Retinoic Acid and Histone Deacetylases Regulate Epigenetic Changes in Embryonic Stem Cells

Background: Retinoic acid (RA) and RA receptors (RARs) are major regulators of transcription in stem cells.

Results: Histone deacetylase (HDAC) 1, 2, and 3 bind different RAR target gene regulatory regions blocking histone acetylation, whereas RA treatment removes HDACs and increases histone acetylation levels.

Conclusion: HDACs differentially repress RA target genes.

Significance: Identification of new repression mechanisms for RA target genes.

Abstract

All-trans-retinoic acid (RA) is a vitamin A metabolite that plays major roles in regulating stem cell differentiation and development. RA is the ligand of the retinoic acid receptor (RAR) family of transcription factors, which interact with retinoic acid response elements (RAREs) within target gene proximal promoters and enhancers. Although RA-mediated gene activation is well understood, less is known about the mechanisms for repression at RA-regulated genes. Using chromatin immunoprecipitation experiments, we show that in embryonic stem cells in the absence of RA, histone deacetylases (HDACs) differentially bind to various RAREs in proximal promoters or enhancer regions of RA-regulated genes; HDAC1, HDAC2, and HDAC3 bind at RAREs in the Hoxa1 and Cyp26a1 gene regulatory regions, whereas only HDAC1 binds at the RARβ2 RARE. shRNA knockdown of HDAC1, HDAC2, or HDAC3 differentially increases the deposition of the histone 3 lysine 27 acetylation (H3K27ac) epigenetic mark associated with increases in these three transcripts. Importantly, RA treatment differentially mediates the removal of HDACs from the Hoxa1, Cyp26a1, and RARβ2 genes and promotes the deposition of the H3K27ac mark at these genes. Overall, we show that HDACs differentially bind to RA-regulated genes to control key epigenetic marks involved in stem cell differentiation.

Introduction

Epigenetic modification of the genome is a major regulator of embryonic stem cell (ESC)² differentiation and development. Stem cells have the ability to self-renew and to differentiate along multiple cell lineages, and epigenetic regulation of gene expression is essential for proper stem cell maintenance and differentiation (1, 2). Examination of these specific epigenetic changes and their modulators has unraveled a complex circuitry that finely regulates stem cell differentiation.

Genes are transcriptionally regulated in part at promoter regions where proteins involved in transcription, including RNA polymerase II, bind to initiate transcription. However, genes are also regulated by cis-regulatory elements termed enhancers, located upstream or downstream of transcription initiation sites (3–6). Epigenetic regulation at these enhancer regions is required for gene expression (5, 6). At both promoter and enhancer regions, histone tails are targets for extensive post-transcriptional modifications, such as acetylation and methylation, that regulate many aspects of gene transcription (7). Methylation of histone tails, especially trimethylation of histone 3 lysine 27 (H3K27me3) and histone 3 lysine 9 (H3K9me3), is generally associated with gene repression. Conversely, methylation of H3K4 generally marks active enhancer regions and acetylation of H3K4, H3K9, and H3K27 is highly correlated with transcriptional activation (8–10). The presence of the H3K27ac mark distinguishes active enhancers as compared with poised enhancers in ESCs, suggesting that enhancers with the H3K27ac marks are a better indicator of genes that play major roles during different stages of development (3, 11). Removal of histone methylation and acetylation is mediated by demethylases (KDMs/PRDMs) and histone deacetylases (HDACs), respectively (12).

All-trans-retinoic acid (RA) is a vitamin A metabolite that plays a major role in regulating ESC differentiation by acting as the ligand for the retinoic acid receptor (RAR) family of transcription factors (13). The RARα, β, and γ receptors bind retinoic acid response elements (RAREs) within target gene promoters as heterodimers with the retinoid X receptor family, and we have shown that RARγ specifically interacts with a subgroup of RAREs that are located at either the promoters or enhancers of many target genes in ESCs (14, 15). RARs regulate the deposition of both activating and repressive epigenetic marks on histones at target gene promoters. We have shown that upon RA exposure at RAR-regulated genes, the H3K27me3 repressive marks decrease, whereas the H3K4/9ac activation marks increase, allowing for transcriptional activation of target gene expression (14, 15, 17, 18). Although our lab and others have performed extensive studies focused on understanding how RARs recruit co-activators to promote gene expression, the co-repressors that interact with RARs and how RA regulates them are not as well understood.

HDACs are typically present within larger, co-repressor multiprotein complexes, where they mediate the removal of an acetyl group from lysines on histone tails (19). There are several subgroups of HDACs, and HDAC1, HDAC2, and HDAC3 are all members of the class I subfamily based upon their homology with the founding member in budding yeast, Rpd3 (12, 20). HDAC3 is a deacetylase present in complexes that contain co-repressors that have been reported to interact with the RARs, the nuclear receptor co-repressor (NCoR1) and the silencing mediator of retinoic acid and thyroid hormone receptor (SMRT, also known as NCoR2) (19). HDAC1 and HDAC2 are deacetylase components of the nucleosome remodeling and deacetylation (NuRD), CoREST, and Sin3 co-repressor complexes (21).

To date, the specific interactions of various HDACs and their effects on histone acetylation at different RAR target gene promoters have not been examined. Additionally, whether RARs and their interacting co-activators and co-repressors can regulate H3K27ac levels at RA-regulated enhancer regions is unknown.

Here we determined which HDACs bind to RA-regulated genes and the role of RA in regulating HDAC interactions. We also examined HDAC regulation of H3K27ac mark deposition at known RA target gene promoters and enhancers and found that different HDACs bind and regulate histone acetylation in a gene-specific manner. Overall, these studies highlight the complex regulation of RA-mediated gene transcription during stem cell differentiation.

EXPERIMENTAL PROCEDURES

Chemicals and Antibodies

Cells were treated with a final concentration of 1 μm all-trans-retinoic acid (Sigma), which was dissolved in 100% ethanol vehicle under dim light and kept at 4 °C (the final concentration of ethanol was <0.1%). The HDAC1 specific inhibitor JNJ-26481585 was purchased from Selleck Chemicals (S1096, Houston, TX), and the HDAC3 specific inhibitor pimelic diphenylamide 106 was a kind gift from Dr. Joel Gottesfeld (26, 27). Both inhibitors were dissolved in 100% DMSO to a final concentration of 0.11 or 0.22 nm (for JNJ-26481585), and 1, 2, or 5 μm (for 106 inhibitor) as indicated. These concentrations were optimized to achieve histone acetylation. HDAC1 (06-720; Millipore, Billerica, MA), HDAC2 (H-54; Santa Cruz Biotechnology, Santa Cruz, CA) (for Westerns), HDAC2 (7029; Abcam, Cambridge, MA) (for ChIP), HDAC3 (Ab7030; Abcam), H3K27ac (Ab4729, Abcam, Cambridge, MA), H3K4/19ac (06-599; Millipore), and actin (MAB1501; Millipore) antibodies were used.

Derivation and Culture of the Murine ESC Lines and Lentiviral shRNA Infection

CCE WT ESCs were cultured as previously described (28). HEK293T cells were transfected with 2 μg of PLKO.1 TRC cloning vectors with shRNA plasmids directed specifically toward murine HDAC1 (TRCN0000039402; Sigma), HDAC2 (TRCN00000393950; Sigma), or HDAC3 (TRCN00000393890; Sigma), 1 μg of pVSV-G envelope plasmid, and 1 μg of pCMVΔR8.9 packaging plasmid using Lipofectamine 2000 (Invitrogen). Lentiviral particles were collected from the supernatant. CCE WT ESCs were infected with shRNA lentiviral particles that were treated with polybrene at a 4 μg/ml final concentration. After 24 h, cells were selected with 1 μg/ml puromycin for 7 days. Colonies derived from single cells were picked and treated with 1 μg/ml puromycin for an additional 3 days. Cells lines were screened for efficient knockdown by Western blotting.

Chromatin Immunoprecipitation (ChIP)

Experiments were performed as previously described (28, 29). Briefly, 2.5 × 10⁶ CCE WT ESCs were cultured with or without 1 μm RA and/or 106 inhibitor for 1, 8, or 24 h as indicated. Sonicated and precleared lysates containing 15–45 μg of DNA were immunoprecipitated using 0.5–1.0 μg of antibodies specific for HDAC1, HDAC2, HDAC3, H3K27ac, or IgG (negative control). 3 μl of purified DNA was used for qPCR analysis using primers specific for designated gene promoter and enhancer regions. 3 μg of input DNA was used to normalize immunoprecipitated DNA, and all values were normalized to the IgG negative control, which in each experimental plot was set to 1.

Western Blotting

CCE WT and shRNA knockdown ESCs were harvested in SDS lysis buffer, boiled, and resolved on SDS-PAGE gels, followed by Western blotting using antibodies specific for HDAC1 (1:2000), HDAC2 (1:1000), HDAC3 (1:10000), H3K27ac (1:5000), H3K4/19ac (1:5000), or actin (1:10000).

RNA Isolation and Reverse Transcription

RNA was extracted using TRIzol (Invitrogen) according to the manufacturer's instructions. Quantified RNA was reverse transcribed using the qScript cDNA synthesis kit (Quanta Biosciences, Gaithersburg, MD).

Real Time qPCR Analysis

Reactions were carried out using SYBR Green quantitative PCR master mix and a Bio-Rad iCycler. Primers specific for each proximal promoter (pp), RARE, or gene are shown (Table 1). For mRNA quantitation, 36B4 was used as an internal control mRNA to which all values were normalized.

TABLE 1.

Primer sequences used for qPCR and predicted product size

Target	Forward (5′ → 3′)	Reverse (5′ → 3′)	Product size
			bp
Hoxa1 RARE	TCTTGCTGTGACTGTGAAGTCG	GAGCTCAGATAAACTGCTGGGACT	268
Hoxa1 PP	ATTGGCTGGTAGAGTCACGTG	GAAAGTTGTAATCCCATGGTCAGA	276
Cyp26a1 RARE1	CCCGATCCGCAATTAAAGATGA	CTTTATAAGGCCGCCCAGGTTAC	87
Cyp26a1 RARE2	TTCACTGAGATGTCACGGTCC	TTCCCAATCCTTTAGCCTGA	64
RARβ RARE	TGGCATTGTTTGCACGCTGA	CCCCCCTTTGGCAAAGAATAGA	284
Hoxa1 mRNA	TTCCCACTCGAGTTGTGGTCCAAGC	TTCTCCAGCTCTGTGAGCTGCTTGGTGG	220
Cyp26a1 mRNA	GAAACATTGCAGATGGTGCTTCAG	CGGCTGAAGGCCTGCATAATCAC	272
RARβ mRNA	GATCCTGGATTTCTACACCG	CACTGACGCCATAGTGGTA	247
36B4 mRNA	AGAACAACCCAGCTCTGGAGAAA	ACACCCTCCAGAAAGCGAGAGT	448

Statistical Analysis

Statistical analysis was performed on at least three separate, independent experiments (n = 3 or >3) using the Graph Pad Prism 6.0 software. The means ± S.E. were determined, and Student's t test was used to compare two independent populations, where p < 0.05 was considered statistically significant.

RESULTS

HDACs Differentially Bind to the Promoter and Enhancer Regions of RA-inducible Genes in the Absence of RA

Because HDAC1, HDAC2, and HDAC3 play major roles in embryonic development, we first addressed their roles at RA-regulated genes in murine ESCs. We asked which HDACs interact at the enhancer and/or pp regions of the RA-regulated genes, including Hoxa1, Cyp26a1, and RARβ2, in ESCs cultured without RA. A schematic of the gene promoter and enhancer regions is shown in Fig. 1A. The Hoxa1 gene has an enhancer region with an RARE located ∼4.6 kb downstream of the Hoxa1 pp (30). The Cyp26a1 gene possesses two RAREs, one in the pp and a second in an enhancer ∼2 kb upstream from the Cyp26a1 pp (31, 32). Finally, the RARβ2 gene possesses an RARE within its pp (33).

FIGURE 1.

Differential binding of HDACs to regulatory regions of RA-regulated genes in WT ESCs cultured without RA. A, schematic representation of gene structures of RA-regulated genes (from Ref. 17). B–D, lysates were collected for ChIP experiments using antibodies specific for HDAC1 (B), HDAC2 (C), and HDAC3 (D). qPCR was then used to quantitate HDAC association with RAREs and pp regions. The respective regions of each gene are depicted below the graphs. Immunoprecipitations (IP) using IgG were included as negative controls. Binding is expressed relative to the IgG negative control, with the IgG data for experiments set to 1 after normalizing to pre-IP input DNA. Error bars represent standard errors of independent experiments where n = 3 for biological repeats. *, p < 0.05.

We performed ChIP experiments using antibodies specific for HDAC1, HDAC2, or HDAC3 to assess HDAC binding. By qPCR, we detected HDAC1 binding at the Hoxa1 RARE, Cyp26a1 pp/RARE1, Cyp26a1 RARE2, and the RARβ2 pp/RARE (Fig. 1B). HDAC2 also bound at the Hoxa1 RARE, Cyp26a1 pp/RARE1, and Cyp26a1 RARE2, but not at the RARβ2 pp/RARE (Fig. 1C). In contrast, we found that HDAC3 only bound at the Hoxa1 RARE and Cyp26a1 RARE2 regions (Fig. 1D). These results demonstrate that the binding of the three HDACs is not uniform at the three genes we analyzed. Furthermore, although only HDAC3 has been reported to interact with the RARs as the HDAC component of the NCoR/SMRT complex (34), we detected HDAC1 rather than HDAC3 at the RARβ2 pp/RARE and Cyp26a1 pp/RARE1 (Fig. 1B). HDAC2 and HDAC3 were not bound at the RARβ2 pp/RARE (Fig. 1, C and D).

HDACs Differentially Regulate the Deposition of the H3K27ac Mark in the Absence of RA

The H3K27ac epigenetic mark identifies active enhancers in ESCs (3). We next determined whether HDAC1, HDAC2, or HDAC3 specifically regulated H3K27ac deposition by using shRNAs specific for each HDAC to generate ESC lines with stable knockdowns (k.d.) of HDAC1, HDAC2, or HDAC3 (Fig. 2, a–c). As previously reported (35), we observed a compensatory increase in HDAC2 and HDAC1 proteins in HDAC1 k.d. or HDAC2 k.d, respectively (Fig. 2, a and b). We found that global H3K27 and H3K9/14 acetylation increased in all HDAC knockdown lines (Fig. 2, d and e), demonstrating that all three HDACs participate in regulating histone acetylation in ESCs.

FIGURE 2.

Stable knockdown of HDACs in ESCs. HDAC1, HDAC2, or HDAC3 were stably knocked down in ESCs using lentiviral delivery of specific shRNAs. Western blotting was performed to analyze the knockdowns (a–c). Changes in global histone acetylation were determined by assessing H3K9/14ac and H3K27ac using specific antibodies (d and e). Actin served as a loading control (f). A representative blot is shown where n = 3 for biological repeats. IB, immunoblot.

We next performed ChIP experiments in the HDAC k.d. cell lines using an antibody for H3K27ac. If a specific HDAC prevented the deposition of the H3K27ac mark, we would expect to observe an increase in H3K27ac levels when that HDAC was knocked down. First, we detected no significant increase in H3K27ac levels at the Hoxa1 pp (Fig. 3aA), which is expected because we showed that there are no HDACs at this pp region (Fig. 1, B–D). Second, we observed an ∼3-fold increase in H3K27ac level at the Hoxa1 RARE when HDAC1, HDAC2, or HDAC3 was knocked down as compared with WT or “empty vector” (E.V.) cells (Fig. 3aB). This result is in line with our findings that HDAC1, HDAC2, and HDAC3 all bind to the region with the Hoxa1 RARE (Fig. 1, B–D). Additionally, knockdown of HDAC1, HDAC2, or HDAC3 caused a 2.5-fold increase in Hoxa1 mRNA levels (Fig. 3bA).

FIGURE 3.

HDACs differentially regulate H3K27ac levels at RA-regulated genes. a, ChIP experiments were performed using an anti-H3K27ac antibody in WT (Parental), empty vector (E.V.), and various shRNA HDAC knockdown cell lines. qPCR analysis was used to determine H3K27ac levels at gene regulatory regions. A, Hoxa1 pp. B, Hoxa1 RARE. C, Cyp26a1 pp/RARE1. D, Cyp26a1 RARE2. E, RARβ2 pp/RARE. Binding is expressed relative to the IgG negative control set to 1 after normalizing to pre-IP input DNA. In each panel, IP: IgG represents the negative control ChIP performed with a nonspecific IgG antibody. b and c, RNA was reverse transcribed and qRT-PCR analysis was used to determine mRNA levels compared with 36B4 control mRNA levels. b, A, Hoxa1. B, Cyp26a1; C, RARβ2. c, A, Wnt3a (Hoxa1 regulated gene). B, Spry2 (RARβ regulated gene). Error bars represent standard error of independent replicate experiments where n = 4 for biological repeats. *, p < 0.05.

We detected no significant changes in H3K27ac levels at the Cyp26a1 pp/RARE1 in the HDAC k.d. lines (Fig. 3aC), but knockdown of HDAC1 or HDAC3 was associated with a 4-fold increase, and knockdown of HDAC2 was associated with a 2-fold increase in H3K27ac at the Cyp26a1 RARE2 as compared with the parental WT cells (Fig. 3aD). Consistent with this result, knockdown of HDAC1, HDAC2, and HDAC3 resulted in 4-fold increases in Cyp26a1 mRNA levels (Fig. 3bB). Thus, HDAC1, HDAC2, and HDAC3 all bind near the Cyp26a1 RARE2 and negatively regulate Cyp26a1 mRNA levels in ESCs.

We detected a 3-fold increase in H3K27ac mark deposition at the RARβ2 pp/RARE in the HDAC1 k.d. cells (Fig. 3aE) and an increase in RARβ2 mRNA level (Fig. 3bC). Strikingly, neither HDAC2 nor HDAC3 affected the deposition of the H3K27ac mark at the RARβ2 pp/RARE (Fig. 3aE).

We also assessed the expression of genes important for proper ESC differentiation that are regulated by Hoxa1 or RARβ2. Wnt3a is a major regulator of the developing nervous system and is essential for many facets of neuronal differentiation (36). Our lab has shown that Wnt3a is regulated by Hoxa1, because Wnt3a expression is greatly reduced in Hoxa1^−/− ESCs (37). Interestingly, we found that Wnt3a transcripts increased in HDAC1, HDAC2, and HDAC3 k.d. cells as compared with control (Fig. 3cA). This is consistent with our finding that HDAC1, HDAC2, or HDAC3 k.d. increased Hoxa1 mRNA levels (Fig. 3bA). The Sprouty family of genes are regulators of neuronal and endothelial cell differentiation (38), and we observed a 2-fold decrease in Spry2 expression in RARβ2^−/− ESCs.³ HDAC1 k.d. induces ∼6-fold increase in Spry2 mRNA levels (Fig. 3cB), which is again in line with our finding that RARβ2 transcriptional activation is primarily regulated by HDAC1 (Fig. 3bC).

Overall, these experiments show that different HDACs regulate the deposition of the H3K27ac epigenetic mark at different genes. HDAC1, HDAC2, and HDAC3 have roles at the Hoxa1 RARE and Cyp26a1 RARE2, whereas only HDAC1 plays a role at the RARβ2 pp/RARE. These findings have implications for proper stem cell differentiation because downstream targets of Hoxa1 and RARβ2 that are important for specific lineage differentiation are also differentially affected by specific HDAC knockdown.

Pharmacological Inhibition of HDAC1 or HDAC3 Regulates H3K27ac Deposition in the Absence of RA

We treated WT ESCs with the HDAC1 specific inhibitor JNJ-26481585 (39) at various concentrations and performed ChIP experiments to assess the levels of the H3K27ac mark. We found that JNJ-26481585 treatment did not cause changes in the deposition of the H3K27ac at the Hoxa1 pp (Fig. 4aA) but caused significant increases in H3K27ac levels at the Hoxa1 RARE, Cyp26a1 pp/RARE1, Cyp26a1 RARE2, and the RARβ2 pp/RARE (Fig. 4a, B–E). This is in line with our finding that HDAC1 does not bind to the Hoxa1 pp but interacts at the other gene regulatory regions to regulate H3K27ac levels (Fig. 1B). When either HDAC1 or HDAC2 was knocked down, we did not observe a significant increase in the H3K27ac levels at the Cyp26a1 pp/RARE1 (Fig. 3aC). Our result using the JNJ-26481585 inhibitor suggests that the increase in HDAC2 protein level in HDAC1 k.d. cells (Fig. 2a) can partially compensate for HDAC1 at this promoter region.

FIGURE 4.

Pharmacological inhibition of HDAC1 or HDAC3 promotes H3K27ac deposition in WT ESCs in the absence of RA. WT ESCs were treated with varying concentrations of the HDAC1 specific inhibitor JNJ-26481585 (a) or the HDAC3 specific inhibitor 106 (b) for 24 h. Lysates were collected for ChIP experiments, and a H3K27ac antibody was used for immunoprecipitation (IP). qPCR was performed using primers specific for Hoxa1 pp (A), Hoxa1 RARE (B), Cyp26a1 pp/RARE1 (C), Cyp26a1 RARE2 (D), or RARβ2 pp/RARE (E). ChIP with an IgG antibody (negative control) is included in each panel. The level of H3K27ac represents immunoprecipitated DNA relative to the IgG negative control set to 1 after normalizing to pre-IP input DNA. Error bars represent standard error of independent experiments where n = 3 for biological repeats. *, p < 0.05.

Similarly, we treated WT ESCs with the HDAC3 specific inhibitor pimelic diphenylamide 106 (26, 27) and performed ChIP experiments. We found that 106 treatment did not change H3K27ac levels at the Hoxa1 pp (Fig. 4bA), but 106 inhibitor increased the levels of H3K27ac by 3-fold compared with the untreated cells, at the Hoxa1 RARE (Fig. 4bB). Similarly, 106 treatment did not cause changes in the H3K27ac mark at the Cyp26a1 pp/RARE1 (Fig. 4bC) but resulted in increased levels of H3K27ac near the Cyp26a1 RARE2 (Fig. 4bD). In contrast, we did not observe any change in the level of the H3K27ac mark at the RARβ2 pp/RARE after 106 treatment (Fig. 4bE). These results demonstrate that H3K27ac is a substrate of HDAC3 and that HDAC3 levels are inversely associated with H3K27ac levels at the Hoxa1 RARE and Cyp26a1 RARE2, regions where we showed binding of HDAC3 (Fig. 1D).

RA Differentially Regulates HDAC Removal

We next hypothesized that the addition of RA would result in the removal of HDACs from these regulatory regions. We again performed ChIP experiments using HDAC antibodies in extracts from WT ESCs cultured in the presence or absence of RA. We found that RA had no effect on HDAC1 binding at the Hoxa1 RARE and the Cyp26a1 pp/RARE1, but the addition of RA caused the loss of HDAC1 at the Cyp26a1 RARE2 and the RARβ2 pp/RARE (Fig. 5A). Interestingly, whereas HDAC2 bound to the Hoxa1 RARE, Cyp26a1 pp/RARE1, and Cyp26a1 RARE2, RA had no effect on its binding (Fig. 5B). We also showed that RA caused a loss of HDAC3 binding at the Hoxa1 RARE and Cyp26a1 RARE2 (Fig. 5C). These results demonstrate that RA regulates the binding of HDAC1 and HDAC3 but not HDAC2.

FIGURE 5.

RA differentially regulates HDAC binding. WT ESCs were cultured without or with 1 μm RA for 24 h. Lysates were collected for ChIP experiments. Immunoprecipitation (IP) was performed using antibodies specific for HDAC1 (A), HDAC2 (B), or HDAC3 (C), followed by qPCR to quantitate HDAC association at the promoter and enhancer regions of RA-regulated genes. The respective regions of each gene are identified below the graph. Untreated cells and immunoprecipitation using IgG were included as negative controls. Binding is expressed relative to the IgG negative control with IgG set to 1 after normalizing to pre-IP input DNA. Error bars represent standard error of independent replicate experiments where n = 3 for biological repeats. *, p < 0.05.

RA Increases H3K27ac Levels at Both Proximal Promoters and RAREs

Because H3K27ac was recently identified as a specific mark of active enhancers (3), we next determined whether RA could affect the deposition of the H3K27ac mark at the RAREs (Fig. 1A). We found that in untreated WT ESCs, the H3K27ac mark was present at all pp regions and RAREs compared with the IgG negative controls (Fig. 6a, A–E). After RA treatment, the increase in the H3K27ac mark at the Hoxa1 pp was ∼3-fold (Fig. 6aA), whereas it was ∼5-fold at the Hoxa1 RARE (Fig. 6aB). RA resulted in an 8-fold increase in the H3K27ac level at the Cyp26a1 pp/RARE1 (Fig. 6aC) and a 7-fold increase at the Cyp26a1 RARE2 (Fig. 6aD). At the RARβ2 pp/RARE, increases were ∼10-fold (Fig. 6aE). Additionally, RA treatment caused 40-, 200-, and 10-fold increases in Hoxa1, Cyp26a1, and RARβ2 mRNA levels, respectively (Fig. 6b, A–C). Overall, these data demonstrate that RA regulates changes in the H3K27ac mark, thus identifying another mechanism by which RA regulates ESC differentiation.

FIGURE 6.

RA increases H3K27ac levels at RAREs in WT ESCs. a, ChIP experiments were performed at various time points (1, 8, and 24 h) with or without treatment with 1 μm RA using an antibody specific for H3K27ac or IgG (negative control). qPCR was performed using primers specific for Hoxa1 pp (A), Hoxa1 RARE (B), Cyp26a1 pp/RARE1 (C), Cyp26a1 RARE2 (D), or RARβ2 pp/RARE (E). Fold change represents immunoprecipitated DNA relative to the IgG negative control set to 1 after normalized to pre-IP input DNA. b, RT-PCR analysis was used to determine mRNA levels compared with 36B4 internal control mRNA levels of Hoxa1 (A), Cyp26a1 (B), and RARβ2 (C). Error bars represent standard error of independent experiments where n = 4 for biological repeats. *, p < 0.05 comparing RA treatment with untreated control samples.

DISCUSSION

Although prior publications (24, 25) have shown that RARs can interact with HDAC3, we have found that different HDACs bind to the promoter and enhancer regions of RA regulated genes. We show here that in addition to HDAC3, HDAC1 and HDAC2 play major roles in regulating transcriptional repression of RA-regulated genes (Figs. 1 and 3). Importantly, we have found that the RARβ2 pp/RARE is regulated primarily by HDAC1 (Fig. 1B). We also show that in the absence of RA, k.d. of HDAC1, HDAC2, or HDAC3 causes a specific increase in the H3K27ac mark at RAREs, proving that H3K27 is a substrate for HDAC1, HDAC2, and/or HDAC3 deacetylation, depending on the gene (Fig. 3a, A–E). Furthermore, we have found that RA regulates HDAC binding in a gene-specific manner. A summary of these findings is depicted in a model (Fig. 7).

FIGURE 7.

Model for HDAC regulation of RA-regulated genes. A, there is no HDAC binding at the Hoxa1 proximal promoter, but in the absence of RA, HDAC1, HDAC2, and HDAC3 are bound near/at the Hoxa1 RARE. At the Hoxa1 RARE HDAC1, HDAC2, and HDAC3 inhibit H3K27ac, but upon RA exposure only HDAC3 is removed. B, in the absence of RA, HDAC1 and HDAC2 are bound to the Cyp26a1 pp/RARE1, and HDAC1, HDAC2, and HDAC3 are bound at the Cyp26a1 RARE2. Upon RA treatment, HDAC1 and HDAC2 remain bound at the Cyp26a1 pp/RARE1, whereas HDAC3 is removed from the Cyp26a1 RARE2. C, HDAC1 binds to the RARβ2 pp/RARE where it inhibits H3K27ac deposition and RARβ2 transcription, and RA treatment mediates the removal of HDAC1.

RAREs are comprised of two direct repeats (DR) of the AGGTCA sequence with 5 base pairs (DR5) or two base pairs (DR2) between the repeats. Both the Hoxa1 and RARβ2 genes possess identical DR5 RAREs (30, 33). The fact that HDAC3 is recruited to the Hoxa1 RARE but not to the RARβ2 RARE is extremely interesting, because it suggests that despite similar binding patterns of the RARs and co-activators to these RAREs, other interacting transcriptional regulatory proteins at these two RAREs are different. A previous study found that RARs interact with 462 target loci in ESCs, whereas only 47 interact with the classical DR5 repeat, and only a subset of these are responsive to RA (40). Comparing these genes with those reported in a study that examined binding of HDAC1 in ESCs (41), we determined that only 7 out of 47 RAR-bound genes also interact with HDAC1. This would further suggest that the circuitry of epigenetic gene regulation that allows for precise control of stem cell differentiation is quite complex.

HDAC1, HDAC2, and HDAC3 are all class I HDACs and can sometimes act redundantly (12, 42–44). However, they have been shown to be functionally different, particularly during development (12, 35, 45). We found that HDAC1, HDAC2, and HDAC3 exert different functions in terms of regulating H3K27ac levels on different genes (Fig. 3a, A–E) and are differentially regulated by RA in the context of different genes (Fig. 5, A–C). These results have two major implications. Our data indicate first that these HDACs are functionally different in terms of RA-regulated ESC differentiation and second that RARs may interact with various co-repressor complexes that contain these different HDACs.

Despite the fact that HDAC1 and HDAC2 are nearly identical in structure and sequence, our finding that they are functionally different in terms of RA- and RAR-regulated transcription is in line with other studies showing that HDAC1 and HDAC2 have different functions. Although germ line deletion of HDAC1 is early embryonic lethal, HDAC2 germ line deletions result in viable mice, but with low viability after birth (35, 42, 46–48). Using conditional knockouts of HDAC1 or HDAC2 in ESCs, Dovey et al. (35) showed that HDAC1, not HDAC2, is the major deacetylase that regulates both deposition of the H3K56ac mark and differentiation along specific lineages (12). Similarly, HDAC1 has different functions compared with HDAC3 during development (49). Maroni et al. (49) found that knockdown of HDAC1, and not HDAC3, caused an increase in expression of the Runx2 gene and increased alkaline phosphatase activity, indicative of early osteoblast maturation. Conversely, HDAC3, and not HDAC1, increased late osteogenic markers, including calcium and collagen deposition (49).

To date, RARs have been shown to interact with the NCoR/SMRT (NCoR1/NCoR2) co-repressor complex, and the HDAC associated with this complex is HDAC3 (24, 25). We detected HDAC1 and HDAC2 binding at various RAREs (Fig. 1, B and C), suggesting that RARs also interact with other co-repressor complexes such as NuRD, Sin3, and/or REST/CoREST. The NuRD co-repressor complex is comprised of multiple proteins that mediate repression, including HDAC1 and HDAC2, chromatin remodelers, and histone KDMs (19). Similarly, HDAC1 and HDAC2 are the deacetylases present in the Sin3 and REST/CoREST co-repressor complexes, and like the NuRD complex, the Sin3 and REST/CoREST complexes are essential for proper development (21, 50). That RARs interact with other co-repressor complexes in addition to NCoR/SMRT is supported by a study that showed that the NuRD complex scaffolding protein MBD3 facilitates PML-RARα fusion protein-mediated epigenetic changes in acute promyelocytic leukemia cells (51). Furthermore, the interaction between MBD3 and PML-RARα at the RARβ2 promoter was lost after RA treatment (51). This result suggests that wild-type RARs may interact with the NuRD complex in other cell types, including ESCs, and that identification of the co-repressor complexes that interact with RARs in stem cells will significantly add to our understanding of how RARs regulate stem cell differentiation.

It should also be noted that in addition to mediating histone deacetylation, HDACs may perform other important roles at RA-regulated genes. For example, HDACs can mediate the deacetylation of other substrates in addition to histone tails, and it has been shown that co-activators such as p300 and CBP are catalytically inactive unless they are acetylated (52, 53). Alternatively, HDACs may be essential for the formation and maintenance of co-repressor complexes, where they could serve as scaffolding proteins (54). A recent study has shown that mutations in HDAC3 that abolish its catalytic activity can increase gene expression of target genes but do not increase histone acetylation at the target gene promoters (55). Instead, ablation of HDAC3 interaction with the NCoR complex was important for HDAC3 function (55), suggesting that in some instances, HDAC3 plays an important role as a scaffolding protein. Therefore, in our system, in addition to histone deacetylation, HDAC k.d. may affect the activities of other transcriptional regulators or the formation of co-repressor complexes.

Our lab has previously shown that RA can mediate the removal of the polycomb repressive complex 2 component, Suz12, which is responsible for deposition of the H3K27me3 mark and inhibition of gene transcription (14–18, 29). Others have shown that the NuRD complex mediates deacetylation of H3K27, thereby allowing recruitment of polycomb repressive complex 2 and deposition of the H3K27me3 mark. This two-step process mediates gene repression (56). Because in this study we have shown that HDACs regulate deposition of the H3K27ac mark at RAREs (Fig. 1, B–E), it will be interesting to determine whether HDACs are required for polycomb repressive complex 2 recruitment at RAREs in the absence of RA.

The NuRD complex has been identified as a regulator of Suz12 recruitment to gene promoters, which suggests that interaction of RARs with the NuRD complex could be a mechanism linking histone deacetylation and histone methylation at RAREs. Additionally, in ESCs, RA promotes the recruitment of the UTX/KDM6A demethylase to the Hox4A–D gene cluster promoters, where it specifically demethylates H3K27me2 and H3K27me3 (57, 58). Finally, another important demethylase that may play a major role in RA-mediated transcription is lysine-specific-demethylase 1 (LSD1/KDM1A), which specifically demethylates epigenetic marks indicative of activation, including H3K4/9me1 and H3K4/9me2 (59). LSD1 is part of the CoREST co-repressor complex where it functionally interacts with HDAC1 and HDAC2, and LSD1 plays important roles during ESC differentiation (60–62). In future studies, it will be important to address the question of whether RARs can regulate NuRD, UTX/KDM6A, or LSD1/KDM1A recruitment to RA-regulated genes and to define the roles of HDACs in regulating their recruitment.

In our studies we have also highlighted the fact that enhancer RARE regions may be more important than proximal promoters for the regulation of gene expression. First, we found that in the absence of RA, there was no HDAC interaction at the Hoxa1 pp (Fig. 1, B–D), but we did observe an increase in H3K27ac levels at the Hoxa1 pp after exposure to RA. This is in line with previous studies from our lab that found that the Hoxa1 pp was RA-responsive despite a lack of RAR binding (17). Overall, these studies further support interactions between the Hoxa1 pp and RARE as previously discussed (17). However, we also found that the Hoxa1 and Cyp26a1 enhancers exhibited much more dynamic changes in the levels of H3K27ac after HDAC k.d. (Fig. 3a, B and C) and changes in HDAC binding in response to RA (Fig. 5, A–C) as compared with the promoter regions. The other major epigenetic mark at enhancer regions is H3K4me1, which is often a marker of a “poised” enhancer (3). Although we found that RA promotes an increase in the epigenetic H3K27ac mark at active enhancers (Fig. 6), we found that RA treatment has little effect on H3K4me1 at these same enhancer regions (data not shown). These results suggest that RA has a greater role in the deposition of activating epigenetic marks, such as H3K27ac, that convert poised enhancers to “active” enhancers.

Collectively, our studies have identified novel interactions between HDACs and RA-regulated genes and have defined some of the mechanisms by which HDACs and RA regulate the deposition of the H3K27ac mark. We have shown that HDAC1, HDAC2, and HDAC3 regulate the Hoxa1 and Cyp26a1 RARE enhancer regions, whereas HDAC1 is the major HDAC at the RARβ2 RARE. We also found that these HDACs and RA are major regulators of the deposition of the H3K27ac active gene enhancer mark during differentiation. Because RA and HDAC inhibitors are being examined for potential combinatorial stem cell and cancer therapies that regulate cell growth and differentiation (22, 23), our findings that HDACs differentially regulate the transcription of RA target genes provide major insights into the usefulness of HDAC1, HDAC2, or HDAC3 specific inhibitors for future therapies.