Differential modulation of transcriptional activity of estrogen receptors by direct protein-protein interactions with the T cell factor family of transcription factors.

Two major signaling pathways, those triggered by estrogen (E(2)) and by the Wnt family, interact in the breast to cause growth and differentiation. The estrogen receptors ER(alpha) and ER(beta) are activated by binding E(2) and act as ligand-dependent transcription factors. The effector for the Wnt family is the Tcf family of transcription factors. Both sets of transcription factors recognize discrete but different nucleotide sequences in the promoters of their target genes. By using transient transfections of reporter constructs for the osteopontin and thymidine kinase promoters in rat mammary cells, we show that Tcf-4 antagonizes and Tcf-1 stimulates the effects of activated ER/E(2). For mutants of the former promoter, the stimulatory effects of ER(alpha)/E(2) can be made to be dependent on Tcf-1, and for the latter promoter the effects of the T cell factors (TCFs) are dependent on ER/E(2). Direct interaction between ERs and Tcfs either at the Tcf/ER(alpha)-binding site on the DNA or in the absence of DNA is established by gel retardation assays or by coimmunoprecipitation/biosensor methods, respectively. These results show that the two sets of transcription factors can interact directly, the interaction between ERs and Tcf-4 being antagonistic and that between ERs and Tcf-1 being synergistic on the activity of the promoters employed. Since Tcf-4 is the major Tcf family member in the breast, it is suggested that the antagonistic interaction is normally dominant in vivo in this tissue.

The mammary gland grows and develops in response both to systemic hormones, which circulate in the body, and to locally produced signaling molecules in the mammary gland itself (1). An example of the former is estrogen (E 2 ) 1 , which interacts with the estrogen receptors ER ␣ or ER ␤ to produce activated transcription factors (2,3), and an example of the latter is the Wnt family of proteins, the effector of which is the transcription factor family of T cell factors (Tcfs) (4,5). Activated ER ␣ and ER ␤ receptors recognize specific DNA response elements (ERE), DNA sequences located within the regulatory regions of target genes (6), whereas the Tcf family of transcription factors recognize a different set of consensus sequences typified by the (A/T)(A/T)CAAAG sequence for Tcf-1 in lymphocytes (7). However, it is unknown if these two sets of transcription factors can interact with each other to modify the transcriptional output from target genes.
To investigate if the E 2 and Wnt signaling pathways in mammary development can interact at the level of their effector transcription factors, we have chosen the target gene of osteopontin (OPN) and its cognate promoter for such a study. OPN, despite its occurrence in bone, is thought to play a key role in mammary development, since it is specifically overexpressed during pregnancy and lactation (8), and targeted inhibition of its expression causes suppression of lobular alveolar structures and lactational deficiencies in transgenic mice and cultured mammary epithelial cells (9). Moreover, OPN is secreted by epithelial cells of mammary origin in response to inter alia activation of ER (10) and can also be produced by sequestration of Tcf-4 (11) by CAAAG-containing DNA fragments in the cell line rat mammary 37 (Rama 37) (12). The 2.3-kbp promoter for the rat OPN contains two half-EREs or SF1 response elements (SFREs) (13,14), which are activable as in the mouse (15), and also three Tcf recognition sequences (16), which act to suppress its activity in the Rama 37 cells in culture (17). We now use transient transfections into the Rama 37 cell line of reporter constructs for the OPN promoter to investigate the effects of interaction of the ERs and Tcfs at a promoter that contains sites for both sets of transcription factors.
The 2.3-kbp fragment of the 6-kbp rat OPN promoter (16) from Dr. A. Ridall, University of Texas, was amplified by PCR with Taq polymerase using AAG CCT GGA TGT CCT TCT CTG CTT and GTC GAC ACT GCA AAG CCA AGG ATG as forward and reverse primers, cloned into PCR 2.1 vector (Invitrogen, Carlsbad, CA), released by digestion with HindIII and SalI and then coupled to a firefly luciferase reporter construct as described by the manufacturer (Promega, Madison, WI). The OPN promoter firefly-luciferase reporter constructs with point mutated inactive half-estrogen response elements (SFREs) (13,14) were generated with the QuikChange Site-directed Mutagenesis kit (Stratagene) using for the mutated SFRE at position 1551 (OPNS 1 M) ATG AGG TTC GTG TCT CTA GAG CTC AGT GGA GGC ACG AGA GGA AT and ATT CCT CTC GTG CCT CCA CTG AGC TCT AGA GAC ACG AAC CTC AT as forward and reverse primers, respectively, and using for the mutated SFRE at position 1579 (OPNS 2 M) GAG GCA CGA GAG GAA TTC GGG CTC ACT GTG TGC TTT GTG CAG AT and ATC TGC ACA AAG CAC ACA GTG AGC CCG AAT TCC TCT CGT GCC TC as forward and reverse primers, respectively. The OPN promoter reporter construct containing both mutated SFREs (OPNS 1 S 2 M) was generated by a combination of the above. Tcf-4 and Tcf-1 cDNAs were obtained from Prof H. Clevers, University of Utrecht, Holland (5), and were separately cloned into the pBK-CMV expression vector as above. All the latter Tcf-related oligonucleotides and vectors have been characterized previously (12,17). The expression vector for the transcription factor protein c-Fos was obtained from Professor I. Verma (Salk Institute, CA). This recognizes the AP1-binding site TGA GTC AG (26).
Cell Lines and Transient Transfections-Rama 37 cell line (27) was cultured in Dulbecco's modified Eagle's medium, 10% (v/v) fetal calf serum, 100 g/ml penicillin, 100 g/ml streptomycin (all Life Technologies, Inc.), harvested, and seeded in multiwell plates at 2.5 ϫ 10 5 cells/3.5-cm well in 1 ml of serum-free medium. After 24 h, cells were cotransfected with the predetermined optimum amounts of the following where indicated: 200 ng of ER ␣ , HEG19 (AF-2), HE15 (AF-1), L540Q (ER ␣ M), ER ␤ expression vector, 500 ng of Tcf-4, Tcf-1, c-Fos protein expression vector, 20 ng of pM12 or pM12/D-DNA vector, 1 g of OPN or related mutant promoter firefly-luciferase reporter construct (DLR, Promega), 1 g of ERE.TK or 2ERE.TATA promoter chloramphenicol acetyltransferase (CAT) reporter construct. The assays for the firefly luciferase-linked promoter also contained a control expression vector of 5 ng of pRL Renilla luciferase (DLR, Promega), whereas those for the CAT-linked promoters contained 150 ng of pSV-␤-galactosidase control expression plasmid. The concentration of pRL Renilla luciferase was reduced to avoid interference with the firefly luciferase assay in accordance with the manufacturer's instructions. 10 Ϫ8 M E 2 (Sigma) and 10 Ϫ6 M of the pure antiestrogen ICI 164384 (gifts of Dr. Alan Wakeling, AstraZeneca Pharmaceuticals, Macclesfield, UK) were added, where appropriate, to the culture medium after a further 24 h (28). Cells were incubated for a further 24 h and harvested in 300 l of Reporter Lysis Buffer (Promega), and either firefly luciferase and control Renilla luciferase were assayed as described in the Dual Luciferase Reporter Assay System (Promega), or CAT and control ␤-galactosidase activities were assayed as described previously in 100-and 150-l aliquots, respectively (12,22). Firefly luciferase activity was normalized to Renilla luciferase activity, and CAT activity was normalized to ␤-galactosidase activity; maximum activity was reached by 48 h, and results at 72 h were similar. Control key experiments using OPN promoter-CAT reporter constructs (17) gave the same results as the OPN promoter firefly-luciferase reporter constructs used here.
Gel Shift Assays-These were performed (7) by incubating 0.5 ng of 32 P-pM12-DNA, pM12/D-deleted DNA or SFRE-DNA (specific activity ϳ5 ϫ 10 8 cpm/g) with 4 l of protein extract from a reticulocyte cell-free transcription-translation protein synthesizing lysate pre-treated with 10 Ϫ6 M E 2 for 1 h at 20°C (12,22), in 30 l 0.03 M KCl, 1 mM Na 2 HPO 4 , 0.01 M HEPES (pH 7.9), 0.25 mM dithiothreitol, 10% (w/v) glycerol, 1.4 g of poly(dI-dC) (Amersham Pharmacia Biotech), 1 g of single-stranded salmon sperm DNA, 0.01% (w/v) SDS for 40 min at 0°C. Where necessary mouse monoclonal antibody (mAb) to ER ␣ or to ER ␤ was mixed with 4 l of lysate containing the indicated translation product for 30 min at 0°C, and then the combinations were added to the reaction mixtures and the incubations continued for a further 40 min at 0°C. Samples were electrophoresed through nondenaturing 4% (w/v) polyacrylamide gels, which were dried, exposed to Fuji X-Omat film for 18 h with an intensifying screen, and processed for autoradiography. The reticulocyte lysate was incubated with the expression vectors for ER ␣ , for ER ␤ , for Tcf-4, for Tcf-1 or with buffer alone in a cell-free protein-synthesizing transcription-translation system to generate the relevant protein as described previously (12,22). The mAbs to human ER ␣ and ER ␤ (15) were obtained from Santa Cruz Biotechnology.
Complex Detection by Immunoprecipitation-For immunoprecipitation in vitro, products were generated in a coupled transcription-translation cell-free protein-synthesizing reticulocyte lysate with [ 35 S]methionine and the requisite expression vectors and immunoprecipitated with mAbs to the different transcription factors. For coimmunoprecipitation in vitro, potential complexes between [ 35 S]methionine-labeled Tcf or control c-Fos products and nonradioactive ERs, usually pretreated with 10 Ϫ6 M E 2 for 1 h at 20°C, were immunoprecipitated with mAbs to the ERs above (29). In detail templates (ER ␣ /mutants, ER ␤ , c-Fos, Tcf-1, and Tcf-4) were prepared using a Hybaid kit (Hybaid, Middlesex, UK) and resuspended in RNase-free distilled water. An in vitro coupled transcription and translation kit (T7/T3-TNT; Promega, Madison, WI) was used according to the manufacturer's instructions. After completion of the 90-min reaction, samples were kept on ice. One ml of immunoprecipitation buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.2 mM Na 3 VO 4 , 0.5% (w/v) Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 25 g/ml leupeptin, 25 g/ml aprotinin, and 25 g/ml pepstatin) was added to each sample, mixed, and incubated on ice for 30 min. Twenty l of protein A/G-agarose (PAGA), pre-washed three times in immunoprecipitation buffer, were added to each sample and incubated for an additional 4 h at 4°C with rotation to remove any proteins that had interacted nonspecifically with PAGA. PAGA was removed by centrifugation at 14,000 rpm for 3 min in a benchtop centrifuge. The supernatant was incubated with 2.5 g of the requisite antibody (ER ␣ , ER ␤ , Tcf-1, or Tcf-4 antibody) overnight at 4°C with rotation. Twenty l of PAGA were added to each sample and incubated at 4°C for an additional 60 min. PAGA antibody conjugates were recovered by centrifugation at 14,000 rpm for 3 min, resuspended in 1 ml of Buffer A (phosphate-buffered saline, 0.2% (w/v) Triton X-100, and 350 mM NaCl), and centrifuged. Samples were resuspended in 1 ml of Buffer B (phosphate-buffered saline and 0.2% (w/v) Triton X-100), centrifuged, and resuspended in SDS sample buffer. Samples were resolved by electrophoresis on 10% (w/v) polyacrylamide gels at 200 V for 45 min with equal amounts of protein being loaded per lane. The gel was fixed for 30 min in 10% (v/v) propanol, 10% (v/v) acetic acid, dried under vacuum, and exposed to Kodak X-Omat AR X-ray film for 6 -24 h before developing the film.
Production and Analysis of Biotinylated Products-The biotinylated Tcf proteins were produced by adding 1 l of Transcend tRNA kit (Promega) containing biotinylated ⑀-amino lysyl-tRNA to 50 l of the same TNT transcription-translation coupled system as above primed by the expression vectors for Tcf-4 or Tcf-1, incubated at 30°C, and then cooled on ice for 60 min (29). For analysis of biotinylated products, 2-l samples of the reactants were immunoprecipitated by mAbs to Tcf-4 or Tcf-1, electrophoresed through SDS-polyacrylamide gels as above, and the separated proteins transferred by blotting onto Immobilon membranes (Millipore Corp., Watford, UK) (30). The membranes were incubated with blocking buffer containing 0.02 M Tris-HCl (pH 7.0), 0.9% (w/v) NaCl, 0.5% (v/v) Tween 20 for 1 h with gentle shaking and then with 1:10,000 dilution of the streptavidin-horseradish peroxidase conjugate. The protein bands were visualized with Super Signal West Pico chemiluminescence system (Pierce and Wariner, Chester, UK) and photographed by exposure to Kodak X-Omat AR X-ray film for 2 min, and then the film was developed.
Complex Detection by Direct Binding-Binding reactions were carried out in an Iasys two channel resonant mirror biosensor at 20°C on planar aminosilane surfaces (Affinity Sensors, Saxon Hill, Cambridge, UK). Biotinylated Tcfs were captured directly onto streptavidin-derivatized aminosilane surfaces from 20 l of the above proteinsynthesizing reticulocyte lysates. Alternatively, recombinant human ER ␣ was captured on the aminosilane surface by cross-linking with Bis(sulfosuccinimidyl)suberate (Pierce and Wariner). The distribution of the immobilized Tcfs and of the bound ER ␣ and ER ␤ on the surface of the biosensor cuvette was inspected by examination of the resonance scan, which showed that at all times these molecules were distributed uniformly on the sensor surface and therefore were not microaggregated. A single binding reaction consisted of adding 5 l of lysate to the surface of a cuvette containing 25 l of phosphatebuffered saline with 0.02% (v/v) Tween 20 (31). The amount of protein bound (ϮS.E.) during the association phase was calculated using the non-linear curve-fitting FastFit software (Affinity Sensors) provided with the instrument. Regeneration of the surfaces was achieved with three 50-l washes with 20 mM HCl but caused a 10 -40% loss of ER ␣ /ER ␤ binding capacity. Experiments were repeated on three independently prepared surfaces.

Effect of ERs/Tcfs on the OPN Promoter-Report Construct-
Transient transfection of the rat OPN-promoter luciferase-reporter construct into Rama 37 cells was undertaken with predetermined optimum concentrations of expression vectors for various ER/Tcf-related transcription factors and of the same vector containing a 20-bp oligonucleotide containing the recognition sequence for the Tcf family of transcription factors (12,17). Since these mammary cells contain only Tcf-4 and not Tcf-1, the effect of addition of this oligonucleotide was to sequester preferentially Tcf-4 and prevent it working in the transient transfection assay (12,17). Individual expression vectors containing ER ␣ cDNA in the presence of E 2 (ER ␣ /E 2 ), containing an active Tcf-binding site (pM12-DNA), or containing Tcf-1 cDNA significantly stimulated (3-6-fold), and a vector containing Tcf-4 cDNA significantly reduced (2-fold) the activity of the OPN promoter-reporter construct (Student's t test, p Յ 0.015) ( Table I). These results are consistent with the fact that inhibitory Tcf-4 and not the stimulatory Tcf-1 occurs in the Rama 37 cells (12,17) and hence sequestration of any Tcfs by pM12-DNA results in a net stimulation of the OPN promoter-reporter construct. The stimulation produced by the expression vector for ER ␣ /E 2 was significantly increased by a further 1.5-or 2-fold, respectively, when vectors containing pM12-DNA or Tcf-1 DNA or reduced by about 10-fold when a vector containing Tcf-4 cDNA were included in the reaction mixtures (p Յ 0.02) ( Table I). The pM12/D-DNA without a core Tcf binding sequence CAAAG was without effect (not shown).
The stimulation achieved by the expression vector for ER ␣ / E 2 , either alone or in combination with other additives, was abolished completely when the anti-estrogen ICI 164384 (ICI) was present in the culture medium, but ICI was without effect on the stimulation achieved by vectors containing pM12-DNA or Tcf-1 cDNA (Table I). The lack of effect on the OPN promoter-reporter construct of ICI in the presence of pM12-DNA indicates that the increase produced by pM12-DNA is not due to ER activity. In contrast the expression vector for ER ␤ /E 2 failed to stimulate significantly the OPN promoter-reporter construct alone or enhance its activity with vectors containing pM12-DNA or Tcf-1 cDNA (p Ն 0.3). These results suggest that Tcf-4 and Tcf-1 alter the activity of the OPN promoter-reporter construct at a site different to that for ER ␣ /E 2 , which in turn is not recognized by ER ␤ /E 2 .
Effect of ERs/Tcfs on the ERE Promoter-Reporter Constructs-Transient transfection of promoter CAT reporter constructs containing only sequences recognized by ER ␣ /E 2 and not the Tcf CAAAG core recognition sequences gave different results. Thus the activity of the TK promoter-reporter construct, which contained a consensus estrogen responsive element (ERE.TK), and the minimal estrogen-responsive promoter-reporter construct (2ERE.TATA) were not altered significantly in Rama 37 cells by cotransfection of optimal amounts of Tcf-1, Tcf-4 expression vectors, nor by pM12-DNA (p Ն 0.3). However, cotransfection of Rama 37 cells with optimal amounts of Tcf-1, Tcf-4 expression vectors, or of pM12-DNA in the presence of ER ␣ /E 2 significantly altered the activities of these two promoter-reporter constructs over those achieved by expression vectors for ER ␣ /E 2 alone (p Յ 0.01). Vectors for Tcf-1 and pM12-DNA increased these promoter-reporter activities by 1.3-1.5-fold and for Tcf-4 suppressed their activities by 5-10-fold (Table I). The stimulated activities of the two ERE promoter-reporter constructs produced by cotransfections of expression vectors for ER ␣ /E 2 and for Tcf-1, ER ␣ /E 2 and pM12-DNA were this time completely abolished by ICI in the culture medium (p Ն 0.7). Once again the stimulations achieved by cotransfection of Rama 37 cells with pM12-DNA were lost when it was replaced by the CAAAG-deleted pM12/D-DNA (not shown). The results for the activities of the ERE.TK promoterreporter construct were similar for cotransfection of Rama 37 TABLE I Effect of transcription factors on activation of promoters Rama 37 cells were transiently cotransfected with the luciferase promoter reporter gene coupled downstream from the OPN promoter or with the CAT reporter gene coupled downstream from the ERE.TK or downstream of the 2ERE.TATA together with optimal concentrations of expression vectors for ER ␣ , ER ␤ , Tcf-4, Tcf-1, and/or pM12-DNA. Transfections were also conducted without or with estradiol (E 2 ) and without or with ICI in the culture medium. The mean Ϯ S.D. of the promoter activity of the promoter-reporter constructs with various additions relative to the promoter-reporter constructs alone for three separate experiments is shown for the optimum inputs of the different additives.
cells with the ER ␤ expression vector, and the stimulation achieved by ER ␤ /E 2 was altered in the same way by cotransfection with Tcf-1, Tcf-4 expression vectors or with pM12-DNA (stimulation 1.6 -1.8-fold, inhibition 5-fold), and the stimulated activities were abolished completely by ICI (not shown). The expression vector for ER ␤ /E 2 , however, was unable to stimulate the activity of the 2ERE.TATA promoter-reporter construct under any conditions (Table I). These results show that the effects with Tcf-1, Tcf-4, and pM12-DNA on promoters containing only EREs are totally dependent on the presence of the ER and E 2 and suggest that the relevant Tcfs may bind to the ERs themselves for transactivation. Specificity of the Effect of ER ␣ /Tcf-1 on the OPN Promoter-Reporter Construct-The estrogen receptor consists of several functional domains including the ligand-binding site and the DNA-binding site (18,21). However, only the expression vector HEG19 with both functional binding domains, AF-2 with E 2 , increased the activity of the OPN promoter-luciferase reporter construct by about the same 6 -7-fold value as the expression vector for the complete ER ␣ with E 2 (Student's t test, p ϭ 0.32). Expression vectors for the transactivation inactive L540Q mutant and for HE15 with a functional DNA binding domain without an active hormone-binding site AF-1 were without significant effect (p Ն 0.24). The expression vector for a transcription factor without a recognition site on the OPN promoter (16), c-Fos (26), was also included in the transient transfection assays. By itself the expression vector for c-Fos was without effect, and it did not modify the stimulatory activity of the OPN promoter-reporter construct produced separately by the expression vector for ER ␣ /E 2 nor by the expression vector for Tcf-1 (Student's t test p Ն 0.25) ( Table II). The enhanced 2.5-3.5-fold stimulation of the OPN promoter-reporter con-struct produced by simultaneous addition of expression vectors for ER ␣ /E 2 and Tcf-1 over those obtained by each expression vector alone was lost completely without E 2 (p ϭ 0.25), showing the dependence for this stimulatory effect on the presence of the ligand E 2 .
The rat OPN promoter (16) contains two possible estrogen control sites in the form of two potential SF1 response elements (SFREs) (13,14). When transfected with the expression vector for ER ␣ /E 2 in Rama 37 cells, the OPN promoter-reporter constructs mutated in the first (OPNS 1 M) or second (OPNS 2 M) SFRE produced significant 61% (p ϭ 0.010) or near significant 26% (p ϭ 0.058) decreases in activity, respectively, compared with the nonmutated promoter-reporter construct. The activities of these two promoter-reporter constructs with the expression vector for ER ␣ /E 2 were also significantly different from one another (p ϭ 0.026), and that for the double mutant with no active SFREs (OPNS 1 S 2 M) was not significantly different from control (p ϭ 0.51) ( Table II). As anticipated, this doubly mutated OPN operator-reporter construct was still stimulable to the same degree as the wild-type construct by the expression vector for Tcf-1 (p ϭ 0.57). When transfected simultaneously with the expression vectors for ER ␣ /E 2 and for Tcf-1, the activity of the doubly mutated OPN promoter-reporter construct was still increased by 2.5-fold over that with the Tcf-1 expression vector alone (Table II). This increase in activity was not significant without the E 2 ligand (p ϭ 0.12). Moreover, when expression vectors for the different domains of ER ␣ and for Tcf-1 were transfected with the doubly mutated OPN promoterreporter construct, only the vector expressing the AF-2 domain produced a similar 2.3-fold stimulation over that with the expression vector for Tcf-1 alone (p ϭ 0.21), the expression vectors for the point mutated ER ␣ and for AF-1 were without significant effect over the control (p Ն 0.12) (Table II). This stimulation of the doubly mutated OPN promoter-reporter construct by the expression vector for AF-2 in the presence of Tcf-1 was also dependent on E 2 (p ϭ 0.007) and was also not significant without E 2 (p ϭ 0.13) (Table II). These results suggest that ER ␣ can interact with Tcf-1 through its AF-2 domain in an E 2 -dependent manner to activate the OPN promoter-reporter construct at its Tcf recognition sites.

Effect of ERs/Tcfs on Gel Mobilities of CAAAG-and SFREcontaining DNA-To test whether ERs could interact with the
Tcfs when the latter were bound to the promoter region through their CAAAG core recognition sequence, the 32 Plabeled 20-bp double-stranded oligonucleotide 32 P-pM12-DNA was incubated with reticulocyte lysates in which ER ␣ /E 2 and/or Tcf-4 was expressed from the appropriate vectors in a coupled transcription-translation system. Without the expression vectors the reticulocyte lysate failed to retard the electrophoretic mobility of pM12-DNA on polyacrylamide gels (Fig. 1A). When the Tcf-4 expression vector was included in the protein-synthesizing lysate, a single retarded band for pM12-DNA was produced (Fig. 1A, lanes 1 and 8). When both Tcf-4 and ER ␣ /E 2 expression vectors were included in two separate lysates, a further reduction in the mobility of pM12-DNA was observed, and this component now split into two bands (Fig. 1A, lanes 2  and 9). When the 32 P-pM12-DNA was replaced by the CAAAGdeleted pM12/D-DNA and incubated with ER ␣ and Tcf-4-containing lysates, no retarded bands were observed (Fig. 1A,  lanes 4 and 12), and no retarded bands were observed when pM12-DNA was incubated with lysates containing only ER ␣ /E 2 (Fig. 1A, lane 5). Antibodies to ER ␣ but not to ER ␤ , when added to incubations containing 32 P-pM12-DNA and combined ER ␣ /E 2 and Tcf-4 lysates produced a similar-sized single band as with Tcf-4 lysates alone (Fig. 1A, lanes 10 and 11). The same results were obtained when lysates contained ER ␤ instead of TABLE II Specificity of ER ␣ /Tcf-1 in activation of the OPN promoter Rama 37 cells were transiently cotransfected with the luciferase reporter gene coupled downstream from the OPN promoter or one of its mutated forms with one (OPNS 1 M or OPNS 2 M) or both (OPNS 1 S 2 M) of its SFREs missing and optimal concentrations of expression vectors for ER ␣ , L540Q mutant of ER ␣ (ER ␣ M), HE15 an ER ␣ mutant containing only AF-1, HEG19 an ER ␣ mutant containing only AF-2, c-Fos, and/or Tcf-1. Transfections were also conducted without or with E 2 in the culture medium.
The mean Ϯ S.D. of the activity of the OPN promoter-reporter constructs with various additions relative to that for the OPN promoterreporter construct alone for 2 or 4 a separate experiments is shown for the optimum inputs for the different additives.
To test the reciprocal interaction between Tcf-4 and ER ␣ when the latter was bound to the promoter region through its SFRE recognition sequence, the 32 P-labeled 29-bp doublestranded oligonucleotide 32 P-SFRE-DNA was incubated with similar reticulocyte lysates expressing ER ␣ and/or Tcf-4 to those above. Without the expression vectors, the reticulocyte lysate failed to retard the electrophoretic mobility of SFRE-DNA on polyacrylamide gels (Fig. 1B). When the ER ␣ expression vector with E 2 was included in the lysate, a major retarded band for SFRE-DNA was observed (Fig. 1B, lane 1). When the products from both ER ␣ /E 2 and Tcf-4 expression vectors were included from the two separate lysates, a further reduction in the mobility of SFRE-DNA was observed (Fig. 1B, lane 2). No retarded bands were observed when SFRE-DNA was incubated with lysates containing only Tcf-4 (Fig. 1B, lane 3). These results suggest that ERs/E 2 and Tcfs can interact directly when bound to DNA.
Coimmunoprecipitation of ERs and Tcfs-To investigate whether the ERs and Tcfs could interact directly without the necessity to bind to DNA, coimmunoprecipitation of the relevant molecules produced by suitable expression vectors in coupled transcription-translation reticulocyte lysates was attempted. That the correct molecules were produced in these lysates was confirmed by separately using the expression vectors for ER ␣ , ER ␤ , Tcf-4, and Tcf-1 to direct protein syn-thesis with [ 35 S]methionine and obtaining the correct sized 35 S-labeled proteins of 65, 62, 65, and 33 kDa, respectively, on polyacrylamide gels after immunoprecipitation with only the cognate (Fig. 2A, lanes 1, 3, 5, and 7) and not the noncognate antibody ( Fig. 2A, lanes 2, 4, 6, and 8). 35 S-Tcf-4 or 35 S-Tcf-1-containing lysates were then incubated with nonradioactive ER ␣ -and ER ␤ -containing lysates with E 2 , immunoprecipitated with the relevant anti-ER, and analyzed on polyacrylamide gels (Fig. 2B). In the presence of ER ␣ /E 2 an antibody to ER ␣ coimmunoprecipitated a 35 S-labeled protein of 65 kDa from Tcf-4-producing lysates (Fig. 2B, lane 1) and a 35 S-labeled protein of 33 kDa with Tcf-1-producing lysates (Fig. 2B, lane 2). These proteins corresponded to the major radioactive proteins in their respective lysates and possessed the same molecular weights as those reported for human Tcf-4 and Tcf-1, respectively (7,32). The same results were obtained when ER ␤ replaced ER ␣ -containing lysates, and the resultant complexes were immunoprecipitated with an antibody to ER ␤ (Fig. 2B, lanes 3 and 4). No such 35 S-labeled proteins were detected in the resultant immunoprecipitates when lysates containing ER ␣ /E 2 or ER ␤ /E 2 were omitted from the incubations (Fig. 2B, lanes 5-8). Sometimes smaller, considerably less abundant 35 S-labeled protein bands were also observed on the gels (e.g. Fig. 2B); these minor products  1-4) was incubated with proteins produced in a reticulocyte lysate by an expression vector for ER ␣ treated for 1 h with 10 Ϫ6 M E 2 (lanes 1 and 2), for Tcf-4 (lanes 2 and 3), and with proteins produced in the reticulocyte lysate alone (lane 4). The products of the incubations were analyzed on polyacrylamide gels. The unbound 32 P-pM12-DNAs or 32 P-SFRE-DNA is located at the bottom of the gels.

FIG. 2. Immunoprecipitation of combinations of proteins synthesized in cell-free extracts by expression vectors for transcription factors.
A, identification of single proteins synthesized in cell-free extracts. 35 S-Labeled proteins were produced in cell-free protein-synthesizing reticulocyte lysates by expression vectors for ER a (lanes 1 and 2), for ER ␤ (lanes 3 and 4), for Tcf-4 ( lanes 5 and 6), and for Tcf-1 (lanes 7 and 8). The resultant proteins were immunoprecipitated with a monoclonal antibody to ER ␣ (mER ␣ ) (lanes 1 and 4), to ER ␤ (mER ␤ ) (lanes 2 and 3), to Tcf-4 (mTcf-4) (lanes 5 and 8), and to Tcf-1 (mTcf-1) (lanes 6 and 7) and analyzed on polyacrylamide gels. The autoradiogram shows the position of the 35 S-labeled proteins. The positions of authentic ER ␣ , Tcf-4, ER ␤ , and Tcf-1 are shown by arrows on the left-hand side and of standard marker proteins on the right-hand side of the autoradiogram. B, coimmunoprecipitation of Tcfs and ERs. 35 S-Labeled proteins (*) produced in reticulocyte lysates by expression vectors for Tcf-4 (lanes 1, 3, and 5-8) or for Tcf-1 (lanes 2 and 4) were incubated with nonradioactive proteins produced in reticulocyte lysates by expression vectors for ER ␣ (lanes 1, 2, and 5) or for ER ␤ (lanes 3, 4, and 6) and treated for 1 h with 10 Ϫ6 M E 2 . Any resultant complexes were immunoprecipitated with a monoclonal antibody to ER ␣ (mER ␣ ) (lanes 1, 2, 5, and 7) and to ER ␤ (mER ␤ ) (lanes 3, 4, 6, and 8) and were analyzed on polyacrylamide gels. The positions of authentic Tcf-4 (7 ) and Tcf-1 (4) are indicated by arrows. might have arisen from proteolytic cleavage of the major proteins during incubation.
The specificity of the immunoprecipitated complexes formed between ERs and Tcfs was tested with an unrelated transcription factor, c-Fos, and with mutants in the activating functions of ER ␣. When 35 S-labeled c-Fos or 35 S-Tcf-4-containing lysates were incubated with nonradioactive ER ␣ /ER 2 -containing lysates and then immunoprecipitated with anti-ER ␣ , no radioactive c-Fos protein of 55 kDa (26) was observed on the resultant polyacrylamide gels (Fig. 3A, lanes 2, 5, and 6), despite the presence of radioactive Tcf-4 (Fig. 3A, lanes 1, 3 and 4). When the expression vectors for the ER ␣ deletion mutants HE15 (AF-1) and HEG19 (AF-2) were transcribed/translated in the reticulocyte lysate, they produced proteins of about 45 kDa consistent with their sizes (Fig. 3B, lanes 1 and 4); whereas that for the point mutated L540Q was similar to the parental ER ␣ (not shown). However, only the expression vector HEG19 with both ER ␣ functional domains (AF-2) produced a product that in the presence of Tcf-4-containing lysates was precipitable with mAb to Tcf-4; the expression vectors HE15 with only the DNA-binding domain (AF-1) and the transactivation-inactive L540Q failed to do so (Fig. 3B, lanes 2, 3, 5-7, 9, and 10). Moreover, the amount of 35 S-HEG19 precipitable product de-pended on the presence of Tcf-4 lysates and was enhanced by the inclusion of E 2 (Fig. 3B, lanes 6 -8). These results suggest that ER ␣ can interact specifically with Tcf-4 through its AF-2 domain in an E 2 -dependent manner.
Binding of ERs to Immobilized Tcfs-Binding of ERs to immobilized Tcfs was also explored in an optical biosensor. Biotinylated ⑀-amino lysyl-tRNA was used to synthesize biotinylated proteins in a reticulocyte lysate directed by an expression vector for Tcf-4 or Tcf-1. Immunoprecipitation of 35 S-labeled protein aliquots directed separately by Tcf-4 and Tcf-1 and their analysis on polyacrylamide gels confirmed that they produced the same molecular mass proteins of 65 and 33 kDa, respectively, with the cognate but not with the noncognate antibodies as authentic Tcf-4 and Tcf-1, and these proteins could react with streptavidin peroxidase (not shown). Three different surfaces were prepared by immobilizing the in vitro synthesized biotinylated proteins on streptavidin-derivatized biosensor surfaces for endogenously synthesized proteins and for approximately equal quantities of Tcf-4 or Tcf-1 proteins (8 -9 fmol/ mm 2 ) produced by expression vectors in the reticulocyte lysate. Addition of lysates directed by expression vectors for ER ␣ or ER ␤ to the control surface of endogenously synthesized proteins failed to increase the response over that obtained with endogenously synthesized proteins alone (Fig. 4A). In contrast, addition of ER ␣ -or ER ␤ -containing lysates to the Tcf-4 surface elicited a greater than 2-fold increase in response (Fig. 4B). Addition of ER ␣ -or ER ␤ -containing lysates to the Tcf-1 surface produced a 2-5-fold increase in binding, respectively (Fig. 4C). When a second aliquot of ER ␣ -or ER ␤ -containing lysate was added, binding only increased by a further 20 -50%, indicating that each interaction ER ␣ with Tcf-4 and Tcf-1, and ER ␤ with Tcf-4 and Tcf-1 was saturable (not shown). Within the limits of the loss of binding observed after regeneration (10 -40%) (33), inclusion of 9 nM estradiol had little effect on the interaction between ERs and Tcfs. In controls for testing the interaction the other way around, nonbiotinylated pure recombinant ER ␣ was cross-linked to the biosensor surface. Addition of lysates directed by an expression vector for Tcf-1 elicited a 2.4-fold increase in response over that with a nonprogrammed lysate, and lysates containing c-Fos were without any effect (Fig. 4D). The 5-or more fold differences in the amount of ER ␣ coating the cuvette compared with the biotinylated Tcfs reflected the different methods of attachment. These results suggest that whichever partner ER or Tcf was attached to the surface of the biosensor cuvette, they can still interact in a specific manner.

DISCUSSION
Both ER/E 2 , like other nuclear receptors (34) and the Tcf family of transcription factors (32), interact separately with accessory proteins to modulate transcription of presumptive target genes. ERs can also bind to other transcription factors and modulate their activity at non-ERE sites in promoters (35,36). Here we show that the ER/ER 2 and the Tcfs that recognize entirely different sequences in the promoter can interact and form a complex using two independent methods, coimmunoprecipitation and refractive index changes in a biosensor. We also show that this complex can assemble at a Tcf recognition sequence on a small 20-bp oligonucleotide and at an SFRE recognition sequence on another separate 29-bp oligonucleotide, as judged by gel mobility shift experiments. Although the ERs and Tcfs used in the experiments reported herein have been synthesized from expression vectors in cell-free coupled transcription-translation systems, control experiments demonstrate that they are the major products synthesized and that there are no complexes formed with material from unprimed cell-free systems alone. Moreover, reticulocyte protein-synthesizing lysates that contain a transcription factor c-Fos, with no  1, 3, and 4) or for c-Fos (lanes 2, 5, and 6) were applied directly to the gel (lanes 1 and 2) or were incubated with nonradioactive proteins produced in reticulocyte lysates by the expression vector for ER␣ pretreated with 10 Ϫ6 M E 2 for 1 h (lanes 4 and 6). Any resultant complexes were immunoprecipitated with mAbs to ER␣ (lanes 3-6) and analyzed on polyacrylamide gels. The positions of authentic Tcf-4 and c-Fos are indicated. B, specificity of ER␣/Tcf-4 interaction. 35 S-Labeled proteins (*) produced in reticulocyte lysates by expression vectors HE15 for AF-1 (lanes 1-3), HEG19 for AF-2 (lanes 4 -8), or for the point mutated ER ␣ L540Q (ER␣M) (lanes 9 and 10) were applied either directly to the gel (lanes 1 and 4) or were incubated with nonradioactive proteins produced in reticulocyte lysates by the expression vector for Tcf-4 (lanes 3 and 7-10). The products of HEG19 lysates (AF-2) and of L540Q lysates were also preincubated with 10 Ϫ6 M estradiol (E 2 ) before inclusion (lanes 6, 8 and 10). Any resultant complexes were immunoprecipitated with mAb to Tcf-4 (mTcf-4) (lanes 2, 3, and 5-10) and analyzed on a polyacrylamide gel. The positions of the products of HE15 (AF-1) and HEG19 (AF-2) are indicated by the arrow. recognition sites on the rat OPN promoter-reporter construct (37,38), fail to increase its activity with or without ER ␣ /Tcf-1 and also fail to bind to ER ␣ /Tcf in either the coimmunoprecipitation or biosensor assays. This result eliminates the possibility of nonspecific interaction with any transcription factor. However, the identification of protein complexes in vitro does not necessarily prove that such complexes occur in vivo and direct recourse to other types of experiments that assay for complex formation in vivo, such as the yeast one-hybrid system (39,40), will be necessary to establish this fact unequivocally.
The functional significance of the interaction between ERs and Tcfs, either off or on the DNA, is supported by the fact that effects of Tcf-1, Tcf-4, and pM12-DNA can occur on promoters containing only EREs and not any Tcf recognition sites and that these Tcf-related effects are totally dependent on an activated ER. Moreover, when the two SFRE recognition sites are removed by standard inactivating mutations from the OPN promoter-reporter construct leaving only the Tcf recognition sites, expression vectors for ER ␣ /E 2 are still capable of stimulating this doubly mutated construct in an E 2 -dependent manner in the presence of the expression vector for Tcf-1. This result confirms that only Tcf-1-occupied Tcf sites are required for further stimulation of the OPN promoter-reporter construct by ER ␣ /E 2 . The functional results have been obtained by transient transfection of promoter-reporter constructs together with cotransfection of expression vectors for ERs and Tcfs in the Rama 37 cell system. These results are likely to reflect the activity of the endogenous promoters in the Rama 37 cells, at least for OPN, since the activity of the same promoter-reporter construct in the Rama 37 cells accurately reflects the levels of endogenous OPN mRNA produced (12,17). Moreover, control experiments have demonstrated that all the effects on the activities of the promoter-reporter constructs produced by transient cotransfection of the expression vectors for the ERs and Tcfs are dose-dependent, up to a maximum input of a particular expression vector and reflect also the levels of the appropriate ER/Tcf synthesized (not shown). Since the Rama 37 cells can mimic some of the differentiation processes observed in the normal mammary gland (41), it is possible that the results obtained by transient transfections of this cell line in culture may reflect control of transcription of endogenous genes like OPN in vivo.
At first sight it may be surprising that a vector with one Tcf-binding site, when transfected into the Rama 37 cell line, can produce a discernible titration effect on the level of available Tcf molecules in the nucleus of the resultant transfectants. However, transient transfection of the same optimum concentration of vector containing pM12-DNA results in a dramatic fall in the level of nuclear Tcf-4 protein (12,17), similar to that observed in cells permanently transfected with about 100 copies of the same DNA (17,29). Both methods of transfection presumably yield enough copies of pM12-DNA to sequester a sufficient number of Tcf molecules in the nucleus so that the majority of those bound to the OPN promoter are removed, thereby stimulating its transcription (12,17). Moreover, the stimulatory effect on the OPN promoter-reporter construct of the vector containing the Tcf-binding site can be gradually reversed by increasing concentrations of the expression vector for Tcf-4 (12,17), consistent with this idea. The stimulatory effect with pM12-DNA on the ER/E 2 -activated ERE promoters is entirely dependent on the Tcf recognition sequence contained within the pM12-DNA, since there is no effect when this Tcf site is deleted. This result suggests that, in the Rama 37 cells, it is Tcf-4 rather than Tcf-1 that is normally complexed with the ERs, thus inhibiting the activity of the two ERE promoterreporter constructs used here. Sequestration of Tcf-4 by pM12-DNA in the Rama 37 cells can then stimulate the activity of the two ERE promoter constructs still further. The fact that Tcf-4 can inhibit and Tcf-1 can stimulate the promoter-reporter activity of ER/E 2 in transient transfection assays in the Rama 37 cell system may be due to subtle differences in binding, as reported for E 2 and antiestrogens (42).
The rat OPN promoter (16) contains no classical ERE sequences but two potential SF1 response elements (SFREs), a single half-site preceded by a consensus trinucleotide (13,14). That ER ␣ /E 2 but not ER ␤ /E 2 can transactivate the rat OPN promoter-reporter construct is consistent with the report that ER ␤ /E 2 fails to transactivate the mouse OPN promoter, which also contains only SFREs (17), and with our results and the results of others (43) for the minimal 2ERE.TATA. In contrast, the more complex ERE.TK is capable of transactivation by both ER ␣ /E 2 and ER ␤ /E 2 , as reported by others (43). In addition, the FIG. 4. Biosensor detection of complexes between ERs and Tcfs. The biosensor cuvettes contained streptavidin surfaces coated with biotinylated proteins produced in reticulocyte lysates by no expression vector (A), expression vector for Tcf-4 (B), and expression vector for Tcf-1 (C). The biosensor cuvette also contained surfaces coated with nonbiotinylated pure ER ␣ (D). Their response to additions of proteins produced in the reticulocyte lysate alone or proteins produced in the lysate by expression vectors for ER ␣ or ER ␤ is shown in arc seconds (arc s). The bars represent the S.E. for each set of measurements. The amount of biotinylated material immobilized on the streptavidin surfaces corresponded to 13 (A), 305 (B), 184 (C), and 1500 (D) arc s; 600 arc s ϭ 1 ng of protein/mm 2 . rat OPN promoter (17) also contains three (A/T)(A/T)CAAAG consensus recognition sequences for the Tcfs (12). The presence of these two potential DNA-binding sites raises the possibility that ER ␣ and the Tcfs could both bind to the DNA and to each other in the OPN promoter to regulate the final signal. Since Tcf-4 and not Tcf-1 is present in the Rama 37 mammary cells (12), and sequestration of the Tcfs by transient or permanent transfection by CAAAG-containing DNAs leads to elevated levels of endogenous OPN mRNA (12), the dominant signal produced normally by the Tcfs in vivo, at least on the OPN promoter in the Rama 37 cell line, is inhibitory, and this would oppose positive stimulation by ER/E 2 .
The estrogen receptor ␣ consists of several functional domains. The amino-terminal region exhibits a hormone-independent transactivation function (AF-1), the central region is principally involved in interactions with the DNA, and the carboxyl-terminal region is involved with hormone-dependent transactivation functions (AF-2) (18,21). The results with the doubly mutated OPN promoter-reporter construct with both SFREs inactivated suggests that ER ␣ can interact with Tcf-1 through its AF-2 domain to transactivate this promoter via Tcf-binding sites on its DNA. Similar conclusions have been obtained from direct ER ␣ and Tcf-4 interactions observed by immunoprecipitation techniques. In both cases the interactions require E 2 for full effect. However, there is a very limited transactivation of the doubly mutated OPN promoter-reporter construct via Tcf-1 with lysates containing complete ER ␣ or the AF-2 product without E 2 . This limited interaction in the absence of E 2 between ER ␣ and Tcf-4 is also observed to a more marked extent when measured by immunoprecipitation techniques and is most marked when measured in the biosensor where little dependence on E 2 is preserved. However, transactivation of the OPN promoter-reporter construct is conducted inside the rat mammary cell line Rama 37, whereas immunoprecipitation is performed in the presence of reticulocyte lysates that are even more diluted in the biosensor due to the immobilization and washing processes. Thus the different conditions of assay may reflect decreasing concentrations of an inhibitor for ER ␣ whose inhibition is abrogated by binding E 2 ; one such example is heat shock protein 90 (44).
Transgenic mouse studies have separately implicated the E 2 /ER (45) and Wnt/Tcf signaling systems (46) in various developmental processes; different family members have more or less defined functions. Now we have shown that some of the different family members of the two sets of downstream transcription factors ER ␣ /ER ␤ and Tcf-1/Tcf-4 can interact directly and thereby modulate the activity of promoters for potential target genes. The precise modulation, however, depends on the identity of the family member in a particular cellular environment. In the case of the mammary gland, the major Wnt family member Wnt-4 is up-regulated in early pregnancy from a low level in virgin mice (47,48) and in reconstituted glands causes secondary branching of ducts that terminate in very incomplete and lactationally deficient alveolar-like structures (49). Similar secondarybranching structures are also seen in transgenic mice in which OPN expression is specifically inhibited in the mammary glands of pregnant/lactating animals (9), and this early developmental period in normal pregnant mice corresponds to the period of absolute minimum expression of OPN mRNA (20). Thus it is possible that Wnt and E 2 can interact to control this switch from secondary branching to lobular alveolar development through OPN in vivo, as we have shown their effector transcription factors Tcf-4 and ER ␣ can interact to control OPN promoter activity in cultured mammary cells in vitro.