Binding of Retinoic Acid Receptor Heterodimers to DNA

The retinoic acid signaling pathway is controlled essentially through two types of nuclear receptors, RARs and RXRs. Ligand dependent activation or repression of retinoid-regulated genes is dependent on the binding of retinoic acid receptor (RAR)/9-cis-retinoic acid receptor (RXR) heterodimers to retinoic acid response element (RARE). Although unliganded RXR/RAR heterodimers bind constitutively to DNA in vitro, a clear in vivoligand-dependent occupancy of the RARE present in the RARβ2 gene promoter has been reported (Dey, A., Minucci, S., and Ozato, K. (1994) Mol. Cell. Biol. 14, 8191–8201). Nucleosomes are viewed as general repressors of the transcriptional machinery, in part by preventing the access of transcription factors to DNA. The ability of hRXRα/hRARα heterodimers to bind to a nucleosomal template in vitro has therefore been examined. The assembly of a fragment from the RARβ2 gene promoter, which contains a canonical DR5 RARE, into a nucleosome core prevented hRXRα/hRARα binding to this DNA, in conditions where a strong interaction is observed with a linear DNA template. However, histone tails removal by limited proteolysis and histone hyperacetylation yielded nucleosomal RAREs able to bind to hRXRα/hRARα heterodimers. These data establish therefore the role of histones NH2termini as a major impediment to retinoid receptors access to DNA, and identify histone hyperacetylation as a potential physiological regulator of retinoid-induced transcription.

The retinoic acid signaling pathway is controlled essentially through two types of nuclear receptors, RARs and RXRs. Ligand dependent activation or repression of retinoid-regulated genes is dependent on the binding of retinoic acid receptor (RAR)/9-cis-retinoic acid receptor (RXR) heterodimers to retinoic acid response element (RARE). Although unliganded RXR/RAR heterodimers bind constitutively to DNA in vitro, a clear in vivo liganddependent occupancy of the RARE present in the RAR␤2 gene promoter has been reported (Dey, A., Minucci, S., and Ozato, K. (1994) Mol. Cell. Biol. 14, 8191-8201). Nucleosomes are viewed as general repressors of the transcriptional machinery, in part by preventing the access of transcription factors to DNA. The ability of hRXR␣/hRAR␣ heterodimers to bind to a nucleosomal template in vitro has therefore been examined. The assembly of a fragment from the RAR␤2 gene promoter, which contains a canonical DR5 RARE, into a nucleosome core prevented hRXR␣/hRAR␣ binding to this DNA, in conditions where a strong interaction is observed with a linear DNA template. However, histone tails removal by limited proteolysis and histone hyperacetylation yielded nucleosomal RAREs able to bind to hRXR␣/hRAR␣ heterodimers. These data establish therefore the role of histones NH 2 termini as a major impediment to retinoid receptors access to DNA, and identify histone hyperacetylation as a potential physiological regulator of retinoid-induced transcription.
Core histones H2A, H2B, H3, and H4 are the main protein components of chromatin around which DNA is wrapped in 146-bp 1 segments, forming stable nucleosomal structures. Nucleosome spacing and histones post-translational modifications, and most notably acetylation, varies greatly from one chromosomal domain to another (with hyperacetylated his-tones being preferentially associated to transcriptionally active chromatin) and have strong effects on gene activity (reviewed in Refs. 2 and 3). Beside these long range effects, chromatin assembly on regulatory regions of eukaryotic promoters affects directly the transcriptional activity of genes. Due to the organization of these DNA sequences into precisely positioned arrays of nucleosomes, transcription factors access to their cognate response elements is in most cases restricted and nucleosome assembly is therefore viewed as a general cellular strategy to repress transcription. In vitro assembly of nucleosomes on short DNA fragments documented this type of effect for many DNA-binding proteins, including Gal4 (4), SP1 (5), nuclear factor 1 (6), heat shock factor (7), and TATA-binding protein (8), and genetic experiments in yeast using the PHO5 transcription unit have established a link between transcriptional activation and chromatin structure disruption (reviewed in Ref. 9). However, chromatin organization is also, in some instances, a mean to favor transcriptional activity of genes, as described for the estrogen-regulated vitellogenin B1 gene (10). Therefore, transcription regulation must be viewed as a complex set of interactions involving specific transacting factors, general transcription factors, and coactivators recruitment onto a nucleoprotein complex. The productive interaction of DNA-binding proteins with their cognate response elements is thus conditioned by two parameters: (i) structural features of the nucleosome and (ii) dynamic processes leading to the alteration of chromatin structure by macromolecular complexes such as SWI/SNF (4,11), which is associated to RNA polymerase II under low stringency conditions (12), nucleosome remodeling factor (13), or nucleoplasmin and NAP-1 (14). Structural features of chromatin also implies a competition between chromatin constituents and transcriptional coactivators (some of which bearing strong structural similarities with histones H3 and H4 (15)), which will eventually determine the overall transcriptional activity of genes (reviewed in Ref. 16).
In an effort to better understand the role of chromatin structure in nuclear receptors binding to their response elements, we have used purified components to investigate the impact of nucleosome assembly on a promoter containing a prototypical retinoic acid response element (RARE). Retinoic acid receptors (RARs and RXRs) heterodimers bind, in vitro, to RAREs with high affinity, irrespective of the presence of ligand. On the contrary, Ozato and co-workers (1) established, by in vivo footprinting experiments, that agonist treatment of target cells is an absolute prerequisite for heterodimers binding to the RARE of the RAR␤ gene, as it is to observe biological effects of retinoids in vivo. This agonist-dependent occupancy of hormone response element was also observed with the glucocorticoid receptor (17). The transcriptional and DNA binding activities of these nuclear receptors are therefore controlled in vivo at multiple levels, which include post-translational modifications (see The transcriptional activation observed in the presence of RAR and RXR is triggered by binding of heterodimers to RAREs that consist, in most cases, of a direct repeat (DR) of the sequence PuGGTCA. A direct repeat of the hexanucleotide PuG(G/T)TCA spaced by five nucleotides favors the binding of RXR/RAR heterodimers, whereas a spacing of four, three, or one base converts the direct repeat into a thyroid hormone, vitamin D, and 9-cis-RA or PPAR response element, respectively (Ref. 22, and reviewed in Ref. 23). RXR/RAR heterodimers display the highest affinity for half-sites spaced by 5 bp (DR5) and a lower affinity for half-sites having a 2-bp spacing. DR1 RAREs accommodate heterodimer binding in an opposite polarity (i.e. RAR being the 3Ј-bound receptor) (reviewed in Ref. 24). Nucleosome assembly on DNA sequences containing such direct repeats should therefore introduce two new types of constraints upon RXR/RAR heterodimer binding to DNA, as a consequence of the helical nature of the DNA. The first type of constraint is defined by the translational phasing of the nucleosome, which describes 5Ј and 3Ј boundaries of the octamer core on the linear DNA sequence and thus identify the dyad axis of the nucleosomal core particle. Transcription factors access to their cognate DNA-binding site is facilitated when protein-DNA interactions take place close to nucleosome boundaries, reflecting a looser interaction of DNA with histones, as demonstrated for the glucocorticoid receptor (25). The second type of constraint is the rotational phasing within the nucleosome core, which reflects the orientation of any segment of DNA relative to the core histone surface (26).
In this work, we have first examined the positioning of histone octamers in vitro on a retinoid-regulated promoter, the P2 promoter of the human retinoic acid receptor ␤ (RAR-␤2) that contains a DR5 response element (␤-RARE (27)). We find that nucleosomes assembled spontaneously on DNA fragments from the RAR␤ gene P2 promoter at a position placing the ␤-RARE at the dyad axis of the core particle. Binding of purified hRAR␣/ hRXR␣ heterodimers was prevented by the wrapping of the RARE around the native octamer. On the contrary, heterodimer binding could be observed when histone tails were removed by limited proteolysis and when histones were hyperacetylated, suggesting a critical role of this post-translational modification in the regulation of the DNA binding activity of retinoids receptors in vivo.

Bacterial and Eukaryotic Cell Lines
The JM109 bacterial strain was used for all subcloning procedures, whereas the M15 strain (Qiagen) was used for the overexpression of both His 6 -hRAR␣ and His 6 -Flag-hRXR␣ (18). HeLa cells were used as a source for core histones and grown in Dulbecco's minimal essential medium containing 10% fetal calf serum, penicillin, and streptomycin (100 units/ml) to 80 -90% confluency prior to harvesting and histone octamer extraction.

Core Histone Purification
Core histones were prepared essentially as described in Côté et al. (32). The entire procedure was carried out at 4°C. HeLa cells from 20 T225 flasks were collected in 1 ϫ phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 , 1.4 mM KH 2 PO 4 , pH 7.4) and washed twice with this buffer. 20 pellet volumes of buffer CLB (20 mM Tris-HCl, pH 8.0, 3 mM MgCl 2 , 0.25 M sucrose, 0.5 mM phenylmethylsulfonyl fluoride) with 0.5% Nonidet P-40 were added and cells broken by 10 -15 strokes in a Dounce homogenizer. Lysate was centrifuged for 20 min at 4,000 ϫ g and the nuclei pellet washed twice with buffer CLB. 50 ml of buffer TNE 400 (10 mM Tris-HCl, pH 8.0, 0.4 M NaCl, 1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 5 M leupeptin) was added for 12 mg of DNA and the mixture was stirred gently for 15 min. Nuclei were pelleted and washed once with buffer TNE 400 , then resuspended in buffer PB 600 (50 mM NaPO 4 , pH 6.8, 0.6 M NaCl, 0.5 mM phenylmethylsulfonyl fluoride, 5 M leupeptin). While stirring, 20 g of dry hydroxyapatite (Bio-Gel HTP, Bio-Rad) was added per 12 mg of DNA. The slurry was poured in a Econo column (5.0 ϫ 20.0 cm, Bio-Rad) and the flow-through was discarded. The resin was washed with 15 volumes of buffer PB 600 and core histones were eluted from the matrix with buffer PB 2500 (50 mM NaPO 4 , pH 6.8, 2.5 M NaCl, 0.5 mM phenylmethylsulfonyl fluoride, 5 M leupeptin). OD 280 was monitored and fractions containing more than 2 mg/ml histones were pooled and stored at Ϫ80°C until use. Hyperacetylated histones were prepared as above except that cells were treated for 18 h with 10 mM sodium butyrate (Sigma). Cross-linking experiments were performed using dimethyl suberimidate (Pierce) to a final concentration of 1 mg/ml.
Histones were fractionated on 17% polyacrylamide, 0.9 M acetic acid, 6.25 M urea as described by Panyim and Chalkley (29) on 20-cm gels. Gels were run at 20 mA for 15 h and stained with Coomassie Blue. In this system, the expected order of migration is H1-H3-H2A, H2B-H4 (from top to bottom).

Retinoic Acid Receptors Purification
Full-length hRAR␣ and hRXR␣ were purified by metal chelate affinity chromatography to homogeneity as described previously (30). hRAR␣ was expressed as a NH 2 -terminal fusion protein with a His 6 tag, whereas hRXR␣ was fused to the His 6 tag upstream of the Flag epitope (IBI-Kodak, Rochester, NY). hRXR␣ was eluted from the affinity matrix by enterokinase cleavage, thereby removing the His 6 tag from the protein. Both receptors were expressed in the M15 Escherichia coli strain. Monoclonal antibodies directed against MRGS-His 6 and Flag epitopes were purchased from Qiagen and IBI-Kodak, respectively.

Plasmids and DNA Probes
The bacterial expression vectors pQE9-hRAR␣ and pQE9-His 6 -F-hRXR␣ have been described previously (18). Sequences containing the wild type ␤-RARE response element were isolated from pPro-RAR␤ (27) by restriction enzyme digestion. Milligram amounts of plasmid DNA were cut and selectively dephosphorylated at one end by calf intestine alkaline phosphatase (Promega), and the resulting insert was purified by agarose gel electrophoresis and electroelution. After phenol/chloroform extraction, DNA fragments were ethanol-precipitated and quantified. 10 pmol were labeled with T4 polynucleotide kinase and purified by gel-filtration using standard procedures (31).

Reconstitution of Nucleosomes
Nucleosomes were reconstituted essentially as described in Ref. 32, except that the dialysis step was replaced by a serial dilution in buffer NRB (25 mM Tris-HCl, pH 7.4, 1.2 mM MgCl 2 , 5 mM ␤-mercaptoethanol, 5% glycerol) from 2 M NaCl to 1.5, 1.2, 0.8, 0.6, and 0.3 M NaCl to reach a final volume of 200 l. Typically, 50 -100 ng of end-labeled probe (1-5 ϫ 10 6 cpm, 300 fmol) were mixed with varying core histones amounts, so as to obtain a histone:DNA mass ratio ranging from 0.5 to 3, 0.5 g of salmon sperm DNA, and 0.5 g of bovine serum albumin. Naked DNA controls were obtained by performing similar dilutions in the absence of histones, which were added after the last dilution step. Aliquots of each reconstitute were analyzed on a 4.5% polyacrylamide gel in 0.5 ϫ TBE (1 ϫ TBE is 90 mM Tris base, 90 mM boric acid, 2 mM EDTA) and samples containing more than 90% of reconstituted octamer-DNA complexes were used in further experiments. The final concentration of nucleosome core particles was around 5-10 fmol/l.

DNase I and Exonuclease III Protection Assays
Reconstitutes or naked DNAs were treated with either DNase I or exonuclease III prior to DNA extraction and resolving of DNA fragments on 6% urea denaturing polyacrylamide gels. 15 l of reconstitution or control mixtures (about 30,000 -40,000 cpm) were brought to 5 mM MgCl 2 and CaCl 2 and 1 unit of DNase I (Worthington, Freehold, NJ) was added. Digestions were for 0 to 8 min at room temperature (ϳ22°C), and stopped by addition of 100 l of stop buffer (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 1% SDS). 10 g of proteinase K was added per sample and incubated overnight at 37°C. DNAs were extracted with phenol and phenol/chloroform prior to ethanol precipitation. Exonuclease III protection experiments were run similarly except that samples were brought to 4 mM MgCl 2 , and digested with 200 units of exonuclease III (Appligene/Oncor, Strasbourg, France).
Linear amplification of the coding strand was performed using standard polymerase chain reaction conditions (33) and the 19-mer oligonucleotide 5Ј-ACACAGAATGAAAGATTGAATT-3Ј. 1 pmol of 5Ј-end labeled primer was used in a reaction mixture containing 1.5 mM MgCl 2 and ϳ30 femtomoles of DNase I-treated, non-labeled DNA. Thermocycling was performed using 25 cycles of 94°C for 30 s, 41°C for 1 min, 72°C for 30 s in a Perkin-Elmer 2400 thermocycler. Amplification products were phenol-chloroform extracted, ethanol-precipitated, and analyzed on 8% acrylamide, 6 M urea sequencing gels.

Electrophoretic Mobility Shift Assays
Binding reactions and electrophoresis were run as described in Ref. 18. Unless mentioned otherwise, ϳ30,000 cpm of naked DNA or reconstituted nucleosomes (about 10 fmol) were combined in a 20-l reaction containing 20 mM HEPES, pH 7.4, 80 mM NaCl, 1 mM EDTA, 3% glycerol, 0.5 g of salmon sperm DNA, and 1 to 5 pmol of purified His 6 -hRAR␣ and/or F-hRXR␣. Complexes were resolved on a 4% native polyacrylamide gel in 0.5 ϫ TBE at 4°C at 20 volts/cm (8 -10 mA), or when mentioned on 1% agarose gels in 0.5 ϫ TAE buffer (10 V/cm, 35 mA for 3 h at 4°C). 1 ϫ TAE is 90 mM Tris acetate, pH 7.4, 2 mM EDTA. Antibodies used in supershift experiments were those used for Western blot analysis of receptor preparations. Gels were dried and autoradiographed at Ϫ80°C.

Other Techniques
Western Blotting Procedure-Proteins were resolved on a 10% SDS-PAGE and transferred onto a nitrocellulose membrane. Immunodetection of His 6 -hRAR␣ and F-hRXR␣ was performed as described previously using the IBI BioMax system (19).
Protein Assay-The protein content of receptor preparations was assayed by the Bradford assay (34) using bovine serum albumin as a standard.
Sequencing Reactions-Sequencing reactions were run using 32 Plabeled primers and dideoxynucleotides mixes according to the manu-facturer's instructions (Amersham/U. S. Biochemical Corp.). Sequencing reactions were run using the native pPro-RARE␤ plasmid as a template, yielding DNA ladders extending beyond the 5Ј and 3Ј ends of the DNA fragment used in reconstitution experiments.

RESULTS
Nucleosome Assembly on DNA Fragments from the Promoter of the Human RAR␤ Gene-In this study, we used purified core histones from HeLa cells (Fig. 1, A and B) to reconstitute nucleosome core particles following dilution from high salt. As shown by SDS-PAGE fractionation of core histones, histone H1 concentration was less than 3% of total histones (Fig. 1A). Histone oligomers were further characterized by cross-linking to characterize the octameric structure of purified histones. The main cross-linked band appeared to be an octamer, indicating that our starting material for reconstitution experiments is indeed a stable core nucleosome. DNA fragments used in nucleosome reconstitution experiments contain a RARE organized as a direct repeat of two hexanucleotides separated by 5 bp (DR5). An imperfect DR5 RARE (DR5) is also found 14 bp upstream of this RARE, whereas the TATA box is located 6 bp downstream of the DR5 sequence. Additional cis-acting elements are also present in this sequence (Fig. 4). Labeled DNA fragments of varying length (326, 240, and 182 bp, Fig. 1C) containing these functional cis-acting elements were used to investigate whether they can assemble into a nucleosome core in the presence of histone octamers. Similar amounts of each DNA fragment were thus used as a template for nucleosome assembly with increasing amount of histones. As expected, DNA fragments above 200 bp generated complexes displaying discrete electrophoretic mobilities, indicative of the association of several histone octamers on these DNAs, or of distinct octamer positioning (Fig. 1C). More interestingly, the 182-bp frag- Translational and Rotational Phasing of the Reconstituted Mononucleosome-To examine whether this product has features of native nucleosomes, we first optimized reconstitution parameters to obtain Ͼ95% of the probe reconstituted in nucleosome cores by carefully adjusting the histone:DNA mass ratio, which was found to be optimal in the 0.8 to 3.0 range in our conditions ( Fig. 2A). Boundaries of the core particle were determined by the exonuclease III protection assay (Fig. 2B). This enzyme digests from the 3Ј end to the 5Ј end of DNA and its progression along the sequence is strongly, but not totally, inhibited by proteins bound to DNA. Thus translational positioning of the reconstituted nucleosome can be analyzed by the ExoIII assay (Fig. 2B). The protection pattern of the labeled top strand indicated a major histone-induced ExoIII stop at posi-tion ϩ32 (see Fig. 4 for sequence numbering), with weaker protections being observed at positions ϩ42, ϩ52, ϩ 62. On the contrary, no stop to ExoIII progression could be detected for the 5Ј-labeled lower strand, suggesting that histone octamers position preferentially at the 5Ј end of this particular DNA fragment. Thus these data define a ϳ150-bp segment on which nucleosome cores adopt a preferential, but not unique, positioning. The nucleosomal structure of this complex was further characterized by DNase I experiments (Fig. 3). DNase I cleavage sites (which correspond to a maximal accessibility of the DNA minor groove) on reconstitutes were clearly different from these of naked DNA. A periodic pattern extended from Ϫ101 to ϩ2 on the upper strand, but the Ϫ61 to Ϫ20 segment was consistently found to be less accessible to DNase I cleavage for reasons that are not clear to us. Thus DNase I-generated fragments were amplified by a linear polymerase chain reaction (Fig. 3B) to fully characterize the rotational positioning of the double helix over the DR5 RARE and TATA box sequences. Again, a highly 10 -11 bp periodic cleavage pattern was observed, typical for a DNA fragment organized around the surface of a histone octamer. This repetitive pattern was also clearly and reproducibly detected on the bottom strand in reconstituted DNA fragments on a stretch extending from Ϫ87 to ϩ24 (Fig. 3C). Thus the DNase I pattern, together with 5Ј and 3Ј boundaries defined by the ExoIII protection assay, strongly suggest that a histone octamer is positioned from Ϫ112 to ϩ32, with the DR5 RARE lying across the dyad axis of the nucleosome and demonstrate that this precise DNA segment has an intrinsic structure directing a precise translational and rotational positioning of the histone octamer. These features are summarized in Fig. 4.
DNase minor groove cleavage sites indicate that the third and fourth bases of each half-site of the DR5 RARE are highly accessible in the reconstituted mononucleosome. Since, according to molecular modeling (35), DNA-binding domains of RXR and RAR make extensive contacts with the major groove of the first three or four bases of each half-sites, we conclude that major contact points are facing toward the octamer, and therefore are in an orientation unlikely to be compatible with heterodimer binding to the RARE.
Importance of Histones NH 2 Termini in RAR/RXR Heterodimer Binding to Nucleosome Core Particles-To test this hypothesis, full-length, E. coli-expressed hRXR␣ and hRAR␣ were purified to homogeneity using a Ni 2ϩ -nitrilotriacetic acid affinity matrix (Fig. 5A) and their ability to bind to a 20-mer oligonucleotidic probe containing the DR5 RARE was assessed by the electrophoretic mobility shift assay (Fig. 5B). Cooperative binding of purified RXR/RAR heterodimers to this response element was consistently observed in these conditions, indicating that a large fraction of the purified polypeptides is functional and binding to the core RARE with an affinity in the nanomolar range. Binding of purified heterodimers to the DR5 RARE present in the 182-bp DNA fragment used in mononucleosome reconstitution experiments occurred with a similar efficiency (Fig. 6, lanes 1-7). The nature of each complex was characterized by supershift experiments using monoclonal antibodies directed against the NH 2 -terminal tag of each receptor. The electrophoretic mobility of DNA⅐RAR/RXR complexes was clearly decreased in the presence of each immunoglobulin (lanes 8 and 9), demonstrating that RAR/RXR heterodimers are indeed formed in these conditions. This DNA template was therefore assembled into a nucleosome core and tested similarly for its ability to bind RAR/RXR heterodimers. As predicted from the rotational orientation of both half-sites, this template did not accommodate RXR/RAR dimers, even in conditions where more than 90% of the naked probe was bound

FIG. 2. Translational positioning of the 182-bp RAR-␤2 promoter fragment.
A, optimized reconstitution with the HinfI-EcoRI fragment: 5Ј-end labeled DNA was assembled into a nucleosome core using the salt dilution method as described under "Experimental Procedures." The result of a typical experiment yielding more than 90% of reconstituted material is shown. In this case, a histone:DNA mass ratio of 2.0 was used. F, naked DNA; R, reconstituted mononucleosome. B, exonuclease III protection assay of naked and octamer-bound RAR-␤2 promoter fragment. Boundaries of monosomes reconstituted using the HinfI-EcoRI fragment from the RAR-␤2 promoter were determined by ExoIII digestion. Samples containing either the naked DNA (lanes F) or reconstituted monosomes (lanes R) were incubated with 200 units of ExoIII (ExoIII) or not (No ExoIII) for 4 min at room temperature. After DNA extraction, samples were resolved on a 8% acrylamide-urea gel. Major stops are indicated by filled arrowheads. Positions of ExoIII stops were assigned based on the Maxam-Gilbert G sequencing track (G lanes). Numbers indicate the sequence position number of relevant Gly residues along the ␤-RARE promoter. Gly residues are located at positions ϩ65, ϩ64, ϩ63, ϩ62, ϩ61, ϩ60, ϩ54, ϩ47, ϩ43, ϩ42, ϩ41, ϩ39, ϩ38, ϩ35, ϩ34, ϩ32, and ϩ27. C, scheme of the boundaries of the reconstituted core nucleosome. The ellipsoid indicates the major 5Ј and 3Ј boundaries determined by ExoIII protection assays. Salient features of the promoter are also indicated.
(compare lane 5 to lane 14). 10-Fold higher receptor concentrations were also used and failed to evidence an interaction of these nuclear receptors with nucleosomal templates (data not shown). We conclude from these experiments that the organization of retinoic response elements around an histone octamer prevents, in our system, the binding of hRXR␣/hRAR␣ heterodimers to a prototypic response element.
Amino-terminal regions of histones are hyperacetylated in transcriptionally active chromatin. A direct link between these two processes has now been established (reviewed in Ref. 36 -38). This post-transcriptional modification can also be mimicked in vitro by removal of histone tails by trypsin (see Ref. 39 and references therein). We wished to determine whether histone tail removal or histone hyperacetylation could influence binding of RXR/RAR to nucleosomal templates. Therefore, we treated purified core particles with trypsin or extracted core histones from HeLa cells treated with 10 mM sodium butyrate, a potent inhibitor of histone deacetylase activities. Electrophoretic analysis of both histones preparations evidenced the expected structural alterations: limited proteolysis of core histones yielded polypeptides migrating more rapidly than native histones in SDS and acid-urea polyacrylamide gels (Fig. 6A). Sodium butyrate treatment generated histone species that migrated more slowly in SDS-polyacrylamide gels (Fig. 6B), and  Figs. 2 and 3 on the wild type RAR-␤2 promoter are shown. Sequence has been truncated at 5Ј and 3Ј boundaries of the nucleosome (ϩ112/Ϫ34). Dots indicate the nucleosomespecific DNase I-hypersensitive sites detected on the coding strand, whereas triangles depict DNase I-hypersensitive sites detected on the noncoding strand. Positions of the imperfect DR5 element, as well as that of the canonical DR5 RARE and of the TATA box are indicated between the two strands. Additional cis-acting elements such as a cAMP-response element (Ϫ99 to Ϫ92), a AP-1 response element (Ϫ84 to Ϫ78), and a Inr element (Ϫ8 to ϩ5) have also been characterized in this promoter. acid-urea gels characterized at least 7 species with distinct electrophoretic mobilities that were different from those appearing with native histones. This pattern is compatible with the hyperacetylation of histones H3 and H4 which are highly acetylated and usually form triple bands in such a system. These two types of preparation were thus used in reconstitution experiments as described above. Both sources of core histones yielded nucleosome particles with the Ϫ112/ϩ70 RAR␤ P2 promoter fragment, with a slightly increased electrophoretic mobility observed with trypsinized histones. Reconstitutes were then tested for their ability to bind RXR/RAR heterodimers. We observed that histone tails removal by limited trypsin treatment allowed the binding of receptor heterodimers to the mononucleosomal DNA. The specificity of heterodimer binding to the nucleosomal DR5 RARE was assessed by competition with a 100-fold excess of unlabeled DR5 oligonucleotide, which was able to displace the complex. The addition of the anti-RAR antibody disrupted this complex, suggesting that the conformation of the ternary complex RAR/RXR-nucleosomal RARE is distinct from that of the binary complex RAR/ RXR-RARE (compare lane 8 to lane 26). On the contrary, the anti-RXR IgG increased the molecular mass of the ternary complex as observed for the naked DNA fragment, showing that the NH 2 -terminal end of RXR is still fully accessible in this configuration. An identical result was observed when DNA binding assays were carried out with hyperacetylated templates (Fig. 6, left panel). Note that in this typical experiment, a slightly higher affinity of RAR/RXR heterodimers for the nucleosomal template was observed but was not found to be statistically significant. Thus our data identify histone tails as major determinants of hRXR␣ and hRAR␣ access to DR5 retinoic acid response elements, and similar results were obtained using DR2 and DR1 RAREs (data not shown). DISCUSSION The design of an in vitro system using purified components with precisely characterized features enabled us to study whether the association of RAREs with histones is of major importance in regulating the access of retinoid receptors to their cognate DNA-binding sites. We established first that histone octamers are positioned at a preferential site on the RAR-␤2 promoter DNA. Boundaries determined by exonuclease III protection experiments located the nucleosome between Ϫ112 and ϩ34, showing that this DNA fragment has structural properties facilitating DNA bending and therefore curving around the histone octamer. DNase I protection experiments carried out on the wild type DNA reconstituted around native core histones evidenced a typical cleavage pattern alternating every 10 Ϯ 2 bases pairs, and allowed the determination of the rotational setting of the wild type DNA. Minor grooves of both half-sites were found to be oriented facing away from the octamer, forming a closed recognition interface for RXR/RAR heterodimers. Indeed, Rastinejad and colleagues (35) reported that RXR/thyroid hormone receptor (T3R) DBDs dimers, and by extension RXR/RAR dimers, engage the major grooves of the successive half-sites. In contrast, the minor groove of the TATA box was found to be sensitive to DNase I digestion, evidencing a proper exposure of the TATA-binding protein-binding site (40). However, this translational and rotational setting of the TATA box has proven to provide a poor substrate for TATAbinding protein binding within the Xenopus borealis 5 S rRNA gene (41) and to the adenovirus major late promoter TATA box (8). In keeping with these observations, we found that the nucleosome-assembled DR5 response element did not allow cooperative binding of RXR/RAR dimers, in opposition to naked DNA. Glucocorticoid receptor is known to bind to its cognate response elements as a homodimer structurally close to RXR/ T3R and RXR/RAR DBDs dimers. Wrange and colleagues (25) reported that a glucocorticoid response element having a analogous rotational setting (with minor grooves pointing away from the histone octamer) is unable to bind glucocorticoid receptor dimers. By analogy, we would predict that modifying the rotational setting of the DR5 RARE would render the nucleosome permissive to RXR/RAR binding. RXR/T3R dimer binding to the DR4 thyroid response element of the T3R(␤)A promoter forming a nucleosomal template has been reported to occur both in vivo and in vitro (42). RXR/T3R binding was, however, in this system, thyroid-hormone independent; ligand binding led to an alteration of the chromatin structure and coincidental transactivation of the promoter. Alteration of the rotational setting of the wild type T3RE was found be detrimental for the binding of RXR/T3R heterodimers to nucleosomal DNA (43). Ozato and colleagues (1, 44) reported a stringent ligand-dependent occupancy of the RARE of the RAR-␤2 promoter in vivo, as well as other cis-acting elements of the promoter. This FIG. 5. Purification of hRXR␣ and hRAR␣ by affinity chromatography. A, 10 g of hRAR␣ and hRXR␣ purified by immobilized metal affinity chromatography were resolved by a 8% SDS-PAGE. Gels were then either silver-stained (left panel) or proteins were transferred on a nitrocellulose membrane (right panel). hRAR␣ and hRXR␣ were detected by chemiluminescence using an anti-His 6 monoclonal antibody and the anti-Flag M2 monoclonal antibody, respectively. Both immunoglobulins recognize denaturated and native forms of each receptor. Molecular masses are indicated in kDa on the left. B, electrophoretic mobility shift assay of purified hRAR␣ and hRXR␣ binding to a 20-mer oligonucleotide containing the DR5 RARE from the RAR␤ promoter. 300 fmol of purified hRAR␣ or hRXR␣ or both were incubated with ϳ10 fmol of labeled probe and resolved on a 5% polyacrylamide gel. A 100-fold of the same radioinert probe (Spec.) or of a SP1 consensus binding site (Non spec.) was added to the binding mixture to assess the specificity of heterodimer formation. occupancy occurred coincidentally to promoter activation, without inducing major remodeling of the nucleosomal structure of this promoter. This observation is in striking contrast with in vitro assays showing that RXR/RAR dimers bind constitutively to RAREs (i.e. in a ligand-independent manner) (for a review, see Ref. 21, and references therein). We postulate that, based on our observations, these discrepancies could be explained by the specific architecture of the RAR-␤2 promoter leading to a closed conformation of the DR5 element in a nucleosomal context, although its in vivo organization remains to be precisely characterized.
If RXR/RAR heterodimers are excluded from nucleosomal DNAs in our system, then the nucleoprotein structure of this locus has to be altered in some way to allow for RXR/RAR recruitment. Histone hyperacetylation is the most common post-translational modification targeted to the NH 2 termini of these proteins. The reversible charge neutralization of highly conserved lysine residues by acetyl groups reduces the capacity of histones tails to stabilize the path of DNA along the octamer core. Consequently, allosteric changes are thought to occur in the nucleoprotein complex and renders the nucleosomal DNA more accessible to transcription factors (39), although a role in the disruption of nucleosome-nucleosome interactions can be predicted from recent crystallographic data (45). Histone tails clipping in vitro by limited trypsin treatment mimicks the effects of hyperacetylation and facilitate GAL4, TATA-binding protein, and TFIIIA binding to "chromatinized" DNA (39,41,46). To ascertain whether this post-translational modification could influence RXR/RAR heterodimers binding to nucleosomal DNA, we used both trypsinized histones and histones extracted from sodium butyrate-treated cells in reconstitution experiments (Fig. 6). Our results demonstrate that both sources of core histones yielded nucleosome core particles able to bind receptor heterodimers, identifying histone hyperacetylation as a major control event in the regulation of retinoid receptors access to DNA. We note that the yeast coactivator Gcn5p has been identified as the catalytic subunit of a histone acetyltransferase (HAT) (47). More relevant to the studied phenomenon is the characterization of a mammalian Gcn5 homologue, p/CAF (48). This protein has been shown to interact with CBP/p300, a co-integrator interacting with T3R, RAR, and RXR (49) and to stimulate the transcriptional activity of progesterone and estrogen receptors (50). CBP/p300 is also a coactivator for a number of transcription factors such as cAMP response element-binding protein, AP-1, MyoD, and c-myb (51)(52)(53). Very importantly, CBP/p300 has been shown to possess intrinsic HAT activity (54,55), suggesting that RXR/RAR binding to RAREs could direct HAT(s) to retinoic acid-regulated promoters. Although transcriptional synergy with CBP/p300 is generally characterized using transiently transfected templates for which the importance of the chromatin structure is questionable (56), this is not exclusive of its involvement in the coordinate recruitment of transcriptional regulators altering chromatin structure. Two other nuclear receptor coactivators, SRC-1 and ACTR, have also been shown to possess histone acetylase activity (57,58). Thus several HAT activities may be tethered to hormonally-regulated promoters by liganded receptors and act synergistically. A possible cooperativity between two disscribed above, except that core histones were treated with limiting amount of trypsin. The specificity of the binding reaction and the nature of the complex was probed as described in C. F, hRAR␣/hRXR␣ heterodimers binding to hyperacetylated nucleosome cores. Reconstitution and binding reactions were exactly as described above, except that core histones were extracted from sodium butyrate-treated HeLa cells. The specificity of the binding reaction and the nature of the complex was probed as described in C.
FIG. 6. Effect of histone tails removal or hyperacetylation on the binding of hRXR␣/hRAR␣ heterodimers to nucleosome core particles. A, Coomassie Blue staining of a 15% SDS-PAGE (left) or of a acid-urea gel containing 50 g of native or trypsinized octamers. Molecular masses markers are shown on the left. B, Coomassie Blue staining of a 15% SDS-PAGE (left) or acid-urea gel containing 50 g of native or hyperacetylated histones. Molecular masses markers are shown on the left. C, D, E, and F, electrophoretic mobility shift assay of hRXR␣/hRAR␣ heterodimers binding to naked or reconstituted nucleosomal templates. C, labeled HinfI-EcoRI (182 bp) was used as a probe in electrophoretic mobility shift assay binding reactions and incubated in the presence of increasing amounts of purified hRAR␣ or hRXR␣ or both receptors. The specificity of the binding reaction was assessed by competition with a 50-fold excess of the same unlabeled probe (lanes 6, 15, 24, and 33) or of an unrelated oligonucleotide (SP-1 consensus binding site, lanes 7, 16, 25, and 34). The identity of proteins contained in the heterodimeric complex was confirmed using anti-RAR (anti-His tag) and anti-RXR (anti-Flag) monoclonal antibodies, which generated complexes of lower electrophoretic mobility (lanes 8 and 9, respectively). A similar characterization of complexes was carried out in lanes 17, 18, 26, 27, 35, and 36. In a typical 20-l binding reaction, 10 fmol of the labeled probe was combined with 0.1 to 0.3 pmol of purified receptor(s), 0.5 g of nonspecific DNA and incubated for 60 min at 4°C. 5 l of ascites fluid was added for supershift assays (lane 8, anti-RAR antibody; lane 9, anti-RXR antibody). Complexes were resolved on a 1% agarose gel at 4°C in 0.5 ϫ TAE.  6 -9 contain the same amount of RAR and RXR than lane 5. D, electrophoretic mobility shift assay of hRXR␣ and hRAR␣ binding to mononucleosome reconstituted with native histones. The affinity of hRAR␣, hRXR␣, or hRAR␣/hRXR␣ heterodimers for this nucleosomal template was assayed as described in C, using ϳ30 -50 fmol of nucleosomal template. The specificity of the binding reaction and the nature of the complex was probed as described above. E, hRAR␣/hRXR␣ heterodimers binding to trypsinized nucleosome cores. Reconstitution and binding reactions were exactly as de-tinct chromatin-modifying complexes is underlined by the physical interaction between the ADA5⅐Gcn5 complex and the SWI⅐SNF complex (36), which potentiates glucocorticoid receptor-mediated transcription in yeast (59) and for which mammalian counterparts have been identified (60). Consistent with these observations, the histone deacetylase HDAC1 was found to be associated with SMRT, a nuclear receptor corepressor binding to unliganded RARs (61), and inhibition of cellular deacetylase activities by trichostatin A led to potentiation of retinoic acid-induced transcription and cellular differentiation (62,63). All the studies above therefore establish a direct link between nuclear receptor-mediated transcriptional activation and histone acetylation.
Several predictions can be made from these and our data, assuming that retinoid-controlled promoters are organized in nucleosomal arrays: (i) histone hyperacetylation will be of more or less importance to retinoic acid-induced transcription activation, depending on the rotational and translation positioning of the response element. As a consequence of the observed in vitro positioning of the RAR-␤2 promoter, we would predict a strong dependence of its transcriptional activity on HAT activities. (ii) CBP and/or associated HAT (p/CAF, Gcn5p homologues) and other factors yet to be identified are likely candidates for chromatin-dependent RXR/RAR coactivators. Transcription factors like E1A (59), or agents (anti-AP1 retinoids (64)) known to modulate CBP-retinoid receptors interaction may have a direct influence on HAT activities targeted to retinoid-controlled promoters, and thus on RA-mediated transcriptional activity. Further experiments in progress in our laboratory comparing the in vitro and in vivo organization of RA-controlled promoters will provide new insights into the importance of chromatin remodeling in retinoid receptors-mediated transcriptional activation, in conjunction with the use of powerful molecular tools such as specific ligands and mutated receptors.