Interaction between HMGA1 and Retinoblastoma Protein Is Required for Adipocyte Differentiation*

It is generally accepted that the regulation of adipogenesis prevents obesity. However, the mechanisms controlling adipogenesis have not been completely defined. We have previously demonstrated that HMGA1 proteins play a critical role in adipogenesis. In fact, suppression of HMGA1 protein synthesis by antisense technology dramatically increased growth rate and impaired adipocyte differentiation in 3T3-L1 cells. Furthermore, we showed that HMGA1 strongly potentiates the capacity of the CCAAT/enhancer-binding protein β (C/EBPβ) transcriptional factor to transactivate the leptin promoter, an adipocytic-specific promoter. In this study we demonstrate that HMGA1 physically interacts with retinoblastoma protein (RB), which is also required in adipocyte differentiation. Moreover, we show that RB, C/EBPβ, and HMGA1 proteins all cooperate in controlling both Id1 and leptin gene transcriptions, which are down- and up-regulated during adipocyte differentiation, respectively. We also demonstrate that HMGA1/RB interaction regulates CDC25A and CDC6 promoter activities, which are induced by E2F-1 protein during early adipocyte differentiation, by displacing HDAC1 from the RB-E2F1 complex. Furthermore, by using Hmga1−/− embryonic stem cells, which failed to undergo adipocyte differentiation, we show the crucial role of HMGA1 proteins in adipocyte differentiation due to its pivotal involvement in the formation of the RB-C/EBPβ complex. Altogether these data demonstrate a key role of the interaction between HMGA1 and RB in adipocyte differentiation.

The HMGA proteins, including HMGA1a, HMGA1b, and HMGA2, are chromatinic proteins that do not have transcriptional activity per se but are able to regulate the transcription of several genes by protein/DNA and protein/protein interactions (1)(2)(3)(4)(5)(6)(7). They have an important role in the process of adipogen-esis. In fact, the targeting of the Hmga2 gene causes a drastic reduction in fat tissue, whereas its activation by the deprivation of its carboxyl-terminal tail results in large accumulation of fat tissue in ectopic areas (8,9). At odds with the Hmga2 gene, the Hmga1 gene seems to have a negative role in adipocytic cell growth control. In fact, it has been shown that suppression of the protein synthesis in the preadipocytic 3T3-L1 cells stimulates cell proliferation (10), and that the impairment of even one allele of the Hmga1 gene by homologous recombination results in the accumulation of fat patches in the abdominal/pelvic region in mice (11). These effects are consistent with a critical role of HMGA1 and HMGA2 alterations in the generation of human lipomas (11).
Our group has previously demonstrated that HMGA1 proteins bind in vivo and in vitro the CCAAT/enhancer-binding proteins (C/EBPs) 4 (10), which are transcription factors required for the transcriptional activation of adipocyte-specific genes (12)(13)(14). Moreover, we have shown that HMGA1 proteins stimulate the C/EBP␤ DNA binding activity during adipocyte differentiation and that HMGA1 strongly potentiates the capacity of C/EBP␤ to transactivate the leptin promoter, an adipocyte-specific promoter (10).
It is known that transcriptional regulation of adipocyte differentiation requires the concerted activity of several transcription factors, which control growth arrest as well as the coordinated expression of adipocyte-specific genes. The retinoblastoma proteins are thought to be critical in controlling cell cycle and terminal differentiation. It has been described that RB is able to induce cell cycle arrest by negatively regulating the E2F family of transcriptional factors (15,16). RB also controls terminal differentiation by binding to and regulating the activity of several tissue-specific trans-activators (17). It has also been demonstrated that RB and C/EBP␤ physically interact and functionally cooperate to activate several promoters during adipogenesis (18). Moreover, it has been shown that HMGA2 * This work was supported in part by grants from the Associazione Italiana interacts with RB and induces E2F1 activity in mouse pituitary adenomas by displacing HDAC1 from the RB-E2F1 complex (19). These data prompted us to investigate the hypothesis that HMGA1 proteins might also bind the RB proteins giving rise to a multiprotein complex able to regulate adipocyte differentiation.
Here we report that HMGA1 protein physically interacts with RB during adipocyte differentiation. This interaction is direct because in vitro translated RB binds GST-HMGA1.
Moreover, we show that RB, C/EBP␤, and HMGA1 proteins cooperate in controlling the Id1 and the obese promoter activity, which are down-regulated and up-regulated, respectively, during adipocyte differentiation. Additionally, we demonstrate that the interaction between HMGA1 and RB is important in the regulation of CDC25A and CDC6 promoters, which are also controlled by the E2F-1 protein, by displacing HDAC1 from the RB-E2F1 complex.

EXPERIMENTAL PROCEDURES
Cell Cultures-The mouse NIH 3T3-L1 cells used in this study were generously donated by Dr. E. Santos (NCI, National Institutes of Health). Cell cultures were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% calf serum (Invitrogen). Induction of adipocyte differentiation in 3T3-L1 cells was essentially performed as described elsewhere (20). Briefly, confluent 3T3-L1 cells were grown in DMEM supplemented with 10% calf serum until confluency. Two days later, they were grown in DMEM supplemented with 10% fetal calf serum and 0.5 mM 1-methyl-3-isobutylxanthine, 10 Ϫ6 M dexamethasone, and 10 g/ml insulin for 48 h. Cells were further cultured in the same culture medium devoid of dexamethasone and methylisobutylxanthine for 6 days. 293 cells were maintained in DMEM containing 10% fetal calf serum (Invitrogen).
Plasmids-To construct the Hmga1b expression plasmid (pCMV/Hmga1b), the full-length Hmga1b cDNA was subcloned into the HindIII site of the expression vector pRc/CMV (Invitrogen). To construct the glutathione S-transferase (GST) fusion genes, the entire Hmga1 coding sequence was amplified by PCR with pairs of primers linked to restriction sites (EcoRI and BamHI) and cloned in the pGEX-2T plasmid (Promega) (pGST/Hmga1b). Hemagglutinin tagged Hmga1 expression plasmids (pHA-A1) containing the entire and various portions of the Hmga1 coding sequence were amplified and inserted into the pCEFL-HA expression vector as follows: pHA-A1 (amino acids 1-96) is constituted by the entire coding sequence; pHA-A1/T is constituted by the first 79 amino acids, including the three AT-hook domains; pHA-A1/1-63 is constituted by the first 63 amino acids, including the first two AT-hook domains and the region between the second and the third AT-hook domain; pHA-A1/1-53 is constituted by the first 53 amino acids, including the first two AT-hook domains; pHA-A1/1-43 is constituted by the first 43 amino acids, including the first AT-hook domain and the region between the first and the second AT-hook domain; pHA-A1/ 23-96 contains the Hmga1b coding sequence deprived of the first 23 amino acids; pHA-A1/31-96 contains the Hmga1b coding sequence deprived of the first 31 amino acids, which include the first AT-hook domain. Plasmids encoding C/EBP␤ and the reporter vector pfluc/B1 are described elsewhere (21,22). The p(Ϫ161)ob-luc plasmid, containing 161 bp of the obese gene promoter driving a luciferase gene is described elsewhere (21). Vectors encoding the wild-type and the mutant RB proteins (R661W and 13S) have been described previously (23). To in vitro translate RB protein, the entire Rb cDNA was obtained by PCR amplification and cloned in pBluescript vector (Stratagene, La Jolla, CA). The CDC25A-luc and CDC6-luc constructs, containing respectively, the CDC25A (Ϫ755, ϩ423) and the CDC6 promoters (Ϫ1534, ϩ225) cloned into the pGL3 basic plasmid upstream of the luciferase reporter gene, were a generous gift from K. Helin.
Bacterial Expression, Protein Purification, in Vitro Translation, and Protein-Protein Binding-For in vitro binding assays, a 900-bp EcoRI-BamHI fragment generated by PCR, including the complete coding sequence Hmga1b cDNA was subcloned in pGEX2T and expressed as GST fusion protein in Escherichia coli strain BL21 bacteria. Stationary phase cultures of E. coli BL21 cells transformed with the pGEX-2T or pGST-Hmga1b plasmids were diluted five times with 400 ml of LB with ampicillin (100 g/ml), grown at 30°C to an A 600 of 0.6, and induced with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside. After an additional 2 h at 30°C, the cultures were harvested and resuspended in 10 ml of cold phosphate-buffered saline (PBS: 140 mM NaCl, 20 mM sodium phosphate (pH 7.4)), 1 mM phenylmethylsulfonyl fluoride, and protease inhibitors (Roche Applied Science). The cells were broken by a French press. The lysate was rocked at 4°C for 20 min with Triton X-100 to 1% and clarified by centrifugation at 12,000 rpm for 10 min at 4°C. Bacterially expressed GST and GST-HMGA1b proteins were bound to glutathione-agarose (Sigma). The beads were washed, and the size and purity of the bound proteins were evaluated by Coomassie staining of SDS-polyacrylamide gel. The recombinant proteins were eluted with a buffer containing PBS, 10 mM reduced glutathione, and 10% (v/v) glycerol. For the His-HMGA1b protein, the supernatant was purified by using nickel-agarose beads supplied with the His-Trap purification kit (Amersham Biosciences) following the manufacturer's instructions, eluted with 500 mM imidazole, and dialyzed in PBS. Equal amounts of GST and GST-Hmga1b proteins (5 g) were used for binding assays. For the RB recombinant protein, transcription and translation reactions were performed with the T7-rabbit reticulocyte lysate kit (Promega, Madison, WI), as suggested by the manufacturer. The in vitro translated RB was allowed to associate with glutathione-agarose-bound GST or GST-HMGA1b for 2 h in lysis buffer at 4°C (18). The pellets were washed four times in lysis buffer; the proteins were dissociated by boiling in loading buffer and electrophoresed on SDS-7.5% polyacrylamide gel. The gel was dried and autoradiographed. The HDAC1 full-length recombinant protein was from Abnova (P01).
Immunoblotting and Immunoprecipitation-Nuclear extracts were prepared as follows. Cells were washed twice in PBS and resuspended in 10 volumes of a solution containing 10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.1 mM EGTA, 0.5 mM DTT (homogenization solution). The cells were disrupted by passage through a 26-gauge needle. Nuclei were collected by centrifugation at 1500 rpm and resuspended in 1.2 ml of extraction solution containing 10 mM Hepes, pH 7.9, 0.4 M NaCl, 1.5 mM MgCl 2 , 0.1 mM EGTA, 0.5 mM DTT, 5% glycerol to allow elution of nuclear proteins by gentle shaking at 4°C. Nuclei were pelleted again by centrifugation at 12,000 rpm, and the supernatant was stored at Ϫ70°C until use. The protease inhibitors leupeptin (5 mM), aprotinin (1.5 mM), phenylmethylsulfonyl fluoride (2 mM), pepstatin A (3 mM), benzamidine (1 mM) were added to both homogenization and extraction solutions. Total protein extracts from human 293 cells were prepared with Nonidet P-40 lysis buffer (1% Nonidet P-40, 50 mM Tris-HCl, pH 8.0, 150 mM NaCl) with protease inhibitors on ice for 15 min. Total extracts from terminally differentiated or undifferentiated NIH 3T3-L1 fibroblasts were prepared as described previously (18). Protein concentration was determined by the Bradford protein assay (Bio-Rad). The antibodies used for immunoprecipitation, Western blotting, and ChIP were as follows: anti-C/EBP␤ (C-19) rabbit polyclonal antibodies (Santa Cruz Biotechnology, Santa Cruz, CA); anti-HA 12CA5 mouse monoclonal antibodies (Roche Applied Science); anti-RB (C-15) rabbit polyclonal antibodies (Santa Cruz Biotechnology); anti-E2F1 (C-20) rabbit polyclonal antibodies (Santa Cruz Biotechnology); anti-acetylated E2F1 polyclonal antibody was raised against a synthetic E2F1 peptide acetylated at position 117, 120, and 125 after conjugation to keyhole limpet hemocyanin (24); anti-HDAC1 (06-720) rabbit polyclonal antibodies (Upstate Biotechnology Inc., Lake Placid, NY); anti-HMGA1 polyclonal antibody was raised against a synthetic peptide located in the NH 2 -terminal region (25,26). To confirm equal loading, the same Western blots were incubated with antibodies versus ␥-tubulin (Sigma). For co-immunoprecipitation experiments, antigens and antibodies were incubated for 1 h before the addition of protein A-Sepharose beads (GE Healthcare). After another 1 h, the beads were collected and washed five times with lysis buffer. The beads were then boiled in SDS loading buffer for immunoblotting analysis. The protein extracts separated by SDS-PAGE were transferred to Immobilon-P transfer membranes (Millipore). Membranes were blocked with 5% nonfat milk proteins and incubated with antibodies at the appropriate dilutions. Bound antibodies were detected by the appropriate horseradish peroxidase-conjugated secondary antibodies followed by enhanced chemiluminescence (Amersham Biosciences).
Transient Transfections and Luciferase Activity Assays-Expression vectors encoding for RB (23), C/EBP␤ (21), and HA-Hmga1 (10) or the same amount of the empty vectors were transfected into 293 or 3T3-L1 pre-adipocytes by calcium phosphate precipitation (27) or using Lipofectamine (Invitrogen), respectively. Reporter constructs (0.2 g) were as follows: p(Ϫ161) ob-luc (22), pfLUC/B1 (21), CDC6 (p[Ϫ1534,ϩ225]) (28), pGL3 CDC25A (29), or pGL3 plasmids. CMV-␤-gal was used for normalization. Cells were harvested 36 h after transfection and lysed, and the protein extracts were used for Western blots or immunoprecipitation assays or luciferase activities (luminometer Lumat LB9507; Berthold). The relative activities were calculated by dividing the normalized activities by the basal activity of the reporter constructs, which were considered to be equal to 1. The data represent the average of three independent experiments, performed in duplicate, with standard deviations.
RT-PCR Analysis-Total RNA was extracted by RNAzol (Tel-Test, Inc., Friendswood, TX). 1 g of total RNA, digested with free-RNase DNase, was reverse-transcribed using random hexanucleotides as primers (100 mM) and 12 units of avian myeloblastosis virus reverse transcriptase (Invitrogen). Subsequent PCR amplification was as follows: 200 ng of cDNA were amplified in a 25-l reaction mixture containing TaqDNA polymerase buffer, 0.2 mM dNTPs, 1.5 mM MgCl 2 , 0.4 mM of each primer, 1 unit of TaqDNA polymerase (PerkinElmer Life Sciences). The PCR amplification was performed for 25 cycles (94°C for 30 s, 55°C for 1 min, and 72°C for 1 min). The primers used for aP2 gene expression were 5Ј-GATGC-CTTTGTGGGAACCTGG-3Ј and 3Ј-TTCATCGAATTCC-ACGCCCAG-5Ј (30). In addition, a set of primers specific for the enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was added to each reaction after 20 cycles of PCR to serve as an internal control for the amount of cDNA tested. The GAPDH-specific primers were 5Ј-ACATGTTCCAATAT-GATTCC-3Ј (forward) and 5Ј-TGGACTCCACGACGTACT-CAG-3Ј (reverse). The reaction products were analyzed on a 2% agarose gel and then transferred by blotting to GeneScreen plus nylon membranes (PerkinElmer Life Sciences). The membranes were hybridized with aP2 cDNA probes. The relative level of aP2 expression was assessed by comparison with the level of GAPDH in the same sample. Quantification of the hybridization signal was performed using a PhosphorImager from GE Healthcare. The images recorded by the PhosphorImager were analyzed by volume integration with the ImageQuant software.
ES Culture and Differentiation of Embryoid Bodies-Stem cell lines used throughout the study were the AB2.1 ES cells (31). ES cells were grown on a layer of mitomycin D-inactivated fibroblasts (SNL76/7). ES cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Invitrogen), 16% fetal bovine serum (Hyclone), 1000 units/ml leukemia inhibitory factor (Chemicon), and 5 ϫ 10 Ϫ5 M ␤-mercaptoethanol. Medium was changed daily, and ES cells were split every 2-3 days. We previously generated wild-type or Hmga1 ϩ/ϩ , Hmga1 ϩ/Ϫ , and Hmga1 Ϫ/Ϫ ES cells (32); Rb Ϫ/Ϫ ES cells were kindly provided (33,34). For differentiation, ES cells were cultured as described (35). Briefly, hanging drops containing 10 3 cells in 20 l of culture medium were maintained for 2 days under the lids of bacteriological dishes filled with phosphatebuffered saline. The embryoid bodies formed were then transferred into bacteriological plates and maintained for 3 days in suspension in cultivation medium supplemented either with 0.1% DMSO or with all-trans-retinoic acid (RA) (10 Ϫ8 M). Medium was changed every day. EBs were maintained 2 days more in suspension in cultivation medium and then were allowed to settle onto gelatin-coated plates in the presence of differentiation medium. This medium consists of cultivation medium supplemented with 85 nM insulin, 2 nM triiodothyronine, and 10% selected fetal calf serum (Invitrogen). Medium was changed every 2 days (35). Attached EBs were harvested after 0, 4, and 8 days for pro-tein analyses and after 13 days for RNA extraction. Alternatively, after 13 days, EBs were stained with Oil Red O to assess adipocyte differentiation.
Electrophoretic Mobility Shift Assay-Protein/DNA binding was determined by electrophoretic mobility shift assay (EMSA), as described previously (25). The E2F1 oligonucleotide (sc-2507, Santa Cruz Biotechnology) was mutated as follows (mutated bases in boldface type) in the AT-mut oligonucleotide: 5Ј-ACTTGGGTTTCGCGCCCTTTCTCAA-3Ј. The interleukin-2 oligonucleotide was previously described (36). The DNA-protein complexes were resolved on 6% (w/v) nondenaturing acrylamide gel and visualized by exposure to autoradiographic films.
Chromatin Immunoprecipitation and Reprecipitation-ChIP was carried out with an acetyl-histone H3 immunoprecipitation assay kit (Upstate Biotechnology) according to the manufacturer's instruction. For ChIP experiments with the antibody anti-RB (C-15) rabbit polyclonal antibodies (Santa Cruz Biotechnology), anti-acetylated E2F1 (24), anti-HDAC1 (06-720) rabbit polyclonal antibodies (Upstate Biotechnology Inc., Lake Placid, NY), and anti-HMGA1 (polyclonal antibody raised against a synthetic peptide located in the NH 2 -terminal region), the conditions were as reported previously (37). For Re-ChIP experiments, complexes were first eluted by incubation for 30 min at 37°C in 250 l of Re-ChIP elution buffer (2 mM DTT, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl, 20 mM Tris-HCl, pH 8.1) and diluted 4-fold in Re-ChIP dilution buffer (1% Triton X-100, 2 mM EDTA, 150 mM NaCl, 20 mM Tris-HCl, pH 8.1) and subjected again to the ChIP procedure. Cross-linking was reversed by incubating samples overnight at 65°C with 20 l of 5 M NaCl. Samples were then incubated in proteinase K solution (10 mM EDTA, 40 mM Tris-HCl, pH 6.5, 40 mg/ml proteinase K) for 1 h at 45°C. DNA was purified with phenol/ chloroform/isoamyl alcohol and precipitated by adding 2 volumes of ethanol and tRNA. PCRs were carried out by standard procedures for a number of cycles optimized to ensure product intensity within the linear phase of amplification. The PCR products were separated on a 2% agarose gel, stained with ethidium bromide, and either scanned using a Typhoon 9200 scanner or blotted and hybridized with specific probes. Immunoprecipitated chromatin was amplified by PCR using the following primers: Mu-cdc6-pr-up 5Ј-AGACCTGGGGCTG- Oil Red O Staining-After 20 days of differentiation treatment, culture medium was removed, and ES cells were washed twice with PBS. Cells were then fixed with 4% formalin/PBS. After cells were washed three times with PBS and incubated in 60% isopropyl alcohol for 10 min, the cells were stained with 1.8% Oil Red O in 60% isopropyl alcohol for 10 min. After washing three times, EBs were scored for adipocyte differentiation and photographed.

HMGA1 Protein Binds RB during Adipocyte Differentiation-To examine the possible interaction between HMGA1
and RB proteins during adipocyte differentiation, we performed a co-immunoprecipitation assay. To this aim, 3T3-L1 cells were differentiated into adipocytes by exposure to fetal bovine serum and differentiating agents (dexamethasone, methylisobutylxanthine, and insulin), as described previously (38). Cells were harvested at time 0 (2 days post-confluence) and at different times during differentiation. Protein extracts were prepared and immunoprecipitated with anti-RB or anti-HMGA1 antisera and immunoblotted with the reciprocal antibodies. The interaction was detected already at 6 h of differentiation and decreased at day 2 (Fig. 1A). The levels of the HMGA1 and RB proteins were evaluated by Western blotting (Fig. 1B).
The interaction between RB and HMGA1 was further investigated by in vitro and in vivo experiments. The cDNA encoding HMGA1b protein was inserted into GST expression plasmid, and the fusion protein GST-HMGA1b was expressed in bacteria. GST-HMGA1b fusion protein was mixed with protein extracts from RB expressing 293 cells ( Fig. 2A). GST-HMGA1b ( Fig. 2A, 2nd lane), but not GST alone (3rd lane), bound specifically to the hypophosphorylated isoform of RB.
For the in vivo experiments, 293 cells were transfected with pCMV-Rb alone or with Hmga1 expression vectors. Cellular extracts were immunoprecipitated with either anti-RB or anti-HMGA1 antisera and immunoblotted with the reciprocal antibodies. Fig. 2B shows that co-expression of RB and HMGA1 proteins resulted in reciprocal co-immunoprecipitation of the two proteins. Cellular extracts derived from 293 cells used for co-immunoprecipitation assays were analyzed by Western blotting (Fig. 2C).
HMGA1-RB Binding Is Direct-To further evaluate the interaction between HMGA1 and RB, a pulldown assay was performed by incubating the two proteins. To this aim, full-length RB labeled with [ 35 S]methionine produced by translation in reticulocyte extracts was allowed to bind to HMGA1b fused to glutathione S-transferase (GST-HMGA1b). The complex was immobilized on a glutathione-Sepharose matrix, and the bound proteins were analyzed by SDS-PAGE followed by Western blotting with the RB antibodies. As shown in Fig. 2D, 2nd lane, GST-HMGA1b was able to precipitate the RB protein. Therefore, we concluded that the two proteins interact and that the interaction is direct. The specificity of the interaction was confirmed by the observation that RB did not adhere to GST resin devoid of HMGA1b under identical conditions (Fig. 2D, 3rd  lane).
C-pocket of the RB Protein Is Necessary for the Binding to HMGA1 Proteins-At least three distinct protein binding regions have been identified in RB as follows: the large A/B pocket (amino acids 395-876), which is the minimal functional domain, corresponding to the binding site for E2F heterodimers; the A/B pocket binds the LXCXE peptide motif in several proteins; the C-pocket binds the nuclear c-Abl tyrosine kinase (23). To define which region of RB protein is required for binding to HMGA1, vectors encoding either wild-type or mutant RB proteins were expressed in 293 cells, and then cellular lysates were mixed with GST-HMGA1b in pulldown assays. The RB constructs used for this experiment are shown in Fig. 3A. One of the mutants used was R661W, a naturally occurring, low penetrance RB allele, which contains a substitution of Trp for Arg-661 in the "B" region of RB. This mutant is defective for the LXCXE and E2F binding. The other mutant, RB 13S has been previously generated and characterized; it contains mutations in the A/B pocket and in the C pocket. This mutant is defective for c-Abl binding (23). After the incubation, the complexes were separated by SDS-PAGE and analyzed by Western blotting with anti-RB antibodies. We determined that the carboxyl-terminal domain of RB is required for an efficient HMGA1 binding (Fig. 3B, upper panel, compare lanes 3 and 4, 13S RB and R661W RB, respectively, with lane 2, wild-type RB). A comparable level of RB expression was revealed by Western blotting by anti-RB antibodies (Fig. 3B, lower panel).
Mapping of the HMGA1 Region Responsible for Binding to RB-To map the HMGA1 region required for binding to RB, we generated a series of progressive carboxyl-terminal deletions of the Hmga1 gene (Fig. 4A). The resulting cDNAs were tagged with the influenza HA epitope into the pCEFL-HA expression vector. These mutants were tested for their interaction in vivo with RB in co-immunoprecipitation assays. Each Hmga1 plasmid was transiently transfected in 293 cells together with a  pCMV-Rb expressing vector. Forty eight hours after transfection, cells were harvested, and protein extracts were immunoprecipitated with anti-HA antibodies. As shown in Fig. 4B, deletion of the carboxyl-terminal tail (pHA-A1b/1-79) and of the third AT-hook domain (pHA-A1b/1-63) did not impair the binding of HMGA1 to RB. Conversely, removal of the region between the second and the third AT-hooks (amino acids 54 -63, construct pHA-A1b/1-53) completely abrogates the interaction of HMGA1 with RB (compare wild-type HMGA1 with the pHA-A1b/1-53 mutant). Moreover, deletion of the region that contains amino acids 1-23 significantly impairs the HMGA1b/RB interaction, as demonstrated by using the 23-96 and 31-96 plasmids, in which the 1-23 region (pHA-A1b/23-96) and the first AT-hook (pHA-A1b/31-96) are deleted, respectively. These results demonstrate that the carboxyl-terminal tail and the third AT-hook domain are not essential for this interaction, whereas regions 54 -63 and 1-23 are. Immunoblotting analysis with anti-HA antibodies showed that equal amounts of wild-type and mutant HMGA1 proteins were produced (Fig. 4C).

Role of the Interaction between HMGA1 and RB in Adipocyte
Differentiation-In the attempt to investigate the role of the interaction between HMGA1 and RB with C/EBP␤ in the activation of adipocyte-specific genes, we analyzed the activity of two promoters regulated by C/EBP␤ in the presence or absence of these proteins. The obese gene codes for leptin and its expression was detected only in adipocyte cells. The Id1, Id2, and Id3 genes are negative regulators for the basic helix-loop-helix transcription factors. It has been shown that the expression of at least one of these genes, Id1, is regulated by C/EBP␤ (21) and down-regulated during adipocyte differentiation (39).
First, we analyzed the obese minimal promoter (Ϫ161 obluc), which contains the C/EBP motif, and it is a natural target of C/EBP transcription factors (22). We showed that the ob gene promoter is activated 4 -5-fold in 293 cells by C/EBP␤ and 18 -20-fold by the cooperative action of C/EBP␤ with HMGA1 (Fig. 5A). Conversely, no activation was observed in the presence of HMGA1 alone (10). Here we show that the presence of HMGA1 and RB together leads to a further enhancement of the ob promoter by C/EBP␤ (Fig. 5A).
To evaluate the effect of RB and HMGA1 on C/EBP-mediated transcription of the Id1 promoter, we used the luciferase reporter vector, pfluc/B1, which contains the B1 enhancer upstream from a heterologous promoter. This enhancer, derived from the Id1 gene, binds to and is specifically regulated by C/EBP␤ (21). Transfection of pfluc/B1 vector in 293 cells resulted in a certain basal activity, because of endogenous C/EBP␤, which was increased when the exogenous protein was co-expressed with the reporter (Fig. 5C). This activity was not observed when a mutated form of this vector, pfluc/B1s, which contains a mutation in the C/EBP␤-binding site, was used (data not shown). Co-expression of C/EBP␤ with either RB or HMGA1 decreased luciferase activity about 2-3-fold (Fig. 5C). The transfection of both RB and HMGA1 together with C/EBP␤ resulted in a further decrease in luciferase activity. The same results have been obtained in pre-adipocytic 3T3-L1 cells (not shown). These experiments demonstrate that RB and HMGA1 can cooperate in the regulation of C/EBP␤-mediated transcription. Fig. 5, B and D, show the levels of the RB, C/EBP␤, and HMGA1 proteins in the transfected cells.
HMGA1 Enhances E2F-1 Activity-Cell proliferation and differentiation have been considered to be mutually exclusive events; however, a close relationship has been established between both cellular processes during the adipocyte differentiation program (40). Re-entry into cell cycle is one of the key events taking place in early adipogenesis. As in most cells, the transition from growth-arrested pre-adipocytes into the S-phase likely depends on the reactivation of the G 1 cyclins/ cyclin-dependent kinases and the retinoblastoma protein RB-E2F pathway that controls the G 1 /S transition of the cell cycle. Association of E2Fs with proteins of the RB family facilitates active repression through recruitment of histone deacetylases (41,42). Upon re-entry into cell cycle of these growth-arrested preadipocytes, the members of the retinoblastoma family are phosphorylated by the cyclin/cyclin-dependent kinase holoenzymes, releasing the E2F complex, resulting in the activation of the E2F target genes (43). To better define the role of the RB/HMGA1 interaction in the regulation of E2F-1 transcriptional activity, we analyzed the activity of two E2F responsive promoter genes, such as CDC25A (44) and CDC6 (45), fused to a luciferase reporter gene, in the presence of RB and HMGA1 proteins. 3T3-L1 cells were used as recipient cells. As shown in Fig. 6, A and C, RB represses CDC25A and CDC6 promoter activity (a reduction to 10 and 50% of the initial signal), whereas HMGA1 does not reduce the promoter activity. When RB was co-transfected with increasing levels of HMGA1, repression by RB of the CDC25A and CDC6 promoter activity was significantly antagonized by the presence of HMGA1 proteins. Western blot analysis showed that the transfected cells expressed adequate levels of the RB and HMGA1 proteins (Fig. 6, B and D).
HMGA1 Proteins Displace HDAC1 from the RB-E2F1 Complex-One of the mechanisms by which the RB-E2F complex represses transcription is the recruitment of the HDAC1 to the E2F-regulated promoters by RB (42). Histone deacetylases and histone acetyltransferases are two counteracting enzyme families whose enzymatic activity controls the acetylation state of proteins, in particular histones. Acetylation of histones regulates gene transcription through its influence on chromatin conformation. In addition, several non-histone proteins are modified in their stability or biological function by the acetylation state of their specific lysine residues.
We thus asked whether the interaction of HMGA1 with RB could displace HDAC1 from the RB-E2F complex increasing E2F1 activity as recently demonstrated for HMGA2 (19). To this aim, we performed co-immunoprecipitation assays in the 3T3-L1 preadipocytic cells transiently transfected with RB in combination or not with wildtype and mutated HMGA1 expression vectors. As shown in Fig. 7A, HMGA1 is able to reduce the binding of RB to HDAC1, although the mutant of HMGA1-(1-43), unable to bind RB, does not exert the same effect suggesting that the interaction between HMGA1 and RB proteins plays a crucial role in the displacement of HDAC1 from RB. The expression of RB, HDAC1, and HMGA1 proteins was verified by West-FIGURE 5. Role of HMGA1 and RB in C/EBP␤-regulated adipocyte-specific genes. 293 cells were transfected as described under "Experimental Procedures" with p(Ϫ161)ob-luc (A) and pfluc/B1 (C) and the indicated plasmids. The relative activities were calculated by dividing the normalized activities with the activity of the empty constructs, which have been considered equal to 1. The data represent the average of three independent experiments, performed in duplicate, with standard deviations. B and D, after transfection, cells were harvested, and cell lysates were divided into 2 aliquots. One of these aliquots was used for Western blot (WB) analysis as a control of protein expression. Protein extracts were separated by SDS-PAGE, transferred to Immobilon-P, and immunoblotted with the indicated antibodies. ern blotting analysis. ␥-Tubulin expression was used to equalize protein loading (Fig. 7B).
To verify this result, we used a cell-free system in which RB and HDAC1 recombinant proteins were incubated with or without increasing amounts of a recombinant wild-type HMGA1b protein. As shown in Fig. 7C, HDAC1 was displaced in a dose-dependent manner by the binding to RB in the presence of 5 and 10 g of HMGA1b recombinant protein. This result demonstrates that HMGA1b directly interferes with the binding between RB and HDAC1. In Fig. 7D, Western blotting analysis shows the amount of recombinant proteins used in the co-immunoprecipitation assay shown in Fig. 7C.
HMGA1 Proteins Bind to E2F1 Target Promoters and Displace HDAC1-HMGA1 proteins allow the assembly of multiprotein complexes by directly binding to the DNA in AT-rich sequences. We used an E2F consensus oligonucleotide (E2FRE) that has an AT stretch compatible with HMGA1 binding in an EMSA with a recombinant GST-HMGA1b protein. As shown in Fig. 8A, 3rd lane, GST-HMGA1b was able to bind to the E2F-responsive oligonucleotide but not to the same oligonucleotide mutated in the AT-stretch flanking the E2F-consensus sequence (AT-mut). Binding specificity was also demonstrated by competition experiments showing a loss of binding with the addition of 200-fold molar excess of unlabeled interleukin-2 promoter region that contains AT-rich HMGA1-binding sites (36).
We next evaluated whether HMGA1 is able to bind Cdc25A and Cdc6 promoters, two E2F-responsive promoters, and whether it is able to displace HDAC1 in vivo. To this aim, we performed ChIP assays in 3T3-L1 cells transfected or not with a vector encoding HMGA1b protein. Immunoprecipitated chromatin was then analyzed by semiquantitative PCR, using primers spanning the E2F binding regions of the Cdc25A and Cdc6 promoters. Occupancy of these regions by HMGA1 was clearly FIGURE 6. Regulation of E2F-responsive CDC25A and CDC6 promoters by RB and HMGA1 in pre-adipocytic cells. 3T3-L1 cells were transfected as indicated under "Experimental Procedures." Histograms show the luciferase activities of extracts from 3T3-L1 cells co-transfected with the CDC25A-luc (A) or CDC6 (C) reporters and the indicated RB and HA-Hmga1b plasmids. The relative activities were calculated by dividing the normalized activities with the activity of the empty vectors, which have been considered equal to 1. The data represent the average of three independent experiments, performed in duplicate, with standard deviations. After transfection, cell lysates were divided into 2 aliquots. One of these aliquots was used for transactivation assays, and the other for Western blot (WB) analysis as a control of protein expression. Protein extracts were separated by SDS-PAGE, transferred to Immobilon-P, and immunoblotted with the indicated antibodies (B and D). detectable in anti-HMGA1-precipitated chromatin from untransfected or from HMGA1-transfected 3T3-L1 cells, being HMGA1 proteins endogenously expressed in these cells (Fig. 8B).
Moreover, in the samples overexpressing HMGA1 proteins, a drastic reduction of HDAC1 binding on both promoters was detected (Fig. 8B). By Re-ChIP assays we confirmed the role of HMGA1 in the displacement of HDAC1 from RB. In fact, the anti-RB complexes were released, reimmunoprecipitated with anti-HDAC1, and then analyzed by semiquantitative PCR (Fig.  8C). This assay demonstrates that the HDAC1/RB binding was decreased in the presence of HMGA1 on the Cdc25A and Cdc6 promoters. The expression of HDAC1 and HMGA1 proteins was verified by Western blotting analysis. ␥-Tubulin expression was used to equalize protein loading (Fig. 8D).
HMGA1 Overexpression Promotes E2F1 Acetylation-We have shown that HMGA1 displaces HDAC1 from RB at the E2F-binding sites on the Cdc25A and Cdc6 promoters, a process that could account for the decreased RB activity following HMGA1 overexpression. In fact, HDAC1 controls the stability and biological function of proteins, involved in gene transcription, by their acetylation status. An example of this is E2F1, whose acetylation enhances DNA binding and stabilizes the protein (46).
To analyze the E2F1 acetylation status on the Cdc25A and Cdc6 promoters and correlate it with the interaction between HMGA1 and RB, we transfected 3T3-L1 cells with an empty vector or a vector encoding HMGA1b protein and subjected the lysates to a ChIP assay with a specific anti-acetylated E2F1 antibody. As shown in Fig. 9A, E2F1 acetylation was enhanced by the HMGA1 overexpression. This result, consistent with our previous data demonstrating the role of the interaction of HMGA1 with RB in displacing HDAC1 from the E2F-responsive promoters, was also confirmed by an additional experiment in which total lysates from cells transfected as in Fig. 9A were analyzed by Western blot to evaluate the amount of E2F1 acetylation (Fig. 9B). Taken together, these results suggest that HMGA1 overexpression promotes the E2F1 acetylation status.
Hmga1-null ES Cells Fail to Undergo Adipocyte Differentiation and Are Required for the Induction of the RB-C/EBP Complex during Adipocyte Differentiation-To better define the role of the HMGA1 proteins in differentiation, we generated wild-type, Hmga1 Ϫ/Ϫ , Hmga1 ϩ/Ϫ ES cells (32). We investigated the ability of wild-type, Hmga1 Ϫ/Ϫ , Hmga1 ϩ/Ϫ , and Rb Ϫ/Ϫ ES cells (34) to undergo adipocyte differentiation following RA and hormonal treatment, as described (35,47). Rb Ϫ/Ϫ ES cells were chosen as a control of impaired adipocyte differentiation (18). After 20 days of differentiation treatment, EBs were stained with Oil Red O to evaluate the adipocyte differentiation. Clusters of cells filled with lipid droplets appeared in the wild-type EBs, whereas we observed a dramatic reduction in the capacity to achieve adipogenesis in Hmga1 Ϫ/Ϫ EBs. This deficiency was characterized by both a drastically reduced percentage of Oil Red O-positive Hmga1 Ϫ/Ϫ EBs compared with wildtype (33 and 100%, respectively) and a drastically reduced number of adipocytes per EB in Hmga1 Ϫ/Ϫ EBs compared with wild-type (see Oil Red O staining panel in Fig. 10). As expected, no cells showing adipocyte differentiation appeared in the Rb Ϫ/Ϫ EBs (Fig. 10A) and in control wild-type, Hmga1 Ϫ/Ϫ , and Rb Ϫ/Ϫ EBs treated with DMSO. Several Hmga1 null clones were analyzed for morphological differentiation and expression of adipocyte-specific markers by semiquantitative RT-PCR assay. In particular, in Hmga1 ϩ/ϩ EBs, aP2 expression was already detectable in control, DMSO-treated, cells and increased after differentiation (Fig. 10B, 1st and 2nd lanes). Conversely, in Hmga1 Ϫ/Ϫ EBs, aP2 was absent in untreated cells and remained absent or was only partially induced in few clones after differentiation (Fig. 10B, 3rd to 6th lanes, and data  not shown). Because all the clones showed a similar behavior in failing a complete terminal adipocyte differentiation, only two of them are shown in the figure. As expected, aP2 gene expression was undetectable in Rb Ϫ/Ϫ cells (Fig. 10B, 7th and 8th  lanes). No bands are seen in nonreverse-transcribed RNAs, excluding a possible DNA contamination (data not shown).
Terminal adipocyte differentiation requires the interaction between the hypophosphorylated RB and the transcription factors C/EBPs. This interaction is necessary, because Rb Ϫ/Ϫ fibroblasts cannot be induced to differentiate into adipocytes (18). We investigated whether the presence of the HMGA1 protein is necessary for this complex formation. First, we evaluated the induction of HMGA1, C/EBP␤, and RB expression by Western blotting in wild-type Hmga1 ϩ/ϩ , heterozygous FIGURE 7. HMGA1 proteins compete with HDAC1 for RB binding. A, lysates from 3T3-L1 cells transfected with plasmids expressing RB in combination or not with wild-type HMGA1b and truncated HMGA1b-(1-43) were subjected to immunoprecipitation (IP) with anti-RB antibody and then blotted against HDAC1, as indicated on the right. Western blot (WB) with RB antibodies was used as a control of the equal loading of the immunoprecipitated RB protein. B, lysates from cells transiently transfected as in A were analyzed by Western blot with specific antibodies to verify protein expression levels. ␥-Tubulin expression was used as a control for equal protein loading. C, HDAC1 and RB recombinant proteins were co-immunoprecipitated in the presence of 5 or 10 g of recombinant HMGA1b protein. The samples were separated by SDS-PAGE and transferred onto Immobilon-P membranes. The filters were probed with either HDAC1 or RB antibodies, as indicated. D, Western blotting analysis shows the amount of recombinant proteins used in C.
Hmga1 ϩ/Ϫ , and homozygous Hmga1 Ϫ/Ϫ ES cells induced to differentiate into adipocytes. To this aim, nuclear proteins were extracted at different times of adipocyte differentiation (0 h, 4 days, and 8 days). The expression of these proteins in this system was different from that shown in the pre-adipocytic 3T3-L1 cells. In fact, HMGA1 protein levels are normally high during embryonic development and in ES cells (48). Indeed, its expression was already high at time 0, which corresponds to the beginning of the hormonal treatment, and it did not change with the treatment (Fig. 10C). C/EBP␤ expression was already high, because its maximal induction is reached upon retinoic acid treatment, which precedes hormonal induction of differentiation in this system. RB protein was present in both its phosphorylated and hypophosphorylated isoforms at time 0, but it almost exclusively accumulated in its hypophosphorylated isoform following the differentiating treatment (Fig. 10C).
We then investigated whether the interaction between RB and C/EBP␤ could take place in the absence of HMGA1 during adipocyte differentiation. Equal amounts of protein extracts were immunoprecipitated with anti-C/EBP␤-specific antisera. The immunocomplexes were transferred on membrane and immunoblotted with the anti-RB antibodies. The results are shown in Fig. 10C; the interaction between RB and C/EBP␤ was detected only in the wild-type Hmga1 ϩ/ϩ ES cells. Consistently, a significant reduction in the RB-C/EBP␤ complex was observed in the Hmga1 ϩ/Ϫ ES clones, whereas it was completely absent in the homozygous Hmga1 Ϫ/Ϫ ES cells. Moreover, we have demonstrated that HMGA1-C/EBP␤ coprecipitated in the RB Ϫ/Ϫ ES cells, indicating that the interaction between these two proteins can also occur in the absence of RB (data not shown).

DISCUSSION
Adipocyte differentiation requires the concerted activity of several transcription factors, which control growth arrest and the coordinated expression of adipocyte-specific genes. Among these factors, an important role has been ascribed to C/EBP and RB proteins. In fact, C/EBP knock-out mice showed severe defects in adipose tissue formation (49), and the various member of the C/EBP family are expressed at specific periods during the differentiation of 3T3-L1 cells (12,14). Hormonal stimulation causes C/EBP␣ and C/EBP␤ levels to increase and induce the expression of the transcription factor peroxisome proliferator-activated receptor ␥ (50). Peroxisome proliferator-activated receptor ␥, in turn, increases the expression of C/EBP␣, which promotes the induction of several adipocyte-specific genes, such as the fatty acidbinding protein 422/aP2 (51) and the obese gene (52). A critical role in adipogenesis is also taken by RB. In fact, Rb Ϫ/Ϫ fibroblasts fail to differentiate into fat-storing cells (18).
Previously, our group has proposed a critical role for the HMGA1 proteins in adipocyte differentiation of the 3T3-L1 cells, and this role was played by the interaction with the C/EBP proteins (10). Therefore, the aim of this work was to investigate a possible interaction of HMGA1 with RB. Here we demonstrate that HMGA1 binds RB in 3T3-L1 cells and in 293 cells by reciprocal co-immunoprecipitation experiments. We also FIGURE 8. HMGA1 proteins bind to E2F1 target promoters and displace HDAC1. A, EMSA performed incubating 100 ng of GST or GST-HMGA1b recombinant proteins with a radiolabeled E2F-responsive oligonucleotide (E2FRE). To assess the specificity of the binding, GST-HMGA1b protein was incubated with a radiolabeled E2F-responsive oligonucleotide mutated in the region rich in AT bases (AT-mut) or in the presence of a 200-fold excess of unlabeled interleukin-2 oligonucleotide used as competitor. B, lysates from 3T3-L1 cells transfected with empty vector or HMGA1 expressing plasmids were subjected to ChIP using specific polyclonal antibodies, as indicated. Immunoprecipitates from each sample were analyzed by PCR using primers that cover a region of mouse Cdc25A and Cdc6 promoters that contain the E2F1-binding site. To assess the specificity of the binding, PCR amplification of the immunoprecipitated DNAs were performed also using primers for the mouse gapdh gene promoter. C, in Re-ChIP experiments, soluble chromatin from transfected 3T3-L1 cells as described in B was immunoprecipitated with anti-RB and anti-HDAC1 antibodies, eluted, and reimmunoprecipitated with anti-HDAC1 and anti-RB respectively. The purified DNA was used as template for PCR with primers that amplify the mouse Cdc25A promoter region that contains the E2F1-binding site. D, cellular extracts derived from untransfected or HMGA1 overexpressing 3T3-L1 cells used for ChIP and Re-ChIP experiments were analyzed by Western blotting (WB). retinoic acid-induced wild-type ES cells. This would suggest the requirement of HMGA1 proteins for the interaction between RB and C/EBP␤ proteins. In the homozygous Rb Ϫ/Ϫ ES cells, we were instead able to coprecipitate C/EBP␤ and HMGA1 proteins, indicating that the interaction between these two proteins can occur also in the absence of RB. These data, taken together, indicate that HMGA1 proteins are required for the formation of the RB-C/EBPs protein complex that is essential for the expression of several adipocyte-specific genes.