Retinoid-dependent Recruitment of a Histone H1 Displacement Activity by Retinoic Acid Receptor*

Targeted recruitment of histone acetyltransferase (HAT) activities by sequence-specific transcription factors, including the retinoic acid receptors (RARs) and retinoid X receptors (RXRs), has been proposed to lead to destabilization of nucleosomal cores by acetylation of core histones. However, biochemical evidence indicates that destabilization and depletion of linker H1 histones must also occur at the promoter regions of actively transcribing genes. Mechanisms by which nuclear receptors and other transcription factors affect the removal of histone H1 from transcriptionally silent chromatin have not been previously described. In this report, we show that RARs interact in a ligand-dependent manner with HMG-I, which is known to displace histone H1 from chromatin. We further show that HMG-I and a novel related protein, HMG-R, also interact with other transcription factors. Using sense and antisense constructs of HMG-I/R in transient transfection assays with a retinoid responsive reporter, we also demonstrate that HMG-I/R is important for retinoid dependent transcriptional activity of RAR. These findings suggest a step wise mechanism by which RARs and other transcription factors can cause a targeted unfolding of compact chromatin as a first step in transcriptional activation, which would then be followed by recruitment of HAT activity and subsequent events.

Retinoic acid receptors and retinoid X receptors (RAR and RXR ␣, ␤, and ␥) are sequence-specific, ligand-dependent transcription factors belonging to the superfamily of steroid/thyroid/vitamin D 3 nuclear receptors (1). RAR-RXR 1 heterodimers induce gene expression in a ligand dependent manner through RA responsive elements (RAREs) present in the promoter regions of responsive genes (2). Recently, CBP/p300, Sug1/Trip1, TIF1, SRC-1/N-CoA1, TIF2/GRIP1, and ACTR have been identified as co-factors, which interact with RARs and other nuclear receptors in a ligand-dependent manner (3,4). Biochemical evidence supports models involving depletion of the nucleosomal core as well as H1 histones at the promoter regions of actively transcribing genes. CBP/p300, its associated protein p/CAF, SRC-1, and ACTR have intrinsic histone acetyltransferase (HAT) activity (5)(6)(7). It has been proposed that recruitment of HAT activity by sequence-specific transcription factors leads to acetylation of core histones and a destabilization of the nucleosomal core, thereby facilitating transcriptional activation (8 -10). However, an earlier obligatory step in transcriptional activation involves an unfolding of the compact, 30-nm chromatin fiber, which results only from a displacement of the potent transcriptional repressor, histone H1, from chromatin (11,12). Mechanisms by which nuclear receptors and other transcription factors affect this obligatory removal of histone H1 from transcriptionally silent chromatin have not been described previously. In this report, we provide evidence that RARs interact in a ligand-dependent manner with HMG-I, which is known to displace histone H1 from chromatin (13). Additionally, we identify an HMG-I-related protein, HMG-R, which also interacts with RARs in a ligand-dependent manner. We further show that HMG-I and HMG-R interact with RXR␣, PPAR␥, c-Jun, and CBP, thus indicating recruitment of HMG-I/R by various transcription factors as a common mechanism for enhancer-dependent transcriptional activation. Finally, using transient transfections, we demonstrate that HMG-I/R is required for retinoid-dependent transactivation of a reporter construct by RAR, thus showing the functional consequences of RAR-HMG-I/R interactions.
For transient transfections, the RA-responsive reporter, RARE3-tk-CAT, containing three copies of the canonical DR5 motif in pBLCAT8ϩ, was kindly provided by Dr. S. Mader. The HMG-I sense, pHMG-I(S) and HMG-I antisense (pHMG-I(AS)) expression vectors were prepared by PCR amplification of HMG-I cDNA from pACT2-HMG-I using primer paires 5Ј-AGGGATCCACCATGAGTGAGTCGAGCTCGAAG-3Ј and 5Ј-AGGAATTCTCACTGCTCCTCCTCCGAGGA-3Ј. The PCR-amplified product was cloned into T overhangs of pTarget expression vector (Promega, Madison, WI), and sense and antisense expression vectors were obtained by sequencing the cloned fragments to determine their directionality. Similarly, HMG-R sense, pHMG-R(S) and HMG-R antisense, pHMG-R(AS), expression vectors were prepared by PCR amplification of HMG-R cDNA from pACT2-HMG-R using primer pairs 5Ј-AGGGATC-CACCATGAGTGAGTCGAGCTCGAAG-3Ј and 5Ј-AGGAATTCAGTGGG-ATGTTAGCCTTGTCCAGG-3Ј and subsequently cloning the PCR-amplified product into T overhangs of pTarget expression vector. The directionality of the expression vectors was determined by sequencing the cloned fragment. pACT2-HMG-I and pACT2-HMG-R were identified as RAR interacting proteins from HaCaT keratinocyte cDNA library.
PCR Amplification of HMGs-HMG amplification was performed on keratinocyte cDNA library and reverse transcribed keratinocyte and skin raft RNA using the primer pairs 5Ј-AGGGTACCATGAGTGAGTC-GAGCTCGAAGTCC-3Ј and 5Ј-AGAAGCTTTCACTGCTCCTCCTCCG-AGGACTC-3Ј encompassing HMG-I from the ATG to the stop codon. The amplified products were resolved on a NuSieve gel (4%), visualized by ethidium bromide staining, excised, cloned into TA cloning plasmid, and sequenced.
Transformation of Yeast-Transformation of yeast was carried out by using the lithium acetate method (14).
␤-Galactosidase Assays-For qualitative evaluation of RAR interactions, ␤-galactosidase filter lift assays were carried out. Yeast colonies that grew on selective media were streaked on fresh selective plates in the presence of the retinoid for 3 days at 30°C. Cells were transferred onto Whatman No. 5 paper, submerged in liquid nitrogen for 10 s, placed on a filter paper presoaked in Z buffer (100 mM sodium phos- Italicized letters correspond to common amino acids among HMGs; normal and bold letters represent amino acids shared between HMG-I and R; italicized and bold letters represent amino acids shared between HMG-I and -Y, and "-" denotes deletion of an amino acid. C, PCR amplification of HMG-R from keratinocytes and skin rafts. Reversetranscribed RNA from foreskin keratinocytes (lanes 2 and 3), skin rafts (lane 4), or HaCaT library cDNA (lanes 1 and 5) was PCR amplified using HMG-I oligonucleotides described under "Materials and Methods." phate, pH 7.0, 10 mM KCl, 1 mM MgSO 4 ) supplemented with 50 mM ␤-mercaptoethanol and 0.07 mg/ml 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside. Filters were then incubated at 30°C and checked for the appearance of blue colonies, thus indicating interacting proteins. For quantitative studies, yeast colonies, positive in lift assay, were grown in ϪLeu, ϪTrp medium (5 ml) overnight at 30°C. Overnight cultures (2 ml) were inoculated in fresh YPD medium (8 ml) in the presence or absence of the retinoid, TTNPB (1 M), grown for 5 h at 30°C, and A 600 recorded. Cells (1 ml) were pelleted, suspended in Z buffer (100 l) and lysed by freeze-thawing. Z buffer (700 l) containing ␤-mercaptoethanol was added, followed by addition of 4 mg/ml o-nitrophenyl-␤-D-galactopyranoside (160 l) and the mixture incubated for 5 min to 1 h at 30°C. The reaction was stopped by the addition of 1 M Na 2 CO 3 (400 l) and A 420 recorded. ␤-Galactosidase activity was calculated using the formula, 1000 ϫ A 420 /time (minute) ϫ A 600 . Interacting protein expression vector was isolated, cDNA sequenced, and compared with the BLAST algorithm.
HAT Activity-CBP or p300 antibodies (Santa Cruz) were added to HeLa cell nuclear extract and incubated at 4°C for 2 h. Protein A/G-Sepharose (1:1 mix, 25 l) was added and the tubes rotated overnight at 4°C. Immunoprecipitated CBP or p300 were pelleted and washed thrice with 1 ml lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% Nonidet P-40, 0.1 mM phenylmethylsulfonyl fluoride, aprotinin, and leupeptin). After the final wash, 25 l of lysis buffer was added along with 1 l of bovine serum albumin, calf thymus histones (Sigma) or recombinant, purified HMG-I (10 g each), and 1 l of [ 3 H]acetyl-CoA, and incubated at 30°C for 45 min. Acetylation was quantitated using P-81 filters as described previously (5).
HMG-RAR Interaction in Vitro-Baculovirus produced RAR␣ or RAR␥ were immunoadsorbed on protein A/G-Sepharose beads (1:1 mix, 10 l) using RAR␣ or -␥ antibodies or normal rabbit IgG as nonspecific antibodies (Santa Cruz). 35 S-Labeled Gal4-HMG-I was synthesized by transcription and translation in vitro and incubated with immobilized RAR␣ or -␥. After washing with radioimmune precipitation buffer, the labeled proteins were analyzed by SDS-polyacrylamide gel electrophoresis and PhosphorImager.
Transfections and CAT Assays-HeLa cells, grown in Dulbecco's modified Eagle's medium containing charcoal-stripped fetal calf serum (10%), were transfected using the GenePORTER transfection procedure (Gene Therapy Systems, San Diego, CA). Cells were plated 18 h before transfection at 40% confluence (40,000 cells/well) in a 24-well plate. The cells were transfected with the RA-responsive reporter, RARE3-tk-CAT (1 g), and 0.1 g of either pHMG-I(S) or pHMG-R(S) or varying concentrations of pHMG-R(AS) expression vector, along with 5 l of GenePORTER reagent for each well in a total volume of 0.5 ml. For assays involving co-transfections of pHMG-R(S) and pHMG-R(AS) expression vectors, HeLa cells were transfected with 1 g of RARE3-tk-CAT and 0.01 g of each of the expression vector constructs. A renilla luciferase expression vector, pRL-SV40 (10 ng) (Promega), was used as an internal control in transfections to normalize for variations in transfection efficiency. The cells were transfected in quadruplicate for 5 h, and retinoids were added 18 h post-transfection. The CAT activity was quantified by counting the amount of 3 H-acetylated forms of chloramphenicol using a liquid scintillation counter. The renilla luciferase activity was quantified using dual luciferase reporter assay system (Promega).

RESULTS AND DISCUSSION
To gain a better understanding of the mechanism of retinoid action, we used a yeast two-hybrid system to identify and characterize proteins that interacted with human RAR␥ in a ligand dependent manner. RAR is a modular protein containing six functional regions, namely, A through F. A/B region contains a ligand-independent transactivation function, C region contains two zinc fingers and corresponds to the core of the DBD, and E region contains the ligand binding domain (LBD), heterodimerization domain, and ligand-dependent transactivation function. The sequence of RAR␥ encompassing the C-F regions (amino acids 90 -454) was used as bait ( Fig. 1) to isolate cDNAs encoding interacting proteins from a HaCaT keratinocyte cDNA library. Thus, the bait construct (pAS2-RAR␥⌬AB) contained both the DBD and the LBD of RAR fused to the Gal4-DBD in an yeast expression vector pAS2-1. A schematic representation of all the bait constructs used herein is presented in Fig. 1. Stably transfected pAS2-RAR␥⌬AB yeast cells were transformed with the second vector that expressed the Gal4-AD fused to HaCaT keratinocyte library cDNAs. In addition to known RAR interacting proteins, such as RXRs, Sug1, and RIP 140, we isolated complete cDNAs for HMG-I protein and a novel HMG-I-related protein (HMG-R). Neither pAS2-RAR␥⌬AB nor pACT2-HMG-I/HMG-R was active when expressed alone (Fig. 2A). Interaction in yeast cells was observed only when pAS2-RAR␥⌬AB was expressed with HMG-I/HMG-R cDNA fused to the Gal4-AD. HMG-I and HMG-R proteins interacted specifically with RAR␥⌬AB in the presence but not in the absence of the RAR-specific agonist, TTNPB (15) ( Fig. 2A), demonstrating the ligand dependence of RAR/HMG interaction. TTNPB-induced RAR␥⌬AB interaction with both HMG-I and HMG-R was dose-dependent (data not shown). Upon sequencing, HMG-R was found to be a variant of HMG-I and showed a deletion of a stretch of 67 nucleotides from the HMG-I sequence. This deletion resulted in a frameshift so that the two proteins are identical in their first 65 amino acids but differ thereafter. A comparison of the putative amino acid sequence of HMG-R with those of HMG-I and another related member HMG-Y (16) is shown in Fig. 2B.
To prove that HMG-R species is present in other RNA samples apart from the HaCaT cDNA library from where it was isolated, RT-PCR was performed on total RNA prepared from cultured keratinocytes and skin rafts (three-dimensional cultures of fibroblasts and keratinocytes). In addition to a major amplified fragment of 320 base pairs (HMG-I), another amplified fragment of 250 base pairs (HMG-R) was observed in all the samples (Fig. 2C). The identity of these bands was confirmed by sequencing the TA-cloned PCR fragments. HMG-R is a novel form, which appears to be produced from the HMG-I/Y gene by alternative splicing using noncanonical splice donor and acceptor sites.
To demonstrate that the full-length RAR␥ also interacts with HMG-I/HMG-R in a ligand-dependent manner in vivo, stably transformed RAR␥ (pAS2-RAR␥) yeast cells were further transformed with pACT2-HMG-I/HMG-R and assayed for ␤-galactosidase activity in the absence or presence of TTNPB. Fulllength RAR␥, by virtue of its intrinsic activation functions (2), elicited a low level of ␤-galactosidase activity in the presence of TTNPB, which was further induced approximately 25-fold in the presence of pACT2-HMG-I and pACT2-HMG-R (Fig. 3A). pACT2-HMG-I/R also showed ligand-dependent interaction with RAR␣ and RAR␤ (Fig. 3, B and C). RAR␥ full-length protein interacted more avidly with HMG-I/R than RAR␥⌬AB (Fig. 3, B and C). To determine the regions of RAR involved in interaction with HMG in vivo, pGBT-RAR␥-DEF, pAS2-RAR␥⌬AB, and pAS2-RAR␥ transformed yeast cells were used. pACT2-HMG-R interacted poorly with RAR␥DEF and RAR␥⌬AB but interacted strongly with RAR␥ (Fig. 3D). These results demonstrate that the LBD, which is present in the E region of RAR, alone is not sufficient for interaction with HMG, but the ligand-independent A/B transactivation function, AF-1 (2), is also required for full interaction. Full-length RAR␥ interacted with both HMG-I and HMG-R in a retinoid dose-dependent manner (Fig. 3E). Similar dose responsiveness has been observed for the retinoid-dependent induction of a number of endogenous RA-responsive genes.
HMG-I/Y are non-histone chromosomal proteins that bind to A/T-rich sequences in the minor groove of the DNA helix, through a DNA binding motif called the "A⅐T hook" (17,18). Three A⅐T hook motifs are present in the HMG-I sequence (amino acids 21-31, 53-63, and 78 -88), whereas HMG-R contains the first two A⅐T hook motifs of HMG-I. Linker histone H1 is a generalized repressor, which plays a pivotal role in com-paction of chromatin into transcriptionally silent fiber (13,19). Transcriptional activation involves remodeling of the silent chromatin fiber through a reduction in the associated H1 complement (20,21). HMG-I/Y proteins, by virtue of their ability to displace H1 from their binding sites in scaffold-associated regions, can derepress the H1-mediated inhibition of transcription (13). HMG-I/Y proteins are associated with transcriptionally active H1-depleted chromatin (13) and are overexpressed in hyperproliferative cancer cells as compared with normal cells (22)(23)(24)(25). In accordance with its role in active transcription, HMG-I has been shown to interact with a number of other transcription factors such as NF-B, ATF-2, Elf-1, Oct-2, Oct-6/Tst-1, and PU.1 (26 -32). In addition to RAR, HMG-I and HMG-R also interacted with RXR␣ (Fig. 4A), PPAR␥ (Fig. 4B), c-Jun (Fig. 4C), and CBP (Fig. 4D). Unlike in the case of RARs, the interaction between PPAR␥ and HMG-I/R was PPAR␥ agonist-independent (Fig. 4B). These data taken together suggest that transcription factors in general may direct HMG-I to promoter regions of target genes. Since acetylation of HMG would destabilize its interaction with DNA, and RARs associate with HATs, we asked the question whether CBP/p300 could potentially acetylate HMG-I. Our data demonstrate that although HMG-I interacted with CBP, it was not acetylated by either CBP or p300 (Fig. 5A), demonstrating its potential to bind to its DNA site in the context of the RAR-CBP/p300 complex.
To further characterize the RAR-HMG interaction observed in the yeast two-hybrid system, we studied their association in vitro. RAR antibodies were immobilized on protein A-and G-Sepharose beads and used for adsorbing baculovirus-produced RARs. The RAR-bound beads were then mixed with radiolabeled in vitro translated Gal4-HMG-I in the presence or absence of TTNPB (1 M). RAR␣ and RAR␥ antibodies specifically immunoprecipitated HMG-I in a ligand-dependent manner (Fig. 5B).
If HMG-I/R proteins are important in nuclear hormone receptor-mediated transcription, then their effects should be quantifiable in transient transfection experiments. Accordingly, we next examined the effect of HMG-I/R on the transcriptional activity of RAR. HeLa cells were transiently transfected with RARE3-tk-CAT, an RA-responsive reporter, with or without HMG-I or HMG-R expression vectors in the presence or absence of the retinoid agonist, TTNPB (1 M). Transfection with either HMG-I or HMG-R sense expression vector resulted in approximately 2-fold induction in the retinoid-dependent expression of RARE3-tk-CAT (Fig. 6A). Since HMG-I is already highly expressed in transformed cells (22)(23)(24)(25), the observed modest induction of retinoid-dependent expression of RARE3tk-CAT by HMG-I/R co-transfection is an expected outcome. In order to abrogate the levels of endogenous HMG-I/R proteins, an antisense HMG-R expression vector, pHMG-R(AS), was constructed and transiently transfected into HeLa cells with the reporter, RARE3-tk-CAT, in the presence of TTNPB. As shown in Fig. 6B, TTNPB (1 M) induced the expression of RARE3tk-CAT, and this activity was inhibited by co-transfection with pHMG-R(AS) in a dose-dependent manner. Finally, we tested whether co-transfection with pHMG-R(S), the HMG-R sense expression vector, rescues the pHMG-R(AS)-mediated inhibition of RARE3-tk-CAT expression. As shown in Fig. 6C, TT-NPB (1 M) induced the expression of RARE3-tk-CAT through the endogenous repertoire of RARs present in HeLa cells, and this activity was inhibited by approximately 50% by co-transfection with 0.01 g of pHMG-R(AS). Furthermore, co-transfection with pHMG-R(S) relieved the inhibitory activity of pHMG-R(AS) on RARE3-tk-CAT expression. These results demonstrate that HMG-I/R is involved in RAR-mediated transactivation in cells.
RARs and other nuclear receptors recruit HAT activities to their site of action by interacting with CBP/p300, ACTR, and SRC-1 (3)(4)(5). These enzymatic activities would acetylate histones H2A, H2B, H3, and H4 (5) and thus partially release the constrained negative DNA supercoiling of the nucleosomal core. However, these activities should be preceded by removal of the linker histone (H1), which constrains the active 10-nm chromatin fiber containing nucleosomal cores into a compact and transcriptionally silent 30-nm fiber. Histone H1 is not acetylated in its DNA binding N-or C-terminal tails (19), suggesting that its affinity to DNA cannot be reduced by HAT activity and that removal of H1 has to be effected by an alternate mechanism. Our data indicate that RARs and other transcription factors are capable of recruiting the H1 displacing activities of HMG-I/HMG-R in a targeted and ligand-dependent manner (Fig. 7). Since HMG-I protein has 19 putative acetylation sites (lysine residues), including two in the DNA binding A⅐T hook domain (Fig. 2B), the acetylation of HMG-I by CBP/p300 could potentially decrease its ability to displace H1. However, in accordance with its proposed role as a H1 displacement factor, either CBP or p300 (Fig. 5A) did not acetylate HMG-I.
In order to destabilize compact higher order chromatin structures and to liberate DNA from the nucleosome at the site of active transcription and facilitate RNA polymerase II complex assembly, both linker and chromosomal histones need to be released by transcription factors. Our model predicts that nu-clear receptors and other transcription factors may recruit HMG-I/HMG-R and HATs in a targeted step wise manner, thereby unraveling linker and core histone assembly and facilitating the formation of transcriptionally competent DNA in the promoter regions of target genes.