Oct-3/4 and Sox2 regulate Oct-3/4 gene in embryonic stem cells.

Oct-3/4 is a key transcriptional factor whose expression level governs the fate of primitive inner cell mass and embryonic stem (ES) cells. Previously, an upstream 3.3-kb distal enhancer (DE) fragment was identified to be responsible for the specific expression of mouse Oct-3/4 in the inner cell mass and ES cells. However, little is known about the cis-elements and trans-factors required for DE activity. In this study, we identified a novel cis-element, called Site 2B here, located approximately 30 bp downstream from Site 2A, which was previously revealed in DE by an in vivo chemical modification experiment. Using the luciferase reporter assay, we demonstrated that both Site 2A and Site 2B are necessary and sufficient for activating DE in the contexts of both the native Oct-3/4 promoter and the heterologous thymidine kinase minimal promoter. In an electrophoretic mobility shift assay we showed that Site 2B specifically binds to Oct-3/4 and Sox2 when ES-derived cell extracts were used, whereas Site 2A binds to a factor(s) present in both ES and NIH 3T3 cells. Furthermore, we showed that the physiological level of Oct-3/4 in ES cells is required for Site 2B-mediated DE activity using the inducible knock-out system of Oct-3/4 in ES cells. These results indicate that Oct-3/4 is a member of the gene family regulated by Oct-3/4 and Sox2, as reported before for the FGF-4, UTF1, Sox2, and Fbx15 genes. Thus, Oct-3/4 and Sox2 comprise a regulatory complex that controls the expression of genes important for the maintenance of the primitive state, including themselves. This autoregulatory circuit of the Sox2.Oct-3/4 complex may contribute to maintaining robustly the precise expression level of Oct-3/4 in primitive cells.

Totipotent cells are defined by their ability to produce both somatic and germ cells as well as extraembryonic tissues in mammals. The maintenance of totipotent cells lies at the heart of the continuity of life from one generation to the next. Although the molecular mechanism ensuring the totipotency in cells, in other words, preventing differentiation to a specific lineage of cells, is not fully understood, several key molecules have been identified. The transcription factor Oct-3/4 (encoded by the Pou5f1 locus) is one of such molecules and has been studied extensively (1)(2)(3)(4)(5).
Mouse Oct-3/4 is expressed during early development in cells that have totipotent or pluripotent differentiation ability (1, 6 -9). Oct-3/4 protein is present in the nuclei of the eight-cell embryo and all cells in the subsequent morula stage. When the embryos develop into the blastocyst, the trophectoderm is formed in the outer layer of the embryos as the first cell lineage specification. Oct-3/4 expression is extinguished in the trophectodermal cells, whereas it is continued in the inner cell mass (ICM) 1 that maintains the pluripotency to generate nontrophoblast extraembryonic tissues and all fetal cell types including germ cells. After implantation, Oct-3/4 expression is observed in the epiblast, which is subsequently down-regulated during gastrulation. In the later stages, Oct-3/4 is expressed only in primordial germ cells. In vitro, Oct-3/4 is expressed in undifferentiated embryonic stem (ES) and embryonal carcinoma (EC) cells and is down-regulated when these cells are induced to differentiate by retinoic acid (RA) treatment or the removal of leukemia inhibitory factor (2,7). These remarkable expression patterns of Oct-3/4 during early mouse development and the undifferentiated cell lines suggest that Oct-3/4 plays an important role in maintaining the toti-or pluripotency of cells.
The essentiality of Oct-3/4 in producing totipotent cells was demonstrated by the gene targeting technique. Oct-3/4-deficient knock-out embryos develop to the blastocyst stage, but the internally existing cells that form the pluripotent ICM cells in wild-type blastocysts are differentiated into the trophectoderm lineage, and no ICM cell is established (9). Similar results were obtained when the Oct-3/4 gene was knocked-down by the small interfering RNA method in early mouse development (10). Oct-3/4 does not simply control the pluripotency in a binary on-off fashion; rather, the precise level of Oct-3/4 determines the fate of ES cells (11). A less than 2-fold increase from the normal expression level causes differentiation into ectoderm and mesoderm, whereas a reduction to less than 50% leads to dedifferentiation into trophectoderm. Therefore, it is necessary to understand the mechanism controlling Oct-3/4 expression in a quantitative manner.
Oct-3/4 is a member of the POU (for Pit, Oct, and Unc) family of transcription factors. The POU family proteins share the POU domain, which is a bipartite DNA binding domain consisting of the ϳ75-amino acid POU-specific domain (POU S ) and the ϳ60-amino acid POU homeodomain (POU H ). These two subdomains, connected by a variable linker region, form independent structures and individually make sequence-specific contacts to cognate DNA target sequences. The spatial combi-nation of the two DNA binding subdomains provides a particular POU protein with the overall specificity and affinity for DNA binding and probably interaction with other proteins, such as coactivators. In this respect, the POU proteins can be viewed as intramolecularly fused heterodimers composed of POU S and POU H . However, the binding specificity of the POU proteins is not strict; they recognize a diverse set of DNA sequences (12). Oct proteins comprise a subfamily of the POU family and bind to the octamer motif, the consensus sequence of which is ATGCAAAT. POU S and POU H bind to the two halves of the consensus sequence, ATGC and AAAT, respectively (13).
Oct-3/4 regulates several genes, including FGF-4, UTF1, Sox2, Fbx15, Rex-1 and osteopontin, through distinct mechanisms (14 -20). For example, the osteopontin enhancer is characterized by the presence of an octamer sequence and a nearby copy of a palindromic POU H -recognized sequence. Oct-3/4 binds to these motifs to activate the gene as a monomer or a dimer (16). However, Oct-3/4 frequently activates its target genes synergistically with a heterologous transcription factor, Sox2. The FGF-4 gene, which is expressed in ICM and essential for the survival of postimplantation mouse embryo (21), is the first gene demonstrated to be regulated by Oct-3/4 and Sox2 (15). The FGF-4 enhancer in the untranslated region of exon 3 consists of a closely juxtaposed Oct-3/4 binding site and Sox2 binding site (15,22). Sox factors are a family of transcriptional factors characterized by the presence of an ϳ80-amino acid Sry-related high mobility group domain, and Sox2 is specifically expressed in undifferentiated cells and several differentiated cells (23). Oct-3/4 and Sox2 form a complex that cooperatively recognizes the bipartite Oct-3/4-and Sox2-binding elements present in the FGF-4 enhancer. It has been proposed that the combinatorial recognition of the bipartite cis-elements confers the specificity of the interacting trans-factors on the target gene. FGF-4, however, is not a unique target of the Sox2⅐Oct-3/4 complex. At least three other genes, UTF1, Fbx15, and Sox2 itself, are regulated by the Sox2⅐Oct-3/4 (18,20,24). Interestingly, these target genes are also expressed in pluripotent cells, and thus it is speculated that the Sox2⅐Oct-3/4 complex plays an important role in the transcriptional regulation in pluripotent cells.
Because the specific expression of Oct-3/4 during early mouse development is required for the correct maintenance of pluripotent cells, the regulatory control of the Oct-3/4 expression has been studied extensively. Previous studies identified two upstream regions, proximal enhancer (PE) and distal enhancer (DE), as the cis-elements controlling Oct-3/4 expression in different types of mouse cells. DE is responsible for activating Oct-3/4 in ICM cells, ES cells and primordial germ cells, whereas PE is functional in the epiblast and EC cells (8). Although the significance of these two regions in vivo has been well established, the molecular basis of the enhancer activity remains to be clarified. Here, we studied the cis-elements important for DE activity and found that Oct-3/4 and Sox2 activate Oct-3/4 expression by binding to a novel cis-element identified in DE.

MATERIALS AND METHODS
Cell Culture-NIH 3T3 and 293T cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, and 1ϫ Antibiotic-Antibiotics (Invitrogen). D3 ES cells were maintained in Dulbecco's modified Eagle's medium (high glucose) (Invitrogen) supplemented with 15% fetal bovine serum, 2 mM glutamine, 1 mM sodium pyruvate, 0.1 mM 2-mercaptoethanol, 1ϫ nonessential amino acids, 1ϫ Antibiotic-Antibiotics, and 100 units/ml murine leukemia inhibitory factor (ESGRO, Chemicon, CA) on 0.1% gelatincoated dishes. For differentiation experiments, ES cells were treated with 1 M all-trans-RA (Sigma) in the same medium without leukemia inhibitory factor and harvested at different time points for protein extraction. ZHBTc4 ES cells were cultured according to Niwa et al. (25).
Plasmid Construction-A series of deletion mutants of the upstream region of the Oct-3/4 gene was constructed by PCR from the 23-kb SpeI/SalI fragment. For reporter assays, the mutant fragments were inserted into the pGL3-Basic plasmid (Promega) or the pGL3-TK plasmid. pGL3-TK contains the minimal promoter of the herpes simplex thymidine kinase (TK) gene derived from pBLCAT5. Mouse Oct-3/4 and Sox2 cDNAs were generated by PCR from mRNA derived from undifferentiated R1 ES cells, modified to produce N-terminally HA-tagged protein, and subcloned into pcDNA3 (Invitrogen). Nucleotide sequences were confirmed by ABI sequencers.
Luciferase Reporter Assay-Cells (1 or 5 ϫ 10 4 cells) were plated on a 12-well plate. 24 h later, 100 ng of reporter plasmids along with 10 ng of the pGL3-RL/TK control plasmid was transfected using Lipofectamine (Invitrogen). 2 days after the transfection, luciferase activities were measured with the dual luciferase assay system (Promega).
In Vitro Translation-Recombinant proteins were produced in the rabbit reticulocyte lysate using the TNT Quick Coupled Transcription/ Translation Systems (Promega).
Antibodies-Antibodies were derived as follows: anti-Oct-3/4 antibodies were from Santa Cruz Biotechnology and Dr. H. Hamada, Osaka University. Anti-Sox2 antibodies were from Santa Cruz Biotechnology and Drs. Y. Kamachi and H. Kondo, Osaka University. Anti-actin, anti-GATA-4 and anti-normal-rabbit IgG antibodies were from Santa Cruz Biotechnology, and anti-HA antibody was from Sigma.
Electrophoretic Mobility Shift Assay (EMSA)-Oligonucleotide probes in 1ϫ annealing buffer (10ϫ annealing buffer is 200 mM Tris-HCl, pH 8.0, 100 mM MgCl 2 , and 500 mM NaCl) were heated at 85°C for 5 min and then gradually cooled to room temperature. The annealed oligonucleotides were digested with BamHI overnight. The concentrations of the probes were adjusted to 1 g/l. 30 g of whole cell extracts was first incubated with 6 g of herring sperm DNA or 0.5 g of poly(dG-dC) as nonspecific competitors in 23 l of 1ϫ binding buffer (5ϫ binding buffer contains 50 mM Hepes, pH 7.8, 50 mM KCl, 5 mM EDTA, 25 mM MgCl 2 , 50% glycerol, 25 mM dithiothreitol, 3.5 mM phenylmethylsulfonyl fluoride, and 1ϫ Complete) with or without antibodies, and then incubated on ice for 10 min (for the simple EMSA) or 1 h (for the antibody supershift experiments). In the supershift experiments, 3 g of the antibody was included in the reaction. Then, 32 P-labeled oligonucleotide probes were added to the reaction, and the incubation was continued for an additional 30 min at room temperature. Protein-DNA complexes were analyzed on 5% polyacrylamide gel in 0.5ϫ TBE (1ϫ TBE contains 89 mM Tris-HCl, 89 mM boric acid, and 2 mM EDTA). The electrophoresis was conducted by applying 200 V to 15 ϫ 15-cm gels for 90 min at 4°C. The gels were allowed to dry, and then the signals were detected by autoradiography.

Oct-3/4 DE Regulates ES-specific Activity of the Gene-To
dissect the potential elements present in DE, we performed the reporter assay using the luciferase gene. First, we constructed a plasmid containing the 4.8-kb upstream region of the mouse Oct-3/4 gene (the upstream end corresponds to the XhoI site at nt Ϫ4760; the numbering of nucleotides in this study is according to Ref. 26, i.e. the position relative to the translational start site) fused with the luciferase gene (pOct4.8-luc, Fig. 1A). In this construct, the initiation codon of the luciferase gene was precisely overlapped with that of the Oct-3/4 gene so that the native TATA-less Oct-3/4 promoter was used for transcribing the reporter. Previously, it was demonstrated that DE and PE are contained, respectively, in the 3.29-kb and 1.39-kb BamHI restriction fragments (Fig. 1A) (8). Because pOct4.8-luc contains both BamHI fragments, it was expected that pOct4.8-luc recapitulates the expression pattern of the Oct-3/4 gene in vivo. We transfected this plasmid into terminally differentiated NIH 3T3 cells and undifferentiated D3 ES cells, together with the control reporter plasmid containing the Renilla reniformis luciferase. The control plasmid served as an internal control for calibrating transfection efficiency. When the luciferase expression level shown by pBasic-luc, which does not possess a promoter, was set at 1 unit, the activity shown by pOct4.8-luc was 23.7 in D3, but 0.3 in NIH 3T3 cells (Fig. 1A). These results reproduced those of a previous study (8) and demonstrated that pOct4.8-luc contains the region necessary for the activation of the Oct-3/4 gene in ES cells and for the repression in NIH 3T3 cells.
A sequence of CCCCTCCCCCC (Ϫ2080/Ϫ2070) was identified previously in DE as a region protected in vivo from chemical modification specifically in Oct-3/4-expressing EC and ES cells (27). Interestingly, this sequence, called Site 2A, is within the region Ϫ2136/Ϫ1939. To examine whether Site 2A is important for DE activity, we constructed another set of upstream deletion mutants focusing on the sequence around Site 2A (Fig.  1C). Upstream deletions of up to Ϫ2082 did not result in a notable reduction in the enhancer activity compared with pOct4.8-luc (Fig. 1C, bars 2-8). However, an additional deletion of up to Ϫ2078 reduced the activity significantly (ϳ50%). The deletion of up to Ϫ2078 resulted in the partial removal of Site 2A. When Site 2A was further (Ϫ2075, Ϫ2072) or completely (Ϫ2069) removed, the enhancer activity was also reduced gradually (bars 9 -12). Site 2A was originally noted to have the sequence motif CCCTCCCCC, which was first identified in PE and called Site 1A (nt Ϫ1171/Ϫ1163) (27,28). CCCTCCCC is a consensus sequence of the Sp1/Sp3 binding site, and these sequences in Sites 2A and 1A are conserved in the human, mouse, and bovine Oct-3/4 genes (26), suggesting a role of these transcription factors in Sites 2A and 1A. The gradual reduction of DE activity during increasing deletions at Ϫ2078CCTCCCCAϪ2069 (Fig. 1C, bars 9 -12) is consistent with the idea that the putative Sp1/Sp3 binding sequence in Site 2A is important for DE activity.
To examine the role of the CCCTCCCC motif in Site 2A more vigorously, we introduced point mutations at or around the sequence in pOct2.4-luc, which contained the complete Site 2A and showed an enhancer activity comparable with that of pOct4.8-luc. When the distal or proximal half of CCCTCCCC (CCCT and CCCC, respectively) or the whole CCCTCCC was mutated to unrelated sequences (Fig. 1D, mutants A-2, -3, and -4, respectively), the enhancer activity in D3 ES cells was unexpectedly not affected compared with the wild-type pOct2.4-luc (Fig. 1D, compare bars 5, 6, and 7 with bar 3). Similarly, when the CTGC sequence present immediately upstream of the CCCTCCC motif was replaced with an unrelated sequence, the enhancer activity was not reduced either ( Fig.  1D, mutant A-1 and Fig. 1D, bars 3 and 4). Therefore, we concluded that the exact sequence of the CCCTCCC motif is not essential for the function of Site 2A.
DE Contains a Novel Motif Site 2B-To delineate the region necessary for the enhancer activity of DE in ES cells more precisely, we introduced a series of internal deletions in pOct2.2-luc which contained the complete Site 2A. These mutants contained regions ranging from Ϫ2173 to various proximal ends, namely Ϫ2173/Ϫ2038, Ϫ2173/Ϫ1934, and Ϫ2173/ Ϫ1277 ( Fig. 2A). The Oct-3/4 promoter region (Ϫ284/Ϫ1, hereafter called Fragment P) is included in all of the mutant constructs. In contrast, PE, which is known to be dispensable for Oct-3/4 expression in ES cells (8), is not present in the mutant constructs. As shown in Fig. 2A, pOct2.2-luc exhibited an enhancer activity that was comparable with that of pOct4.8luc ( Fig. 2A, bars 2 and 3). When the internal region Ϫ2037/ Ϫ285 was deleted (mutant Ϫ2173/Ϫ2038ϩP), the enhancer activity was reduced significantly (bar 4). However, similar mutants possessing smaller internal deletions (mutants Ϫ2173/Ϫ1938ϩP and Ϫ2173/Ϫ1277ϩP) showed increased DE activities compared with pOct2.2 (bars 5 and 6). These results led us to propose two hypotheses. First, the region missing in Ϫ2173/Ϫ2038 but present in Ϫ2173/Ϫ1938 (i.e. region Ϫ2037/ Ϫ1938) contains an activating element for DE. Because this region does not overlap with and lies proximally to Site 2A, this putative activating element should be distinct from Site 2A and is called Site 2B hereafter. Second, the region deleted in both Ϫ2173/Ϫ1938 and Ϫ2173/Ϫ1277 (i.e. region Ϫ1276/Ϫ285) con-tains an inhibitory element for DE. PE may be a candidate for this inhibitory element. The latter possibility was not investigated further in this study.
To map the location of Site 2B, we constructed a second series of internal deletion mutants of pOct2.2-luc. Starting with mutant Ϫ2173/Ϫ2038, which showed significantly reduced enhancer activity, we included proximally extending regions and tested these mutants for their enhancer activity in D3 ES cells (Fig. 2B). mutant Ϫ2173/Ϫ2029ϩP showed minimal activity, as did mutant Ϫ2173/Ϫ2038ϩP (bar 5). In contrast, mutant Ϫ2173/Ϫ2024ϩP showed significant activity comparable with that of pOct2.2-luc (bar 6). This result suggested the presence of an important element in the region Ϫ2028/Ϫ2024. A second increase in activity was noted when mutants Ϫ2173/Ϫ2015ϩP and Ϫ2173/Ϫ2009ϩP were compared, suggesting the presence of another activating element in the region Ϫ2014/Ϫ2009. We therefore tentatively mapped the proximal boundary of Site 2B at Ϫ2024. It was noted that pOct1.9-luc, which contained neither Site 2A nor 2B, did not show any noticeable transcription (bar 16).
We next determined the distal boundary of Site 2B. We first examined whether the intervening sequence between Sites 2A and 2B is important for DE. For this purpose, successive 8 or 9 nucleotides present within the intervening region were replaced with the XhoI linker (CTCGAG) in mutant Ϫ2173/ Ϫ1991ϩP (Fig. 2C, IN-aϳIN between them is not essential for DE activity, and the distal boundary of Site 2B was tentatively mapped at Ϫ2040 (the distal end of the sequence contained by the enhancer-proficient IN-d). Accordingly, Site 2B corresponds to the region Ϫ2040/ Ϫ2024. This region is highly conserved among human, bovine, and mouse (26).
We examined whether there are any requirements for the length of the intervening sequence between Sites 2A and 2B. When the region Ϫ2063/Ϫ2044 (20 bp) was replaced with the XhoI linker (6 bp), DE activity was totally abolished (Fig. 2C,  IN-f). However, when the distal or proximal half of the same region (9 bp each) was replaced separately with the linker, the activity was not reduced (Fig. 2C, IN-a and IN-c). Therefore, we concluded that there should be a minimum length requirement for the linker region to activate DE. On the other hand, when an unrelated sequence of 5, 10, or 30 nucleotides was inserted into the linker sequence, the DE activity was largely unaffected (Fig. 2C, IN-g, IN-h, and IN-i). Therefore, it appears that the length of the linker region does not have a maximum limitation as examined here.
Both Sites 2A and 2B Are Necessary and Sufficient for Specific Activation of Native and Heterologous Promoters in ES Cells-We analyzed the relative importance of Sites 2A and 2B in ES cell-specific expression of the Oct-3/4 gene. To this end, a short DNA fragment, 2Aϩ2B, 2A, or 2B, was inserted at the upstream of luciferase gene driven by the minimal TK promoter (Fig. 3A). 2Aϩ2B, 2A, and 2B contained both Sites 2A and 2B, Site 2A only, and Site 2B only, respectively. Note that 2A contained the distal 3 bp of Site 2B but lacked the critical nucleotides identified in Fig. 2 (Fig. 2B, bar 5). These reporter plasmids were transfected into D3 ES cells or NIH 3T3 cells, and the enhancer activity of each construct was quantitated by setting the value for pBasic-luc (which does not contain a promoter) at 1. The addition of either Site 2A or Site 2B to pTK-luc did not significantly increase the absolute activity or the relative activities in ES cells and NIH 3T3 cells of pTK-luc (Fig. 3A, bars 7-10; ϳ2-fold higher activity in ES cells than in NIH 3T3 cells). In contrast, when both Sites 2A and 2B were fused to pTK-luc, significant activation specifically in D3 ES cells (ϳ8-fold compared with pTK-luc in D3 cells, bar 5), but not in NIH 3T3 cells (bar 3), was observed.
Together with the results obtained with mutants lacking Site 2A or Site 2B (Figs. 1 and 2), these results clearly demonstrated that both Sites 2A and 2B are necessary and sufficient for the specific activation of the native Oct-3/4 promoter and the heterologous minimal TK promoter in ES cells.
Oct-3/4 and Sox2 Recognize Site 2B and Activate DE in ES Cells-Because Sites 2A and 2B are responsible for ES cellspecific DE activity, it is possible that these elements are recognized by ES cell-specific factors. To test this possibility, EMSA was performed using 32 P-labeled oligonucleotide probes 2A and 2B, which are 25-nt-long double-stranded DNAs containing Site 2A and 2B sequences, respectively (Fig. 4A, top). Oligonucleotide 2A showed one shifted band in both cases where extracts from D3 or NIH 3T3 cells were used (Fig. 4A,   lanes 1 and 2). In contrast, oligonucleotide 2B showed at least five shifted bands only when extracts from D3 were used (lanes 3 and 4, arrows [1][2][3][4][5]. The addition of a 10-fold or 100-fold excess of cold oligonucleotide inhibited the formation of shifted bands, indicating that the interaction is specific (data not shown for oligonucleotide 2A and Fig. 4A, lanes 5-8, for oligonucleotide  2B). Therefore, the results suggested that oligonucleotide 2B is recognized by a factor(s) specifically present in ES cells, and oligonucleotide 2A is recognized by a factor(s) present in both ES cells and NIH 3T3 cells.
Inspection of the sequence of Site 2B led to the identification of a potential octamer element and a potential Sox binding motif. The consensus sequence of the octamer element, which is recognized by POU proteins, is ATGCAAAT. POU proteins contain two DNA binding domains, POU S and POU H . One half of the octamer element, ATGC, is recognized by POU S and is relatively conserved. The other half, AAAT, is recognized by POU H , is relatively degenerative among different targets, and contributes less to specific binding compared with POU S -recognized sequences (12,18,29). POU S and POU H exert their DNA binding ability independently and in a flexible manner by adopting various overall DNA-protein interactions (30). As such, the relative orientation, direction, and spacing of POU Sand POU H -target half-elements vary among different Oct-protein-target genes.
We found that AGAT GCAT present in Site 2B is close to the octamer motif, wherein the POU S -specific ATGC (represented as the opposite strand sequence GCAT, hereafter called the Oct-S sequence) completely matches the consensus and the putative POU H -specific sequence (AGAT, called the Oct-H sequence) has two base substitutions from the consensus (AAAT). Adjacent to this potential octamer element is AACAAAG (hereafter called the Sox2 sequence), which matches the consensus sequence of the Sox binding site (A/T)(A/T)CAA(A/T)G. Because it is known that Sox factors and octamer-binding factors frequently function coordinately, we focused on the possibility that these binding sites are functional.
We constructed oligonucleotide 2B-derived probes harboring mutations at the Oct-H, Oct-S, Oct-HϩOct-S, Sox2, or Oct-HϩOct-SϩSox2 sequence (Fig. 4B, Oct-mtH, Oct-mtS, Oct-mt, Sox-mt, and Oct/Sox-mt, respectively). When Oct-mtH or Oct-mtS was incubated with D3 extracts (Fig. 4B, lanes 2 and 3), two of the five shifted bands (bands 1 and 3) produced by the wild-type probe disappeared, but other bands (bands 2, 4, and 5) were observed. Oct-mt gave rise to a similar result except that band 1 persisted. When Sox-mt was used as a probe, we observed a reciprocal result for bands 2 and 3: band 2 disappeared, and band 3 persisted (lane 5). Band 1 disappeared in this case. Oct/Sox-mt also failed to produce band 3, and the intensity of band 2 was reduced significantly (lane 6). These results suggested that band 3 is produced by an octamerbinding factor, whereas band 2 is produced by Sox factors.
ES cells contain at least three types of octamer-binding proteins, Oct-1, Oct-3/4, and Oct-6 (31). Oct-1 is ubiquitously present in cells, including NIH 3T3 cells (e.g. see Ref. 32). Because the retarded oligonucleotide 2B bands were not produced by the NIH 3T3 cell extracts, it is unlikely that Oct-1 is involved in the formation of these bands. Here, we focused on Oct-3/4 and performed the supershift assay in EMSA using antibodies specific to Oct-3/4 or Sox2 (Fig. 4B, lanes 7-11). The anti-Oct-3/4 antibodies used here are specific to Oct-3/4 and do not recognize Oct-1 or Oct-6 (33). Similarly, the anti-Sox2 antibodies used here are specific to Sox2 and do not recognize closely related Sox1 or Sox3. 2 Anti-Sox2 and anti-Oct-3/4 eliminated  8 and 9). Moreover, when both antibodies were incubated simultaneously, both bands disappeared (lane 10). These results indicated that bands 2 and 3 correspond to DNA⅐Sox2 and DNA⅐Oct-3/4 complexes, respectively. The finding that bands 1, 4, and 5 were not affected by the inclusion of anti-Oct-3/4 or anti-Sox2 antibodies suggested that these bands are formed by different factors. We did not detect a retarded band corresponding to the DNA⅐Sox2⅐Oct-3/4 ternary complex.

Oct-3/4 and Sox2 in Oct-3/4 Regulation 5312 bands 2 and 3, respectively (lanes
Finally, to confirm that bands 2 and 3 are produced by Sox2 and Oct-3/4, respectively, we prepared HA-tagged Sox2 and HA-tagged Oct-3/4 using the rabbit reticulocyte lysate (Fig. 4C, left) and used these recombinant proteins in EMSA (Fig. 4C,  middle). Upon incubating with the oligonucleotide 2B, HAtagged Sox2 and HA-tagged Oct-3/4 produced bands that showed identical mobilities with band 2 and band 3, respectively (compare lane 1 with lanes 3-5). Band 1, 4, or 5 was not observed, indicating that they are not produced by Oct-3/4 or Sox2. We did not observe a band that appeared only when both Sox2 and Oct-3/4 was incubated with the probe and potentially represented the DNA⅐Sox2⅐Oct-3/4 ternary complex. When a mixture of HA-Sox2 and HA-Oct-3/4 recombinant proteins was incubated with the mutant probes used in Fig. 4B, the results were consistent with this hypothesis: band 2 and band 3 appeared only when the Sox binding motif and the octamer motif were intact, respectively (Fig. 4C, right). Taken together, we concluded that oligonucleotide 2B is specifically recognized by Sox2 (band 2) and Oct-3/4 (band 3).
The functional significance of the Sox2 and Oct-3/4 binding sites was examined in the reporter assay using D3 cells. Five mutant sequences (Oct-mtH, Oct-mtS, Oct-mt, Sox-mt, and Oct/Sox-mt), which were subjected to EMSA (Fig. 4B), were introduced into the reporter plasmid pOct2.4-luc (Fig. 5A). All of these mutant plasmids expressed less luciferase activities compared with the wild-type (Fig. 5A, bars 3-8). Among them, Oct-mtH showed the mildest reduction of expression (ϳ80% of the wild-type), whereas Oct-mtS and Oct-mt produced more profound effects (ϳ40% of the wild-type). This observation is consistent with the previous ones that mutations at POU Srecognized sequences resulted in a more profound effect on gene activation than those at POU H -recognized sequences (e.g. Ref. 20). Oct-mt is a double mutant consisting of Oct-mtH and Oct-mtS. The observation that Oct-mtS and Oct-mt showed similar effects also suggested that Oct-S is more critical than Oct-H in this pathway. Sox-mt produced the most profound effect as a single site mutant (ϳ40% of the wild-type). Because the POU-DNA binding is very flexible, it is possible that there are occult elements in the neighbor of the region we focused on in these experiments which play an important role in the enhancer activity. Mutations immediately downstream (N1mt) or upstream (N2-mt) of the Oct-Sox2 sequences did not have any effect on the enhancer activity (Fig. 5A, bars 13 and  14). Thus, it is unlikely that there is a promiscuous binding site of Oct-3/4 at the surrounding region. Interestingly, the triple mutant, Oct/Sox-mt, which is mutated for all of the Oct-H, Oct-S, and Sox2 sequences, showed an additively lower level of expression (ϳ20% of the wild-type) than Sox-mt or Oct-mt, indicating that Sox2 and Oct sequences function in parallel. These results demonstrated that the Oct-H, Oct-S, and Sox2 sequences at Site 2B are all functional enhancer elements and

FIG. 4. Site 2A is recognized by ubiquitous factors present in both NIH 3T3 cells and D3 ES cells, and Site 2B, by D3-specific factors.
A, labeled oligonucleotides 2A and 2B (top) were incubated with D3 or NIH 3T3 cell extracts and subjected to EMSA (bottom). The specificity of the DNA-protein complexes that were obtained when D3 extracts and oligonucleotide 2B were incubated was examined by adding the indicated excess amounts of cold oligonucleotide 2B to the mixture (lanes 6 and 7). B, oligonucleotide 2B and its mutant forms (top) were incubated with D3 extracts and subjected to EMSA (bottom). In lanes 8 -11, anti-Sox2, anti-Oct-3/4 antibodies or normal rabbit IgG was included as indicated. C, rabbit reticulocyte lysate (RRL) was incubated with pcDNA3 (mock), pcDNA3-HA/Sox2 (HA-Sox2), or pcDNA3-HA/Oct-3/4 (HA-Oct-3/4) (left) and incubated with oligonucleotide 2B (middle). D3 extracts were also used as control. A mixture of HA-Sox2 and HA-Oct-3/4 was incubated with the mutant probes (right).
In cases where Sox2 and Oct-3/4 binding sites function coordinately, these two binding sites are juxtaposed very closely: no intervening nucleotide is present in the cognate sequences of UTF1 and Sox-2 gene enhancers and a 3-bp sequence in the FGF-4 enhancer. The proximity of these two elements is important for the enhancer activity of the sequence because artificial insertion of an unrelated sequence between the Sox2 and Oct-3/4 binding sites reduced the enhancer activity (34), thereby attesting that the specific spatial arrangement of the two factors is important for transcriptionally active complex formation. To test whether Site 2B behaves similarly or not, we inserted 2-, 4-, or 10-bp unrelated sequences between the Sox2 and Oct elements in pOct2.4-luc reporter plasmid (Fig. 5A,  2B-IN2, -IN4, and -IN10). The expression levels produced by the three insertion mutants in D3 cells were ϳ40 -60% of that by the wild-type pOct2.4-luc (Fig. 5A, bars 3, 9 -11). Therefore, it is suggested that the proximity of the octamer motif and the Sox2 binding site is important for Site 2B to activate the Oct-3/4 gene in ES cells.
Site 2B Binding Activity Is Down-regulated by RA-induced Differentiation in ES Cells-The expression of the Oct-3/4 gene is extinguished in ES and EC cells when differentiation is induced by RA (2). We examined the correlation between the Site 2B binding activities (bands 2 and 3) and the expression levels of Oct-3/4 and Sox2 during RA-induced differentiation of D3 ES cells (Fig. 6A). The abundance of Oct-3/4 and Sox2 proteins was reduced significantly on day 1 in the RA-treated cells. Oct-3/4 was negligible on day 2 and thereafter, whereas low levels of Sox2 remained throughout the experiment (Fig. 6A, note that the two bands migrating slightly faster and slower than the Oct-3/4 signal are cross-reactive unrelated proteins). GATA-4, which is known to be up-regulated by RA-induced differentiated ES cells (36), was found to be activated gradually on day 2 and thereafter, indicating that the RA treatment indeed induced the differentiation. We found that bands 2 and 3 were reduced significantly on day 1 and barely detectable on day 2 and thereafter in the RA-treated cells (Fig. 6B). These results are consistent with the notion that bands 2 and 3 are produced by Sox2 and Oct-3/4, respectively, and suggest that the extinction of the Site 2B enhancer is involved in the down-regulation of the Oct-3/4 gene during differentiation.
Requirements of Oct-3/4 for DE Activity-To analyze whether Oct-3/4 is indeed required for activating DE, we utilized the inducible knock-out system of Oct-3/4. In ZHBTc4 ES cells, both alleles of the endogenous Oct-3/4 gene are inactivated, and Oct-3/4 and Sox2 in Oct-3/4 Regulation one copy of tetracycline (Tc)-regulated exogenous Oct-3/4 transgene expresses Oct-3/4. The Oct-3/4 expression is repressed by the addition of Tc (11). We transiently transfected the DE reporter plasmid pOct2.4-luc, as well as a reporter construct containing the minimal TK promoter plus the UTF1 enhancer (Utf-tk) as a positive control, to ZHBTc4 cells in the presence or absence of Tc, and luciferase activity in the harvested extracts was measured 24 h later. It is known that the UTF1 regulatory element is controlled by Oct-3/4 and Sox2 (18). The activity was normalized to the transfection efficiency, and the relative activity of Tc(Ϫ)/Tc(ϩ), i.e. Oct-3/4(ϩ)/Oct-3/4(Ϫ) was calculated for each construct (Fig. 7). This assay is advantageous in that one can analyze the effect of the physiological level of Oct-3/4 on the reporter plasmids. Utf-tk was highly activated by the presence of Oct-3/4 (12.3-fold), as expected (bar 1), whereas the minimal TK promoter alone was not significantly affected by the presence or absence of Oct-3/4 (1.1-fold) (bar 2). pOct2.4luc, which contains both Sites 2A and 2B, showed a significant activation (5.7-fold) in the presence of Oct-3/4, whereas pOct1.9-luc, which lacks Site 2A and 2B, showed only moderate activation (2.3-fold) (bars 3 and 12, respectively). We also found that the different mutations introduced into pOct2.4-luc exhibited very similar behaviors in the assay that involved simple transfection of reporter plasmids to ES cells and the Oct-3/4-on/off experiments (Figs. 5 and 7). These results were not straightforwardly expected. For example, one may expect that Sox-mt, which presumably binds to Oct-3/4 and not Sox2, will be more affected by the absence of Oct-3/4 than Oct-mt, which presumably binds to Sox2 and not Oct-3/4. However, Oct-mt, Sox-mt, and Oct/Sox-mt showed similar degrees of Oct-3/4 dependence (Fig. 7, bars 6 -8). The simplest interpretation of these results is that the Oct-3/4 and Sox2 present in ES cells act together on DE and cannot act independently of each other in this system. Taken together, it was demonstrated that the enhancer activity of Site 2B is dependent on Oct-3/4 present in ES cells. DISCUSSION This study demonstrated that the Oct-3/4 gene itself is among the list of the Sox2⅐Oct-3/4-regulated genes. This conclusion is based on several lines of evidence. We showed that the ES-specific enhancer activity of DE depends on the octamer sequence and the Sox2 binding sequence present at Site 2B. Extracts derived from ES cells, but not those from NIH 3T3 cells, produced complexes with Site 2B oligonucleotide in EMSA. Moreover, the DE activity was reduced significantly when Oct-3/4 was inducibly knocked down in ES cells. Because ES cells differentiated when Oct-3/4 was knocked down (11), the last result may be explained by an indirect effect of the differentiation. However, the reporter activity was measured 24 h after Tc was added to the culture, the earliest time point when the Oct-3/4 protein disappeared and well before the cells showed differentiation phenotypes (11). Indeed, Sox2 and Nanog, a homeoprotein important for the maintenance of the pluripotency (37,38), are expressed at this time point (11). 3 Therefore, it is likely that the effect was caused directly by the absence of Oct-3/4. The artificial assembly of Site 2A and Site 2B with the minimal TK promoter or the native Oct-3/4 promoter indicated that both Site 2A and Site 2B are necessary and sufficient for activating these promoters specifically in ES cells. Therefore, the Oct-3/4 expression in ES cells is regulated at least by two layers of mechanisms: first, the combination of Oct-3/4 and Sox2 at Site 2B, and second, the unknown functions of Site 2A.
Modes of Action of Oct-3/4 and Sox2 at Site 2B-The FGF-4 gene has been most extensively studied to elucidate the molecular mechanism involved in the control of gene activity by Sox2 and Oct-3/4. It was found that Sox2 and Oct-3/4 specifically bind to the enhancer and activate FGF-4 transcription in EC cells, and such related factors as Oct-1, Oct-6 or Sox5 cannot substitute for them (15). Sox2 and Oct-3/4 bind simultaneously to the motifs to form a ternary factor DNA complex and activate the FGF-4 transcription synergistically. Indeed, Oct-3/4 and Sox2 associate directly with each other through a physical interaction between the POU domain of Oct-3/4 and the high mobility group domain of Sox2. This protein-protein interaction confers the cooperative binding of Oct-3/4 and Sox2 to target DNAs possessing a correct spatial arrangement of the cognate binding sites (34). From these observations, it was proposed that the precise arrangement of Oct-3/4 and Sox2 binding sites within the FGF-4 enhancer produces the expression specificity via 2-fold mechanisms. First, the combination of two transcription factors present in one cell contributes to the selection of the target genes. Second, the exact organization of the binding sites in the enhancer determines the spatial arrangement of the two transcription factors bound to the elements and provides a specific platform for the assembly of a transcriptionally active complex and expression specificity (15,34,35  2B does not match the consensus sequence (ATTA/T GCAT) perfectly: the POU S -recognized sequence (AGAT) differs significantly from the consensus ATTA/T, whereas the POU H -recognized sequence is identical with the consensus (GCAT). This is not unprecedented, however, as other Sox2⅐Oct-3/4-regulated genes, UTF1, Sox2, and Fbx15, have atypical POU H -recognized sequences (ACTA GCAT, ATAT GCTA, and TTTA TCAT, respectively). It was shown that the POU H -recognized sequence contributes relatively less to the specific binding of Oct proteins to target DNAs (18,20). Indeed, a recent systematic analysis of a battery of octamer motif mutants demonstrated that Oct-3/4 accommodates a diverse set of mutations at the POU H -recognized site (29). When a POU S -recognized sequence is atypical, Oct proteins can bind to the target via a different mode of nucleotide recognition: The neighboring AT-rich Sox2 binding motif (which resembles the authentic POU S site ATTA/T) is recognized instead of the atypical POU H -recognized sequence and together with the POU S -recognized sequence, contributes to the binding of Oct proteins to target DNAs. In such cases, the Sox2 binding site is occupied by either Oct proteins or Sox2 in a mutually exclusive manner (18). Given that the POU H -recognized sequence at Site 2B is atypical and we could not detect the Sox2⅐Oct-3/4⅐DNA ternary complex, it is possible that Sox2 and Oct-3/4 bind to Site 2B in such alternative modes. However, we think that this possibility is unlikely because we observed the Oct-3/4⅐DNA complex in the EMSA using Sox-mt as the probe (Fig. 4B). We did not find an additional AT-rich sequence in the neighbor of Site 2B which may potentially serve as a POU H -recognized sequence. Moreover, mutations introduced to the adjacent regions of Site 2B did not affect the enhancer activity of Site 2A (N1-mt and N2-mt in Fig.  5). We therefore believe that Sox2 and Oct-3/4 recognize the Sox2 binding motif and the octamer motif within Site 2B, respectively.
Although we could not detect the Sox2⅐Oct-3/4⅐DNA ternary complex in the EMSA, several lines of evidence suggest that Sox2 and Oct-3/4 act together, and not independently, on DE. First, when the DNA length between the Sox binding site and the Oct-3/4 binding site was increased, the enhancer activity was reduced significantly (Fig. 5). Second, Oct-mt, Sox-mt, and Oct/Sox-mt showed similar degrees of Oct-3/4 dependence (Fig.  7). This observation can be explained by the hypothesis that Site 2B is activated by Oct-3/4 only when both Sox2 and Oct-3/4 binding sites are competent for binding to the cognate factors. The reason that we did not detect the Oct-3/4⅐Sox2⅐DNA ternary complex in EMSA is not clear. The complex may be too unstable to be detected in such in vitro experiments and/or may require additional proteins other than Oct-3/4 and Sox2 to be stabilized.
Functions of Site 2A-Because Site 2B alone was not sufficient to activate the homologous and heterologous promoters in ES cells, and both Site 2A and Site 2B were required (Fig. 3), it is clear that Site 2A plays an important role in DE. However, Site 2A tolerated extensive nucleotide exchanges in the reporter assay (Fig. 1D), making the significance of the primary sequence enigmatic. Site 2A was originally identified by its close resemblance to the Site 1A sequence present in PE. In vivo footprinting experiment showed that Site 2A exhibited strong chemical protection in undifferentiated p19 EC cells and D3 ES cells, but not in RA-induced differentiated cells (27). We detected apparently the same complexes with oligonucleotide 2A in the EMSA when incubated with ES cell-derived and NIH 3T3 cell-derived extracts and did not find any ES-specific complex (Fig. 4A). The nucleotide sequences around Site 2A are very GC-rich. 35 of the 47 nucleotides (74%) of the region Ϫ2109/Ϫ2063 are GC (Site 2A is Ϫ2080/ Ϫ2069). If factors recognizing Site 2A do so by recognizing GCrich sequences in a relaxed manner, it would be difficult to pinpoint the responsible sequence by means of a simple mutation assay. Alternatively, Site 2A may contribute to DE by adopting a specific conformation of DNA, such as bent and triplex DNAs. The exact role played by Site 2A should be addressed in future studies.
Significance of Oct-3/4 Gene Regulation by Sox2 and Oct-3/4-Sox2 and Oct-3/4 expressions overlap during early embryogenesis, and both are important for the maintenance of the pluripotent state (39). Sox2 expression in EC cells is regulated by Sox2 itself and Oct-3/4, suggesting the possibility that Sox2 is activated in primitive cells by a positive autoregulatory loop (20). Here we showed that Oct-3/4 expression in ES cells is regulated by Sox2 and Oct-3/4. Therefore, it is tempting to speculate that the same positive feedback loop maintains the expression of Sox2 and Oct-3/4 together and that the Sox2 and Oct-3/4 are regulated coordinately. However, previous observations suggest that these genes are regulated in a more complex manner. First, it was reported previously that when embryonic ectodermal cells are converted into mesodermal cells, UTF1 and Sox2 expressions are extinguished rapidly whereas Oct-3/4 expression is reduced rather gradually (20). The difference in the kinetics of Sox2 and Oct-3/4 expression suggests that these two genes are regulated at least in part in a different manner. Second, weak but some level of Oct-3/4 expression was found in Sox2 Ϫ/Ϫ blastocyst cultures, suggesting that Sox2 may not be required for this low level of Oct-3/4 transcription (39). However, it is not possible to exclude the possibility in this case that the maternally derived Sox2 was responsible for the Oct-3/4 transcription. Third, it was reported that the transcript derived from the endogenous inactivated Oct-3/4 allele was expressed when the ectopic functional Oct-3/4 transgene was repressed in ZHBTc4 ES cells (11). Therefore, although Oct-3/4 plays a role in DE (Fig. 7), it is not essential for maintaining the Oct-3/4 expression, suggesting that Oct-3/4-independent pathways are involved in the control of Oct-3/4 expression.
One possible Oct-3/4-independent pathway is that Oct-1, another Oct protein abundantly expressed in ES cells, activates Site 2B. Oct-1 is larger than Oct-3/4 and is detected as a more slowly migrating band in EMSA. However, we did not observe such a candidate retarded band in EMSA. Alternatively, Site 2A, another cis-element required for DE activity, may perform its function synergistically with Site 2B. Because Sites 2A and 2B are positioned closely, it is possible that the factors tethered to Sites 2A and 2B are assembled into a common, more highly ordered structure. In such a case, the activating complex may be sufficiently robust to maintain the Oct-3/4 expression even after one component (Oct-3/4) is removed from the complex. As Site 2A was recognized by a factor(s) present in both D3 ES cells and NIH 3T3 cells (Fig. 4A), a ubiquitously expressed factor may be involved in this process.
The quantitative level of Oct-3/4 is important for the fate determination of primitive cells (11). To maintain the pluripotent state, it is required that the cells maintain the Oct-3/4 protein level within the narrow window of abundance. Although the regulatory mechanism of Oct-3/4 is expected to be more complicated than a simple autoregulatory loop as discussed, the observation that both Sox2 (20) and Oct-3/4 (this study) are controlled by the Sox2⅐Oct-3/4 complex suggests that this circuit regulation may contribute to a robust maintenance of the steady levels of Oct-3/4 and Sox2 in pluripotent cells.