Epigenetic Regulatory Mechanisms Distinguish Retinoic Acid-mediated Transcriptional Responses in Stem Cells and Fibroblasts*

Retinoic acid (RA), a vitamin A metabolite, regulates transcription by binding to RA receptor (RAR) and retinoid X receptor (RXR) heterodimers. This transcriptional response is determined by receptor interactions with transcriptional regulators and chromatin modifying proteins. We compared transcriptional responses of three RA target genes (Hoxa1, Cyp26a1, RARβ2) in primary embryo fibroblasts (mouse embryonic fibroblasts), immortalized fibroblasts (Balb/c3T3), and F9 teratocarcinoma stem cells. Hoxa1 and Cyp26a1 transcripts are not expressed, but RARβ2 transcripts are induced by RA in mouse embryonic fibroblasts and Balb/c3T3 cells. Retinoid receptors (RARγ, RXRα), coactivators (pCIP (NCOA3, SRC3)), and p300 and RNA polymerase II are recruited only to the RARβ2 RA response element (RARE) in Balb/c3T3, whereas these proteins are recruited to RAREs of all three genes by RA in F9 cells. In F9, RA reduces polycomb (PcG) protein Suz12 and the associated H3K27me3 repressive epigenetic modification at the RAREs of all three genes. In contrast, in Balb/c3T3 cells cultured in the +/−RA, Suz12 is not associated with the Hoxa1, RARβ2, and Cyp26a1 RAREs, whereas slow levels of the H3K27me3 mark are seen at these RAREs. Thus, Suz12 is not required for gene repression in the absence of RA. Even though the Hoxa1 RARE and proximal promoter show high levels of H3K9,K14 acetylation in Balb/c3T3, the Hoxa1 gene is not transcriptionally activated by RA. In Balb/c3T3, CpG islands are methylated in the Cyp26a1 promoter region but not in the Hoxa1 promoter or in these promoters in F9 cells. We have delineated the complex mechanisms that control RA-mediated transcription in fibroblasts versus stem cells.

and CARM1 have been shown to interact with the retinoid receptors and mediate their transcriptional activation in response to the ligand (for review, see Refs. 21). Coactivators of the SRC family and p300/CBP possess histone acetyltransferase activity, which catalyzes the addition of acetyl groups to histones residues wrapped around the DNA, thus opening the compact chromatin structure (22,23). Acetylated lysine residues on histones serve as binding sites for bromodomain-containing chromatin remodeling complexes such as SWI/SNF and Spt-Ada-Gcn5 acetyltransferase (SAGA) complexes (24). Chromatin remodelers utilize the energy from ATP hydrolysis to reposition the nucleosomes and facilitate transcription (25). Thus, histone modifications, in combination with chromatin remodeling, decondense the chromatin structure and thereby facilitate transcription (26).
Moreover, a critical role for polycomb group (PcG) proteins in regulating transcription, primarily via repression, has been reported. PcG proteins function as epigenetic silencers and exist in multiprotein complexes known as polycomb repressive complexes (PRCs) (for review, see Refs. [27][28][29][30]. PRCs have essential roles in embryonic development and differentiation (for review, see Refs. 31 and 32).
Using F9 teratocarcinoma stem cells as a model system, we recently showed that in the absence of RA, the PcG protein, Suz12, was associated with the RAREs present in Hoxa1, Cyp26a1, and RAR␤ 2 genes (33,34). However, upon exposure of cells to RA, PcG proteins are displaced from these genes, and this displacement is accompanied by the recruitment of transcriptional machinery and the activation of transcription (33). These studies demonstrated an important link between some of the key mediators of retinoid signaling and their regulation by the reversible association of PcG proteins. These seminal findings in F9 stem cells have raised an intriguing question concerning the regulation of retinoid target genes by PcG proteins in other cellular lineages, especially those that lack the ability to self-renew. During the course of cellular differentiation, lineage-specific genes are activated, whereas genes involved in pluripotency are silenced (for review, see Refs. 35 and 36). Differentiation is a multistep process involving complex epigenetic modifications, such as histone modifications and changes in DNA methylation (28,(37)(38)(39). The complex and dynamic interplay between a multitude of transcriptional co-regulators and chromatin modifiers determines the cell-specific "gene signature" that governs differentiation. Thus, this study is aimed at determining if the dynamic association of PcG proteins at retinoid-responsive genes is a regulatory mechanism specific to stem cells or if it is a mechanism important in other differentiated cellular lineages.
We utilized both F9 Wt stem cells and immortalized mouse fibroblasts as model systems to assess the binding dynamics of PcG proteins at the RA regulatory elements located in the Hoxa1, Cyp26a1, and RAR␤ 2 genes. Because the transcriptional state of a gene is determined by the overall chromatin structure, we examined the chromatin states of the RAREs (8,9,40,41) that regulate the expression of these genes by chromatin immunoprecipitation (ChIP) assays. We show that the lack of transcriptional activity of Hoxa1 and Cyp26a1 in Balb/c3T3 cells is an outcome of the differential chromatin signatures at these genes as we observed different epigenetic modifications associated with the Hoxa1 and Cyp26a1 RAREs in F9 stem cells.

MATERIALS AND METHODS
Cell Culture-Balb/c3T3 clone A31 (ATCC-CCL163)-immortalized mouse fibroblasts, primary mouse embryonic fibroblasts (MEFs), and F9 mouse teratocarcinoma stem cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 2 mM glutamine. Both cell lines were maintained in an incubator at 37°C, 10% CO 2 , and 95% humidity. All-trans-retinoic acid (Sigma) was dissolved in 100% ethanol under dim light and stored at 4°C. RA was used at a final concentration of 1 M by dilution in the culture media, and all experiments involving RA were performed under dim light conditions. The final concentration of ethanol was kept below 0.1% in these experiments.
RNA Isolation and Reverse Transcription-F9 Wt, Balb/c3T3, and MEF cells were seeded at a density of 1 ϫ 10 6 cells/100-mm plate (plates were gelatin coated in case of F9 Wt cells), and cells were allowed to attach overnight. The next day cells were treated with fresh culture medium containing 1 M RA. Control plates were treated with 0.1% ethanol. After RA treatment, cells were harvested at 8, 24, 48, and 72 h. Fresh RA containing culture medium was replenished in 72 h plates after 48 h. Cells were harvested in 1 ml of Trizol (Invitrogen), and total cellular RNA was extracted according to the manufacturer's protocol. RNA was quantified on a UV spectrophotometer by measuring the optical density at 260 nm. 3 g of RNA was reverse-transcribed to cDNA using 0.5 g of oligo(dT) 12-18 primers (Invitrogen) and 200 units of Superscript II (Invitrogen). The cDNA obtained was diluted 10-fold, and 2 l of diluted cDNA was utilized for PCR reactions.
Western Blot Analysis-F9 Wt and Balb/c3T3 cells were plated and treated with RA as in RNA isolation experiments. The cells were harvested in SDS denaturing buffer followed by a brief sonication step. The samples were boiled, and 60 g of protein sample was separated on a 10% SDS-PAGE gel and transferred to a nitrocellulose membrane. The membrane was blocked by 5% blotto for 2 h at room temperature followed by incubation with the primary antibody (anti-Suz12, 1:1000; anti-␤ actin, 1:1000) at 4°C overnight with shaking. After washing, the blots were incubated with secondary antibody (anti-goat, sc-2020 (1:5000) for ␤ actin and anti-rabbit, sc-2027 (1:5000) for Suz12) at room temperature for about 1 h. The membranes were incubated with Supersignal substrate (Pierce), and signal was detected using Biomax film (Eastman Kodak, Co.).
Chromatin Immunoprecipitation-2.5 ϫ 10 6 F9 Wt cells or 3.2 ϫ 10 6 Balb/c3T3 cells were plated in 150-mm tissue culture dishes (gelatin coated in case of F9 Wt cells) ϳ40 h before formaldehyde fixation. RA treatment was staggered such that all plates with varying times of RA addition (0, 8, and 24 h) were cross-linked and harvested at the same time. Cells were plated in an additional plate for counting to normalize for the number of cells per immunoprecipitation. For the two-step cross-linking protocol, cells were washed with ice-cold cold phosphate buffered saline (PBS) and fixed with 10 ml of 2 mM disuccinimidyl glutarate (Pierce) solution in PBS for 45 min at room temperature with gentle shaking. Subsequently, cells were washed twice with PBS and fixed with 1% formaldehyde solution in PBS at room temperature for 10 min. In the one-step cross-linking procedure, cells were fixed by adding 37% formaldehyde directly to the culture media in the plates such that the final concentration of formaldehyde was 1%. Plates were then incubated at room temperature with gentle shaking for 10 min. In both the one-step and two-step cross-linking protocols, the formaldehyde cross-linking reaction was quenched by adding 1.25 M glycine to a 200 mM final concentration. Cells were then washed with PBS and harvested in PBS. After harvesting, cells were centrifuged and lysed in 350 l of lysis buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA, 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate). A tablet of complete mini protease inhibitor (Roche Applied Science, catalog number 11836153001) was added fresh to 10 ml of lysis buffer. Chromatin was sonicated using a Branson 150 Sonifer (setting 3) for 2 ϫ 15 s, and cellular debris was removed from the sheared chromatin by centrifugation at 14,000 rpm for 10 min at 4°C. For each immunoprecipitation (IP), 3.5 ϫ 10 6 cells were used, and the volume of soluble chromatin needed for the IP was calculated by normalizing to the total number of cells counted using an electron particle counter (Coulter Z, Beckman Coulter, Inc., Fullerton, CA) at the time of harvesting. Depending on the cell density at the time of harvest, ϳ30 -70 l (of a total volume of ϳ350 l) of soluble chromatin was diluted to a total volume of 500 l with lysis buffer, and chromatin was precleared with 25 l of 50% protein A Sepharose, PBS slurry (catalog no. 17-0780-01, Amersham Biosciences) by incubating the samples with gentle shaking at 4°C for 1-2 h. After preclearing, beads were removed by centrifugation, and the precleared chromatin was incubated with 2 g of specific antibody. Immunoprecipitations were performed overnight with shaking at 4°C. Immunocomplexes were recovered by incubating with 50 l of 50% protein A-Sepharose, PBS slurry for ϳ2 h at 4°C. For the RNA polymerase II (Pol II) IP, the immunocomplex was incubated with 5 l of anti-IgM (M8644 -1MG; Sigma) for ϳ1 h before the addition of protein A-Sepharose beads. After incubating immunoprecipitated material with protein A-Sepharose beads, beads were centrifuged at 2800 rpm and washed twice with lysis buffer for about ϳ5-10 min by gentle shaking at 4°C. This was followed by two washes with ChIP wash buffer (50 mM Tris-HCl (pH 8.5), 500 mM LiCl, 5 mM EDTA, 1% Nonidet P-40, 1% sodium deoxycholate) and TE buffer (10 mM Tris-HCl (pH 8.0), 1 mM EDTA) each. After the final wash, the supernatant was completely removed, and immunocomplexes were eluted from the beads by incubating the beads with 100 l of elution buffer (50 mM Tris-HCl (pH 8.0), 1% SDS, 1 mM EDTA) at 65°C for 10 min. Samples were vortexed for 15 s and then centrifuged. The eluted DNA-protein complexes were transferred to new tubes, and NaCl was added to the eluted complexes to a final concentration of 200 mM. Finally, samples were incubated at 65°C overnight for reverse cross-linking. For the input samples, 25 l (5.5% of the starting IP for each antibody) of the precleared chromatin was incubated with 75 l of elution buffer, and NaCl was added to a final concentration of 200 mM. The input samples were also reverse cross-linked with the immunoprecipitated samples overnight. The reverse crosslinked DNA was then purified using a PCR purification kit (Qiagen) according to the manufacturer's protocol. DNA obtained after PCR purification was assayed for real time PCR analysis.
Semiquantitative and Real Time PCR Analysis-Semiquantitative PCR was carried out using commercial Taq polymerase (Invitrogen). PCR reactions were run on a Bio-Rad iCycler using the following protocol; Step 1, 94°C initial denaturation (3 min); Step 2, denature at 94°C (30 s), anneal at 58 -62°C (optimized based on primer pair) for 30 s, extension for 45 s at 72°C. The above protocol was repeated for n cycles such that the PCR was in the linear range for each primer pair monitored (n does not exceed 39). The PCR products were separated by electrophoresis on ethidium bromide (0.4 mg/ml)-stained 1.5% (w/v) agarose gel. Real time PCR was carried out in a total volume of 20 l using USB Hot start-IT SYBR Green quantitative PCR master mix containing 10 nM fluorescein as an internal reference (U. S. Biochemical Corp.). PCR reactions were set up in triplicate with 2 l of cDNA per PCR reaction. The template was denatured at 94°C for 3 min followed by annealing at 60 -62°C (optimized by primer pair) and extension at 72°C for 45 s. An additional 40 cycles of annealing/extension were carried out, and fluorescence was monitored during a 10-s step at 84°C (temperature was modified depending on the primer pair). Real time data were quantified using a linear standard curve generated by calculating the cycle thresholds of the serially diluted cDNA. The cycle threshold (Ct) values of unknown samples in a ChIP experiment were measured, and concentrations of unknown samples were calculated by interpolating the concentrations on the standard curve. Real time PCR for ChIP was carried out on DNA purified from ChIP samples using iQ SYBR Green supermix (Bio-Rad, catalog no. 170-8884). Purified DNA (2 l) was utilized for the PCR reaction in a total volume of 20 l. Sense and antisense primers were added to the reaction mixture at a final concentration of 100 nM each. PCR reactions were carried out using a touchdown protocol. The template was denatured at 94°C for 15 s followed by initial annealing/extension at 70°C while reducing the temperature by 1°C for 10 cycles. Subsequently, an additional 40 cycles of annealing/extension were carried out at 60°C. Fluorescence was monitored during a 10-s step at 77°C (temperature was modified depending on the primer pair) after annealing/extension at 60°C. Real time data were quantified using a linear standard curve generated by calculating the cycle thresholds of the serially diluted input samples of known concentrations. The cycle threshold (Ct) values of unknown samples in a ChIP experiment were measured, and concentrations were calculated by interpolating the percentage input on the standard curve.
The following primers were used in this study, and primer specificity was checked using the University of California Santa Cruz in silico (UCSC Genome Bioinformatics) PCR program. Only a single amplicon is generated at the target locus of interest as verified by running the product on the agarose gel.
DNA Methylation Analysis by Bisulfite Sequencing-Genomic DNA was isolated and purified by phenol chloroform extraction. Purified DNA (1 g) was utilized for the bisulfite conversion reaction to convert cytosines to uracil as recommended by the manufacturer's protocol (EZ DNA methylation-Direct kit, Zymo Research Corp.). The bisulfite modified DNA was used as a template for the PCR reaction. The PCR products were purified using a PCR purification kit (Qiagen), and eluted DNA was ligated using the pGEMT-Easy vector ligation kit (Promega). Ligated product was transformed into ultracompetent (Xl10 strain) bacteria (Stratagene). The transformed bacterial cultures were plated on LB agar plates containing ampicillin and cultured overnight at 37°C. The antibiotic resistant colonies were screened by PCR using bisulfite specific primers to select for the ones containing the correct inserts. DNA was isolated by miniprep from at least 10 such positive colonies and sequenced using a commercial SP6 primer. The following primers were used in amplifying the bisulfite modified DNA, and they were designed using Methprimer.
Statistical Analysis-Statistical analysis (mean Ϯ S.E.) was performed using Graph Pad Prism 4.0 and Microsoft Excel. S.E. represents the error across three or more independent experiments. Significance (p value) was calculated to compare three or more groups of data using the one way analysis of variance or t test. A p value Ͻ 0.05 indicates a statistically significant change.

Differential Expression of Hoxa1, RAR␤ 2 , and Cyp26a1 Transcripts in F9 Teratocarcinoma Stem Cells and Balb/c3T3
Fibroblasts-F9 teratocarcinoma stem cells (denoted as F9 Wt) have been used as a model system to study transcriptional regulation by RA (33,34,42). We monitored the expression of three genes transcriptionally regulated by RA, Hoxa1, RAR␤ 2 , and Cyp26a1, in Balb/c3T3 mouse fibroblasts in comparison to F9 Wt stem cells. Balb/c3T3 fibroblasts are a good model for comparative analysis of transcriptional regulation with respect to F9 Wt cells because they represent a more differentiated cell lineage and are responsive to RA-mediated growth arrest (43,44).
We first monitored Hoxa1, Cyp26a1, and RAR␤ 2 mRNA levels by quantitative real time PCR analysis after treatment of the cells with RA for varying times. As previously reported (33,42), transcripts of all three genes were induced by RA treatment in F9 Wt cells as early as 8 h, with increasing expression up to 72 h ( Fig. 1). However, Hoxa1 (Fig. 1a) and Cyp26a1 (Fig. 1c) transcripts were not induced by RA in Balb/c3T3 cells even after 72 h of RA treatment. RAR␤ 2 transcripts were induced in Balb/ c3T3 cells as early as 8 h after RA addition (ϳ25-fold, p Ͻ 0.01, Fig. 1b). The kinetics of RAR␤ 2 transcriptional activation in Balb/c3T3 cells were different from those in F9 Wt cells (Fig.  1b). We also monitored the transcripts of a gene that encodes a ribosomal protein, 36B4, by real time PCR as a control. As expected, 36B4 levels remained unchanged upon RA treatment in both cell lines (Fig. 1d).
Because the Hoxa1 and Cyp26a1 genes were not expressed in Balb/c3T3 cells, it was imperative to determine whether retinoid receptors (RAR␣ and RAR␥) were expressed in these cells. We showed by semiquantitative reverse transcription-PCR that RAR␣ and RAR␥ transcripts were expressed similarly in F9 Wt and Balb/c3T3 cells (Fig. 1d).
Additionally, we determined Hoxa1, Cyp26a1, and RAR␤ 2 mRNA levels in primary MEFs. This was done to ensure that retinoid signaling was not altered during the immortalization process of the Balb/c3T3 cells. Hoxa1 and Cyp26a1 transcripts were not induced, whereas the RAR␤ 2 gene was transcriptionally activated by RA in primary MEFs (Fig. 1e). Thus, we con-clude that the Hoxa1 and Cyp26a1 genes are not transcriptionally activated by RA in primary and immortalized mouse fibroblasts even though all three retinoic acid receptor isotypes are expressed.
RAR␥ and RXR␣ Are Differentially Recruited at the Hoxa1, Cyp26a1, and RAR␤ 2 RAREs in Balb/c3T3 Cells-According to the current model of nuclear receptor-mediated transcriptional activation, RAR/RXR heterodimers are bound to the RAREs of their target genes and repress transcription by virtue of their interaction with nuclear co-repressor family members (NCoR/ SMRT) (45). In line with this model, previous studies have shown that RAR/RXR heterodimers are constitutively bound to Hoxa1, Cyp26a1, and RAR␤ 2 RAREs in F9 Wt cells (33). Because we found that Hoxa1 and Cyp26a1 transcripts are not expressed in Balb/c3T3 cells ( Fig. 1, a-c), we hypothesized that this could result from the absence of a functional RAR/RXR heterodimer at the Hoxa1 and Cyp26a1 RAREs in Balb/c3T3 cells. Alternatively, this lack of expression of Hoxa1 and Cyp26a1 could result from the lack of displacement of co-repressors. Thus, we employed a two-step ChIP assay to monitor the association dynamics of RAR␥ and RXR␣ in the absence and presence of RA at 8 and 24 h post-RA addition. The two-step cross-linking ChIP protocol provides increased sensitivity for detection of RAR␥ (33).
We monitored the association of RAR␥ and RXR␣ at the Hoxa1, Cyp26a1, and RAR␤ 2 RAREs. The Hoxa1 RARE is located ϳ2 kb 3Ј of the Hoxa1 gene ( Fig. 2a) (8,46). In contrast, the RAR␤ 2 RARE is located in the proximal promoter region about 55 bp upstream of the transcription start site (TSS) (9). The Cyp26a1 gene has two well characterized RAREs (denoted as R1 and R2), and both are located in the proximal promoter region. The R1 RARE is located ϳ70 bp upstream of TSS (40), and the R2 RARE region is at ϳ1950 bp from the TSS (41). We also monitored the recruitment of RAR␥ and RXR␣ at the proximal promoter region of the Hoxa1 gene (denoted as Hoxa1 PP). The ChIP data are plotted as -fold enrichment at each gene, where "fold enrichment" is defined by the percentage input determined by the real time PCR at the given gene divided by the percentage input at a nonspecific intergenic region, located Ϫ18 kb downstream of the Hoxb1 gene (Ϫ18 kb Hoxb1 3Ј). In addition, all our ChIP assays were performed with IgG as a negative control to rule out any signal due to nonspecific antibody binding. As previously shown (33), we observed recruitment of RAR␥ and RXR␣ at the Hoxa1 RARE region in F9 Wt cells (Fig. 2b). The receptors were present at the Hoxa1 RARE before RA addition, and the receptor levels increased after RA exposure (Fig. 2b). We observed a statistically significant (ϳ12-fold) enrichment of RAR␥ after 8 h of RA treatment in F9 Wt cells as compared with the untreated controls.
In contrast to what we observed in F9 Wt cells, RAR␥ and RXR␣ failed to associate at the Hoxa1 RARE in Balb/c3T3 cells, thus indicating that receptors are absent from this region. There was also no enrichment of either receptor above background levels at the Hoxa1 proximal promoter (PP) region (Fig. 2c).
We found that RAR␥ and RXR␣ associated with the RAR␤ 2 RARE in both F9 Wt and Balb/c3T3 cells (Fig. 2d). Both receptors were present at the RARE in the absence and presence of RA, and the levels remained unchanged after exposure to RA.
RAR␥ and RXR␣ were present at both of the RAREs located in Cyp26a1 gene (R1 and R2) in F9 Wt cells both in the absence and presence of RA. The levels of RXR␣ rose significantly after and reverse-transcribed to form cDNA. The cDNA obtained was utilized for semiquantitative PCR analysis. PCR products were visualized by running them on a 1.5% agarose gel stained with ethidium bromide. enrichments, respectively, at the R1 RARE in F9 Wt cells (Fig.  2e). RAR␥ recruitment also increased by ϳ6-fold at the Cyp26a1 R1 RARE after 8 h of RA treatment (Fig. 2e).
No association of RAR␥ and RXR␣ was observed at either of the RAREs in the Cyp26a1 gene in the Balb/c3T3 cell line. Thus, we conclude that in Balb/c3T3 cells RAR␥ and RXR␣ are absent from the Hoxa1 and Cyp26a1 RAREs, whereas both receptors are present at the RAR␤ 2 RARE.
The Coactivators, pCIP and p300, Are Recruited to the RAR␤ 2 RARE in Balb/c3T3 Cells-Because we found that the RAR/ RXR heterodimers associated with the RAR␤ 2 RARE in Balb/ c3T3 cells, we assessed the recruitment of coactivators pCIP and p300 using the conventional ChIP assay. pCIP was associated with the RAR␤ 2 RARE before RA treatment, as shown by the ϳ4-fold enrichment at the RAR␤ 2 RARE relative to the Ϫ18-kb Hoxb1 negative control intergenic region in Balb/c3T3 cells (Fig. 3c). These data in Balb/c3T3 cells are similar to the basal enrichment levels of pCIP (ϳ5-fold) in F9 Wt cells under control conditions (Fig. 3c). These results are also in agreement with a previous study which demonstrated that several components of the transcriptional machinery, including RNA polymerase II, were assembled at the RAR␤ 2 promoter in the absence of the ligand in P19 cells (47,48). After the addition of RA to the Balb/c3T3 cells, the levels of pCIP at the RAR␤ 2 RARE increased, as indicated by the ϳ15-fold enrichment at 8 and 24 h after RA treatment (Fig. 3c). In contrast, pCIP levels showed only a smaller increase at the RAR␤ 2 RARE in F9 Wt cells (Fig. 3c). At the Hoxa1 RARE, we observed an ϳ10-fold increase in pCIP levels in F9 Wt cells at 8 and 24 h of RA treatment (Fig. 3a). However, in Balb/c3T3 cells we did not see any changes over background levels in the -fold enrichment of pCIP at the Hoxa1 RARE (Fig. 3a). This is consistent with the lack of RA induction of Hoxa1 mRNA (Fig. 1a) and the absence of receptors (RAR␥ and RXR␣) at the Hoxa1 RARE gene in Balb/ c3T3 cells (Fig. 2a). At the Hoxa1 PP region, pCIP levels were comparable with the IgG controls in both cell lines (Fig. 3b). and then cross-linked using the two-step cross-linking procedure. The cells were first incubated with 2 mM disuccinimidyl glutarate for 45 min with gentle shaking followed by fixation with 1% formaldehyde. Cross-linked cells were harvested and lysed in lysis buffer and sonicated to obtain soluble chromatin. Chromatin from ϳ3.5 ϫ 10 6 cells was used for each IP, and chromatin samples were precleared with protein A-Sepharose beads. Precleared chromatin was immunoprecipitated with RAR␥, RXR␣, and IgG antibodies overnight. Bound chromatin was pulled down by protein A-Sepharose beads, washed, eluted, reverse cross-linked, and purified. Immunoprecipitated DNA (2 l) was used for real time PCR. Experiments were repeated at least three times starting with cells, and quantitative PCR was performed in triplicate. The data are plotted as -fold enrichment, which is defined as the percentage of input DNA immunoprecipitated at a specific locus divided by the percentage of input DNA immunoprecipitated at the nonspecific region (Hoxb1-18 kb 3Ј) in the same IP. Data were statistically analyzed using Graphpad Prism 4.0, and error bars indicate S.E. over independent replicate experiments. The asterisks depict significance, where 1 asterisk denotes p Ͻ 0.05 and 2 asterisks denote p Ͻ 0.01. p values indicate statistically significant changes after RA treatment within each cell line compared with the untreated controls for that cell line. Note the differences in the y axis scale in panel d between F9 Wt and Balb/c3T3 cells.
In F9 Wt cells we found that there were ϳ6and ϳ7-fold enrichments of pCIP at the Cyp26a1 R1 and R2 RAREs, respectively, 8 h after RA addition. In contrast, in Balb/c3T3 cells we did not observe any enrichment of pCIP at these RAREs (Fig. 3,  d and e).
The recruitment pattern of another coactivator protein, p300, at all the RAREs investigated in this study was comparable with that for pCIP. The coactivator p300 was present at the RAR␤ 2 RARE in F9 Wt cells, and the levels did not increase upon RA treatment, as shown by the ϳ5-fold enrichment of p300 over the Ϫ18-kb Hoxb1 region in both the absence and presence of RA (Fig. 3c). We observed an ϳ3-fold enrichment of the p300 signal over background (Ϫ18-kb Hoxb1) in the absence of RA at the RAR␤ 2 RARE in the Balb/c3T3 cell line, and the p300 levels showed an ϳ6-fold enrichment at 8 h of RA treatment (Fig. 3c).
The levels of p300 at the Hoxa1 RARE increased in F9 Wt cells after 8 and 24 h of RA treatment, as shown by the ϳ6.5-fold enrichment over background (Fig. 3a). This is in contrast to the lack of recruitment of p300 at the Hoxa1 RARE in Balb/c3T3 cells. p300 was not present at the Hoxa1 PP region in either cell line (Fig. 3b).
In F9 Wt cells we observed an ϳ5-fold enrichment of p300 over background at the Cyp26a1 R2 RARE 8 h after RA addition, whereas in Balb/c3T3 cells p300 levels remained low and comparable with the background signal at the Cyp26a1 R1 and R2 RAREs even 24 h after RA addition (Fig. 3, d and e).
Based on these results, we conclude that RA signaling leads to the recruitment of the coactivator proteins pCIP and p300 only at the transcriptionally active RAR␤ 2 RARE in Balb/c3T3 cells. In contrast, in F9 Wt cells coactivator proteins are recruited in response to RA at all of the RAREs tested in this study, consistent with prior results (33,34).
RNA Polymerase II Only Associates with Transcriptionally Active RAREs in Balb/c3T3 Cells-We next decided to assess the association of Pol II at all of the RAREs. We utilized an antibody that specifically recognizes the phosphorylated serine 5 of the CTD of Pol II (34). The Pol II (pCTDser5) was present at the RAR␤ 2 RARE before RA treatment, as shown by the 3.7and 3.1-fold enrichments of the RNA Pol II signal over the background (Ϫ18-kb Hoxb1) in untreated F9 Wt and Balb/ c3T3 cells, respectively. Pol II occupancy increased at the RAR␤ 2 RARE upon RA treatment in both cell lines. We found a 7.2-fold enrichment of Pol II at the RAR␤ 2 RARE in F9 Wt cells and an ϳ18.8-fold enrichment of Pol II in Balb/c3T3 cells after 8 h of RA addition (Fig. 4c).
In F9 Wt cells, Pol II levels rose by 4.8-fold at the Hoxa1 RARE by 24 h of RA treatment, whereas Pol II was not present at the Hoxa1 RARE in Balb/c3T3 cells (Fig. 4a). This is in agreement with the lack of recruitment of coactivators at this RARE in Balb/c3T3 cells. However, high levels of Pol II were associated with the Hoxa1 PP region in the absence and presence of RA in Balb/c3T3 cells (Fig. 4b). In F9 Wt cells we observed an ϳ18-fold enrichment of Pol II over background at the Hoxa1 PP region in the absence of RA and enrichments of ϳ23and 34-fold over background at 8 and 24 h of RA treatment, respectively (Fig. 4b).
We observed 17-and ϳ35-fold enrichments of Pol II at the Cyp26a1 R1 and R2 RAREs, respectively, in F9 Wt cells upon 24 h of RA treatment (Fig. 4, d and e). In contrast, we did not observe recruitment of Pol II at the Cyp26a1 RAREs in Balb/ c3T3 cells (Fig. 4, d and e). Thus, we have shown that Pol II fails to be recruited at the RAREs of the transcriptionally inactive  Coactivator recruitment (pCIP and p300) at Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs in F9 Wt and Balb/c3T3 cells upon RA treatment. F9 Wt and Balb/c3T3 cells were treated with RA for different times (0, 8, and 24 h) and fixed with 1% formaldehyde. Cells were lysed and sonicated to obtain soluble chromatin. Chromatin was precleared and immunoprecipitated with anti-pCIP or anti-p300 antibodies, and the immunoprecipitated DNA was quantitated using real time PCR. Each experiment was repeated at least three times, starting with cells, and quantitative PCR was performed in triplicate. The data are plotted as -fold enrichment, which is defined as the percentage of input DNA immunoprecipitated at a specific locus divided by the percentage of input DNA immunoprecipitated at the nonspecific region (Hoxb1-18 kb 3Ј) in the same IP. Mean -fold enrichment was calculated, and error bars represent the S.E. over independent replicate experiments. 1 asterisk denotes p Ͻ 0.05, and 2 asterisks denote p Ͻ 0.01. p values indicate significant changes after RA treatment within each cell line compared with the untreated control for that cell line.
genes, Hoxa1 and Cyp26a1, in Balb/c3T3 cells. However, Pol II is present at the proximal promoter of Hoxa1 in Balb/c3T3 cells.
Acetylation of Histone H3 Is Not Sufficient for Transcriptional Activation of the Hoxa1 Gene in Balb/c3T3 Cells-It has been well established that different modifications of histone residues wrapped around the core nucleosomes modulate the chromatin structure, which in turn determines the transcriptional outcome (49,50). Because the addition of acetyl groups to the histone tails opens up the chromatin structure, making it poised for transcription, we specifically examined the acetylation levels of histone 3, a mark of "active" chromatin, at the Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs. Antibody to acetylated lysine residue 9 (K9) and lysine 14 (K14) of the histone H3 tail, hereafter denoted as H3K9,K14ac, was used in ChIP assays. At the Hoxa1 RARE in Balb/c3T3 cells, we observed an ϳ1% of input immunoprecipitation using anti-H3K9,K14ac antibody in the absence and presence of RA compared with ϳ 0.3% of input immunoprecipitation at the intergenic Hoxb1-18 kb downstream region (Fig. 5a). We also observed statistically significant enrichment in the H3K9,K14ac signal compared with the nonspecific IgG controls in Balb/c3T3 cells (Fig. 5a). F9 Wt cells also displayed similar H3K9,K14ac levels (ϳ0.5% of the input immunoprecipitated) at the Hoxa1 RARE (Fig. 5a). H3K9,K14ac levels increased at the Hoxa1 PP region upon RA treatment of F9 Wt cells, as shown by the immunoprecipitation and fixed with 1% formaldehyde. Cells were lysed and sonicated to obtain soluble chromatin. Chromatin was precleared and immunoprecipitated with anti-pCTDser5 antibody, and the immunoprecipitated DNA was quantitated using real time PCR. Experiments were performed at least three times, starting with cell culture, and quantitative PCR was performed in triplicate. The data are averaged and plotted as a percentage of the immunoprecipitated DNA, with the reference to the percentage of input DNA at each locus. The error bars represent S.E. over independent replicate experiments. The percentage of input DNA immunoprecipitated in the same IP at Hoxb1-18 kb 3Ј region is also graphed for comparison as a negative control. 1 asterisk denotes p Ͻ 0.05, and two asterisks denotes p Ͻ 0.01. p values indicate significant changes after RA treatment within each cell line compared with the untreated controls for that cell line. Note that at the Hoxa1 PP region the levels of Pol II are similar in Balb/ c3T3 cells cultured in the presence and absence of RA but are statistically significantly higher than those in the Hoxb1-18 kb 3Ј region (p Ͻ 0.05).

FIGURE 5. Histone acetylation status at Hoxa1, RAR␤ 2 , and Cyp26a1
RAREs in F9 Wt and Balb/c3T3 cells in response to RA treatment. F9 Wt and Balb/c3T3 cells were treated with RA for different times (0, 8, and 24 h) and fixed with 1% formaldehyde. Cells were lysed and sonicated to obtain soluble chromatin. Chromatin was precleared and immunoprecipitated with anti-H3K9,K14ac, and immunoprecipitated DNA was quantitated using real time PCR. Experiments were repeated at least three times, starting with cell culture, and quantitative PCR was performed in triplicate. The data are averaged and plotted as a percentage of immunoprecipitated DNA with reference to the percentage of input DNA at each locus. The error bars represent the S.E. over independent replicate experiments. The percentage of input immunoprecipitated in the same chromatin extracts with nonspecific IgG negative control and the percentage of input immunoprecipitated at the Hoxb1-18 kb 3Ј region is also graphed for comparison. 1 asterisk denotes p Ͻ 0.05, two asterisks denote p Ͻ 0.01, and three asterisks denote p Ͻ 0.0005. p values indicate significant levels of H3K9,K14ac as compared with IgG control at each locus (including Hoxb1-18 kb 3Ј region) within each cell line.
of ϳ2% of input in the absence of RA and ϳ4.5% of input at 8 h of RA treatment (Fig. 5b). However, H3K9,K14ac levels at the Hoxa1 PP region did not increase upon RA treatment in Balb/ c3T3 cells (Fig. 5b).
The RAR␤ 2 RARE exhibited high levels of H3K9,K14ac in the absence and presence of RA in both cell lines. We found that ϳ1% and ϳ2.5% of the input immunoprecipitated with the H3K19/K14 ac antibody in F9 Wt and Balb/c3T3 cell lines, respectively (Fig. 5c).
High levels of H3K9,K14ac were associated with the Cyp26a1 R1 and R2 RAREs in F9 Wt cells, and this acetylation mark increased 8 and 24 h after RA addition (Fig. 5, d and e). In contrast, in Balb/c3T3 cells the Cyp26a1 R1 and R2 RAREs displayed statistically insignificant levels of the H3K9,K14ac mark compared with the IgG control (Fig. 5, d and e). The Hoxb1-18 kb intergenic region was also associated with moderate levels of this acetylation mark in both cell lines, as shown by the ϳ0.2%-0.8% of input immunoprecipitated under similar treatment conditions (Fig. 5). However, this Hoxb1-18 kb region of DNA still served as a useful second control in most of our experiments in addition to our use of negative IgG control. We observed only minimal immunoprecipitation (ϳ0.01% or less) with most other antibodies, such as Pol II, in the Hoxb1-18 kb region (Fig. 4) (33, 34). Therefore, we have demonstrated that acetylation of histone H3 in F9 Wt cells precedes transcriptional induction mediated by RA at the genes monitored in this study. There is only a moderate increase (as observed at the Cyp26a1 R1 and R2 RAREs and Hoxa1 PP region) in H3K9,K14ac levels upon RA treatment. In Balb/c3T3 cells, high levels of histone 3 acetylation are associated with transcriptionally active genes such as the RAR␤ 2 RARE (Fig. 5c). However, the presence of H3K9/K14 ac is not sufficient to initiate transcription, as shown by our data at the Hoxa1 RARE and Hoxa1 PP in Balb/c3T3 cells (Fig. 5, a  and b). Even though the Hoxa1 RARE and Hoxa1 PP region displays high levels of the H3K9/K14 ac mark in Balb/c3T3 cells, the Hoxa1 gene is not transcriptionally activated by RA in these cells (Fig. 5, a and b).
Binding of the PcG Protein, Suz12, Is Reduced at RAREs upon RA Addition in F9 Wt Cells, whereas Suz12 Association Remains Low and Unchanged in Balb/c3T3 Cells-Having demonstrated the absence of retinoid receptors (RXR␣ and RAR␥), coactivators, and RNA polymerase II at the Hoxa1 and Cyp26a1 RAREs in Balb/c3T3 cells, we next sought to determine the mechanism(s) of repression. Because the PcG protein Suz12, one of the core protein components of the PRC2 complex, is present at the Hoxa1, Cyp26a1, and RAR␤ 2 RAREs in F9 teratocarcinoma stem cells (33), we speculated that PcG-mediated silencing could be a mechanism of repression in Balb/c3T3 fibroblasts. We hypothesized that RA is unable to dissociate PcG proteins from the Hoxa1 and Cyp26a1 RAREs in Balb/ c3T3 cells. To test our hypothesis, we performed ChIP assays in F9 Wt and Balb/c3T3 cells using an antibody that recognizes the PcG protein Suz12. In the absence of RA, Suz12 is present at the Hoxa1, RAR␤ 2 , and Cyp26a1 R1 and R2 RAREs in F9 Wt stem cells (Fig. 6). We observed a 10.6-fold higher Suz12 level (over the IgG control) at the Hoxa1 RARE in the absence of RA in F9 Wt cells (Fig. 6a). RA treatment reduced the Suz12 levels at the Hoxa1 RARE to 4-fold over the IgG control at 8 h, and by 24 h of RA treatment, Suz12 levels were reduced to the level of the IgG negative control (Fig. 6a).
The Hoxa1 PP region showed a similar loss of Suz12 occupancy after RA exposure in F9 Wt cells. There was a ϳ33.7-fold enrichment of Suz12 over the IgG negative control in the absence of RA at the Hoxa1 PP. Suz12 levels progressively dropped to an ϳ8-fold enrichment over the IgG control and further to the level of the IgG negative control after 8 and 24 h of RA treatment, respectively, at the Hoxa1 PP in F9 Wt cells (Fig.  6b). In contrast, Balb/c3T3 cells did not show a significant enrichment of Suz12 over the negative IgG control at the Hoxa1 RARE and Hoxa1 PP region in either the absence or presence of RA (Fig. 6, a and b).
Studies from our laboratory have shown that Suz12 is also associated with the RAR␤ 2 and Cyp26a1 RAREs in F9 Wt cells in the absence of RA (Fig. 6) (33). We observed an ϳ14-fold enrichment over the negative IgG control at the RAR␤ 2 RARE in F9 Wt cells, and over a 24-h RA exposure Suz12 levels decreased to background levels (Fig. 6c). However, in Balb/ c3T3 cells Suz12 levels at the RAR␤ 2 RARE were low and similar to the negative IgG control, thus indicating that Suz12 is not present at the RAR␤ 2 RARE in the Balb/c3T3 cells in the absence of RA (Fig. 6c). Suz12 levels were ϳ180-fold higher than the negative IgG control at the Cyp26a1 R1 RARE in F9 Wt cells in the absence of RA, and Suz12 levels were reduced to a level ϳ8-fold over the negative IgG control after a 24-h RA exposure (Fig. 6d). In Balb/c3T3 cells Suz12 levels were low, just marginally above background IgG levels, as shown by the ϳ0.03-0.05% of input immunoprecipitated with anti-Suz12 antibody both in the absence and presence of RA (Fig. 6d). In F9 Wt cells cultured without RA, the Cyp26a1 R2 RARE showed Suz12 binding dynamics similar to those of the R1 RARE, although the levels of Suz12 were lower at the Cyp26a1 R2 RARE, as shown by the ϳ0.27% of input immunoprecipitated in the absence of RA (Fig. 6e). In contrast, Suz12 was not present at the Cyp26a1 R2 RARE in Balb/c3T3 cells (Fig. 6e). Thus, we conclude that in Balb/c3T3 cells, the PcG protein Suz12 is not associated with the Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs in the absence of RA.
We also performed semiquantitative PCR and Western blot analysis to determine the mRNA and protein expression of Suz12 in Balb/c3T3 cells compared with F9 cells. This was done to rule out the possibility that the lack of Suz12 binding at the RAREs tested in Balb/c3T3 cells results from a lack of Suz12 expression in fibroblasts. Although Suz12 mRNA levels are similar in the two cell lines (Fig. 6f), F9 Wt cells show higher expression of Suz12 protein compared with the Balb/c3T3 cells (Fig. 6g).
The "H3K27me3" Epigenetic Mark Associated with the PRC2 Complex Is Present at the Hoxa1 and Cyp26a1 RAREs in Balb/ c3T3 Cells-The PRC2 complex is composed of three core components; Ezh2, Suz12, and Eed (32). The Ezh2 protein in this multiprotein complex possesses the methyltransferase activity for the lysine 27 (K27) residue of histone 3 (H3), and incorporation of the trimethyl moiety at the lysine 27 residue of H3 is associated with repression of transcription (30,51,52). Furthermore, this silencing mark is generally erased from the genes activated upon lineage determination in response to external differentiation stimuli (28,53).
In F9 Wt cells, the H3K27me3 levels were 45.4-fold higher than the negative IgG control at the Hoxa1 RARE in the absence of RA. We observed a ϳ6.3-fold enrichment of the H3K27me3 modification over the negative IgG control at the same locus in Balb/c3T3 cells (Fig. 7a). Upon RA treatment for 24 h, H3K27me3 levels were reduced to background level in F9 Wt cells. The H3K27me3 mark was enriched by ϳ11-fold over the negative IgG control at the Hoxa1 RARE in Balb/c3T3 cells after 24 h of RA treatment (Fig. 7a). The Hoxa1 PP region showed a progressive decrease in this modification, as shown by the reduction from ϳ32.5-fold enrichment over the negative IgG control at 0 h to levels just marginally over the IgG negative control after 24 h of RA treatment in F9 Wt cells (Fig. 7b). We did not observe enrichment of the H3K27me3 chromatin modification over the background IgG levels at the Hoxa1 PP region in Balb/c3T3 cells (Fig. 7b).
At the RAR␤ 2 RARE, the H3K27me3 mark displayed an enrichment of ϳ5.7-fold in the absence of RA, and after 24 h of RA treatment, the H3K27me3 levels were reduced to the levels of the negative IgG control in F9 Wt cells (Fig. 7c). In Balb/c3T3 cells, the H3K27me3 levels at the RAR␤ 2 RARE were unchanged by RA addition (Fig. 7c). This is in contrast to the decrease in Suz12 and H3K27me3 levels in response to RA treatment in F9 Wt cells.
At the Cyp26a1 R1 RARE, the H3K27me3 levels were ϳ62.2fold higher than the negative IgG control in the absence of RA, and the H3K27me3 modification levels progressively declined to those of negative IgG control after 24 h of RA treatment in F9 Wt cells. However, moderate levels of H3K27me3 were associated with the Cyp26a1 R1 RARE, as assessed by the ϳ12.2and ϳ15.5-fold enrichments of the H3K27me3 mark over the IgG control in the absence and presence (24 h) of RA, respectively, in Balb/c3T3 cells (Fig. 7d). The Cyp26a1 R2 RARE also displayed moderate enrichment of the H3K27me3 mark in both the absence and presence of RA in Balb/c3T3 cells (Fig. 7e).
In summary, we have shown that the Suz12 levels remained low and unchanged by RA at all of the RAREs investigated in Balb/c3T3 cells (Fig. 6). However, the H3K27me3 mark was enriched (albeit at lower levels in comparison to untreated F9 Wt cells) at the Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs and was not altered by the addition of RA to Balb/c3T3 cells (Fig. 7). FIGURE 6. a-e, Suz12 recruitment at the Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs in F9 Wt and Balb/c3T3 cells in response to RA treatment is shown. F9 Wt and Balb/c3T3 cells were treated with RA for different times (0, 8, and 24 h) and fixed with 1% formaldehyde. Cells were lysed and sonicated to obtain soluble chromatin. Chromatin was precleared and immunoprecipitated with anti-SUZ12 antibody, and immunoprecipitated DNA was quantitated using real time PCR. Experiments were repeated at least three times, starting with cell culture, and quantitative PCR was performed in triplicate. The data are averaged and plotted as a percentage of immunoprecipitated DNA with reference to the percentage of input DNA at each locus. The error bars represent the S.E. over independent replicate experiments. The percentage of input immunoprecipitated in the same chromatin extracts with nonspecific IgG is also graphed for comparison as a negative control. 1 asterisk denotes p Ͻ 0.05, and two asterisks denote p Ͻ 0.01. p values indicate statistically significant changes after RA treatment within each cell line compared with the untreated controls for that cell line. f, Suz12 mRNA levels in F9 Wt cells and Balb/c3T3 cells are shown. RNA was harvested from F9 Wt and Balb/c3T3 cells treated with 1 M RA for different times (0, 8, 24, 48, and 72 h) and reversetranscribed to form cDNA. The cDNA obtained was utilized for semiquantitative PCR analysis. PCR products were visualized by running them on a 1.5% agarose gel stained with ethidium bromide. g, expression of Suz12 protein in F9 Wt and Balb/c3T3 cells is shown. Cell lysates were harvested from F9 Wt and Balb/c3T3 cells treated with 1 M RA for different times (0, 8, 24, 48, and 72 h). Whole cell extracts (60 g) were separated on 10% SDS-PAGE followed by Western blot analysis with anti- Suz12 (1:1000). Immunoblot with the ␤ actin antibody (1:1000) served as the loading control. MAY 7, 2010 • VOLUME 285 • NUMBER 19

Transcription and Epigenetic Regulation by Retinoic Acid
Methylation of the CpG Islands in the Cyp26a1 Promoter in Balb/c3T3 Cells-Methylation of DNA is another epigenetic mechanism associated with gene silencing (54). Recent studies have revealed methylation of pluripotency-associated genes in embryonic stem cells (ESCs) and embryonal carcinoma cells upon differentiation (55,56). Thus, we hypothesized that the Cyp26a1 and Hoxa1 genes could be silenced in Balb/c3T3 cells during the process of differentiation by virtue of promoter DNA methylation.
We utilized bisulfite modification of genomic DNA in conjunction with DNA sequencing to determine the methylation status of the CpG islands in the Cyp26a1 and Hoxa1 genes. Using an Emboss CpG plot and University of California Santa Cruz genome browser, we determined that the Cyp26a1 and Hoxa1 proximal promoter regions contain CpG islands. The genomic DNA from untreated F9 Wt and Balb/c3T3 cells was bisulfite-modified to convert cytosines to uracil (see details under "Materials and Methods"). The bisulfite-treated DNA was then used to PCR the Hoxa1 and Cyp26a1 PP regions spanning the CpG islands. The amplified PCR products were cloned and sequenced. Sequencing of at least 10 independent bisulfiteamplified clones showed that the CpG islands spanning the proximal promoter region and exon 1 in the Cyp26a1 gene are methylated in Balb/c3T3 cells but not in F9 Wt cells (Fig. 8a).
We also assessed the promoter methylation status of the Hoxa1 gene using primers spanning the 350-bp region upstream of the TSS (refer to Fig. 2a for schematic of gene structure). The CpG islands in the promoter region of the Hoxa1 gene were unmethylated in both F9 Wt and Balb/c3T3 cells (Fig. 8b). Thus, our data show that methylation of CpG islands in the Cyp26a1 promoter correlates with its lack of transcriptional activation in response to RA in Balb/c3T3 cells but that this is not the case for the Hoxa1 proximal promoter.

DISCUSSION
We have delineated the differences in RA-mediated transcriptional responses and underlying epigenetic status of Hoxa1, Cyp26a1, and RAR␤ 2 in F9 teratocarcinoma stem cells and fibroblasts. The findings of our study are depicted in a model (Fig. 9). RA signaling leads to the dissociation of PcG protein Suz12 from the RAREs of Hoxa1, Cyp26a1, and RAR␤ 2 genes and the Hoxa1 PP region in F9 Wt cells only. Retinoid receptors (RAR␥ and RXR␣), coactivators (pCIP and p300), and RNA Pol II are not recruited at the Hoxa1 RARE and Cyp26a1 RAREs in Balb/c3T3 cells. The CpG islands in the Cyp26a1 gene are methylated only in Balb/c3T3 cells, correlating with a lack of transcriptional response to RA. RAR␥ and RXR␣ are present at the RAR␤ 2 RARE in both F9 Wt and Balb/c3T3 cells in both the absence and presence of RA. The coactivators pCIP and p300 and RNA Pol II are recruited to the RAR␤ 2 RARE in Balb/c3T3 cells, similar to the situation in F9 Wt cells, and this corresponds to the transcriptional activation of RAR␤ 2 by RA in both cell types.
Retinoid Receptors (RAR␥ and RXR␣) Are Not Recruited at the Hoxa1 and Cyp26a1 RAREs in Balb/c3T3 Cells-We have shown that RAR␥ and RXR␣ do not associate with the Hoxa1 and Cyp26a1 RAREs in Balb/c3T3 cells even after the addition of RA (Fig. 2, a, d, and e). This is in contrast to the presence of retinoid receptors (RAR␥ and RXR␣) at the RAR␤ 2 RARE in Balb/c3T3 cells in the absence and presence of RA (Fig. 2c). Our data are in agreement with a recent study which showed that MEFs and mouse ESCs share only a small subset of RAR␥and RAR␣-bound genes. Although RAR␤ 2 was bound by RAR␥ in both MEFs and ESCs, the Hoxa and Hoxb clusters were differentially occupied by RARs in MEFs and ESCs (57). FIGURE 7. H3K27me3 levels at Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs in F9 Wt and Balb/c3T3 cells in response to RA treatment. F9 Wt and Balb/c3T3 cells were treated with RA for different times (0, 8, and 24 h) and fixed with 1% formaldehyde. Cells were lysed and sonicated to obtain soluble chromatin. Chromatin was precleared and immunoprecipitated with anti-H3K27me3 antibody, and immunoprecipitated DNA was quantitated using real time PCR. Experiments were repeated at least three times starting with the cell culture, and quantitative PCR was performed in triplicates. The data are averaged and plotted as a percentage of immunoprecipitated DNA, with reference to the percentage of input DNA at each locus. The error bars represent the S.E. over independent replicate experiments. The percentage of input immunoprecipitated in the same IP with nonspecific IgG is also graphed for comparison as a negative control. p values indicate significant changes after RA treatment within each cell line compared with the untreated controls. Note that at the Hoxa1, RAR␤ 2 , and Cyp26a1 RAREs, the levels of H3K27me3 are similar in Balb/c3T3 cells cultured in the presence and absence of RA and are statistically significantly higher than those in the IgG negative controls (p Ͻ 0.05).
Retinoid receptors and other nuclear hormone receptors regulate gene transcription by their ability to reversibly activate and repress genes in a ligand-dependent manner (58). Repression is reported to be mediated by the direct association of the receptor with the corepressors NCoR and SMRT (45,59). However, the absence of retinoid receptors (RAR␥ and RXR␣) at the Hoxa1 and Cyp26a1 RAREs in Balb/c3T3 cells cultured in the absence of RA rules out this nuclear corepressor-mediated transcriptional repression in this model system (Fig. 2, a, d, and  e). We suggest that the lack of receptor recruitment is a result of differential chromatin architecture spanning the Hoxa1 and Cyp26a1 genes in Balb/c3T3 cells. Moreover, the lack of association of retinoid receptors at these RAREs correlates with the lack of recruitment of the coactivator proteins pCIP and p300 in response to initiation of retinoid signaling in Balb/c3T3 cells (Fig. 3). Our data underscore the conclusion that receptor recruitment at the RARE is a key step in the process of transcriptional activation by RA.
PcG Proteins Differentially Regulate Retinoid-responsive Genes in F9 Stem Cells and Fibroblasts-We found that Suz12 failed to associate with any of the RAREs investigated in this study in Balb/ c3T3 cells cultured in the absence of RA. This is in contrast to the RA-mediated displacement of Suz12 in F9 Wt cells (Fig. 6). Interestingly, we discovered that even though Suz12 failed to bind to the RAREs investigated, the epigenetic mark H3K27me3 incorporated by the PRC2 complex (composed of core subunits, SUZ12, EZH2, and EED) was associated, albeit at lower levels, with the Hoxa1 and Cyp26a1 RAREs in untreated Balb/c3T3 cells. Moreover, it is noteworthy that the levels of this repressive modification remained unchanged after exposure of the Balb/c3T3 cells to RA (Fig. 7, a, d, and e). Our data are supported by a published study which showed that Suz12 and the H3K27me3 silencing mark are not associated with the early Hoxa and Hoxb cluster genes in human embryonic fibroblasts (29). The same study also showed that enrichment of H3K27me3 at the Hoxa1 promoter in MEFs was significantly lower than the enrichment observed in mouse ESCs.
Recruitment of Pol II at the Hoxa1 PP Region in Balb/c3T3 Cells despite the Repressed State of the Hoxa1 Gene-We found that in Balb/c3T3 cells, RNA Pol II was recruited to the Hoxa1 PP region even though Pol II failed to associate with the Hoxa1 RARE, which is located ϳ2 kb downstream of the 3Ј end of the Hoxa1 gene. Lack of Pol II recruitment at the Hoxa1 RARE is in agreement with the absence of retinoid receptors at this region and the repressed state of the Hoxa1 gene even in the presence of the ligand RA. However, the fact that Pol II was able to bind to the Hoxa1 PP region in Balb/c3T3 cells suggests that recruitment of Pol II at the proximal promoter region is independent of the association of retinoid receptors with the Hoxa1 RARE. In contrast, in F9 Wt cells Pol II levels increased steadily up to 24 h after RA addition at the Hoxa1 PP (Fig. 4b). Increasing levels of Pol II were associated with the Hoxa1 RARE in F9 Wt cells after RA induction (Fig. 4a). Together, our observations suggest the possibility of communication between the RARE and the proximal promoter region of the Hoxa1 gene after initiation of retinoid signaling in F9 Wt cells.
Additionally, it is noteworthy that the levels of histone modification (H3K9/K14 ac) usually associated with active transcription remained high at the Hoxa1 RARE and Hoxa1 PP in Balb/c3T3 cells after RA addition (Fig. 5, a and b). Thus, acetylation of H3K9/K14 ac alone is insufficient for the transcriptional activation of the Hoxa1 gene by RA. Our data indicate that additional events underlie activation of transcription. These events include coactivator (pCIP and p300) recruitment as early as 8 h after RA addition (Fig. 3) as well as chromatin remodeling activity (15,60). Our data also find support from a published study which reported that Hivep3, which is regulated by RA in mouse ESCs but remains silenced in MEFs, shows significant levels of H3K9 acetylation in MEFs (57).
Methylation of CpG Islands in the Cyp26a1 Promoter in Balb/ c3T3 Cells-CpG islands are CG-rich regions in the mammalian genome with unexpectedly higher frequencies of CG dinucleotides compared with the entire genome (61)(62)(63). DNA methylation by virtue of methylation of cytosines in the CpG islands is a well established epigenetic mechanism of gene silencing in physiological processes of X-inactivation (64) and genomic imprinting (65). Aberrant DNA methylation has been implicated in the silencing of several tumor suppressor genes in cancers (66,67). Several studies have reported hypermethylation as the cause of epigenetic silencing of RAR␤ 2 , a putative tumor suppressor, in several malignancies (68 -70). Reprogramming of DNA methylation patterns has been observed during fertilization and embryogenesis. Although the embryo undergoes a rapid erasure of genome-wide methylation during the preimplantation phase, the methylation pattern is re-established after implantation of the embryo (61,71,72). More recently, several groups have reported changes in the DNA methylation status of pluripotency genes upon differentiation (55,56,73,74). DNA methyltransferases (Dnmt 3a and Dnmt 3b) were shown to methylate the Oct3/4 and Nanog promoters synergistically in embryonal carcinoma and ES cells upon differentiation (75). Another global promoter DNA methylation mapping study showed that several proteins associated with pluripotency, such as Rex1 and Nanog, are characterized by demethylated promoters in ES cells. However, these genes are silenced and methylated in differentiated mouse fibroblasts (55). In addition, several studies have demonstrated that methylation of CpG islands contributes to the restricted expression of gene(s) in specific cellular lineages (73, 76 -78).
Our study demonstrates that CpG islands located in the promoter region of Cyp26a1 gene are differentially methylated in Balb/c3T3 cells versus F9 Wt cells (Fig. 8a). This is the first study demonstrating DNA methylation of the Cyp26a1 promoter in differentiated mouse fibroblasts (Balb/c3T3). DNA methylation of the Cyp26a1 promoter region in differentiated fibroblasts is in agreement with growing evidence for a regulatory role for DNA methylation during cellular differentiation. We were able to show that in Balb/c3T3 cells, the Cyp26a1 promoter region is associated with minimal levels of the histone acetylation mark (H3K9,K14ac), a mark usually associated with transcriptionally active chromatin (Fig. 5, d and e). It is known that one of the mechanisms by which DNA methylation can mediate repression is via the recruitment of histone deacetylase complexes by methyl DNA-binding proteins (79 -81).
The order of these converging epigenetic events cannot be determined from this study. However, our data suggest that several layers of repression, including CpG methylation (Fig.  8a), deacetylated histones (Fig. 5, d and e), and PcG-mediated H3K27me3 modification (Fig. 7, d and e), contribute to the silenced state of the Cyp26a1 gene in a more differentiated cell type (Balb/c3T3 fibroblasts). This is similar to the silencing of the pluripotency gene Oct4, which undergoes PcG mediated repression, DNA methylation, and H3K9 methylation (a mark of constitutive heterochromatin) upon cell differentiation (75,82,83).
Hoxa1 Remains Transcriptionally Inactive in Response to RA in Fibroblasts-The lack of transcriptional response of the Hoxa1 gene in Balb/c3T3 cells correlates with the presence of moderate and unchanged levels of the repressive H3K27me3 modification in the absence and presence of RA at the Hoxa1 RARE. DNA methylation is generally considered to be a more stable mechanism of repression, and the lack of CpG island methylation at the Hoxa1 PP region is indicative of the plasticity at this locus. Fibroblasts are of mesoderm origin and possess a more limited differentiation potential compared with F9 stem cells. Hoxa1 is a retinoid-responsive gene important during early embryonic development and in stem cell differentiation in adults (84 -87). It is also possible that the Balb/c3T3 cells lack factors known to act in concert with RARs to transcriptionally activate the Hoxa1 gene in mesenchymal tissue (88).
Thus, in this study we show that RA differentially regulates the transcription of Hoxa1, Cyp26a1, and RAR␤ 2 in undifferentiated F9 stem cells and differentiated fibroblast cells. Our data show that MEFs and Balb/c3T3 cells fail to express Hoxa1 and Cyp26a1 transcripts upon RA addition. We suggest that the lack of transcriptional activation of Hoxa1 and Cyp26a1 in FIGURE 9. Model for differential transcriptional regulation of three RA regulated genes in F9 Wt and Balb/c3T3 cells. Upon initiation of RA signaling, the PcG protein Suz12 is displaced from the RAREs of Hoxa1, Cyp26a1, and RAR␤ 2 genes and the Hoxa1 PP region in F9 Wt cells. However, low levels of the PcG-mediated H3K27me3 modification are associated with the Cyp26a1 and Hoxa1 RAREs in Balb/c3T3 cells (not depicted). There is no association of RAR␥ and RXR␣, coactivators, and RNA Pol II at the Hoxa1 RARE and Cyp26a1 RAREs in Balb/c3T3 cells, in contrast to the situation in F9 Wt cells. This is consistent with the RA-mediated transcriptional activation of Hoxa1 and Cyp26a1 in F9 Wt cells but not in Balb/c3T3 cells. The CpG islands in the Cyp26a1 gene are methylated only in Balb/c3T3 cells, correlating with the lack of transcriptional response. RAR␥ and RXR␣ are bound at the RAR␤ 2 RARE in both F9 Wt and Balb/c3T3 cells in the absence and presence of RA. The coactivators pCIP and p300 and RNA Pol II are recruited to the RAR␤ 2 RARE in Balb/c3T3 cells, similar to the case in F9 Wt cells, and this corresponds to transcriptional activation by RA in both cell types. response to RA is a result of the differential chromatin signatures at these genes in Balb/c3T3 cells, as we observed different epigenetic modifications associated with the Hoxa1 and Cyp26a1 RAREs in F9 Wt cells. Although it was known that RAR␤ 2 is transcriptionally activated upon RA addition in Balb/ c3T3 cells (44), we are the first to investigate the transcriptional response and regulation of the Hoxa1 and Cyp26a1 genes in differentiated mouse fibroblasts. More importantly, our research demonstrates that differences in chromatin modifications modulate the binding of RARs, thus contributing to cell-specific RA-mediated transcriptional regulation.