Synergy of SF1 and RAR in activation of Oct-3/4 promoter.

The Oct-3/4 transcription factor is expressed in the earliest stages of embryogenesis, and is thus likely to play an important role in regulation of initial decisions in development. For the first time, we have shown that SF1 and Oct-3/4 are co-expressed in embryonal carcinoma (EC) P19 cells, and their expression is down-regulated with very similar kinetics following retinoic acid (RA) induced differentiation of these cells, suggesting a functional relationship between the two. Previously, we have shown that the Oct-3/4 promoter harbors an RA-responsive element, RAREoct, which functions in EC cells as a binding site for positive regulators of transcription, such as RAR and RXR. In this study we have identified in the Oct-3/4 promoter two novel SF1-binding sites: SF1(a) and SF1(b). The proximal site, SF1(a), is located within the RAREoct, and the distal site, SF1(b), is located between nucleotide -193 and -209 of the Oct-3/4 promoter. Both sites contribute to activation of Oct-3/4 promoter in EC cells, with SF1(a) playing a more crucial role. SF1, and its isoforms ELP2 and ELP3 bind to both SF1 sites and activate the Oct-3/4 promoter. This activation depends on the presence of SF1 DNA-binding domain. Thus, Oct-3/4 is the first EC-specific gene reported that is regulated by SF1. Interestingly, SF1 and RAR form a novel complex on the RAREoct sequence that synergistically activate the Oct-3/4 promoter. Both RARE and SF1 cis regulatory elements, as well as the SF1 DNA-binding domain, are needed for this synergism. SF1 and Oct-3/4 transcription factors play a role in the same developmental regulatory cascade.

Transcription factors play a critical role in embryonic development and cellular differentiation. One family of transcription factors which exhibits developmental functions is the POUspecific family. Members of this family share two regions of homology: a highly conserved amino-terminal domain, designated the POU-specific domain, and a more divergent carboxylterminal homeodomain (reviewed in Refs. 1 and 2).
The Oct-3/4 gene product is a member of the POU-specific family of transcription factors (3). It is expressed in totipotent and pluripotent stem cells of the pregastrulation embryo, primordial germ cells, and oocytes (4 -7). Oct-3/4 is also highly expressed in embryonic stem (ES) 1 and embryonal carcinoma (EC) cell lines. Oct-3/4 expression is down-regulated in the developing embryo upon differentiation to endoderm and mesoderm, and in EC and ES cells which are induced to differentiate in vitro following treatment with retinoic acid (EC/RA, ES/RA) (4,5,(7)(8)(9). Expression of Oct-3/4 is crucial to the establishment of pluripotent stem cells in the mammalian embryo, since Oct-3/4-deficient embryos develop to the blastocyst stage, but the inner cell mass cells are not pluripotent (10).
Studies of the down-regulation of Oct-3/4 gene expression suggest an involvement of several cis acting elements. An enhancer element (located 1.2 kb upstream of the initiation site, designated RARE1) has been characterized (11). This enhancer is necessary for a high level of promoter activity in P19 cells before RA treatment and for RA-mediated repression. Another DNA element located farther upstream (ϳ2 kb 5Ј to the initiation site) and designated the distal enhancer has also been identified (12,13). This element confers strong enhancer activity in ES cells, but not EC cells, and in mice is considered to be active specifically in the germ line lineage. We and others have previously detected an RA-responsive element, RAREoct, present in the promoter region of the Oct-3/4 gene (14 -16). This RAREoct motif functions in EC cells as a binding site for positive regulators of transcription, and in RA-differentiated cells as a binding site for negative regulators (15)(16)(17). Unlike the RARE1 region located in the proximal enhancer, the RAREoct promoter element contains a typical recognition sequence for RA receptors (RAR).
Vitamin A (retinol) and its biologically active derivatives (retinoids), most notably RA, exert pleiotropic effects on vertebrate development, cell differentiation, and homeostasis (18,19). The action of diverse ligands (including retinoids, vitamin D, steroid hormones, and thyroid hormone) is mediated by members of the nuclear hormone receptor superfamily. RA exerts its action through two distinct groups within the nuclear hormone receptor superfamily: the retinoid nuclear receptors, RARs (isoforms ␣, ␤, and ␥), and the RXRs (isoforms ␣, ␤, and ␥). The complexity of retinoid signaling is further increased by the fact that, at least in vitro, RARs bind to their polymorphic cis acting response elements, as RAR:RXR heterodimers. Moreover, RXRs are also heterodimeric partners for other nuclear receptors such as thyroid hormone, vitamin D 3 , and peroxisome proliferator-activated receptors, as well as the nerve growth factor I-B (NGFI-B) (reviewed in Refs. 20 and 21), and the OR1 orphan receptors (22). The orphan members of the nuclear receptor family of transcription factors are those for which the activating ligands have not been identified. Two of the well characterized orphan receptors, ARP-I and COUP-TFI, were shown to have the capacity to block RAR:RXR-mediated transactivation (23,24). Previous work in our laboratory indicated inhibition of Oct-3/4 gene expression by the ARP-I and COUP-TFI orphan receptors through the RAREoct site (17).
The RAREoct sequence includes a potential site for the binding of another orphan receptor, steroidogenic factor-1 (SF1). SF1 has emerged as a key regulator of endocrine function within the hypothalamic-pituitary-gonadal axis and adrenal cortex and as an essential factor in sex differentiation. SF1 plays a role in the regulation of genes in steroidogenic cells, Sertoli cells, and gonadotropes. Analysis of the role of SF1 in vivo by targeted gene disruption showed that SF1 knockout mice exhibited adrenal and gonadal agenesis, male-to-female sex reversal of the internal and external genitalia, impaired gonadotrope function, and deletion of a specific region of the hypothalamus (reviewed in Ref. 25). The SF1 protein is one of four known isoforms (SF-1, ELP1, ELP2, and ELP3) generated by alternative promoter usage and 3Ј-splicing of the same gene (26). The SF-1/ELP transcription factors and their closely resembled Drosophila fushi tarazu factor 1 are believed to interact with their recognition site 5Ј-PyCAAGGPyCPu-3Ј or 5Ј-PuPuAGGTC-3Ј as monomers (27)(28)(29).
We show that the SF1/ELP transcription factor(s) is expressed not only in the cell lineages described above, but also in P19 EC cells, and its expression is down-regulated following RA treatment. SF1 activates the Oct-3/4 promoter via two SF1-binding sites. Moreover, we point out a possible functional interaction of the SF1/ELP protein(s) with the RAR proteins to synergistically activate the Oct-3/4 gene expression. The possible effect of SF-1/ELP on the Oct-3/4 gene and its synergism with members of the RAR family of transcription factors, adds sources of diversity to regulation of a RA-responsive gene.

EXPERIMENTAL PROCEDURES
Cells-Murine P19 EC cells, L cells, and COS-1 cells were maintained in Dulbecco's medium supplemented with 10% fetal calf serum, 2 mM glutamine, 100 units of penicillin, and 100 g of streptomycin/ml. For RA treatment, P19 cells were subjected to 1 M RA.
Plasmids and Oligonucleotides-The p0.4 (designated in our previous articles p0.4oct-CAT, Refs. 14 and 17) was constructed by inserting the XbaI-MspI fragment spanning the Ϫ412 to ϩ39 positions of the Oct-3/4 promoter region upstream to the chloramphenicol acetyltransferase (CAT) reporter gene in pJFCAT1 (30). The p0.4R* construct (designated in our previous articles as p0.4octR*-CAT), is identical to the above described plasmid except for four point mutations inserted in the RAREoct site (the mutated sequence is GGGCCcGtcGTgAAG-GCTAGA, with the lowercase letters denoting mutated nucleotides). The p0.4-SF1(a)* construct is identical to the p0.4 plasmid except for four point mutations inserted in the SF1 site, present in the RAREoct motif (the mutated sequence is GGGCCAGAGGTCgctcCTAGA). The p0.4-SF1(b)* construct is identical to the p0.4 plasmid except for four point mutations inserted in the SF1 site present in the newly identified upstream promoter element, UPE (the mutated sequence is GAG-GAgagcGAGGGTTG instead of GAGGACCTTGAGGGTTG). The p0.4-SF1(a,b)* harbors both four point mutations of the SF1 sites present in p0.4-SF1(a)* and in the p0.4-SF1(b)* sequence (the mutated sequences are described above).
The p0.22 construct was generated by subcloning the StuI-MspI fragment spanning the Ϫ178 to ϩ39 positions of the Oct-3/4 promoter region upstream to the CAT reporter gene in pJFCAT1 (30). The p0.22R* is identical to the above described plasmid except for four point mutations inserted in the RAREoct site. The mutated sequence is identical to that described for the p0.4R*. The p0.22-SF1(a)* is identical to the p0.22 plasmid except for four point mutations inserted in the SF1 site present in the RAREoct motif. The mutated sequence is identical to that described above, for the p0.4-SF1(a)*. Site-directed mutagenesis was performed as described previously (31).
The SF1 expression vector coding for a protein deleted in its DNAbinding domain (denoted SF1⌬DBD) has been constructed by a deletion of amino acid numbers 6 to 48 (a deletion of the BsiWI-ApaLI 129-bp DNA fragment), and analyzed by sequencing and SDS-polyacrylamide gels.
DNA Transfections-P19 and L cells were transfected by the calcium phosphate precipitation method (36). P19 cells (5 ϫ 10 5 ) were plated 24 h before transfection and then transfected with 10 g of CAT reporter plasmids together with 5 g of vectors expressing full-length cDNAs of the indicated nuclear receptors, and 2 g of ␤-galactosidase reference plasmid (pCMV-␤-gal), to correct for differences in transfection efficiency. Medium was refreshed 16 -20 h after transfection. After an additional 20 -24 h, cells were harvested for CAT assays. L cells were plated at a density of 1-2 ϫ 10 6 cell/plate and transfected by the calcium phosphate precipitation method as described above. Four h following the transfection, L cells were subjected to a glycerol shock (20% glycerol in Dulbecco's modified Eagle's medium for 1 min), washed for three times by phosphate-buffered saline, and maintained in medium. After additional 44 -48 h cells were harvested for CAT assays.
CAT activity was measured by using [ 14 C]chloramphenicol (53 mCi/ mmol; Amersham Pharmacia Biotech) as the substrate, in the presence of acetyl-coenzyme A at 37°C for 16 h. Chloramphenicol was separated from its acetylated forms by silica thin-layer chromatography and quantitated on PhosphorImager by using ImageQuant software. COS-1 cells were transfected by the DEAE-dextran method (37) with 10 g of plasmid containing the expression vector of the indicated nuclear receptor.
Nuclear Receptors Binding Analyses-For analysis in electrophoretic mobility shift DNA binding assays, whole cell extracts (WCEs) were prepared from transfected COS-1 cells by lysing the cells in 100 l of high salt extraction buffer (0.4 M KCl, 20 mM Tris-HCl, pH 8.0, 2 mM dithiothreitol, 20% (v/v) glycerol), containing the following protease inhibitors: 1 mM phenylmethylsulfonyl fluoride, 0.3 g/ml antipain, 1 g/ml leupeptin, and 0.5 g/ml trypsin inhibitor. The WCEs were frozen for at least 20 min, thawed, and then centrifuged at 10,000 ϫ g for 15 min at 4°C to remove cell debris. One to 10 g of WCE were incubated in a 20-l reaction mixture containing 0.3 ng of end-labeled oligonucleotide (30,000 cpm), in the presence of 10 mM Tris-HCl (pH 7.8), 14% glycerol, 1 mM dithiothreitol, and 2 g of poly(dI-dC).
For the competition and supershift experiments, 1-10 g of the WCEs were mixed with 5 l of 4 ϫ binding buffer in the presence of 2 g of poly(dI-dC). The 20-l reaction mixture was incubated for 5 min at room temperature. Subsequently, 30 ng of unlabeled oligonucleotide or 1 l of the indicated antibody was added for 10 min. Finally, 32 Plabeled oligonucleotide was added and incubation continued for a further 20 min at room temperature. Protein-DNA complexes were analyzed on pre-run 4% polyacrylamide 0.25 ϫ TBE (1 ϫ TBE is 89 mM Tris borate, 89 mM boric acid, 2 mM EDTA) gels. Gels were dried and exposed to x-ray film with an intensifying screen, at Ϫ70°C.
DNase I Footprinting Assays-For DNase I footprinting assays, the indicated DNA fragments were labeled using the Klenow fragment and [␣-32 P]dCTP, to a specific activity of greater than 10,000 cpm/ng of DNA. Probes were incubated with 30 g of protein of WCE in 40 l of reaction mixture containing 10 mM Tris-HCl (pH 7.8), 14% glycerol, 1 mM dithiothreitol, and 100 ng of poly(dI-dC). After incubation for 30 min at room temperature, DNase I (0.5 to 1 unit; Roche Molecular Biochemicals) diluted in 50 mM MgCl 2 , 10 mM CaCl 2 was added for 1 min. The reaction was stopped by the addition of 150 l of stop solution containing 200 mM NaCl, 20 mM EDTA, 1% SDS, and 33 g of yeast tRNA/ml. DNA was extracted with phenol-chloroform, ethanol precipitated, and analyzed on denaturing 6% polyacrylamide gel. Gels were dried and autoradiographed with an intensifying screen at Ϫ70°C. Sequencing lanes of the same probes were generated by the Maxam-Gilbert procedure (38).

SF1 Specifically
Binds the RAREoct Site-We have previously reported the characterization of the Oct-3/4 promoter (14,17). These studies have shown that the Oct-3/4 promoter contains a RA-responsive cis regulatory element, designated RAREoct. This motif has been described as the consensus halfsite structure of receptors that belong to the RAR (40 -43).

SF1 and RAR Regulate the Oct-3/4 Promoter
Examination of the RAREoct sequence also reveals a potential binding site for the steroidogenic factor-1, SF1. This region contains the motif TCAAGGCTA, similar to the consensus reported SF1-binding site (PyCAAGGPyCPu) (44 -50). To find whether SF1 plays a role in regulation of the Oct-3/4 gene, the proteins that bind to the RAREoct sequence were investigated by electrophoretic mobility shift assay using nuclear extracts prepared from undifferentiated P19 cells (Fig. 1A). Incubation of this extract with the labeled RAREoct probe resulted in formation of several complexes, the fastest migrating complex previously designated complex number 4 (14), being the prominent one (lane 1). This complex was competed for by unlabeled RAREoct, and RAREoct*, which is mutated in the first two direct repeats (lanes 2 and 3). It was barely competed by the RARE isolated from the RAR␤ gene, the unrelated OCTA sequence, and most interestingly by a RAREoct oligonucleotide mutated in the third direct repeat, designated SF1(a)* (lanes 4 -6).
To directly test whether the SF1 protein can bind the RAREoct sequence, we transfected COS-1 cells with ELP/SF1 expres- The lowercase letters represent the four mutations inserted in the SF1-binding site. C, RNA (15 g) isolated from P19 cells treated for 0, 3, 6, 20, 48, and 96 h with RA were electrophoresed on 1% agarose-formaldehyde gels, transferred to Nytran filters, and hybridized with labeled Oct-3/4 and ␤-actin cDNA probes, and with a polymerase chain reaction product of the SF1 cDNA (as described previously (91)). The blot was initially hybridized with the Oct-3/4 probe and then stripped and sequentially rehybridized with the additional indicated probes. sion vectors. The amino-terminal portion of ELP1 and ELP2 is 77 amino acids longer than that of SF1 or ELP3. The carboxylterminal of ELP1 protein is 57 amino acids in length, whereas the carboxyl-terminal shared by ELP2, ELP3, and SF1 is 131 amino acids long (26). ELP3 and SF1 are identical throughout the coding sequence and thus we present results demonstrating SF1 effects, only. Analysis of the binding properties of the proteins extracted from COS-1 cells transfected with ELP1, SF1, or ELP2 expression vectors showed that all three cell extracts formed very prominent retarded complexes with RAREoct oligonucleotides (lanes 9 -11). The comigrating complex was absent with COS-1 WCE (lane 8). The SF1, ELP2, and ELP1 faster migrating complex comigrated with the complex generated between the RAREoct oligonucleotide and WCE prepared from P19 cells (compare lane 7 with lanes 9 -11). The comigrating complex present in P19 WCE indeed contains the SF1 transcription factor, since a polyclonal anti-SF1 antibody, which recognizes all three isoforms (lane 14), significantly reduces this complex (compare lane 13, containing an irrelevant antibody, anti-Oct-3/4, with lane 14). This antibody is directed against the DNA-binding domain (51) and thus, efficiently blocks accessibility of the SF1 protein to the DNA. These data strongly indicate that SF1 binds the RAREoct sequence and is present in P19 cells.
To further characterize the binding site involved in SF1-DNA interactions we performed DNase I footprinting experiments using two probes; the wild-type Oct-3/4 promoter (WT,  4). This region also includes, in addition to the Sp1 site, the full RAREoct sequence, as well as the SF1 binding element, and 6 bp 3Ј to it. The COS-1/ELP1 extract protected an identical region (data not shown). Using the SF1-mutated Oct-3/4 promoter fragment (SF1(a)*) as a probe, WCE from transfected and untransfected COS-1 cells protected an identical region between nucleotides Ϫ55 and Ϫ35, harboring the Sp1-binding site (lanes 8 and 9). Thus, mutation in the SF1-binding site clearly eliminated the specific protection pattern observed with the wildtype probe. This finding suggests that mutations in the SF1binding site prevent SF1 protein from binding, but still allow Sp1 to bind. Taken as a whole, these data clearly demonstrate that the RAREoct region is recognized by SF1.
SF1 and Oct-3/4 Are Coexpressed in P19 Cells-To study the physiological relevance of the observations described above, we have followed the expression patterns of Oct-3/4 and SF1 in undifferentiated versus RA-differentiated P19 cells. As shown in the Northern blot analysis in Fig. 1C, Oct-3/4 and SF1 mRNAs are coexpressed in P19 cells and in differentiated P19 cells which were treated with RA for 3, 6, and 20 h. Following 48 h of RA treatment, both mRNAs (as well as proteins, data not shown) decrease to undetectable levels. Thus, treatment of P19 cells with RA causes a concomitant repression of Oct-3/4 and SF1 expression, with a very similar kinetics, raising the possibility of a functional relationship between the two. SF1 Activates the Oct-3/4 Promoter-The results describing the specific binding of SF1 to RAREoct, and the similar kinetics of SF1 and Oct-3/4 down-regulation following RA treatment, suggest that SF1 may affect Oct-3/4 expression. However, these results did not address the role of the SF1 site in regulating Oct-3/4 promoter activity. We, therefore, used site-directed mutagenesis to mutate 4-bp (depicted in Table I) in the SF1-binding site, within the context of the 220-bp Oct-3/4 promoter driving the CAT reporter gene expression (p0.22-SF1(a)*). These wild-type and mutated constructs were transfected into the physiologically relevant cells (P19) that coexpress Oct-3/4 and SF1 proteins. As shown in Fig. 2A the mutation in the SF1-binding site decreased promoter activity in P19 cells to approximately 40% of the level of the wild-type promoter, revealing a role for this element in Oct-3/4 regulation. Thus, the mutation that eliminated DNA⅐SF1 complex formation (Fig. 1, A and B), also markedly reduced promoter activity.
To investigate the role of SF1 in activation of the Oct-3/4 promoter further, we carried out transient transfection experiments, expressing SF1 in L fibroblast cells, which lack SF1. We co-transfected the p0.22 construct, with the control plasmid pCMV-␤-gal, and either the entire wild-type SF1 or the DNAbinding domain-deleted SF1 (SF1⌬DBD) cDNA cloned into an expression vector. As illustrated in Fig. 2B, co-transfection of SF1 expression vector into L cell up-regulates CAT activity driven by the Oct-3/4 promoter by approximately 2.5-fold. Noteworthy, the magnitude of the activation by SF1 in L cells (ϳ2.5-fold) is similar to the inhibitory effect upon mutating the SF1-binding site, in P19 cells (ϳ40%, Fig. 2A). Identical results were obtained with ELP3 expression vector encoding the same protein product as SF1 expression vector (data not shown), and similar results were obtained with ELP2 expression vector (data not shown).
To further study the possibility that activation of the Oct-3/4 promoter by SF1 was through the SF1-binding site, we cotransfected the p0.22-SF1(a)* construct, containing the CAT reporter gene driven by the Oct-3/4 promoter mutated in the SF1 site, with or without the SF1 expression vector. As described above, this mutated version of the Oct-3/4 promoter fails to bind the SF1 protein. The results shown in Fig. 2B FIG. 2. The SF1 protein activates the 220-bp Oct-3/4 promoter fragment. A, the p0.22 and the p0.22-SF1(a)* CAT reporter plasmids (10 g) were transiently transfected with a ␤-galactosidase-containing reference plasmid (2 g) into P19 cells. After 48 h, cells were harvested and lysed, and CAT activities were determined. The percent conversion to the acetylated forms of each separate transfection was normalized to the ␤-galactosidase activity. The relative CAT activity represents CAT activity relative to that obtained from p0.22 which was arbitrarily set to 1. The values for the relative CAT conversion, presented as mean Ϯ S.D. corresponding to p0.22-SF1(a)* is 0.44 Ϯ 0.12. The bar graphs are representative of three independent transfections. B, the p0.22 and p0.22-SF1(a)* CAT reporter plasmids (10 g) were transiently transfected with a ␤-galactosidase-containing reference plasmid (2 g) into L cells, in the absence or presence of the indicated nuclear receptor expression vector (5 g). The amount of DNA added to each transfection was equalized by adding empty vector. Transfection efficiency and CAT activity were monitored and assayed as described above (A). The relative CAT activity represents CAT activity relative to that obtained from either p0. 22  clearly demonstrate that mutation of the SF1 site significantly diminished activation of the Oct-3/4 promoter by SF1. Interestingly, the SF1⌬DBD expression vector was unable to stimulate the wild-type 220-bp Oct-3/4 promoter, suggesting that the interaction between SF1 protein and the cis regulatory sequence is needed for the transcriptional activation. We have also analyzed the effect of ELP1 on Oct-3/4 promoter activity. It was previously shown that ELP1 acts as a repressor of the promoter of the long terminal repeat of the Moloney murine leukemia virus (52). Similarly, we have found that it represses Oct-3/4 promoter activity through the SF1(a) site (data not shown). However, since the physiological relevance of the ELP1 interactions are not clear, we focus on the effects of SF1 and ELP2 on Oct-3/4 expression.
The Oct-3/4 Promoter Contains Two SF1-binding Sites-Although the 0.4-and 0.22-kb Oct-3/4 upstream sequences direct similar levels of CAT expression in P19 cells, the 400-bp fragment consistently directs slightly higher reporter activity (14). Therefore, we analyzed the effect of SF1 on the p0.4 reporter gene (previously designated p0.4oct-CAT (14,17)) in L cells (Fig. 3). As is the case with the p0.22 construct, SF1 activates CAT expression driven by the 0.4-kb Oct-3/4 promoter fragment, and mutation in the SF1(a)-binding site, in this context, decreased its ability to activate this promoter fragment. Similar results were obtained with ELP2 expression vector (data not shown). Close inspection of the sequence of the 400-bp fragment revealed an additional potential SF1-binding site. To assess this possibility we performed DNase I footprint experiments using the p0.4-kb Oct-3/4 promoter as a probe and WCE isolated from COS-1 cells transfected with SF1 expression vector. Interestingly, this WCE protected a region between nucleotides Ϫ209 and Ϫ193 of the Oct-3/4 wild-type promoter (Fig. 4A, lane 3). This region includes the CCTGGAACT sequence which is very similar to the consensus binding site of SF1 (PyCAAGGPyCPu), but in the reverse orientation. WCE from untransfected COS-1 cells did not yield a footprint in this region (lane 4). Examination of the sequence adjacent to this site also revealed a potential binding site for an additional orphan receptor, COUP-TFI. Indeed, DNase I footprint experiments using WCE prepared from COS-1 cells transfected with COUP-TFI expression vector yielded a larger protected region between nucleotides Ϫ230 and Ϫ193 of the Oct-3/4 promoter (data not shown).
To confirm that this region harbors a binding site for SF1 and/or COUP-TFI, and to study it in a greater detail, we synthesized an oligonucleotide corresponding to nucleotide Ϫ235 to Ϫ189 in the Oct-3/4 promoter (designated UPE, for upstream promoter element, depicted in Table I of a prominent complex (Fig. 4B, lane 3) which was competed by a 100-fold molar excess of RAREoct and UPE (lanes 5 and 6), but not by the OCTA-binding site (lane 4). Interestingly, an SF1-mutated version of the UPE sequence, designated UPE-SF1(b)*, in which 4 bp were mutated (depicted in Table I 2 and 3). Moreover, antibodies directed against SF1 eliminated this complex (compare lane 8 with lane 9). To obtain more information regarding the precise nucleotide sequence required for binding of SF1 to UPE, we compared the proteins that bind to UPE with those that associated with UPE-SF1(b)*. As shown in Fig. 4C, the three isoforms, ELP1, SF1, and ELP2 bind UPE, and mutations in the SF1(b) site dramatically reduced their ability to bind (Fig. 4C, compare lanes 1-3 with  lanes 6 -8). Interestingly, these mutations barely interfere with COUP-TFI complex formation (compare lanes 5 and 10), indicating that SF1 and COUP-TFI bind to nonoverlapping UPE sequences.
To address the relative contribution of the two SF1-binding sites to Oct-3/4 promoter activity, we introduced, within the context of the Oct-3/4 400-bp promoter fragment, mutation in SF1(a) and SF1(b) sites, separately. The CAT reporter constructs were designated p0.4-SF1(a)* and p0.4-SF1(b)*, respectively. As shown in Fig. 5A, whereas mutation in SF1(a) lowered promoter activity to approximately 30%, mutation of the SF1(b) site lowered it to 80% as compared with the wild-type promoter (p0.4). Consistent with these results, SF1 and ELP2 up-regulated the activity of the Oct-3/4 promoter mutated in the SF1(b)-binding site, in L cells (Fig. 5B). Due to a low level of promoter activity in P19 cells, attempts to accurately quantify the activity of the Oct-3/4 promoter mutated in both SF1 sites, as compared with mutation in SF1(a) site only, were unsuccessful. However, we were able to show that mutations in both SF1 sites reduced the ability of SF1 to activate the 400-bp Oct-3/4 promoter fragment, even more drastically than a single mutant in SF1(a) site (compare Fig. 5B to Fig. 3). Collectively, these results suggest that both sites of SF1 contribute to activation of the Oct-3/4 promoter, with SF1(a) playing the  SF1(b). A, the p0.4, p0.4-SF1(a)*, and p0.4-SF1(b)* CAT reporter plasmids (10 g) were transiently transfected with a ␤-galactosidase containing reference plasmid (2 g) into P19 cells. Transfection efficiency and CAT activity were monitored and assayed as described in the legend to Fig. 2A. The relative CAT activity represents CAT activity relative to that obtained from p0.4 which was arbitrarily set to 1. The values for the relative CAT conversion, presented as mean Ϯ S.D., corresponding to p0.4-SF1(a)* or p0.4-SF1(b)* are 0.32 Ϯ 0.09 and 0.82 Ϯ 0.08, respectively. The bar graphs are representative of three independent transfections. B, the p0.4-SF1(b)* and the p0.4-SF1(a,b)* CAT reporter plasmids (5 g) were transiently transfected with a ␤-galactosidase containing reference plasmid (2 g) into L cells, in the absence (Ϫ) or presence of the indicated nuclear receptor expression vector (2.5 g). The amount of DNA added to each transfection was equalized by adding empty vector. Transfection efficiency and CAT activity were monitored and assayed as described in the legend to Fig.  2A. The relative CAT activity represents CAT activity relative to that obtained from p0.4-SF1(b)* or p0.4-SF1(a,b)* which were arbitrarily set to 1. The values for the relative CAT conversion, presented as mean Ϯ S.D., corresponding to p0.4-SF1(b)* in the presence of SF1 or ELP2 are 3.30 Ϯ 0.40 and 3.20 Ϯ 0.37, respectively. The relative CAT conversion, corresponding to the p0.4-SF1(a,b)* in the presence of SF1 or ELP2 are 1.16 Ϯ 0.24 and 1.60 Ϯ 0.12, respectively. The bar graphs are representative of four independent transfections.

FIG. 6. SF1 and RAR synergistically activate the Oct-3/4 promoter.
The p0.4 CAT reporter plasmid (5 g) was transiently transfected with a ␤-galactosidase-containing reference plasmid (2 g) into L cells, in the absence (Ϫ) or presence (ϩ) of the indicated nuclear receptor expression vector (2.5 g). The amount of DNA added to each transfection was equalized by adding empty vector. Transfection efficiency and CAT activity were monitored and assayed as described in the legend to Fig. 2A. The relative CAT activity represents CAT activity relative to that obtained from p0.4 which was arbitrarily set to 1 more crucial role. The SF1(b) site may function in regulation of Oct-3/4 expression either through adjacent sites or in other cellular contexts.
The RAR Acts Synergistically with SF1 to Enhance Oct-3/4 Promoter Activity-We have previously shown that RAR:RXR heterodimers activate the Oct-3/4 promoter through the RAREoct sequence (17). The results described above emphasize the ability of SF1 to activate the promoter through the same element. Since these nuclear receptors are also coexpressed in P19 cells, they might either collaborate or interfere with the regulatory effect of each other. A corollary of this prediction is that transcriptional activation of Oct-3/4 promoter by SF1 can be affected by co-transfecting RAR or RXR expression vectors. We therefore transfected L cells (which lack high expression levels of these nuclear receptors) with the p0.4 reporter construct, the control pCMV-␤-gal, and combinations of SF1, RAR, and RXR expression vectors. The total amount of DNA in each transfection reaction was kept constant by compensating with an empty expression vector (Fig. 6). SF1, RAR␣, or RAR␥ alone enhanced the Oct-3/4 promoter activity by 1.5-2.5-fold. However, co-transfection of SF1 together with either RAR␣ or RAR␥ activated the Oct-3/4 promoter by 6-and 11-fold, respectively, suggesting a cooperative interaction between these transcription factors. Similar results were obtained with RAR␤ and SF1 expression vectors (data not shown). In contrast, activation of the Oct-3/4 promoter by SF1 was not enhanced by RXR (␣, ␤, and ␥) expression vectors (data not shown). Taken together, these data indicate that SF1 and RAR (but not RXR) collaborate to synergistically activate the Oct-3/4 promoter.
Complex Formation of SF1 and RAR Proteins on the RAREoct Site-There are several possible mechanisms through which RAR and SF1 can synergistically activate the Oct-3/4 promoter. RAR and SF1 may interact with each other on the DNA either directly or indirectly. To test this possibility, we performed electrophoretic mobility shift assays with the RAREoct-labeled probe and WCE prepared from COS-1 cells transfected with SF1, RAR␣, RAR␤, and RAR␥ separately, or combination of these expression vectors. As shown in Fig. 7A, RAR␣, ␤, and ␥ bind weakly to the RAREoct site yielding a fast migrating complex (lanes [3][4][5]. In contrast, when COS-1 cells were transfected with SF1 together with RAR expression vectors, in addition to the prominent SF1 complex (lane 2), a novel slower mobility SF1⅐RAR complex was detected (designated SF1/RAR) (lanes 6 -8). As shown in Fig. 7B, this slower mobility complex was completely eliminated by excess of RAREoct oligonucleotide (lane 2) and was significantly competed by either an excess of RAREoct mutated in the RAR binding site (RAREoct*) or by mutated SF1(a)* sequence. The SF1 complex was competed with cold RAREoct and RAREoct* oligonucleo- tide (slightly to a lower level), and was not inhibited with the SF1(a)* oligonucleotide (lanes 2-4). These complexes were not competed by an unrelated oligonucleotide (data not shown).
In order to better characterize the novel complex formed in the presence of the SF1 and RAR␥ expression vectors, we have used antibodies directed against SF1 and RAR␥ alone or a combination of these two antibodies. The SF1⅐RAR complex was partially displaced by incubation with anti-SF1 or anti-RAR␥ (Fig. 7C, lanes 5 and 6, respectively), and totally displaced by the presence of both antibodies (lane 4). Similar results were obtained using RAR␣ and RAR␤ together with SF1 (data not shown). Taken together, these results suggest that this novel complex contains both transcription factors (SF1 and RAR), which may interact directly or via an adapter protein, on the RAREoct sequence.
To gain more insight into the mechanism through which SF1 and RAR synergistically activate the Oct-3/4 promoter, we have used three CAT reporter constructs driven by the Oct-3/4 400-bp promoter fragment. In one, the promoter was mutated in the SF1(b) site (p0.4-SF1(b)*), the second was mutated in the first two direct repeats which serve as RAR-binding site (p0.4R*), and the third promoter was mutated in the SF1(a) site (p0.4-SF1(a)*). As illustrated in Fig. 8, mutations in SF1(b)* did not affect the ability of SF1 or RAR␣ to either activate the promoter, or their ability to synergies in Oct-3/4 promoter up-regulation. In contrast, disruption of either the RARE or the SF1(a)-binding sites, located in the RAREoct element, drastically diminished SF1 and RAR␣ ability to synergistically activate the Oct-3/4 promoter. As expected, mutations that affect either RAR or SF1 binding alone, considerably reduced the ability of the respective nuclear receptors to activate the Oct-3/4 promoter. Moreover, a DNA-binding domain-deleted SF1 expression vector was unable to either activate alone or to synergize with RAR in up-regulating the Oct-3/4 promoter (Fig. 6). These co-transfection results corroborate the above described binding studies showing that SF1 and RAR proteins interact on the RAREoct sequence, and emphasize that both RARE and SF1 cis regulatory elements are needed, as well as the SF1 DNA-binding domain, in order to enhance Oct-3/4 promoter activity.

DISCUSSION
The Oct-3/4 transcription factor is expressed in the earliest stages of embryogenesis (4 -7), and is thus likely to play an important role in regulation of initial decisions in development (10). In order to better understand the molecular mechanisms that participate in Oct-3/4 activation in pluripotent cells, we have examined the regulation of the Oct-3/4 promoter activity, in detail. The Oct-3/4 promoter is composed of multiple regulatory elements, including binding sites for Sp1, RAR:RXR heterodimers, COUP-TFI, and ARP-1 (14 -16). In this study we have provided proof that SF1-binding sites, and the SF1 protein play a role in up-regulating Oct-3/4 promoter activity. Furthermore, our data show that SF1 and RAR function cooperatively in transactivation of the Oct-3/4 promoter. Interestingly, we have shown that SF1 is expressed in P19 EC cells, which also express the Oct-3/4 gene, and expression of both transcription factors is down-regulated in EC cells which are induced to differentiate with RA. Considering these data, it seems likely that SF1 positively regulates Oct-3/4 expression in EC cells.
We have identified two novel putative SF1-binding sites in the Oct-3/4 promoter: the proximal site, SF1(a), that is located within the previously identified RAREoct element, and the distal site, SF1(b), which is located between nucleotide Ϫ193 and Ϫ209 of the Oct-3/4 promoter. The RAREoct is a compound element that harbors three direct repeats (R1-R3) of an AGGTCA-like consensus binding site for members of the RAR family of transcription factors, with 1-and 0-nucleotide spacers. The SF1-putative binding site encompasses the R3 and three adjacent nucleotides of the R2, whereas the RAR:RXR heterodimers bind to R1-R2 (16). We have shown that SF1, which is expressed in P19 cells, binds to the RAREoct motif harboring the SF1(a) motif. Moreover, footprinting analysis of the 400-bp Oct-3/4 promoter fragment, using COS-1/SF1 protein extracts revealed a protection pattern similar to that observed in our previous study, using cell extracts prepared from P19 cells (14). The undifferentiated P19 nuclear extracts protected a region between nucleotides Ϫ52 and Ϫ20 of the Oct-3/4 wild-type promoter. This region includes the Sp1, RAREoct, and SF1-binding sites. In contrast, extracts from P19/RA cells protected the region between nucleotides Ϫ52 and Ϫ34. This shorter region lacks the SF1-binding site. Introduction of mutations within the putative SF1(a)-binding site blunted the ability of SF1 to either bind or transactivate the Oct-3/4 promoter. The distal SF1(b) sequence is a putative SF1-binding site, in a reverse orientation CCTGGAACT. Mutation analysis of the proximal and distal SF1 sites indicates that the proximal SF1(a) site is the main site through which SF1 affects the Oct-3/4 promoter, as tested in P19 cells. Interestingly, numerous genes regulated by SF1 contain more than one SF1-binding site in their regulatory elements. The presence, in a promoter, of multiple binding sites for a single transcription factor may reflect different usage in various cellular backgrounds (53)(54)(55).
An Sp1 consensus binding site is located 5Ј to the RAREoct motif. The proteins that take part in activating the Oct-3/4 promoter through the Sp1 site have not yet been identified, but it is likely that there could be unique Sp1 factors characterized by a cell lineage-specific rather than ubiquitous expression pattern. This Sp1 decamer sequence partially overlaps with the R1 sequence. Thus, the Sp1-and SF1-binding sites are closely located in the Oct-3/4 promoter, and moreover, are both protected similarly using P19 (14) and COS/SF1 whole cell extracts. Proximal location of SF1 and Sp1 binding elements were observed in other genes as well (56 -58). An example for cooperative functional interactions between Sp1 and SF1 proteins have been shown in the transactivation of the cholesterol side chain cleavage enzyme (CYP11A) promoter (58). Furthermore, Sp1 and SF1 responsive elements have also been linked and involved in the cAMP-induced expression of various genes (53,55,58,59). The cAMP-induced expression might involve cAMPdependent protein kinase activation, and phosphorylation of the SF1 protein (60). Considering the fact that the Oct-3/4 promoter sequence harbors SF1 and Sp1 neighboring sequences, combined with the data indicating a regulatory effect of cAMP on EC cells (61)(62)(63), it seems likely that Oct-3/4 promoter activity may be regulated through the cAMP signal transduction pathway.
SF1 protein is involved in regulation of expression of genes in the hypothalamic-pituitary-gonadal axis and in the adrenal cortex. SF1 activates genes in steroidogenic cells (steroid hydroxylases, non-cytochrome P450 enzyme, steroidogenic acute regulatory protein, and ACTH receptor), in Sertoli cells (Mü llerian-inhibiting substance and aromatase), and in the gonadotropes (␣-subunit of glycoproteins, ␤-subunit of leutinizing hormone, and the GnRH receptor, reviewed in Ref. 25). In this report we have demonstrated the presence, and indicated a possible functional role, of SF1 protein in a completely different cell lineage, i.e. the EC cells. To our knowledge, this is the first report of an EC-specific gene that is activated by SF1. Moreover, the kinetics of Oct-3/4 repression following RA treatment directly correlates with suppression of the SF1 transcripts. This direct correlation between Oct-3/4 and SF1 expression in EC cells, and down-regulation in RA-treated EC cells, as well as our binding and transfection experiments, may imply a functional relationship between SF1 and Oct-3/4. The DNAbinding domain (DBD) of the SF1 protein is necessary for SF1-mediated activation of Oct-3/4 gene expression. A similar requirement for the SF1 DBD for the activation of the P450ssc gene in ES cells and for differentiation of ES cells into steroidogenic cells, has also been demonstrated (64).
On the basis of our previous experiments which implicate RAR:RXR heterodimers in activation of the Oct-3/4 promoter (17), and in order to get a better understanding of the molecular mechanism underlying activation of the Oct-3/4 promoter via SF1, we have examined the possibility of a functional interaction between SF1 and RAR or RXR nuclear factors. We have shown that RAR acts in concert with SF1 to activate Oct-3/4 promoter expression. Co-transfections of both SF1 and RAR expression vectors into L cells clearly confirmed that the two transcription factors act in synergism on the Oct-3/4 promoter. This synergism was restricted to SF1 and RAR, since co-transfection of SF1 with RXR␣, ␤, or ␥ had no effect on Oct-3/4 expression. SF1 and RAR synergism requires the two respective binding sites, since disruption by site-directed mutagenesis of either site abolishes the synergism. This functional interaction also requires the SF1 DNA-binding domain. It has been previously shown that the RARs alone are inefficient DNA binders and require auxiliary nuclear proteins for effective interactions with responsive elements. The RXRs serve as such auxiliary proteins. It is possible that the SF1 protein is a similar auxiliary factor that facilitates the binding of RAR to DNA, improves the ability of RAR to interact with coactivators, or interferes with the ability of RAR to interact with corepressors. Unlike the RAR:RXR heterodimers that activate the Oct-3/4 promoter in the presence of RA, the RAR and SF1 proteins synergistically activate the Oct-3/4 promoter in the absence of RA added to medium. Moreover, binding assays reveal that RA (10 Ϫ4 to 10 Ϫ7 M) neither activated nor inhibited SF1⅐RAR complex formation (data not shown). In P19 cells Oct-3/4 expression is down-regulated following 24 h of RA treatment, whereas RAR:RXR expression is up-regulated by 3 h of treatment. Interestingly, 24 h of RA treatment results also in up-regulation of ARP-1 and COUP-TFI orphan receptors which repress the Oct-3/4 promoter (15)(16)(17). These orphan receptors bind the RAREoct site with a much higher affinity than the RAR:RXR and most probably also with a higher affinity than RAR and SF1. This high binding affinity provides the orphan receptors with the ability to compete and displace the activating receptors (RAR, RXR, and SF1) from the RAREoct site and subsequently to silence the Oct-3/4 promoter.
Although our studies clearly indicate that SF1 and RAR cooperate to synergistically activate the Oct-3/4 promoter, the nature of RAR-SF1 interaction is not yet clear. Our binding experiments show that SF1 and RAR form a novel complex which contains both transcription factors, on the RAREoct sequence. This complex could be formed through several molecular interactions, such as direct interactions between SF1 and RAR. Alternatively, since SF1 synergies with RAR and not with RXR, they may indirectly interact through a shared coactivator that is specific for this combination. Both were found to interact with multiple coactivators such as SRC-1, CBP, and p300 (65)(66)(67). These two modes of interactions (direct or through coactivators) are not mutually exclusive. In addition, SF1 and RAR can bind to their adjacent sites cooperatively, binding of one transcription factor altering the DNA conformation in a way which increases the local accessibility for the second transcription factor. The effect of SF1 alone on Oct-3/4 promoter activity, although very consistent, is moderate and is enhanced considerably in the presence of RAR. It is possible that tissue-specific expression of genes regulated by SF1 could be reinforced by the presence of other transcription factors expressed in the tissue and act in synergism with SF1. Thus, the ability of RAR to synergize with SF1 in enhancing Oct-3/4 expression, may also contribute to restriction of the activation of Oct-3/4 in appropriate cell types where both transcription factors are expressed.
Similarly to NGFI-B, NURR1, and thyroid hormone, that bind both as monomers or as dimers (reviewed in Refs. 20 and 68), SF1 may also belong to this category of monomer/dimer binding receptors. In fact, several studies have described the effect of protein-protein interactions in SF1-dependent transcription. It has been shown that co-transfection of the zinc finger NGFI-A (Egr-1) and SF1 transcription factors led to synergistic activation of the LH␤ promoter (69 -72). This enhancement of LH␤ transcription most probably results from a direct interaction between Egr-1 and SF1 (71). Similarly, estradiol receptor and SF1 have been found to synergistically regulate the salmon gonadotropin II gene (sGTHII␤) in the presence of estradiol (73). In addition, using the mammalian two-hybrid system, Sp1 and SF1 have been shown to functionally collaborate in the transactivation of the bovine CYP11A (58). SF1 was found to associate and sinergize also with Wilms' tumor Ϫ1 and SOX9 to promote Mü llerian inhibiting substance and anti-Mü llerian hormone expression, respectively (74,75). In contrast to the synergistic activation function observed between SF1 and all the above discussed proteins, the DAX-1, an orphan nuclear receptor, has been shown to interact directly with SF1 in in vitro protein binding studies. The DAX-1 receptor which lacks the conserved zinc-finger DNA-binding domain, retains the protein:protein dimerization region, and recruits nuclear receptor corepressor N-CoR (76), was found to inhibit SF1-mediated transactivation (77).
The RAR proteins are known to exert their biological activities through numerous interactions with other nuclear transcription factors. The best known example is the dimerization of RAR with RXR (reviewed in Ref. 20). Furthermore, RAR protein was found to interact with members of the general basal transcription machinery, such as TFIID, TFIIB, and TFIIH (78 -80). More recently, interactions of RAR:RXR and multiple coactivators (CBP, p300, P/CAF, SRC-1/TIF2 (ACRT)) and corepressors (SMRT, N-CoR, mSin3A, and RPD-3 (HDAC-1)) have been identified. Most interestingly, the coactivator complexes harbor the enzymatic histone acetyltransferase activity, whereas the corepressor complexes exhibit a histone deacetylase activity, indicating the involvement of these complexes in chromatin remodeling (67,81,82). In addition, the recently identified orphan receptor referred to as small heterodimer partner that, like DAX-1, lacks the characterized zinc-finger DNA-binding domain, has been shown to interact and inhibit the activity of the RAR (83). Interestingly, in all our experiments, we have observed a consistent, slightly lower activation of the Oct-3/4 promoter mediated by RAR in the presence of SF1 protein lacking the DNA-binding domain (SF1⌬DBD) as compared with RAR alone. Since both SF1⌬DBD and small heterodimer partner proteins lack their DNA-binding domains, their mechanism of interfering with the transactivation driven by RAR, might be similar. This interference may involve either direct interaction with the RAR protein, or an indirect one, through a critical adapter protein.
The AF-2 transactivation domain located at the carboxyl terminus of many ligand-inducible nuclear receptors is absolutely conserved in all SF1 isoforms. The conservation of this domain raises the possibility that a ligand may mediate SF1dependent transactivation. Recently, it has been demonstrated that SF1 can be activated by oxysterols, and it has therefore been suggested that SF1 is a ligand-activated receptor as well (84). Alternatively, a putative complex formed between SF1 and another nuclear receptor, such as RAR, could be activated by the partner's ligand as in the case of the RAR⅐RXR complex (reviewed in Ref. 20).
We and others have previously shown that the orphan receptors, COUP-TFI and ARP-1, bind with a high affinity to the RAREoct element, and down-regulate Oct-3/4 promoter activity (15)(16)(17). RA treatment of EC cells results in up-regulation of the orphan receptors expression and down-regulation of the SF1 expression. Interestingly, the novel UPE sequence, also identified in this study, harbors binding sites for SF1, COUP-TFI, and ARP-1 as well. However, whereas the binding sites for COUP-TFI/ARP-1 transcription factors and SF1 do not overlap in UPE, they do so in RAREoct. In fact, it was recently found that the murine Dax-1 and the aromatase P450 promoters are regulated by competitive binding of SF1 (that activates) and COUP-TFI (that inhibits) to the same cis-acting element (85,86). Thus, it is reasonable to suggest that COUP-TFI or ARP-1 will inhibit SF1-dependent activation of Oct-3/4 through the RAREoct site. This suggestion is consistent with a model proposing an activation role for RAR and SF1, and a repressive function for COUP-TFI and ARP-1 in regulating Oct-3/4 expression.
In the postgastrulatory embryo, Oct-3/4 is expressed in primordial germ cells and in oocytes (5,7,87). While, SF1 transcripts are detected in the embryonal testis in all compartments: the interstitial region containing fetal Lydig cells, and testicular cords which contain fetal Sertoli cells and primordial germ cells (88). RAR␣, the most ubiquitous and abundant RAR protein expressed during embryonic development, is expressed in the developing gonads and in developing germ cells in the testis (89,90). Thus, it seems likely that the developing germ cells, are the potential in vivo site for a functional interaction between SF1 and RAR␣ to activate Oct-3/4 gene expression.
SF1 is one of several members of the nuclear receptor family that play a dual role both in the development of specific tissues or cell types, as well as in the regulation of multiple differentiating genes (reviewed in Ref. 25). Oct-3/4 is known to participate in early developmental decisions and most probably in maintaining the phenotype of pluripotential cells. Thus, elucidation of the ability of SF1 to activate Oct-3/4 may contribute to the understanding of a possible developmental regulatory cascade.