did intact RXRP, leading usto conclude that the N-terminal domain is not essential foractivating the MHC class I promoter (62). Collectively, thesedata suggest the mechanism of ERE repression by RXR,involves dimerization with an auxiliary protein, associationwith DNA, and interaction of other factors with an intactN-terminal domain.

It is noteworthy that the inhibition observed in the pres-ence of RXR alone did not require the RXR-specific ligand9-cis RA (36, 47). This finding indicates that RXRP canfunction as a transcription factor in the absence of a specificligand (Fig. 2C). In support of these results, we found thatRXR3 and T3Ra can synergistically enhance transcription ofmalic enzyme thyroid response element and MHC class Ireporters in the absence of 9-cis RA (34, 51). A similarunidirectional hormone requirement has been reported forthe VDR/RXR heterodimer (80). This finding is consistentwith the cross-inhibition of T3Rs and RARs in the absence ofa ligand (23).Naar et al. (61) and Lipkin et al. (48) proposed that the

inhibitory effects of nuclear hormone receptors associatedwith binding to several hormone-responsive elements are

caused by conformational constraints of receptor bindingimposed by the target elements. It is possible that suchconformational restraints induced by RXR3 binding to theERE might result in an inhibitory effect mediated by thepreviously described N-terminal or C-terminal transcrip-tional activation domains (TAF-1 and TAF-2 [60, 75, 79]). Ifsuch a mechanism exists, it may be variably dependent onthe specific heterodimer partner. It will be of interest todetermine whether RXRot and RXR-y, which are highlydivergent from RXR, in the N-terminal domain but sharehomology in the DNA and ligand binding domain, can alsoelicit inhibition similar to that demonstrated for RXRwB. Itshould be mentioned that under no condition tested to datehas RXR,B (with or without the T3R) been observed tofunction as a positive transcription factor on an ERE.

Physiologic implications. Physiologic and experimentalstates have been described which may involve RXRB-medi-ated inhibition of estrogen-responsive gene transcriptionsuch as described here. For instance, during normal mam-mary development of the rat, cellular differentiation pro-ceeds to a point at which, despite adequate estrogen expo-sure and the presence of ER, the tissue becomes refractoryto the effect of estrogen (71). Similarly, the modulation ofluteinizing hormone synthesis in the pituitary by steroids iscomplex and cannot be explained solely on the basis ofestrogen and ER concentrations (70). Because the ER actsonly as a transcriptional activator, the existence of a specificopposing factor has been proposed to explain such data.Since RXR, has the capability of interfering with estrogen-induced transcription, RXR, may play a role in these andsimilar instances.

In view of the antagonistic effect of RXRP upon ERfunction, it is possible that RXR, plays a role in theantiproliferative effects of retinoids observed in hormonallyresponsive breast cancer cells (4, 22). A more thoroughunderstanding of the antiestrogenic action of RARs in breastcancer may be clinically relevant since the antiproliferativeeffects of retinoids are enhanced by the anticancer agenttamoxifen (references 2, 66, and 73 and references therein).Further studies are needed to more clearly define the role ofRXR,B in modulation of estrogen-dependent gene transcrip-tion in these physiologic, developmental, and neoplasticstates.

ACKNOWLEDGMENTS

We acknowledge the helpful guidance and technical expertisecontributed by Anne Colston Wentz, John J. Sciarra, Insong Lee,Dave Ennist, Toshi Nagata, and Magda Fgagias. Vera Nikodem andPaul Hallenbeck kindly supplied baculovirus-produced recombinantT3Ra and the T3Ra expression vector. Baculovirus-produced ERwas provided by Myles Brown. The antibody H222 was generouslygiven by Geoffrey Greene. The biotinylated oligomers were synthe-sized by Ettore Appella. We thank Herb Samuels and Barry Formanfor donation of the pEXPRESS-O construct. The specific sugges-tions made by Kim Leslie and C. V. Rao were greatly appreciated.

This work was made possible by support (to J.H.S.) provided bythe Reproductive Scientist Training Program (HD00849-02), theAmerican Fertility Society, the Department of OB-GYN, North-western University, Chicago, Ill., and the Swiss National ScienceFoundation.

REFERENCES1. Bagchi, M. K., S. Tsai, M.-J. Tsai, and B. W. O'Malley. 1987.

Purification and characterization of chicken ovalbumin geneupstream promoter transcription factor from homologous ovi-duct cells. Mol. Cell. Biol. 7:4151-4158.

2. Barbi, G. P., P. Marroni, P. Bruzzi, G. Nicolo, M. Paganuzzi,

MOL. CELL. BIOL.

INHIBITION OF E2-RESPONSIVE GENES BY RXRI 2267

and G. B. Ferrara. 1987. Correlation between steroid hormonereceptors and prognostic factors in human breast cancer. On-cology 44:265-269.

3. Beato, M. 1989. Gene regulation by steroid hormones. Cell56:335-344.

4. Bollag, W., R. Peck, and J. R. Frey. 1992. Inhibition of prolif-eration by retinoids, cytokines and their combination in fourtransformed epithelial cell lines. Cancer Lett. 62:167-172.

5. Borras, T., C. A. Peterson, and J. Piatigorsky. 1988. Evidencefor positive and negative regulation in the promoter of thechicken 81-crystallin gene. Dev. Biol. 127:209-219.

6. Breathnach, R., and B. A. Harris. 1983. Plasmids for the cloningand expression of full-length double-stranded cDNAs under thecontrol of the SV40 early or late gene promoter. Nucleic AcidsRes. 11:7119-7136.

7. Brown, M., and P. A. Sharp. 1990. Human estrogen receptorforms multiple protein-DNA complexes. J. Biol. Chem. 265:11238-11243.

8. Bugge, T. H., J. Pohl, 0. Lonnoy, and H. G. Stunnenberg. 1992.RXRa, a promiscuous partner of retinoic acid and thyroidhormone receptors. EMBO J. 11:1409-1418.

9. Cailleau, R., R. Young, M. Olive, and W. J. Reeves, Jr. 1974.Breast tumor cell lines from pleural effusions. JNCI 53:661-674.

10. Cocuzza, A. J. 1989. A phosphoramidite reagent for automatedsolid phase synthesis of 5'-biotinylated oligonucleotides. Tetra-hedron Lett. 30:6287-6290.

11. Corthesy, B., J.-R. Cardinaux, F.-X. Claret, and W. Wahli.1989. A nuclear factor I-like activity and a liver-specific repres-sor govern estrogen-regulated in vitro transcription form theXenopus laevis vitellogenin Bi promoter. Mol. Cell. Biol.9:5548-5562.

12. Danel, L., G. Souweine, J. C. Monier, and S. Saez. 1983. Specificestrogen binding sites in human lymphoid cells and thymic cells.J. Steroid Biochem. 18:559-563.

13. Darbre, P., J. Yates, S. Curtis, and R. J. B. King. 1983. Effect ofestradiol on human breast cancer cells in culture. Cancer Res.43:349-354.

14. deWet, J. R., K. V. Wood, M. DeLuca, D. R. Helinski, and S.Subramani. 1987. Firefly luciferase gene: structure and expres-sion in mammalian cells. Mol. Cell. Biol. 7:725-737.

15. Dickson, R. B., M. E. McManaway, and M. E. Lippman. 1986.Estrogen-induced factors of breast cancer cells partially replaceestrogen to promote tumor growth. Science 232:1540-1543.

16. Dignam, J. D., R. M. Lebowitz, and R. G. Roeder. 1983.Accurate transcription initiation by RNA polymerase II insoluble extracts from isolated mammalian nuclei. Nucleic AcidsRes. 11:1475-1489.

17. Doucas, V., G. Spyrou, and M. Yaniv. 1991. Unregulatedexpression of c-Jun or c-Fos proteins but not Jun D inhibitsoestrogen receptor activity in human breast cancer derivedcells. EMBO J. 10:2237-2245.

18. Driggers, P. H., B. A. Elenbaas, J.-B. An, I. J. Lee, and K.Ozato. 1992. Two upstream elements activate transcription of amajor histocompatibility complex class I gene in vitro. NucleicAcids Res. 20:2533-2540.

19. Evans, R. M. 1988. The steroid and thyroid hormone receptorsuperfamily. Science 240:889-895.

20. Fawell, S. E., J. A. Lees, R. White, and M. G. Parker. 1990.Characterization and colocalization of steroid binding anddimerization activities in the mouse estrogen receptor. Cell60:953-962.

21. Feavers, I. M., J. Jiricny, B. Moncharmont, H. P. Saluz, andJ. P. Jost. 1987. Interaction of two nonhistone proteins with theestradiol response element of the avian vitellogenin gene mod-ulates the binding of estradiol-receptor complex. Proc. Natl.Acad. Sci. USA 84:7453-7457.

22. Fontana, J. A., A. B. Mezu, B. N. Cooper, and D. Miranda.1990. Retinoid modulation of estradiol-stimulated growth and ofprotein synthesis and secretion in human breast carcinomacells. Cancer Res. 50:1997-2002.

23. Forman, B. M., and H. H. Samuels. 1990. Interactions among asubfamily of nuclear hormone receptors: the regulatory zippermodel. Mol. Endocrinol. 4:1293-1301.

24. Forman, B. M., and H. H. Samuels. 1991. pExpress: a family ofexpression vectors containing a single transcription unit activein prokaryotes, eukaryotes and in vitro. Gene 105:9-15.

25. Fox, H. S., B. L. Bond, and T. G. Parslow. 1991. Estrogenregulates the IFN--y promoter. J. Immunol. 146:4362-4367.

26. Gaub, M.-P., M. Bellard, I. Scheuer, P. Chambon, and P.Sassone-Corsi. 1990. Activation of the ovalbumin gene by theestrogen receptor involves the Fos-Jun complex. Cell 63:1267-1276.

27. Glass, C. K., 0. V. Devary, and M. G. Rosenfeld. 1990. Multiplecell type-specific proteins differentially regulate target sequencerecognition by the a retinoic acid receptor. Cell 63:729-738.

28. Glass, C. K., J. M. Holloway, 0. V. Devary, and J. G. Rosenfeld.1988. The thyroid hormone receptor binds with opposite tran-scriptional effects to a common sequence motif in thyroidhormone and estrogen response elements. Cell 54:313-323.

29. Gorman, C. M., L. F. Moffat, and B. H. Howard. 1982.Recombinant genomes which express chloramphenicol acetyl-transferase in mammalian cells. Mol. Cell. Biol. 2:1044-1051.

30. Green, S., P. Walter, V. Kumar, A. Krust, J.-M. Bornert, P.Argos, and P. Chambon. 1986. Human oestrogen receptorcDNA: sequence, expression and homology to v-erb-A. Nature(London) 320:134-139.

31. Greene, G. L., N. B. Sobel, W. J. King, and E. V. Jensen. 1984.Immunochemical studies of estrogen receptors. J. Steroid Bio-chem. 20:51-56.

32. Gronemeyer, H. 1991. Transcription activation by estrogen andprogesterone receptors. Annu. Rev. Genet. 25:89-123.

33. Grossman, C. J. 1985. Interactions between the gonadal steroidsand the immune system. Science 227:257-261.

34. Hallenbeck, P. L., M. S. Marks, R. E. Lippoldt, K. Ozato, andV. M. Nikodem. 1992. Heterodimerization of thyroid hormone(TH) receptor with H-2RIIBP (RXR,B) enhances DNA bindingand TH-dependent transcriptional activation. Proc. Natl. Acad.Sci. USA 89:5572-5576.

35. Hamada, K., S. L. Gleason, B.-Z. Levi, S. Hirschfeld, E.Appella, and K. Ozato. 1989. H-2RIIBP, a member of thenuclear hormone receptor superfamily that binds to both theregulatory element of major histocompatibility class I genes andthe estrogen response element. Proc. Natl. Acad. Sci. USA86:8289-8293.

36. Heyman, R. A., D. J. Mangelsdorf, J. A. Dyck, R. B. Stein, G.Eichele, R. M. Evans, and C. Thaller. 1992. 9-Cis retinoic acid isa high affinity ligand for the retinoid X receptor. Cell 68:397-406.

37. Kaling, M., T. Weimar-Ehl, M. Kleinhans, and G. U. Ryffel.1990. Transcription factors different from the estrogen receptorstimulate in vitro transcription from promoters containing es-trogen response elements. Mol. Cell. Endocrinol. 69:167-178.

38. Klein-Hitpass, L., M. Schorpp, U. Wagner, and G. U. Ryffel.1986. An estrogen-responsive element derived from the 5'flanking region of the xenopus vitellogenin A2 gene functions intransfected human cells. Cell 46:1053-1061.

39. Klein-Hitpass, L., S. Y. Tsai, G. L. Greene, J. H. Clark, M.-J.Tsai, and B. W. O'Malley. 1989. Specific binding of estrogenreceptor to the estrogen response element. Mol. Cell. Biol.9:43-49.

40. Kliewer, S. A., K. Umesono, D. J. Mangelsdorf, and R. M.Evans. 1992. Retinoid X receptor interacts with nuclear recep-tors in retinoic acid, thyroid hormone and vitamin D3 signalling.Nature (London) 355:446-449.

41. Krust, A., S. Green, P. Argos, V. Kumar, P. Walter, J.-M.Bornert, and P. Chambon. 1986. The chicken oestrogen receptorsequence: homology with v-erb-A and the human estrogen andglucocorticoid receptors. EMBO J. 5:891-897.

42. Kumar, V., and P. Chambon. 1988. The estrogen receptor bindstightly to its responsive element as a ligand-induced homodimer.Cell 55:145-156.

43. Kumar, V., S. Green, G. Stack, M. Berry, J.-R. Jin, and P.Chambon. 1987. Functional domains of the human estrogenreceptor. Cell 51:941-951.

44. Lazar, M. A., and T. J. Berrodin. 1990. Thyroid hormonereceptors form distinct nuclear protein-dependent and indepen-

VOL. 13, 1993

2268 SEGARS ET AL.

dent complexes with a thyroid hormone response element. Mol.Endocrinol. 4:1627-1635.

45. Lazar, M. A., T. J. Berrodin, and H. P. Harding. 1991. Differ-ential DNA binding by monomeric, homodimeric, and poten-tially heteromeric forms of the thyroid hormone receptor. Mol.Cell. Biol. 11:5005-5015.

46. Leid, M., P. Kastner, R. Lyons, H. Nakshatri, M. Saunders, T.Zacharewski, J.-Y. Chen, A. Staub, and P. Chambon. 1992.Purification, cloning, and RXR identity of the HeLa cell factorwith which RAR or TR heterodimerizes to bind target se-quences efficiently. Cell 68:377-395.

47. Levin, A. A., L. J. Sturzenbecker, S. Kazmer, T. Bosakowski, C.Huselton, G. Allenby, J. Speck, C. Kratzeisen, M. Rosenberger,A. Lovey, and J. F. Grippo. 1992. 9-Cis retinoic acid stereoiso-mer binds and activates the nuclear receptor RXRox. Nature(London) 355:359-361.

48. Lipkin, S. M., C. A. Nelson, C. K. Glass, and M. G. Rosenfeld.1992. A negative retinoic acid response element in the ratoxytocin promoter restricts transcriptional stimulation by het-erologous transactivation domains. Proc. Natl. Acad. Sci. USA89:1209-1213.

49. Mangelsdorf, D. J., U. Borgmeyer, R. A. Heyman, J. Y. Zhou,E. S. Ong, A. E. Oro, A. Kakizuka, and R. M. Evans. 1992.Characterization of three RXR genes that mediate the action of9-cis retinoic acid. Genes Dev. 6:329-344.

50. Mangelsdorf, D. J., E. S. Ong, J. A. Dyck, and R. M. Evans.1990. Nuclear receptor that identifies a novel retinoic acidresponse pathway. Nature (London) 345:224-229.

51. Marks, M. S., P. L. Hallenbeck, T. Nagata, J. H. Segars, E.Appella, V. M. Nickodem, and K. Ozato. 1992. H-2RIIBP(RXRP) heterodimerization provides a mechanism for combina-torial diversity in the regulation of retinoic acid and thyroidhormone responsive genes. EMBO J. 11:1419-1435.

52. Marks, M. S., B.-Z. Levi, J. H. Segars, P. H. Driggers, S.Hirschfeld, T. Nagata, E. Appella, and K. Ozato. 1992. H-2RI-IBP expressed from baculovirus vector binds to multiple hor-mone response elements. Mol. Endocrinol. 6:219-230.

53. Marks, M. S., and K. Ozato. Unpublished data.54. Martinez, E., Y. Dusserre, W. Wahli, and N. Mermod. 1991.

Synergistic transcriptional activation by CTF/NF-I and theestrogen receptor involves stabilized interactions with a limitingtarget factor. Mol. Cell. Biol. 11:2937-2945.

55. Martinez, E., F. Givel, and W. Wahli. 1987. The estrogen-responsive element as an inducible enhancer: DNA sequencerequirements and conversion to a glucocorticoid-responsiveelement. EMBO J. 6:3719-3727.

56. Martinez, E., and W. Wahli. 1989. Cooperative binding ofestrogen receptor to imperfect estrogen-responsive DNA ele-ments correlates with their synergistic hormone-dependent en-hancer activity. EMBO J. 8:3781-3791.

57. Martinez, E., and W. Wahli. Unpublished data.58. Maurer, R. A., and A. C. Notides. 1987. Identification of an

estrogen-responsive element from the 5'-flanking region of therat prolactin gene. Mol. Cell. Biol. 7:4247-4254.

59. Metzger, D., J. H. White, and P. Chambon. 1988. The humanoestrogen receptor functions in yeast. Nature (London) 334:31-36.

60. Meyer, M.-E., A. Pornon, J. Ji, M.-T. Bocquel, P. Chambon,and H. Gronemeyer. 1990. Agonistic and antagonistic activitiesof RU486 on the functions of the human progesterone receptor.EMBO J. 12:3923-3932.

61. Naar, A. M., J.-M. Boutin, S. M. Lipkin, V. C. Yu, J. M.Holloway, C. K. Glass, and M. G. Rosenfeld. 1991. The orien-tation and spacing of core DNA-binding motifs dictate selectivetranscriptional responses to three nuclear receptors. Cell 65:1267-1279.

62. Nagata, T., J. H. Segars, B.-Z. Levi, and K. Ozato. 1992.Retinoic acid-dependent transactivation of major histocompati-bility complex class I promoters by the nuclear hormone recep-tor H-2RIIBP in undifferentiated embryonal carcinoma cells.Proc. Natl. Acad. Sci. USA 89:937-941.

63. Nunez, A.-M., M. Berry, J.-L. Imler, and P. Chambon. 1989.The 5' flanking region of the pS2 gene contains a complex

enhancer region responsive to oestrogens, epidermal growthfactor, a tumour promoter (TPA), the c-Ha-ras oncoprotein andthe c-jun protein. EMBO J. 8:823-829.

64. Petty, K. J., B. Desvergne, T. Mitsuhashi, and V. M. Nikodem.1990. Identification of a thyroid hormone response element inthe malic enzyme gene. J. Biol. Chem. 265:7395-7400.

65. Ponglikitmongkol, M., J. H. White, and P. Chambon. 1990.Synergistic activation of transcription by the human estrogenreceptor bound to tandem responsive elements. EMBO J.9:2221-2231.

66. Ratko, T. A., C. J. Detrisac, N. M. Dinger, C. F. Thomas, G. J.Kelloff, and R. C. Moon. 1989. Chemopreventive efficacy ofcombined retinoid and tamoxifen treatment following surgicalexcision of a primary mammary cancer in female rats. CancerRes. 49:4472-4476.

67. Sabbah, M., F. Gouilleux, B. Sola, G. Redeuilh, and E.-E.Baulieu. 1991. Structural differences between the hormone andantihormone estrogen receptor complexes bound to the hor-mone response element. Proc. Natl. Acad. Sci. USA 88:390-394.

68. Seiler-Tuyns, A., P. Walker, E. Martinez, A.-M. Merillat, F.Givel, and W. Wahli. 1986. Identification of estrogen-responsiveDNA sequences by transient expression experiments in a hu-man breast cancer cell line. Nucleic Acids Res. 14:8755-8770.

69. Sharif, M., and M. L. Privalsky. 1991. v-erbA oncogene func-tion in neoplasia correlates with its ability to repress retinoicacid receptor action. Cell 66:885-893.

70. Shupnik, M. A., C. M. Weinmann, A. C. Notides, and W. W.Chin. 1989. An upstream region of the rat luteinizing hormone 1gene binds estrogen receptor and confers estrogen responsive-ness. J. Biol. Chem. 264:80-86.

71. Shyamala, G., W. Schneider, and D. Schott. 1990. Developmen-tal regulation of murine mammary progesterone receptor geneexpression. Endocrinology 126:2882-2889.

72. Somasekhar, M. B., and J. Gorski. 1988. An estrogen-respon-sive element from the 5'-flanking region of the rat prolactin genefunctions in MCF-7 but not in HeLa cells. Gene 69:23-28.

73. Sparano, J. A., and K. O'Boyle. 1992. The potential role forbiological therapy in the treatment of breast cancer. Semin.Oncol. 19:333-341.

74. Tasset, D., L. Tora, C. Fromental, E. Scheer, and P. Chambon.1990. Distinct classes of transcriptional activating domainsfunction by different mechanisms. Cell 62:1177-1187.

75. Tora, L., J. White, C. Brou, D. Tasset, N. Webster, E. Scheer,and P. Chambon. 1989. The human estrogen receptor has twoindependent nonacidic transcriptional activation functions. Cell59:477-487.

76. Umesono, K., K. K. Murakami, C. C. Thompson, and R. M.Evans. 1991. Direct repeats as selective response elements forthe thyroid hormone, retinoic acid, and vitamin D3 receptors.Cell 65:1255-1266.

77. Walker, P., J.-E. Germond, M. Brown-Luedi, F. Givel, and W.Wahli. 1984. Sequence homologies in the region preceding thetranscription initiation site of the liver estrogen-responsivevitellogenin and apo-VLDLII genes. Nucleic Acids Res. 12:8611-8626.

78. Wang, L.-H., S. Y. Tsai, R. G. Cook, W. G. Beattie, M.-J. Tsai,and B. W. O'Malley. 1989. COUP transcription factor is amember of the steroid receptor superfamily. Nature (London)340:163-166.

79. Webster, N. J. G., S. Green, J. R. Jin, and P. Chambon. 1988.The hormone-binding domains of the estrogen and glucocorti-coid receptors contain an inducible transcription activationfunction. Cell 54:199-207.

80. Yu, V. C., C. Delsert, B. Andersen, J. M. Holloway, 0. V.Devary, A. M. Naiair, S. Y. Kim, J.-M. Boutin, C. K. Glass, andM. G. Rosenfeld. 1991. RXRI: a coregulator that enhancesbinding of retinoic acid, thyroid hormone, and vitamin D recep-tors to their cognate response elements. Cell 67:1251-1266.

81. Zhang, X.-K., B. Hoffmann, P. B.-V. Tran, G. Graupner, andM. Pfahl. 1992. Retinoid X receptor is an auxiliary protein forthyroid hormone and retinoic acid receptors. Nature (London)355:441-445.

MOL. CELL. BIOL.