Plasmids and cell culture.

RNAi constructs pSuper-Esrrb-sh1 and pSuper-Esrrb-sh2 were constructed by cloning Esrrb RNAi1 and RNAi2 (16) into pSuper-puro (Oligoengine). pSuper-control contains an oligonucleotide without complementarity to any known mammalian sequence (Dharmacon). Mouse ES cell lines 46C (35) and ZHBTc4 (23) and derivatives of ZHBTc4 were grown on gelatin-coated dishes without feeders on Glasgow minimal essential medium supplemented with leukemia inhibitory factor, 15% fetal bovine serum, 0.25% sodium bicarbonate, 1 mM glutamine, 1 mM sodium pyruvate, nonessential amino acids, 50 μM beta-mercaptoethanol, and penicillin-streptomycin.

Cells from the c6 cell line, called F-Oct4 ES cells from here onwards, were created by electroporating ZHBTc4 cells with linearized pPyCAG (FLAG)₃ Oct4IP, a plasmid in which the Oct4 open reading frame was placed between the N-terminal triple FLAG tag and the internal ribosome entry site (IRES)-puromycin resistance cassette of pPyCAG (FLAG)₃IP (20). Electroporated cells were plated in ES cell medium, and after 24 h, 1 μg/ml puromycin and 1 μg/ml doxycycline were added. After 12 days of selection, puromycin-resistant colonies were picked and tested for FLAG-Oct4 expression by anti-FLAG Western blot analysis. TNG cells have been described previously (6) Puromycin-sensitive TNG-PS cells were derived from TNG cells by excision of the frt-IRES-pac-frt cassette by transient expression of FLPe. Bright field pictures of ES cell cultures were taken using the IX70 inverted microscope (Olympus).

Oct4 purification and mass spectrometry.

F-Oct4 ES cells and control cells (ZHBTc4) were expanded to 50 14-cm dishes, plates were washed once with phosphate-buffered saline, cells were scraped off, and nuclear extracts were prepared (8) and dialyzed to 100 mM KCl (8). A total of 100 μl of anti-FLAG M2 agarose beads (Sigma), equilibrated in buffer C-100 (20 mM HEPES, pH 7.6, 10% glycerol, 100 mM KCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 1× complete EDTA-free protease inhibitor; Roche), was added to 10 ml of nuclear extract, incubated for 3 h at 4°C, transferred into an Eppendorf tube, and washed five times with 1 ml of C-100 buffer plus 0.02% NP-40 (C-100*) and eluted four times with C-100* containing 0.2 mg/ml FLAG tripeptide (Sigma) for 15 min at 4°C. Fractions were loaded onto a 10% sodium dodecyl sulfate (SDS)-phosphonoacetic acid (PAA) gel and silver stained. Elutions 1 and 2, containing the majority of FLAG-Oct4 in purification from the F-Oct4 extract, were concentrated by SpeedVac condensation, loaded onto a 10% SDS-PAA gel, and stained with colloidal Coomassie blue. Gel lanes were cut and subjected to in-gel digestion with trypsin (Promega), essentially as described previously (31). Nanoflow liquid chromatography-tandem mass spectrometry was performed on a 1100 series capillary liquid chromatography system (Agilent Technologies) coupled to an LTQ-Orbitrap mass spectrometer (Thermo), as described previously (27). Database searches to assign proteins to the found peptide fragmentation spectra were performed using Mascot, as described previously (27).

IP.

For immunoprecipitation (IP), 2.5 μg of Oct4 antibody (N19; Santa Cruz Biotechnology) or Esrrb antibody (R&D Systems) was added to 200 μl of 46C ES cell nuclear extract and incubated under rotation for 2 h at 4°C. A total of 1 U Benzonase (Novagen) or 25 μg/ml ethidium bromide was added where indicated. The antibody-extract mixture was added to 20 μl of protein G-Sepharose beads (Amersham) blocked with 1% fish skin gelatin (Sigma) and 0.2 mg/ml chicken egg albumin (Sigma) and rotated for another 90 min. Beads were washed four times with 100 μl of C-100* buffer and boiled in SDS loading dye.

RNAi and assays for Nanog expression.

46C ES cells were transfected with pSuper-Esrrb-sh1, pSuper-Esrrb-sh2, or pSuper-control using Lipofectamine 2000 transfection reagent (Invitrogen). To measure the effect of Esrrb RNAi on the endogenous mRNA and protein levels of Nanog, Oct4, and Esrrb, transfected cells were selected with 1 μg/ml puromycin for 48 h, starting at 24 h posttransfection. RNA was isolated from these samples using Trizol (Invitrogen), and mRNA levels were measured by performing real-time quantitative PCR (qPCR) on an Opticon real-time PCR machine. Protein levels were measured by Western blot analyses using antibodies against Nanog (6), Oct4, Esrrb, and lamin B1 (Santa Cruz Biotechnology).

TNG-PS cells were transfected with pSuper-Esrrb-sh1, pSuper-Esrrb-sh2, or pSuper-control, and transfected cells were selected with puromycin for 48 h, starting at 24 h posttransfection. Green fluorescent protein (GFP) fluorescence of the TNG cells was measured by using a FACSCalibur flow cytometer (Becton Dickinson), as described previously (6).

For measuring the effect of Esrrb RNAi on expression from the Nanog promoter, pNanog-Luc containing a Nanog promoter fragment from −2.5 kb to +50 bp (11) and pRenilla-TK (Promega) were cotransfected with the pSuper constructs. Renilla luciferase assays were done 48 h posttransfection using the dual luciferase reporter system (Promega). The pNanog-Luc mutated Oct binding site (mOS) and its control were described previously (26). pNanog-Luc constructs with a mutant estrogen-related receptor response element (ERRE) contained the mutation GGT to AAC in the ERRE sequence TCTGGGTCA in the Nanog proximal promoter from −230 to +106, compared to the Nanog transcriptional start, and were tested 24 h posttransfection.

ChIP.

ChIP using formaldehyde cross-linking and Oct4 antibodies (sc-8628; Santa Cruz Biotechnology) or Esrrb antibodies (R&D Systems) was done on 46C ES cells, as described previously (2). Dual cross-linking ChIP using formaldehyde and di(N-succinimidyl) glutarate (DSG) was performed on 46C and ZBHTc4 ES cells, as described previously (24). qPCR analysis was performed using DNA Engine Opticon 2. Relative enrichments were calculated by comparing the ChIP efficiency of the region of interest to that of an unrelated region (Amylase). Primers used to amplify the Nanog genomic region are as follows: 5′ (−550 to −462), CACAGGCTCTTTCTTCAGACTTG and TCTTGCTTGCTCTTCACATTGG; Oct-Sox (−215 to −60), TCCCTCCCTCCCAGTCTG and CCTCCTACCCTACCCACCC; and 3′ (+929 to +988), GGTAGAACCAAGAGGCTGCT and CATCACAACACGCACCTGA. Primers used for Zfp42 are as follows: −283 to −117, TGCATCCTCTGCTTGTGTAA and CAGAGCTGTCCCCTTGTCT; Rest (−3216 to −3071), CTCCCCTGGACAATAGCTTC and CGTCCTTCATTTCCTCAGTG; Dppa3 (−1770 to −1550), GATCCAGCTGGTCTGAGCTA and GTGCAGGGATCATAGGAGTG; and Lefty1 (−1264 to −1060), AAGCTGCAGACTTCATTCCA and CGGGGGATAGATGAAGAAAC (21). Primers used for Amylase are as follows: CTCCTTGTACGGGTTGGT and AATGATGTGCACAGCTGAA.

EMSAs.

293T cells were transfected with pPyCAGIP derivatives (20) expressing the cDNAs of FLAG-Esrrb, Oct4, or Sox2. After 24 h, cells were washed and harvested in phosphate-buffered saline, resuspended in lysis buffer (50 mM HEPES, pH 7.9, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40, 1× complete EDTA-free protease inhibitor; Roche) and rotated at 4°C for 20 min. After being microcentrifuged at 13,000 rpm for 10 min, supernatants were divided into aliquots and stored at −80°C. The electrophoretic mobility shift assay (EMSA) was carried out as described previously (7). The antibodies for the supershift were added after the initial incubation for a further 10 min as follows: 2 μg of anti-Oct4 (sc-9081), 2 μg of anti-Sox2 (sc-17320), or 1 μg of FLAG M2 antibody (Sigma).

Immunofluorescence.

46C ES cells were grown on coverslips coated with 0.1% gelatin and stained with antibodies using a standard protocol. In short, cells were fixed in 4% paraformaldehyde and incubated with antibodies against Nanog (6), Oct4 (N19; Santa Cruz Biotechnology), and Esrrb (R&D Systems). Secondary antibodies are from Dako (fluorescein isothiocyanate swine anti-rabbit antibody), Molecular Probes (Alexa Fluor 594 goat anti-mouse antibody) and Jackson ImmunoResearch Laboratories (fluorescein isothiocyanate rabbit anti-goat antibody). Images were taken with an Axio Imager (Zeiss).

The Oct4 protein interacts with Esrrb.

To identify the interaction partners of the Oct4 protein in ES cells, we constructed an ES cell line where, under self-renewing conditions, all Oct4 in the cell has an N-terminal triple FLAG tag (FLAG-Oct4). The parental ZHBTc4 ES cell line (23) has both Oct4 alleles disrupted, and the only Oct4 protein in the cell is transcribed from a doxycycline-suppressible transgene (Fig. 1A). ZHBTc4 cells were transfected with a construct in which the constitutive expression of FLAG-Oct4 is linked through an IRES to puromycin resistance. Simultaneously, doxycycline was added to the medium to repress the inducible Oct4 transgene expression (23). After 12 days of growth, colonies were picked and expanded into cell lines. All cell lines expressed a protein of the same relative molecular weight that reacted with an anti-FLAG antibody on Western blots (data not shown). As the Oct4 level must be tightly regulated to allow continued self-renewal (23), the survival of puromycin-resistant colonies indicates that the FLAG-Oct4 protein is functional. Two of these lines were further tested for their response to doxycycline treatment, and both underwent efficient differentiation at clonal density (Fig. 1B and C). The c6 cell line (F-Oct4 ES cells) was taken forward for biochemical analysis.

F-Oct4 ES cells were expanded, nuclear extracts were prepared, and the FLAG-Oct4 protein was purified using FLAG-affinity technology (see Materials and Methods). Silver stain analysis of a polyacrylamide gel containing the proteins purified from F-Oct4 extracts identified a major band and a minor band running just above the 54,000-molecular-weight marker (Fig. 2A). Both bands are recognized by a FLAG antibody and are not present in purifications from control extracts, suggesting that they represent the FLAG-Oct4 protein (Fig. 2B). No other major bands were observed, indicating that Oct4 does not purify as part of a major stoichiometric complex, despite the mild purification conditions used. Mass spectrometry analysis of two independent purifications identified the presence of Oct4 (10 unique peptides) and Esrrb (5 unique peptides) in F-Oct4 samples but not in control samples. Indeed, Esrrb could be detected by Western blot analysis in the F-Oct4 sample but not in the control (Fig. 2C). The interaction between Esrrb and Oct4 was independently verified by co-IP from extracts of a different ES cell line, 46C, using antibodies against endogenous Oct4 and Esrrb (Fig. 2D and E). Treatment of the extract with Benzonase nuclease or ethidium bromide did not affect the interaction (Fig. 2D and E), indicating that it is not indirectly mediated through DNA.

Moreover, the ability of bacterially expressed GST-Oct4 to pull down FLAG-Esrrb from transfected ES cells (Fig. 2F) indicates that posttranslational modification of Oct4 is not required for interaction between Oct4 and Esrrb.

Esrrb regulates expression of the Oct4 target gene Nanog.

Oct4 regulates expression of a cohort of target genes in ES cells, often acting in concert with Sox proteins (16). To determine whether the binding of Esrrb to Oct4 affected the regulation of gene expression by Oct4, we first examined expression of the Oct4 target gene Nanog. We depleted Esrrb by RNAi using two vectors that express different, previously reported Esrrb short hairpin RNAs (shRNAs) (16) and harbor a puromycin selection marker. After 2 days of puromycin selection, the levels of Nanog mRNA (Fig. 3A) and Nanog protein (Fig. 3B) were reduced in the 46C ES cells treated with either Esrrb shRNA vector compared to the levels of the control. Importantly, this specific depletion of Nanog occurred prior to any reduction in Oct4 expression (Fig. 3A and B) and prior to any morphological evidence of ES cell differentiation (Fig. 3C). This indicates that Esrrb shRNA-induced Nanog depletion is not a consequence of differentiation but occurs prior to differentiation. We also tested the effect of Esrrb depletion on TNG-PS ES cells which have the GFP open reading frame inserted at the start codon of one of the endogenous Nanog alleles (6). Transfection of either Esrrb shRNA vector decreased the mean GFP fluorescence of the population (Fig. 3D) after 2 days of puromycin selection, suggesting that depletion of Esrrb reduces transcription from the Nanog locus. To determine whether Esrrb affects the activity of the Nanog promoter, similar shRNA experiments were performed using a luciferase reporter under the control of a Nanog promoter fragment extending from −2.5 kb to +50 bp compared to the transcription start site. Esrrb shRNA vectors were cotransfected with the luciferase reporters, and luciferase activity was measured 2 days posttransfection. Figure 3E shows that Nanog promoter activity is strongly reduced with either Esrrb shRNA vector compared to activity with a control shRNA vector.

Esrrb regulates Nanog expression using contacts with both Oct4 and a degenerate ERRE.

Oct4 contributes to the regulation of Nanog by binding to an Oct-Sox site (14, 26), located 166 to 180 bp upstream of the mapped transcription start site of the Nanog gene (4, 32). To investigate the relationship between the regulation of Nanog by Esrrb and that by Oct4, we performed ChIP experiments with antibodies for Oct4 and Esrrb in 46C ES cells and examined the precipitates for the presence of the Nanog promoter. A standard ChIP protocol using only formaldehyde as a cross-linking agent confirms that Oct4 binds in the vicinity of the Oct-Sox site (Fig. 4B). Using standard ChIP, we found no enrichment of Esrrb at the Nanog promoter (Fig. 4B), although the Esrrb protein is immunoprecipitated during the ChIP procedure (data not shown). Conventional formaldehyde-based ChIP methods efficiently detect protein-DNA interactions but may not detect the binding of Esrrb to the Nanog promoter if it is stabilized by protein-protein interactions. We therefore used a dual cross-linking ChIP method (XX-ChIP) that uses DSG prior to formaldehyde cross-linking (24). DSG has a longer spacer arm than formaldehyde and has been used to cross-link transcription factor protein-protein interactions on DNA (15). XX-ChIP indeed detects Esrrb at the Oct-Sox site within the proximal Nanog promoter (Fig. 4C). XX-ChIP using the Oct4 antibody also gives a specific enrichment of Oct4 at the Oct-Sox motif (Fig. 4C). Oct4 and Esrrb were also specifically enriched on the Oct-Sox site compared to the input and an immunoglobulin G control (data not shown). We conclude that Esrrb binds to the Nanog promoter in the vicinity of the Oct-Sox motif.

To assess whether Esrrb binding to the Nanog promoter is dependent on Oct4, we made use of the ES cell line ZHBTc4 in which the endogenous Oct4 alleles are disrupted and in which Oct4 is expressed from a doxycycline-suppressible promoter (23). The addition of doxycycline for 12 h removes all Oct4 protein from the cell (Fig. 4D). At this time point, the level of Esrrb is unaffected (Fig. 4D). Esrrb XX-ChIP shows that Esrrb is no longer enriched at the Nanog promoter in the absence of Oct4 (Fig. 4E). To functionally test whether the maintenance of Nanog promoter activity by Esrrb is via Oct4, we tested Nanog promoter-luciferase constructs where the Oct4 binding site is mutated (Nanog mOS) (26). As expected, the Nanog mOS reporter is less active than the wild-type (wt) construct, although still clearly above the background (26) (Fig. 4F). Depletion of Esrrb does not further reduce the activity of Nanog mOS (Fig. 4F), suggesting that the effect of Esrrb on Nanog expression requires Oct4 binding to the Oct-Sox site in the Nanog promoter.

Estrogen-related receptors are thought to act via an ERRE. A consensus ERRE of tcaaGGttca (invariant positions in uppercase) was determined by SELEX and confirmed by in vivo studies (9, 29). Visual inspection of the Nanog promoter identified a degenerate ERRE (sequence TCTGGGTCA) 12 bp upstream of the Oct-Sox motif (Fig. 5A). This sequence is largely conserved in many mammalian species (26). To test the contribution of this putative ERRE to Nanog promoter activity, we mutated the core GGT in this motif into AAC (Fig. 5A) in the context of a Nanog promoter-luciferase construct. Nuclear magnetic resonance structural analysis of the Esrrb DNA binding domain in complex with DNA shows that these bases make a major contribution to DNA binding by Esrrb (10). This mutation strongly reduced Nanog promoter activity (Fig. 5B). Moreover, in contrast to the situation with the unmutated construct, the activity of this mutant could not be further reduced by Esrrb shRNA expression (Fig. 5B).

EMSA was employed to investigate the potential binding of Esrrb to the Nanog promoter. A 57-mer oligonucleotide corresponding to the sequence of the Nanog promoter that includes the putative ERRE and the Oct-Sox site was used. Lysate prepared from 293T cells that were cotransfected with FLAG-Esrrb, Oct4, and Sox2 expression plasmids caused a shift in migration of the probe into two complexes (Fig. 5C, left). The faster migrating complex could be supershifted by antibodies against Oct4 and Sox2 but not by an anti-FLAG antibody. In contrast, the slower migrating band was supershifted by all three antibodies, suggesting that the slower migrating complex is formed by binding of Esrrb, Sox2, and Oct4 (Fig. 5C, left, lanes 2 to 4). Using a second 57-mer oligonucleotide that has the GGT-to-AAC mutation in the putative ERRE, only the faster of the two complexes was observed, and this could be shifted by antibodies against Oct4 and Sox2 but not by anti-FLAG antibody (Fig. 5C, right). Therefore, we conclude that there is an ERRE upstream of the Oct-Sox site in the Nanog promoter that is essential for Esrrb binding and optimal Nanog promoter activity.

To determine whether binding of Esrrb to the Nanog promoter in vitro is dependent upon binding of Oct4-Sox2, EMSAs were performed by mixing cell lysates prepared from individual transfections of Oct4, Sox2, and Esrrb into 293T cells. The addition of the Esrrb lysate caused a weak probe shift (Fig. 5D, lanes 2 and 3), showing that Esrrb alone can bind the Nanog promoter probe. However, in the presence of Oct4 and Sox2, DNA complex formation by Esrrb was enhanced compared to DNA complex formation by Esrrb alone (Fig. 5D, compare lanes 2 and 3 to lanes 6 and 7). This effect was greatest with both Oct4 and Sox2 present, suggesting that an Oct4-Sox2 complex is required for this cooperative effect. Using different combinations of Esrrb, Oct4, and Sox2 lysates leads to the same conclusions (Fig. 5D, bottom). These data, together with the Esrrb ChIP experiments (Fig. 4C and E), suggest a model (Fig. 5E) in which the presence of the Oct4-Sox2 complex bound to the Oct-Sox site in the Nanog promoter strongly enhances the intrinsic capacity of Esrrb to bind to an ERRE located upstream of the Oct-Sox site.

Esrrb binds and regulates other Oct4 target genes.

To determine the generality of the association of Esrrb with Oct4 on Oct4 target genes, we investigate a number of Oct4 target genes with a characterized Oct-Sox DNA element (Fig. 6A), using the ES cell line ZHBTc4. These genes all had a putative ERRE at different distances from the Oct-Sox site and in different orientations (Fig. 6A). XX-ChIP analysis showed that Esrrb binds near the Oct-Sox element in the promoters of Zfp42 (Rex1) and Rest but not near the Oct-Sox site of Dppa3 and Lefty1 (Fig. 6B). Interestingly, removing Oct4 from the promoters by doxycycline treatment (Fig. 4D) prevented the detection of Esrrb at the Rest promoter, whereas Esrrb binding to the Zfp42 promoter was unaffected.

We next determined the contribution of Esrrb to the expression of Zfp42 and Rest by Esrrb knockdown. Expression of Esrrb shRNAs caused a reduction in Zfp42 mRNA expression, whereas Rest mRNA expression was unaffected (Fig. 6C). We conclude that Esrrb binds near the Oct-Sox sites of two other Oct4 targets, Zfp42 and Rest, but that only in the case of Zfp42 does this binding contribute to its expression in ES cells.

Esrrb protein levels correlate with Nanog levels in ES cell colonies.

Oct4-positive (Oct4⁺) ES cell colonies express Nanog in a mosaic fashion (6, 12, 28). The cellular expression of Esrrb was examined by immunostaining ES cells with antibodies against Nanog or Esrrb. Interestingly, 46C ES cell colonies show a mosaic pattern of Esrrb expression within the Oct4⁺ population (Fig. 7A and B). Moreover, when cells are classified according to their relative high, medium, or low levels of Esrrb staining within ES cell colonies (Fig. 7A), there is a good correlation between the levels of expression of Nanog and those of Esrrb (Fig. 7C). This correlation between the cellular levels of Esrrb and Nanog, in combination with our Nanog gene regulation data, suggests that high Nanog expression in the cell may be facilitated by high Esrrb expression.

Transcriptional regulation via transcription factor interactions in ES cell self-renewal.

Here we have provided evidence of a stem cell factor, Oct4, directing its physical interactor, Esrrb, to a target gene, Nanog, to positively regulate transcription. Gene regulation facilitated by complexes of individual transcription factors, like the Oct4-Esrrb complex, may be widespread in ES cells. Visual inspection of the sequences around the Oct-Sox sites of a number of Oct-Sox target genes for homology to the ERRE identified several potential Esrrb targets that were tested for regulation by Esrrb. Of these, ChIP analysis showed Esrrb to be detectable on Zfp42 and Rest but not Lefty1 or Dppa3. The Oct4-independent binding of Esrrb to the Zfp42 promoter may be due to the high match of the Zfp42 ERRE sequence to the consensus (8/9) (Fig. 6A), which may provide sufficient DNA binding affinity. The extreme proximity of the ERRE and Oct sites in Rest coupled with the low match to the ERRE consensus (6/9) could underlie the Oct4-dependent ChIP of Esrrb at Rest. However, the ERRE in Nanog is further removed from the Oct-Sox site and is a better match to the consensus (7/9) than that in Rest, so there is no obvious common feature that can explain the Oct4-dependent binding of Esrrb to each of these sequences. There is also no clear reason why the remaining two Oct-Sox targets do not bind Esrrb. A low match to the consensus could explain the lack of binding to Dppa3 (6/9). However, the consensus match for Lefty1 is the same as that for Nanog (7/9), suggesting that relative spatial disposition and/or distance could play a role. Further experimentation will be required to more deeply understand the relationship of Oct4 and Esrrb binding to DNA and how this affects gene regulation.

Nanog is expressed mosaically within the Oct4⁺ populations in ES cell cultures (6, 12, 28). Moreover, Nanog levels fluctuate in ES cell cultures such that cells expressing a low level or no Nanog can reexpress a high level of Nanog. However, lowered Nanog expression predisposes cells toward differentiation, without marking a commitment event (6). Here we show that the mosaic patterns of Esrrb and Nanog expression in ES cell colonies largely overlap. We also show that Esrrb positively regulates Nanog expression. As Nanog has been reported to positively regulate Esrrb expression (16), Esrrb and Nanog may both act to reinforce expression of the reciprocal gene through a positive feedback loop. How this leads to mosaic and cofluctuating levels of both proteins remains to be determined. Oct4 is not obviously mosaic and appears not to fluctuate, suggesting that it is not a determining factor of fluctuations in Nanog and Esrrb in ES cells. As Nanog levels and, by implication, Esrrb levels regulate the self-renewal efficiency of ES cells, unraveling this regulatory mechanism will be important for a fuller understanding of ES cell self-renewal and the maintenance of pluripotency.