Squelching of ETS2 Transactivation by POU5F1 Silences the Human Chorionic Gonadotropin CGA Subunit Gene in Human Choriocarcinoma and Embryonic Stem Cells

Abstract

The subunit genes encoding human chorionic gonadotropin, CGA, and CGB, are up-regulated in human trophoblast. However, they are effectively silenced in choriocarcinoma cells by ectopically expressed POU domain class 5 transcription factor 1 (POU5F1). Here we show that POU5F1 represses activity of the CGA promoter through its interactions with ETS2, a transcription factor required for both placental development and human chorionic gonadotropin subunit gene expression, by forming a complex that precludes ETS2 from interacting with the CGA promoter. Mutation of a POU5F1 binding site proximal to the ETS2 binding site does not alter the ability of POU5F1 to act as a repressor but causes a drop in basal promoter activity due to overlap with the binding site for DLX3. DLX3 has only a modest ability to raise basal CGA promoter activity, but its coexpression with ETS2 can up-regulate it 100-fold or more. The two factors form a complex, and both must bind to the promoter for the combination to be transcriptionally effective, a synergy compromised by POU5F1. Similarly, in human embryonic stem cells, which express ETS2 but not CGA, ETS2 does not occupy its binding site on the CGA promoter but is found instead as a soluble complex with POU5F1. When human embryonic stem cells differentiate in response to bone morphogenetic protein-4 and concentrations of POU5F1 fall and hCG and DLX3 rise, ETS2 then occupies its binding site on the CGA promoter. Hence, a squelching mechanism underpins the transcriptional silencing of CGA by POU5F1 and could have general relevance to how pluripotency is maintained and how the trophoblast lineage emerges from pluripotent precursor cells.

Chorionic gonadotrophin (CG) is considered as the primary signal for maternal recognition of pregnancy in higher primates, including humans. It acts as a luteotrophic hormone, maintaining progesterone secretion from the corpus luteum and thereby preventing the latter from the functional loss of activity that would normally occur at the end of an ovarian cycle in which a pregnancy was not initiated (1). Human CG (hCG) is first expressed from the trophoblast cells of the preimplantation embryo, beginning about d 7 or 8 after fertilization during the onset of hatching and implantation to the uterine wall (2–4). After implantation, serum concentrations of hCG, which have generally been measured by immune assays that recognize only the β-subunit, rise exponentially before reaching a maximum after 8–10 wk and subsequently falling (5, 6). Although production of the intact hormone drops precipitously in the second trimester, circulating concentrations of the free α-subunit remain high, suggesting that hCG subunit A (CGA) might have an individual role during pregnancy and that expression of the two subunit genes is not particularly well coordinated.

Various transcription factors acting in combination regulate CGA expression (Fig. 1). Two adjacent 18-bp repeat elements, known as cAMP response elements (CRE) spanning −142 to −115 bp in the proximal promoter are crucial (7, 8) and bind a phosphorylated form of CRE-binding protein (9, 10). A region upstream of the most distal CRE also has an important regulatory role and contains binding sites for several transcription factors, including GATA family members, most likely GATA2 (11–14), and TFAP2C (activator protein-2γ) (12, 15). Additionally, ETS2, acting through pair of overlapping ETS2-binding elements that span the −82 to −74 region, is a potent transactivator of CGA expression (16). Mutation of either of the ETS2 sites dramatically reduces the effect of cAMP on CGA promoter activity, whereas mutations within either CRE abolish responsiveness to ETS2. This interdependence of the two control regions emphasizes the importance of ETS2 as a transcriptional regulator of CGA expression and its broad role in the up-regulation of signature genes of trophoblast from a wide range of species (16, 17). Finally, overexpression of the homeobox gene, DLX3, modestly up-regulates the CGA promoter (∼2.5 fold) (18) through a sequence (−114 to −107) that partially overlaps an octamer-binding site placed −117 to −110 bp (Fig. 1).

Fig. 1.

Sequence of the CGA promoter region upstream of the transcription initiation site (−167 to +33) showing proposed regulatory elements along with their sequences. DLX3, Binding site for DLX3; POU5F1, binding site for POU5F1 and possibly other octamer transcription factors. ETS marks the two core binding sequences that are known to bind ETS2. Both sites, one on the sense and the other on the antisense strand are known to be functional (16).

The key transcription factor maintaining pluripotency in the inner cell mass and epiblast of mammalian embryos, POU domain class 5 transcription factor 1 (POU5F1), also may play a role in controlling the expression of the CG subunits. In 1997, for example, Liu et al. (19) showed that POU5F1 reduced the production of CGA-encoded protein by JAr choriocarcinoma cells by approximately 90% and effectively silenced the ability of the CGA promoter to drive a reporter gene and bound to the octamer site described above and to no other sequence in the known control regions of the gene. Intriguingly, however, a mutation that abolished such POU5F1 binding failed to reverse the ability of POU5F1 to silence CGA promoter activity, suggesting that the silencing effect was mediated by either a quenching or squelching mechanism that did not require binding of POU5F1 to DNA. POU5F1 also effectively silenced reporter gene expression driven by promoters from two other genes expressed in trophoblast but not inner cell mass or epiblast, human CGB5 (19, 20) and bovine IFNT. The interferon-τ proteins (IFNT) have an analogous role to hCG, in that they are responsible for rescuing the corpus luteum of pregnancy in ruminant species (21) and, like hCG, are produced in extremely large quantities by the conceptus during the early stages of pregnancy but are not associated with cell types expressing POU5F1. Together such data suggest that POU5F1 might have a broad function in controlling the transcriptional activity of a number of signature genes of the trophoblast before the establishment of that lineage and that its overexpression leads to partial dedifferentiation of the cells. Hence, the mechanism whereby POU5F1 participates in the transcriptional silencing of CGA, the topic of this paper, could have general relevance to how pluripotency is maintained, how the trophoblast lineage emerges, and to the general topic of reprogramming by expression of pluripotency factors.

Materials and Methods

Reporter gene constructs and expression plasmids

Human CGA promoter constructs driving the luciferase (Luc) reporter gene (−255luc), containing the gene control region −255 to +48, were subcloned into pGL2 basic vector (Promega, Madison, WI). A reporter construct with the POU5F1-binding site mutated (CGA-POU5F1mut) was prepared according to Liu et al. (19). Progressively deleted −255luc, −107luc, and −48luc CGA reporters were also created. The expression plasmids for mouse POU5F1 as well as its derivatives pCMV-POU5F1, pCMV-4N-POU4, pCMV-4N, pCMV-POU4, and pCMV-POU4−4 C were provided by H. Schöler (Max Planck Institute for Molecular Biomedicine, Münster, Germany) (21). The expression plasmid for full-length human ETS2 has been described previously (22). One of the two ETS binding sites was mutated by site-directed mutagenesis to create the CGA-ETS2-mutated promoter construct (16). The mouse distal-less homeobox 3 (DLX3) expression plasmid DLX3/pCI-neo was obtained as a gift from M. Morasso (National Institute of Arthritis and, also known as OCT4 Musculoskeletal and Skin Diseases, Bethesda, MD) and has been described elsewhere (18, 23). The POU5F1-binding site mutation also mutated the DLX3-binding site so that the same plasmid was used as the CGA-mut DLX3/POU5F1 construct.

Cell culture and transfections

JAr choriocarcinoma cells (HTB-144; American Type Culture Collection, Manassas, VA) were maintained in RPMI 1640 (Invitrogen, Carlsbad, CA) medium supplemented with 10% fetal bovine serum (Harlan, Indianapolis, IN). The cells were transfected by using Lipofectamine-Plus (Invitrogen) reagents by the manufacturer's instructions. JAr cells were plated in six-well plates (TPP/Midsci, St. Louis, MO) at density 1 × 10⁵ cells/well, incubated overnight and transfected with 1.5 μg of reporter gene construct and selected amounts of expression vector DNA per well in presence of 13 ng of the control plasmid pRSVLTR-βgal (24). Total amount of transfected DNA was kept constant. After 44 h of incubation at 37 C under 5% CO₂, the cells were washed with PBS and lysed with Tropix, Galactolight-plus lysis buffer (Applied Biosystems, Foster City, CA). Luc activity was measured by luciferase assay reagent (Promega) with a 20/20ⁿ luminometer (Turner Biosystems, Sunnyvale, CA). β-Galactosidase activity was measured by using Tropix Galactolight substrate (Applied Biosystems). Extracts were heated at 48 C for 1 h to inactivate endogenous β-galactosidase. The enzymatic activities of each promoter Luc reporter construct were normalized to the control β-galactosidase activity. JAr cell lines stably transfected with pcDNA3-POU5F1 expression vector (clones S1 and S4) and with pcDNA3 empty vector (clones C1 and C2) have been described previously by Liu and colleagues (19, 20). These cell lines were maintained in 90% DMEM supplemented with 10% fetal bovine serum.

Human endometrial stromal cells (hESC) H1 (WA01; WiCell, Madison, WI) were cultured in six-well plates (Nunc, Sigma-Aldrich, St. Louis, MO) on a substratum coated with Matrigel (BD Biosciences, San Diego, CA) in mTeSR1 medium (STEMCELL Technologies, Vancouver, Canada). To examine the regulatory transition from inactive to active CGA expression, the H1 cells were grown in culture medium [80% DMEM/F12 supplemented with 20% KnockOut serum replacement (Invitrogen)/1 mm l-glutamine/0.1 mm 2-mercaptoethanol/1% nonessential amino acids (Sigma-Aldrich)] conditioned by γ-irradiated (7000 cGy) mouse embryonic fibroblast feeder cells and supplemented with recombinant human bone morphogenetic protein (BMP)-4 (10 ng/ml; R&D Systems, Minneapolis, MN), activin A signaling inhibitor, A83–01 (1 μm, Tocris Bioscience, Ellisville, MO), and fibroblast growth factor (FGF) receptor inhibitor, PD173074 (0.1 μm, Sigma-Aldrich) for up to 8 d after cell passage. In parallel, some cells were also grown in conditioned medium supplemented with recombinant human FGF2 (4 ng/ml) with no BMP4, A83–01, and PD173074 to maintain pluripotent status, i.e. inactive CGA expression.

Western blot analysis

Whole-cell lysates were prepared from normal JAr cells, the stably transfected JAr cell lines, S1, S4, C1, and C2, and from undifferentiated and differentiated hESC by using radioimmunoprecipitation assay buffer containing 50 mm Tris-HCl (pH 7.4), 150 mm NaCl, 1 mm phenylmethylsulfonyl fluoride, 1 mm EDTA, 0.1% sodium dodecyl sulfate (wt/vol), 1% Triton X-100 (wt/vol), and a protease inhibitor cocktail (Sigma, St. Louis, MO). Lysates were centrifuged (4000 × g) to remove particulate matter. Cleared cell lysates were analyzed by 12% SDS-PAGE. Prestained dual-label protein ladders (Bio-Rad) were used as molecular weight markers. Proteins were transferred to polyvinylidene difluoride membranes (Immobilon-P; Millipore, Bedford, MA). Affinity-purified sheep anti-DLX3, rabbit anti-POU5F1, or rabbit anti-ETS2 antibodies (17, 24, 25) were diluted 1:1000 before use. Mouse monoclonal anti-POU5F1 and rabbit anti-ETS2 antibodies (sc-5279 at 0.4 μg/ml and sc-351 at 0.2 μg/ml, respectively; Santa Cruz Biotechnology, Santa Cruz, CA) were used in Western blotting, chromatin immunoprecipitation, and coimmunoprecipitation analyses of the hESC. Rabbit anti-β-actin (ACTB) antibody (Cell Signaling Technologies, Danvers, MA) was also diluted 1:1000. Mouse anti-αtubulin antibody (1 μg/ml; GenScript, Piscataway, NJ) was used in one of the figures (see Fig. 9). Secondary antibodies used were horseradish peroxidase-linked antisheep IgG (Jackson ImmunoResearch, West Grove, PA) or horseradish peroxidase-linked antirabbit and antimouse IgG (Cell Signaling Technologies) diluted 1:5000. All antibody dilutions were made in 5% nonfat dry milk (wt/vol). Membranes were developed with Photo-type horseradish peroxidase Western blot detection system (Cell Signaling Technologies) or SuperSignal West Dura chemiluminescent substrate (Thermo Scienctific, Waltham, MA). Images were acquired with the Fuji LAS 3000 Imaging system (Fujifilm Medical Systems, Stamford, CT). Some of the blots were stripped with Restore-Plus Western blot stripping buffer (Thermo Scientific), blocked with 5% nonfat dry milk (wt/vol) and reprobed with a different antibody (manufacturer's protocol).

Fig. 9.

POU5F1, ETS2, and DLX3 expression, interaction, and association with the CGA promoter in hESC before and after exposure to conditions that cause trophoblast differentiation. A, Detection of POU5F1, ETS2, and DLX3 on Western blots of lysates from undifferentiated (hESC/FGF2) (lane 1) and differentiated (hESC/BMP4+PD+A83) H1 hESC (lane 2). The loading control was α-tubulin (TUBA) (lower panel). Analyses were performed on separate blots loaded with the same amounts of protein (30 μg) for each lane. B, Coimmunoprecipitation of POU5F1 and ETS2 from hESC control lysates. Immunoprecipitation was performed with antibody to ETS2 and Western analysis with anti-POU5F1. Lane 1 (input), 8 μg of lysate protein from undifferentiated hESC; lane 2 (control IgG), immune complex collected after addition of a nonspecific IgG; lane 3 (anti-ETS2), immune complex containing POU5F1 collected after addition of ETS2. Molecular weight markers (M_r × 10⁻³) are shown on the right side of the panels. C, Control ChIP analysis to demonstrate the association of histone H3 with exon 3 of RPL30 in chromatin prepared from control hESC (hESC/FGF2). Lane 1 shows the lack of reaction product when no template was used in the PCR reaction. Lane 2 is reaction product after the addition of 7.8 μg of chromatin but performing immunoprecipitation with a nonspecific IgG. Lane 3 is the reaction product with same amount of chromatin after immunoprecipitation with antihistone H3. Lane 4 is the amplification product obtained from 5 ng chromatin without immunoprecipitation. D, ChIP analyses demonstrating the lack of association of POU5F1 with the CGA promoter in undifferentiated (hESC/FGF2) H1 hESC. The experiment was identical with that in C except the reaction products were obtained after immunoprecipitating input chromatin with anti-POU5F1. E and F, ChIP analyses demonstrating the association or lack thereof of ETS2 (E) and DLX3 (F) with the CGA promoter in undifferentiated (hESC/FGF2, left panels) and differentiated (hESC/BMP4+PD+A83, right panels) H1 hESC. In all experiments (C–F), sheared chromatin preparations were exposed to overnight incubation with either control IgG or ChIP quality antibodies. The primers used for PCR were designed to amplify a region of the CGA proximal promoter (−96 to +48) in E and (−189 to −25) in D and F. G, Chromatin prepared from undifferentiated and differentiated hESC. Chromatin was formaldehyde cross-linked and enzymatically digested. DNA was then purified and analyzed by electrophoresis on a 1% agarose gel. Lane 1, DNA (0.24 μg) from control cells (hESC/FGF2); lane 2, DNA (0.14 μg) from differentiated (BMP4+PD+A83) cells. Gels were stained with ethidium bromide. The majority of the enzymatically fragmented chromatin was 150–500 bp in length.

Chromatin immunoprecipitation (ChIP) analysis

ChIP analysis from nuclear extracts of JAr choriocarcinoma cells was performed as described earlier (26). Briefly, sheared chromatin was obtained from stably transfected JAr S4 and C2 cells and incubated with 0.2 ml of protein G-agarose beads (Santa Cruz Biotechnology). One fifth of the total volume was preserved at −80 C as total input control. The remaining chromatin was subdivided into three treatment groups: untreated (no antibody control), exposed to 3 μg of rabbit anti-ETS2, exposed to purified IgG (as a negative control) from a nonimmunized rabbit (Active Motif, Carlsbad, CA). The immune complexes were collected on protein G-agarose beads (Santa Cruz Biotechnology), eluted by using elution buffer, and prepared for PCR analysis. The primers used (ACA CCA AGT ACC CTT CAA TCA, forward; GGA TCC GAA GAG GGA TTT TAG, reverse) were designed to amplify a region of the CGA proximal promoter (−96 to +48) containing the ETS2 binding site. PCR conditions were as follows: 95 C for 2 min for one cycle, 25 cycles of 95 C for 20 sec, 50 C for 20 sec, 72 C for 50 sec, followed by 72 C for 5 min. PCR products were visualized by ethidium bromide staining after electrophoresis on 1.5 or 2% agarose gels. A second set of ChIP assays were performed with sheared chromatin obtained from normal JAr cells transiently transfected at a range of POU5F1 expression vector concentrations. Affinity-purified rabbit anti-ETS2 antibody was used to detect any endogenous ETS2-DNA complex. PCR analysis was performed as above.

To study the interaction between DLX3 and the CGA promoter, a similar protocol was followed. The complex was immunoprecipitated with ChIP grade goat polyclonal DLX3 antibody (N-17-X; Santa Cruz Biotechnology). Amplification of the CGA proximal promoter (−189 to −25) was performed with the following primers (GCT CCA AAC AAA AAT GAC CTA AGG GTT GAA, forward; TAT ACC AGC AGA GTG TTT CCA CCT GCA TCT, reverse) within the CGA proximal promoter (Fig. 1) at an annealing temperature at 52 C. Other conditions were the same as above.

ChIP analyses from extracts of undifferentiated and differentiated hESC (hESC/FGF2 and hESC/BMP4+PD173074+A83, respectively) were conducted with SimpleChIP enzymatic chromatin immunoprecipitation kit (Cell Signaling Technologies). Enzymatically digested chromatin (7.8 μg of DNA per reaction, approximately 150–500 bp in size, see gel image in Fig. 9G) was incubated with 1 μg of the following antibodies: normal rabbit IgG, histone H3 XP rabbit monoclonal antibody (both are from Cell Signaling Technologies), rabbit anti-ETS2 antibody (Santa Cruz Biotechnology), and sheep anti-DLX3 (17). The procedures for immunoprecipitation, elution of chromatin, and DNA purification were followed according to the manufacturer's instructions. DNA (either 5 or 10 ng) from the respective intact chromatin was included as input in the PCR. Primers for the 60S ribosomal protein L30 (RPL30; Cell Signaling Technologies), CGA (−96 to +48), and CGA (−189 to −25) (Fig. 1) were used to analyze the ChIP analysis for histone H3, ETS2, and both POU5F1 and DLX3, respectively.

Coimmunoprecipitation analysis

JAr cells (in 60 mm diameter dishes) were transiently transfected with 3 μg of pCGN-ETS2 and 3 μg of DLX3/pCI-neo expression plasmid DNAs by using Lipofectamine-Plus reagents (Invitrogen). Extracts were prepared from each set of reactions by using radioimmunoprecipitation assay lysis buffer. Cell lysates were cleared by centrifugation at 4000 × g. Immunoprecipitation reactions were started with fresh cell extracts (∼600 μl) containing 1 mg protein. After treating with protein G agarose beads, the lysates were incubated overnight with either 5 μg of affinity-purified anti-ETS2 antibody (sc-351) or purified nonspecific IgG (both from Santa Cruz Biotechnology). The next day, Ig complexes were adsorbed onto 50 μl of swollen, prewashed protein G agarose beads for 6 h. The bound immune complexes were eluted in nonreducing sample buffer (Laemmli buffer without 2-mercaptoethanol) at 80 C for 15 min. Samples were analyzed in 12% SDS-PAGE gels. The immune complexes formed with ETS2 antibody were analyzed by Western blot analysis on 25 μg of protein by using affinity-purified sheep DLX3 Ig [diluted 1:1000 in 5% (wt/vol) nonfat dry milk] as the detection antibody. The bands of DLX3 were visualized by chemiluminiscence with LumiGLO and peroxide reagents purchased from Cell Signaling Technologies. Coimmunoprecipitation studies were also studied after the complex had been immunoprecipitated with affinity-purified DLX3 antibody and Western analysis performed with affinity-purified ETS2 antibody prepared in this laboratory (24).

Whole-cell lysate was prepared from undifferentiated H1 hESC (hESC/FGF2) with a buffer of 50 mm Tris-HCl (pH 8), 50 mm NaCl, 5 mm MgCl₂, 10% glycerol, 1% Nonidet P-40, 1 mm phenylmethylsulfonyl fluoride, protease inhibitor cocktail from Cell Signaling Technologies. Cell lysate containing 136 μg protein was incubated overnight with either 2 μg of anti-ETS2 antibody (sc-351; Santa Cruz Biotechnology) or normal rabbit IgG (Cell Signaling Technologies). Immune complexes were captured by the addition of a 50% slurry (40 μl) of protein G agarose beads (preadsorbed with BSA and sonicated salmon sperm DNA; Cell Signaling Technologies) and incubated for 24 h at 4 C. The beads were then washed three times in the buffer of 50 mm Tris-HCl (pH 8), 50 mm NaCl, 5 mm MgCl₂, 10% glycerol, 1% Nonidet P-40, 1 mm phenylmethylsulfonyl fluoride, protease inhibitor cocktail (Cell Signaling Technologies; 1 ml each). Bound immune complexes were eluted in standard 2× Laemmli buffer at 100 C for 5 min. The immune complexes were analyzed by Western blot analysis by using mouse monoclonal anti-POU5F1 antibody (sc-5279, 0.4 μg/ml) as the detection antibody. The intact lysate (8 μg) was included as input in Western blotting.

Immunoassays

Media collected from hESC cultures every 24 h were stored at −20 C until assay. The concentration of hCG was measured by using solid-phase sandwich ELISA kit (catalog no. BQ 047F; Bio-Quant, San Diego, CA). Standards were supplied by the manufacturer. For all assays, samples were appropriately diluted to fall within the recommended ranges of the standard curves.

Statistical analyses

Transient transfection assays were repeated either two or three times, each with triplicate samples. The data from these independent transfections were analyzed either by pair-wise comparison with a student t test or by one-way analysis of variance followed by Tukey's multiple comparison test (GraphPad Prism version 4; GraphPad Software Inc., San Diego, CA). P < 0.05 was considered significant.

Results

POU5F1 effectively silences CGA promoter activity

JAr cells were transiently cotransfected with the CGA-luc wild-type reporter construct and the ETS2- and POU5F1-expressing constructs, either alone or in combination. ETS2 modestly up-regulated the CGA promoter 3- to 4-fold (P < 0.001) (Fig. 2A). POU5F1, on the other hand, repressed the basal as well as ETS2-mediated transactivation of the CGA promoter by approximately 90% (P < 0.001), confirming previous data (19) (Fig. 2A). To determine the importance of the POU5F1-binding element (−117 ATGGTAAT −110) in controlling expression from the CGA promoter, the site was mutated (−117 ATGGTACG −110, mutPOU5F1). Basal expression from this construct was reduced approximately 90% compared with the wild-type promoter with the intact octamer site (Fig. 2B). Although ectopic ETS2 expression only weakly (1.5- to 2-fold; P < 0.001) up-regulated the activity of the mutPOU5F1(−255luc) promoter, POU5F1 could still effectively repress basal expression from this reporter construct (P < 0.01) despite the fact that POU5F1 was unable to bind to the mutated CGA promoter in vitro (19). Together these results suggest, first, that POU5F1 does not need to bind to its octamer sequence to repress CGA-luc reporter activity and, second, that mutation of the octamer sequence might be interfering with the ability of a second transcription factor to transactivate the CGA promoter. In addition, the experiments imply that POU5F1 silencing of CGA promoter activity involves interference with ETS2-mediated transactivation.

Fig. 2.

POU5F1 silencing of ETS2-mediated CGA promoter activity in JAr cells. A, The −255 CGA promoter-luc reporter construct (−255luc, 750 ng plasmid DNA per milliliter) was cotransfected with ETS2 and POU5F1 plasmid expression vectors (125 ng plasmid DNA per milliliter) either separately or in combination. The Luc activity was normalized relative to β-galactosidase activity from the cotransfected reference reporter, pRSVLTR-βgal (6.5 ng plasmid DNA per milliliter), and the normalized Luc activities are presented as means ± sem from three independent transfections, each run in triplicate (n = 9). If letters above bars are different, there was a significant effect of treatment (P < 0.05). B, Transactivation of the −255luc reporter with its established POU5F1 binding site mutated (mutDLX3/POU5F1) (19) was compared with that of the wild-type reporter. C, Dose dependency of POU5F1-mediated repression of the −255luc. The broad concentration ranges of POU5F1 expression plasmid were cotransfected with fixed concentration of −255luc expression plasmid. Bars marked with different letters represent significant effects of treatment (P < 0.001).

To determine the dose dependency of POU5F1 silencing of CGA promoter transactivation, the ability of ectopic POU5F1 expression to repress the promoter-reporter activity was examined over an extended range of expression plasmid concentrations while keeping the concentrations of −255luc expression plasmid constant (Fig. 2C). The silencing of reporter gene expression was highly sensitive to POU5F1 inhibition, with effects detectable at concentrations as low as 0.5 ng/ml plasmid and half-maximal inhibition noted at 2.5 ng/ml. The latter concentration of plasmid vector was chosen in all subsequent experiments to allow comparisons to be made between different domains of POU5F1 in their relative effectiveness as inhibitors and to study potentially competing effects of other transcription factors.

Domain specificity of POU5F1 required for silencing of CGA promoter activity

POU5F1 protein is comprised of an N-terminal (transactivation) domain [N: 1–126 amino acids (aa)], POU (DNA binding) domain (POU: 127–282 aa), and C-terminal domain (C: 283–352 aa) (Fig. 3A). To determine the role of each domain in silencing CGA promoter activity, expression plasmids lacking various domains, but driven by the same cytomegalovirus promoter, were cotransfected with the ETS2 expression vector and the −255luc reporter. All the POU5F1 (OCT4) vectors were used at 2.5 ng/ml, the concentration required to provide half-maximal inhibition with full-length POU5F1. Consistent with the results shown in Fig. 2A, ectopic expression of ETS2 alone transactivated the promoter approximately 4-fold (Fig. 3A). Whereas the full length POU5F1 construct (at 2.5 ng/ml) again inhibited ETS2-activated reporter expression by about 50%, the POU, the N, or the POU-C domains did not have any inhibitory effect. The constructs observed to cause inhibition were ones that expressed both the N-terminal and POU domains together (N-POU). At 2.5 ng/ml, i.e. 5 ng/dish, the N-POU construct was as an effective a silencer as the one expressing full-length POU5F1, consistent with the hypothesis that both these domains are required for effective silencing of the CGA promoter.

Fig. 3.

Identification of the domains of POU5F1 required for effective silencing of CGA promoter activity. A, The CGA-255luc reporter was transiently transfected either with ETS2 expression vector alone or in combination with various POU5F1 deletion constructs (125 ng plasmid DNA per milliliter) whose structures are represented schematically. Values marked with letter a differed significantly (P < 0.05) from ones with letter b. B, Dose-dependent repression of CGA promoter with the wild type (left panel) and mutated (right panel) the POU5F1-binding element (mutDLX3/POU5F1) sequence by expression of full-length POU5F1 and N-POU deletion constructs. The latter was transfected at four different concentrations, namely 2.5, 50, 250, and 500 ng/ml. Bars are means ± sem (n = 6), and different letters above bars represent significant effect of treatment (P < 0.01). Value of the fourth bar (N-POU 2.5 ng/ml) marked with e* is not different from those letter e at the first (basal) and third (ETS2+POU5F1) (P > 0.05); however, it is different from those at the seventh and eighth bar.

To test this conclusion further, inhibition by the N-POU construct was tested over a series of increasing transfection concentrations. An inhibition curve comparable with that of the full length POU5F1 was observed (Fig. 3B). Expression of N-POU, like POU5F1, silenced ETS2-driven reporter gene expression from the CGA promoter with the mutated octamer site, again stressing that this binding site is not required for POU5F1 inhibition of CGA, even though it is necessary for full ETS2-mediated transactivation.

Regulation of CGA promoter activity by distal-less homeobox 3 (DLX3)

The next objective was to identify the likely candidate transcription factor that interacted with the CGA promoter close to the octamer binding site and that was also implicated in enhancing the ability of ETS2 to drive transactivation of reporter gene expression from the CGA promoter. DLX3 was immediately suspected because it was known to bind within this region and to weakly transactivate CGA (18).

To test whether DLX3 might be the candidate factor, JAr cells were transiently transfected with DLX3 and ETS2 expression vectors (at 125 ng/ml), either alone or in combination. As observed previously, ETS2 up-regulated the promoter 3- to 4-fold, whereas DLX3 alone increased the reporter expression only 2–2.5 fold. However, coexpression of ETS2 and DLX3 synergistically up-regulated the CGA promoter activity 20- to 22-fold (Fig. 4A), suggesting that the transcription factors regulated the promoter cooperatively.

Fig. 4.

Effect of overexpression of DLX3 and ETS2 on CGA promoter activity. A, The −255luc reporter was transiently transfected with ETS2 and DLX3 expression vectors, either alone (125 ng plasmid DNA per milliliter) or in combination. B, Transactivation of the CGA promoter by ETS2 is dependent on DLX3 coexpression. The −255luc reporter was transiently transfected with increasing concentrations of DLX3 expression plasmid with a fixed constant concentration (125 ng plasmid DNA per milliliter) of ETS2 expression plasmid. Bars represent means ± sem (n = 9), and different letters above bars represent significant effect of treatment (P < 0.001).

To determine the dose dependency of DLX3- and ETS2-mediated transactivation of CGA promoter activity, the ability of DLX3 expression to transactivate the promoter-reporter activity was examined over an extended range of expression plasmid concentrations, while keeping, the concentrations of −255luc and ETS2 expression plasmids constant. With increasing concentrations of DLX3 expression vector, up-regulation of CGA promoter activity increased from approximately 39-fold to approximately 110-fold (Fig. 4B).

To determine whether this combinatorial effect of DLX3 and ETS2 is altered upon mutation of either the DLX3 (Fig. 5A) or ETS2 (Fig. 5B) binding sites, separate transfection assays were performed with either the DLX3 binding site or the distal ETS2-binding site mutated. Compared with the wild-type CGA promoter, the synergistic up-regulation was completely abolished with either of the mutated CGA promoters, suggesting that both transcription factors need to bind to their respective binding sequences to transactivate the CGA promoter. Moreover, a reduction in synergism was observed when transient transfection assays were performed with progressively deleted CGA promoter constructs (Fig. 5C). The longest promoter construct, which contained both the intact DLX3 and ETS2 binding sites, was able to up-regulate reporter gene activity approximately 20-fold upon coexpression of DLX3 and ETS2. However, a shorter promoter lacking the DLX3 binding site (−107luc) provided much reduced reporter gene expression. Transient transfection with the shortest promoter (−48luc), lacking both the DLX3 and ETS2 binding sites, gave very low activity and no evidence for synergistic cooperation of the two transactivators. These observations support the previous data and suggest the importance of both an intact DLX3 and ETS2 binding site for full transactivation of the CGA promoter.

Fig. 5.

Effect of mutated the DLX3 and ETS2 binding sites on the ability of DLX3 and ETS2, alone and in combination, to transactivate the CGA promoter. A, The wild-type CGA-255luc reporter or the same construct mutated within its DLX3/POU5F1 binding site (mut DLX3/POU5F1) were cotransfected with the ETS2 and DLX3 expression vectors (125 ng plasmid DNA per milliliter), either alone or together, and Luc activity measured in cell extracts. Value of the sixth bar marked with c* is not different from value of the seventh bar (P > 0.05); however, it is different from the fifth bar. B, It is the same as A, except the ETS2 site on the CGA-255luc reporter had been mutated. C, Effect of truncating the promoter of the CGA-255luc reporter on transactivation by ETS2 and DLX3 either alone or together. −255luc, −107luc, and −48luc refer to lengths of the promoter based on the sequence shown in Fig. 1. D, Effect of increasing transfection concentration of POU5F1 expression plasmid on the ability of ectopically expressed DLX3 to transactivate the CGA-255luc promoter. The DLX3 expression plasmid was transfected under standard conditions (125 ng plasmid DNA per milliliter) either in the absence or presence of increasing concentrations of POU5F1 plasmid. Bars represent means ± sem (n = 6 in A–C; n = 9 in D), and different letters above bars represent significant effect of treatment (P < 0.01).

In addition, the inhibition of DLX3-transactivated CGA promoter by POU5F1 was tested by using varying concentrations of POU5F1 expression plasmid, keeping the concentrations of −255luc and DLX3 expression plasmid unchanged. A moderate 2.5-fold up-regulation of the CGA promoter by DLX3 alone was only partially reversed (∼50%) to the basal level by POU5F1 at the highest concentration of expression plasmid (500 ng/ml) used (Fig. 5D). Note that this concentration is 200-fold higher than that used to down-regulate ETS2-mediated transactivation of the CGA promoter to a similar extent (Fig. 2C) This result suggests that POU5F1 might interfere with DLX3-mediated transactivation by competing for the octamer binding site that is being shared by both transcription factors.

Coimmunoprecipitation of DLX3 and ETS2 from JAr cell extracts

Because we have previously shown that both DLX3 and ETS2 are present in whole-cell lysates of JAr cells and that the addition of anti-ETS2 to the lysates leads to the formation of ETS2/DLX3 immunocomplexes (17), these data are not shown here. Similarly, immunoprecipitation performed with anti-DLX3 and Western blotting with anti-ETS2 yielded complexes that could be immunoprecipitated from the lysates (Fig. 6A, lane 2, anti-ETS2). As expected, the ectopic expression of both ETS2 and DLX3 together markedly increased the amount of ETS2 in the immunocomplexes that formed (Fig. 6A, lane 4). The purified IgG from nonimmunized rabbits was used as a negative control and provided no ETS2 in the immunocomplex (Fig. 6A, lane 3, control IgG). These results demonstrate that DLX3 and ETS2 form a complex stable enough to be immunoprecipitated from JAr cells. Immunofluorescence studies have shown that both proteins colocalize in the JAr cell nucleus (17).

Fig. 6.

Expression of ETS2, DLX3, and POU5F1 in JAr cell lines. A, Coimmunoprecipitation analysis to determine any physical interaction between ETS2 and DLX3. Immunoprecipitation was performed with antibody to DLX3 and Western analysis with affinity-purified ETS2 antibody. Lane 1 (input), Thirty micrograms of control cell lysate; lane 2 (anti-ETS2), ETS2-DLX3 complexes from control cell lysate; lane 3 (control IgG), immunocomplexes collected after the addition of a nonspecific IgG; lane 4 (anti-ETS2), extracts from cells transfected with both ETS2 and DLX3 expression plasmids (3 μg plasmid DNA/dish). B, Detection of ETS2 and DLX3 on Western blots of lysates from the S1 and S4 lines (which stably express POU5F1), from the C1 and C2 lines (which had been transfected with an empty vector at the same time that the S1 and S2 lines had been generated), and a control JAr cell line (JAr). The loading control was ACTB (lower panel). Analyses were performed on separate blots loaded with the same amounts of protein (60 μg) in each lane. C, POU5F1 detection in the S1, S4, C1, and C2 cell lines, with ACTB as a loading control in the lower panel. Molecular weight markers (M_r × 10⁻³) are shown on the left.

Expression of ETS2, DLX3, and POU5F1 in JAr S and C cell lines

Whole-cell lysates were prepared from the nontransfected JAr cells, two cell lines stably expressing POU5F1 (S1 and S4), and two lines that had been mock transfected (C1 and C2) (20). As before, the cell lysates were subjected to Western blotting analysis. When compared with the band of ACTB used as a loading control, it was clear that all of the cell lines expressed similar amounts of ETS2 and DLX3 (Fig. 6B), whereas POU5F1 was detectable only in the S1 and S4 lines but not in the C1 and C2 lines (Fig. 6C). The control JAr cells used in the transfection experiments described earlier were essentially indistinguishable from the C1 and C2 lines (Fig. 6B). This experiment suggested that POU5F1 expression does not alter the expression of DLX3 and ETS2 in JAr cells, only their downstream target genes. This lack of effect on DLX3 and ETS2 transcript levels has been confirmed in microarray analyses of RNA present in the JAr cell lines stably expressing POU5F1 (data not shown).

POU5F1 silences CGA promoter activity by interfering with ETS2 binding

ChIP analyses were performed with chromatin obtained from the S4 and C2 cells to determine whether ETS2 binds to the CGA promoter when POU5F1 is expressed. As anticipated from the earlier mutational analysis, ETS2 failed to form a complex with the CGA promoter in presence of POU5F1 (Fig. 7A, S4, lane 3, anti-ETS2). However, a strong ETS2-DNA complex was observed in the absence of POU5F1 (Fig. 7A, C2, lane 3).

Fig. 7.

ChIP analysis to demonstrate the association of ETS2 with the CGA promoter in presence and absence of POU5F1 expression. A, Sheared chromatin prepared from the S4 line (expressing POU5F1; upper panel) and the C2, nonexpressing line (lower panel) were exposed to overnight incubation with no antibody (lane 1), anti-POU5F1 (lane 2), anti-ETS2 (lane 3), and nonspecific IgG (lane 4). Lane 5 is 10% of total input chromatin. The primers used for PCR in A and B were designed to amplify a region of the CGA proximal promoter (−96 to +48) containing the ETS2 binding site. B, Sheared chromatin prepared from a control JAr cell line after it had been transiently transfected with increasing concentrations (shown on right, 0, 2.5, 10, 50, and 200 ng/ml) of a full-length POU5F1 expression plasmid. Lane 1, No antibody; lane 2, anti-ETS2 antibody; lane 3, nonspecific IgG; lane 4, 10% of total input chromatin.

In addition, ChIP analyses were performed on chromatin from control JAr cells that had been transiently transfected with a range of concentrations of POU5F1 expression plasmids. ETS2 formed a stable complex with the CGA promoter in the absence of POU5F1 (Fig. 7B, lane 2, top panel). The amount of ETS2 associated with the promoter was much reduced after the cells had been transfected with the lowest concentration (2.5 ng/ml) of POU5F1 plasmid (Fig. 7B, lane 2, panel 2) and was lost completely as the plasmid concentration was increased. It should be noted that the primers used in Fig. 7A did not encompass the DLX3/POU5F1 site (Fig. 1) located further upstream, so this experiment was unable to confirm whether POU5F1 was associated with the POU5F1-binding site on the CGA promoter earlier characterized by Liu et al. (19) in EMSA. However, Fig. 7A suggests a lack of POU5F1 association with chromatin close to the ETS2 binding site in the S4 cells. The two experiments described in Fig. 7, A and B, are consistent with the conclusion that ETS2 binding to the CGA promoter is eliminated when POU5F1 is overexpressed and that the mechanism of silencing is via squelching.

POU5F1 interferes with binding of DLX3 to the CGA promoter

To determine whether DLX3 can bind to the CGA promoter in the presence of POU5F1, ChIP analysis was performed with chromatin from S4 and C2 cells. In the absence of POU5F1 expression (C2), a strong DLX3-DNA complex was observed (Fig. 8A, lane 2, anti-DLX3). On the other hand, the amount of DLX3 complexed with DNA was reduced but not eliminated in chromatin isolated from S4 cells that stably expressed POU5F1 (Fig. 8B, lane 2). This experiment was therefore unable to confirm whether there was a competition between POU5F1 and DLX3 for the common DLX3/POU5F1 binding site in the CGA promoter (Fig. 1).

Fig. 8.

ChIP analysis to demonstrate the association of DLX3 with the CGA promoter in presence and absence of POU5F1 expression. Sheared chromatin prepared from C2 cells (A) and S4 cells (B) were separately exposed to overnight incubation with no antibody (lane 1) and anti-DLX3 (lane 2). Lane 3 is 10% of total input chromatin. DLX3 formed a strong complex with the DNA in absence of POU5F1 (A; C2). However, in the presence of endogenously expressed POU5F1 (B; S4), formation of a DLX3-DNA complex was significantly reduced but not completely abolished. The primers used for PCR in A and B were designed to amplify a region of the CGA proximal promoter (−189 to −25) containing both the DLX3 and ETS2 binding sites.

Association of ETS2 and DLX3 with the CGA promoter in hESC before and after differentiation to trophoblast

hESC can be differentiated toward the trophoblast lineage and CG production, with an accompanying down-regulation of POU5F1 by exposure to BMP-4 (27, 28). This process is enhanced if the growth factor FGF-2 is omitted from the medium (29) and if the signaling pathways mediated through MAPK3/1 (ERK1/2) (30, 31) and SMAD-2/3 (32) are inhibited. Accordingly, the hESC/BMP4 model provides a means to study the up-regulation of CGA as POU5F1 expression declines (33).

hESC (H1 line) were grown in feeder-free culture in the absence of FGF2 with 10 ng/ml BMP4, 0.1 μm PD173074, and 1 μm A83–01, an inhibitor of activin A signaling superior in both specificity and potency to SB431542 (35). The culture medium (hESC/BMP4+PD+A83) was replaced daily and CG production measured by ELISA. No CG was detectable until d 5, at which time its concentration averaged 164.5 ±17.4 mIU/ml·d (n = 3). Production increased markedly over the subsequent 3 d, reaching 23,168 ±1,534 mIU/ml (n = 3) at d 8. Control H1 cells cultured in the presence of 4 ng/ml FGF-2 and absence of BMP4, PD173074, and A83–01 maintained an undifferentiated phenotype and failed to produce detectable CG. Cell lysates prepared from both types of culture at d 8 were examined for the presence of POU5F1, ETS2, and DLX3 by Western blotting (Fig. 9A). POU5F1 was strongly expressed in the control cultures maintained on hESC/FGF2 medium but was completely down-regulated in cultures on hESC/BMP4+PD+A83 that had been directed to trophoblast. ETS2 was expressed in both control and differentiated cultures, with a concentration elevated in the latter. In direct contrast to POU5F1, DLX3 was detected only in the cells maintained on hESC/BMP4+PD+A83, i.e. at a time POU5F1 was not expressed.

The interaction between POU5F1 and ETS2 was then examined in the control (hESC/FGF2) cultures in which both proteins were expressed. Immunoprecipitation was performed with anti-ETS2 and Western blotting with anti-POU5F1 (Fig. 9B). Anti-ETS2 was clearly able to immunoprecipitate POU5F1 from the lysates (Fig. 9B, lane 3, anti-ETS2), whereas control rabbit IgG, used as a negative control, provided no POU5F1 in the immunocomplex (Fig. 9B, lane 2). These results demonstrate that POU5F1 and ETS2 form a complex stable enough to be immunoprecipitated from undifferentiated hESC cells.

ChIP analyses were then performed with chromatin obtained from the cells cultured on hESC/FGF2 and hESC/BMP4+PD+A83 for 8 d (Fig. 9G) to determine whether POU5F1, ETS2, and DLX3 could be detected bound to the CGA promoter (Fig. 9, C–F). The ability to detect histone-H3 bound to the hESC chromatin served as the positive control (Fig. 9C). Interestingly, in control hESC, in which POU5F1 is abundant, ChIP assays failed to detect a POU5F1 chromatin complex within the region containing the POU5F1 DNA-binding sequence (Fig. 9D, lane 3, anti-POU5F1), despite the fact that naked DNA representing this region does bind the protein in an apparently specific manner, as assessed by EMSA (19). ETS2 was not associated with the CGA promoter in undifferentiated hESC (hESC/FGF2) (Fig. 9E, left panel, lane 3, anti-ETS2), but a strong ETS2-DNA complex was observed in the cells that had differentiated to trophoblast (Fig. 9E, right panel, lane 3, anti-ETS2). DLX3 also failed to form a complex with the CGA promoter in undifferentiated hESC (Fig. 9F, left panel, lane 3, anti-DLX3), a not unexpected outcome, because Western blots had not detected the presence of any DLX3 protein in the cells under these culture conditions (Fig. 9A). However, a DLX3-DNA complex was observed in the differentiated cells (Fig. 9F, right panel, lane 3, anti-DLX3) in which POU5F1 was not expressed.

Discussion

JAr cells, although derived from a choriocarcinoma, are a particularly useful cell model for studying transcriptional controls operating on CG genes because basal hCG production has not reached its full potential (36). They lack human leukocyte antigen-G (37) and other differentiated markers of extravillous trophoblast (38–41) and do not spontaneously produce multinucleated cells analogous to syncytium (36, 40). Production of hCG by JAr cells is low, with the majority immunohistochemically negative for the hormone, and hCGA output exceeds hCGB by approximately 2:1, a value resembling that of the highly mitotic, placental cytotrophoblast (42). Importantly, when driven to differentiate with reagents such as 8-bromoadenosine, hCG, synthesis not only increases overall, but the ratio of α− to β-product is boosted (42, 43). Accordingly, JAr cells are thought to have arisen from the transformation of cells early in the lineage to mature trophoblast (43), most likely from multipotent progenitors, so that the full complement of factors needed to drive maximal transcription of both CGA and CGB is likely not present. Hence, the cells, although permissive for hCG production, are ideal for examining the progression from relatively undifferentiated to more mature trophoblast as well as the effects of relevant, ectopically expressed transcription factors that are not expressed endogenously in amounts optimal for maximal expression of the CG subunit genes.

The transcription factor ETS2 has long been known to transactivate several signature genes of trophoblast, including CGA, CGB, and IFNT (16, 21, 22, 24, 44), and may even have a role in controlling expression of mouse Cdx2, a gene crucial to the emergence of the trophoblast lineage (45). Ets2 is required for placental development (46, 47) and is expressed in mouse trophoblast stem cells, in which its silencing slows growth and causes terminal differentiation (48). The experiments performed here confirm that ETS2 transactivates the CGA promoter and that this action can be blocked by POU5F1 (Fig. 2, A and B). Only very low concentrations of POU5F1 are needed for this silencing activity (Fig. 2C), and this extreme sensitivity is dependent on the presence of the N-POU domains of the protein, with no requirement for the carboxyl terminal domain (Fig. 3A). This result is consistent with previous studies on IFNT gene regulation in which the same domains of POU5F1 were necessary for silencing (21). As demonstrated by ChIP analysis (Fig. 7, A and B), POU5F1 overexpression interferes with the ability of ETS2 to bind to the CGA promoter, thereby providing the likely explanation for the POU5F1 silencing effect. The ChIP analysis also confirms the great sensitivity of ETS2 binding to low concentrations of POU5F1. Because POU5F1 would appear to sequester free ETS2 rather than ETS2 already bound to the promoter, the mechanism of silencing is by squelching rather than quenching (49). Although we have not investigated the interaction between the two transcription factors in further detail, previous experiments with the in vitro-produced proteins have shown that the POU domain of POU5F1 forms a stable complex at a site located between the POINTED and DNA binding domain and is considered to be the main activation domain of ETS2 (21).

The POU domain transcription factor POU5F1 is considered as the master regulator of stem cell self-renewal and pluripotency and is present in the inner cell mass of mammalian embryos, embryonic germ cells, and cultured embryonic stem cells (50). Maintenance of POU5F1 at a critical concentration is probably required for self-renewal of pluripotent cells. In mouse embryos, overexpression of POU5F1 causes differentiation toward the endoderm lineage, whereas its suppression triggers differentiation toward trophectoderm (51, 52). Similarly, murine and human endometrial stromal cells (ESC) lack a pluripotent capability after knockdown of POU5F1 and instead differentiate to trophectoderm-like cells (53–55). However, in early embryos of all mammals that have been examined, including mouse, human, pig, and cow, POU5F1 is not confined to the inner cell mass but continues to be expressed temporarily in trophectoderm, in which it can presumably continue to influence gene expression (50, 56, 57). Its expression in trophectoderm diminishes as development proceeds and is usually low by the time that the hatching from the zona pellucida is completed and, in the case of the mouse and human, when invasion into the endometrium is initiated (50, 58). This period of development also coincides with the onset of expression of various trophoblast-specific genes, including CGA and CGB in the human and IFNT in ruminant species, both of which are dependent on ETS2 for expression and silenced by POU5F1 (19–21). Hence, it might be expected that CG subunit gene transcription would rise as POU5F1 expression falls in emerging trophoblast and that the expression of the two genes would be extremely low in the JAr cell lines that stably express POU5F1.

According to this squelching model, it seems possible that POU5F1 could have a broad role in preventing the expression of additional genes under the transcriptional control of ETS2 and whose early up-regulation in the inner cell mass might interfere with the capacity of these cells to remain pluripotent. Such a model may parallel the repressive role played by POU5F1 in pluripotent cells in suppressing the expression of genes that are essential for emergence of the three main germ layers and more differentiated lineages (59, 60). Genome-wide studies on human ESC, for example, have identified promoter sequences for more than 1000 genes that bind POU5F1 and demonstrated that many of these regions of DNA also associate with SOX2 and NANOG (61). About 50% of the genes singled out by the three regulatory factors were transcriptionally inactive in embryonic stem cells and were located in transcriptional silent but relatively open regions of chromatin enriched in polycomb repressive complex 2 and H3K27 trimethylated histone (62). Many of these repressed genes were transcription factors, silent in pluripotent cells but poised for expression when the ESC were induced to differentiate. Again, the implication is that sets of transcription factors and their downstream targets would be expressed as the concentration of POU5F1 falls. ETS2, unlike DLX3, however, is expressed in ESC and not among the genes associated with these temporarily quiescent regions of chromatin (Ref. 62 and its Supplemental Tables 7 and 8). Instead, its proximal promoter remains bound to RNA polymerase II (62), and its gene is actively transcribed in hESC (GEO GSE10469) (34). Conceivably, genes with pleiotropic action, such as ETS2, which play a dual role both in maintaining the pluripotent state and in guiding later lineage decisions, exist as complexes with POU5F1 in pluripotent cells and, in that form, have an altered potential to influence transcriptional events.

Another question addressed in this paper was why mutation of the octamer binding site depressed basal as well as ETS2-mediated transactivation of the CGA gene promoter in JAr cells, even though POU5F1 silencing was unaffected. The likely explanation appears to be that this site (Fig. 1) overlaps the one that binds DLX3, a homeobox family transcription factor required for normal placental development in the mouse (18, 23). It has been demonstrated earlier by others that transfected DLX3 up-regulated CGA expression in JAr cells by binding to the promoter at a sequence (−114 to −107) that partially overlaps an octamer-binding site in the so-called junctional regulatory region (18). Nonetheless, DLX3 was observed to be only a weak transactivator, increasing expression of a Luc reporter less than 3-fold. A similar modest (2.5-fold) transactivation of the CGA was noted here (Fig. 5, A–C) when DLX3 was overexpressed. However, the positive effects of DLX3 on CGA promoter activity were greatly increased when it was coexpressed with ETS2, and, because the DLX3 protein concentration was titrated upward by raising the concentration of transfected DLX3, activation of reporter expression could be increased well over 100-fold (Fig. 4B). Full cooperative transactivation therefore depends on the relative concentrations of ETS2 and DLX3, whose binding sites are separated by 26 nucleotides, i.e. roughly 2.6 helical turns, to associate with each other while bound to the promoter and forming a productive complex with the transcriptional machinery. Further experiments are needed to determine how cAMP activation of CRE-binding protein plays into this association. Increased concentrations of relevant transcription factors, such as DLX3, and the continued presence of ETS2, as well as the down-regulation of POU5F1, may explain why CGA transcription rises so rapidly early in pregnancy. Interestingly, POU5F1, although apparently sharing a binding site with DLX3 on the CGA promoter, is not very effective in reversing the ability of DLX3 to act as a transactivator (Fig. 5D). Although the ability of POU5F1 to interact directly with DLX3 has not been examined, there is little evidence to suggest that ectopically expressed POU5F1 has an ability to squelch DLX3. In addition, ChIP assays indicated that POU5F1 did not even associate with the putative DLX3/POU5F1 binding site in ESC (Fig. 9D, lane 3, anti-POU5F1). Accordingly, we believe that it is unlikely that POU5F1 competes with DLX3 for binding to DNA, even if the two are ever coexpressed. Overall, the data support the hypothesis that the silencing effect of POU5F1 on the CGA promoter in JAr cells is largely, if not exclusively, through its association with ETS2 rather than DLX3.

In a final series of experiments (Fig. 9), we were able to provide a partial test of our hypothesis that in pluripotent cells POU5F1, through its ability to squelch the activity of ETS2 and possibly other key transcription factors, can counteract the tendency of such cells to differentiate to trophoblast. In fully pluripotent ES cells, in which both ETS2 and POU5F1 are expressed, neither transcription factor associates with the CGA promoter, but both can be detected in a soluble complex. As POU5F1 itself becomes silenced in response to BMP4 treatment of the cells, the constraints operating over the ability of ETS2 to bind to the CGA promoter appear to be lifted. That, plus the expression, over time, of additional transcription factors, such as DLX3, then permits a major up-regulation of the CG subunit genes and possibly also the emergence of the trophoblast lineage. A role for POU5F1 in suppressing differentiation through transcriptional squelching has not previously received serious consideration.