|Home | About | Journals | Submit | Contact Us | Français|
The HMGA2 protein belongs to the HMGA family of architectural transcription factors, which play an important role in chromatin organization. HMGA proteins are overexpressed in several experimental and human tumors and have been implicated in the process of neoplastic transformation. Hmga2 knockout results in the pygmy phenotype in mice and in a decreased growth rate of embryonic fibroblasts, thus indicating a role for HMGA2 in cell proliferation. Here we show that HMGA2 associates with the E1A-regulated transcriptional repressor p120E4F, interfering with p120E4F binding to the cyclin A promoter. Ectopic expression of HMGA2 results in the activation of the cyclin A promoter and induction of the endogenous cyclin A gene. In addition, chromatin immunoprecipitation experiments show that HMGA2 associates with the cyclin A promoter only when the gene is transcriptionally activated. These data identify the cyclin A gene as a cellular target for HMGA2 and, for the first time, suggest a mechanism for HMGA2-dependent cell cycle regulation.
HMGA2 belongs, together with HMGA1a and HMGA1b, to the HMGA family of nuclear proteins, whose expression has functional implications in the pathogenesis of several human tumors (2, 20, 47, 50). A strong association between the expression of these proteins and the transformed phenotype has been observed since they were first described, in transformed rat thyroid cells (24, 25), and was later found in several human neoplasias (e.g., colon, prostate, cervical, and thyroid carcinomas) (4, 10, 17, 51). Their expression is very low, if not absent, in adult tissues and is restricted to embryogenesis (11, 28, 32, 55).
The first evidence of a direct role played by these factors in tumorigenesis came from transfection of an antisense construct for HMGA2 in normal rat thyroid cells that prevented the neoplastic transformation induced by myeloproliferative sarcoma virus and kirsten murine sarcoma virus (7). More recently, the increased expression of all three HMGA family members was shown to lead to transformation with anchorage-independent cell growth (53, 54), and the overexpression of HMGA1 was also shown to promote tumor progression in human breast epithelial cells (41). Of additional interest, rearrangements of the HMGA2 gene, resulting in the loss of the acidic C-terminal tail, have been frequently detected in benign human tumors of mesenchimal origin (50). Indeed, transgenic mice expressing the truncated HMGA2 protein develop tumors such as lipomas and natural killer lymphomas (1, 3, 6, 18), while mice expressing wild-type HMGA2 develop pituitary adenomas (19).
Despite this evidence, the molecular events and the precise role played by HMGA2 in cell proliferation and tumorigenesis still need to be defined.
HMGA proteins contain about 100 amino acid residues and have three DNA-binding domains which have been named AT hooks because of their ability to interact with the narrow minor groove of AT-rich DNA sequences (42). These nuclear proteins, by binding to DNA and/or to transcription factors, can organize the assembly of nucleoprotein-DNA transcriptional complexes (called enhanceosomes) at the level of enhancers or promoters, resulting in enhancement or repression of transcription. For this reason, they are referred to as architectural transcription factors (42).
Since a critical step in the assembly of enhanceosomes is the ability of HMGA to establish protein-protein interactions with other nuclear factors, we used different approaches to search for new HMGA2-interacting proteins and identified the transcription factor p120E4F. p120E4F is a ubiquitously expressed GLI-Krüppel-related mammalian transcription factor which was first identified as a cellular factor binding to regulatory regions of the adenovirus E4 promoter and which is responsible for E4 regulation of expression in the course of adenoviral infection (21). p120E4F is likely to play a key role in mammalian cell cycle control. In fact, ectopic expression of p120E4F leads to growth suppression, an effect which is mediated by the interaction with the tumor suppressors pRB (15), p14ARF (43), and p53 (46). Moreover, overexpression of p120E4F in NIH 3T3 cells inhibits progression from G1 to S phase by a mechanism that involves the repression of cyclin A (16). This effect is mediated by p120E4F binding to a cyclic AMP-responsive element (CRE) which is required for full transcriptional activation of cyclin A gene (5, 13). The CRE site is also recognized by CREB and ATF family members, raising the question of how the access of these factors to DNA is modulated.
Here we report that HMGA2 associates in vitro and in vivo with p120E4F. Ectopic expression of HMGA2 activates the cyclin A promoter, counteracting the repressing activity of p120E4F exerted through the CRE element. Moreover, we show that HMGA2 is able to induce the expression of the endogenous cyclin A gene. Finally, chromatin immunoprecipitation experiments indicate that association of HMGA2 with the promoter is regulated during the cell cycle, positively correlating with transcriptional activation. The identification of cyclin A, a key factor in cell cycle control, as a functional cellular target for HMGA2 provides for the first time a mechanism that might reveal the oncogenic functions of this factor.
All of the cell lines used were grown at 37°C in a humidified 5% CO2 incubator in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS), 2 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 μg/ml). CHO is a Chinese hamster ovary cell line, 293 is a human embryonic kidney cell line, and SaOS-2 is a human osteosarcoma cell line.
Plasmids pGEX HMGA2, pGEX TRX44/63, and pGEX TRX53/72 expressing the human HMGA2 and two HMGA2 peptides within the Escherichia coli thioredoxin in fusion with the glutathione S-transferase (GST) were already described (36). Plasmids pcDNA3HA-E4F 2.5K, -E4F 694, -E4F 551, -E4F 427, -E4F Δ60-358, -E4F 262, -E4F 162, -E4F Δ60-783, and -E4F Δ350-783, expressing wild-type p120E4F and deletion mutants in frame with a hemagglutinin (HA) epitope used for in vitro translation and transfections, were described previously (46). pcDNA3HA-E4F Δ153-299 was generated by PCR with pcDNA3HA-E4F 2.5K as a template and the primers 5′-CTGCTGAATTCGAGCTGGGAGACGGTGAGA-3′ and 5′-CCAATAGTCTAGATCCAGCTCCAGAACCTG-3′. pGEX E4F Δ60-358 was obtained by cloning the sequence coding for the deletion mutant Δ60-358 into the EcoRI restriction site of the pGEX-4T1 vector. Plasmids pAR HMGA2wt, pAR HMGA294, pAR HMGA283, pAR HMGA2MAD, pAR HMGA273, pAR HMGA254, and pAR HMGA243, used for the expression of recombinant proteins used in far-Western experiments, were described previously (36). The pcDNA3HA-HMGA2 and pcDNA3-HMGA2 plasmids for the expression of wild-type HMGA2 in transfection experiments were obtained by cloning the HMGA2 open reading frame into the BamHI and XhoI restriction sites of the pcDNA3HA vector and the BamHI site of the pcDNA3 vector, respectively.
The CycAwt reporter plasmid was obtained by cloning into the KpnI and HindIII restriction sites of the pGL2Basic vector a 213-bp fragment of the human cyclin A promoter (from bp −165 to +48 relative to the most 3′ transcription initiation site), which was generated by PCR with oligonucleotides 5′-CTCCGGTACCAGCCAGTTTGTTTCTC-3′ and 5′-TGGCAAGCTTAAGACGCCCAGAGATG-3′. The CycAmut reporter plasmid was obtained by using a splice overlap extension PCR technique with the internal oligonucleotides 5′-TTGACTCGAGTCAAGGCCGCGAGCGCTTT-3′ and 5′-CCTTGACTCGAGTCAAGGCGACAGGGTCA-3′, in order to mutate the CRE site with the insertion of an XhoI restriction site (underlined).
Expression and purification of GST fusion proteins for GST pull-down assays were carried out by standard protocols. Proteins were translated in vitro by using a commercial in vitro transcription-translation kit (TNT; Promega), with [35S]methionine (NEN Life Science) according to the manufacturer's instructions. GST pull-down assays were carried out essentially as described previously (36). For far-Western experiments, recombinant HMGA2 proteins were expressed and purified as previously described (33). One microgram of each of the HMGA2 recombinant proteins was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a polyvinylidene difluoride membrane. Far-Western blotting was then carried out as described previously (36).
Subconfluent 293 cells seeded on 100-mm-diameter petri dishes were transfected with the indicated expression vectors by the conventional calcium phosphate procedure. Thirty-six hours later, cells were washed with ice-cold phosphate-buffered saline and then harvested in 1 ml of ice-cold lysis buffer (150 mM NaCl, 50 mM Tris HCl [pH 7.5], 0.5% NP-40, 10% glycerol) supplemented with 1 mM phenylmethylsulfonyl fluoride and protease inhibitors. Lysis was performed at 4°C for 20 min. The lysates were then clarified by centrifugation and incubated for 3 h at 4°C with anti-HMGA2 antibody (36) prebound to 20 μl of protein A-Sepharose CL-4B (Amersham Biosciences). The beads were then washed three times in 1 ml of ice-cold lysis buffer, and the bound proteins were solubilized by addition of 20 μl of SDS-containing sample buffer. Western blot analysis was performed by standard procedures with anti-HA primary antibody (Roche). Bound primary antibodies were visualized by enhanced chemiluminescence.
Transfections were performed by the standard calcium phosphate precipitation method. CHO cells were plated at a density of 0.7 × 106 cells per 60-mm-diameter culture dish and processed 32 h after removal of the precipitates. For luciferase assay, cells were transfected with 1 μg of the reporter construct, the indicated amount of the expression vectors, and 0.1 μg of pRL-CMV Renilla luciferase expression vector (Promega) to normalize for transfection efficiencies. The assay was performed with the dual-luciferase reporter assay system (Promega) according to the manufacturer's instructions.
For immunocytochemistry, SaOS-2 cells were plated 24 h before transfection at a density of 0.5 × 106 cells per 60-mm-diameter culture dish with a glass coverslip and transfected with 5 μg of pcDNA3HA-HMGA2 expression vector. At 32 h after removal of the precipitates, cells were fixed with 3% paraformaldehyde, permeabilized with 0.3% Triton X-100, and incubated with an anti HA-specific mouse monoclonal antibody (Roche) and an anti-cyclin A rabbit polyclonal antibody (Santa Cruz Biotechnology, Inc.). Immunofluorescence was monitored by incubation with an anti-mouse Cy3-conjugated secondary antibody (Sigma) and an anti-rabbit fluorescein isothiocyanate-conjugated secondary antibody (Sigma). Nuclei were stained with DAPI (4′,6′-diamidino-2-phenylindole).
SaOS-2 cells were plated 24 h before transfection in 100-mm-diameter culture dishes and transfected with 5 μg of pcDNA3HA-HMGA2 or pcDNA3HA expression vector. At 36 h after transfection, RNA was extracted by the guanidinium thiocyanate method, and reverse transcription (RT) was performed in a final volume of 50 μl with 2 μg of total RNA, 50 nmol of random nonamer primers (Sigma), 1× reaction buffer, and 200 U of Moloney murine leukemia virus reverse transcriptase (RNase H Minus; Promega). Twenty-five cycles of PCR were performed in a 50-μl volume with 10 μl of the RT mixture, 0.05 μl of [α-32P]dCTP (Perkin-Elmer Life Science), 1× PCR buffer, and 50 pmol of each cyclin A primer or 5 pmol of each GAPDH (glyceraldehyde-3-phosphate dehydrogenase) primer. cDNA templates were incubated for 4 min at 94°C and amplified with cycles of 92°C for 1 min, 45°C for 1 min, and 75°C for 90 s, with a final extension for 5 min at 75°C. In order to compare the PCR products semiquantitatively, 20 to 30 PCR cycles were performed to determine the linearity of the PCR amplification, and the amplified GAPDH cDNA served as an internal control for cDNA quantity and quality. The primers for cyclin A were 5′-GCCATTAGTTTACCTGGACCCAGA-3′ and 5′-CACTGACATGGAAGACAGGAACCT-3′, and those for GAPDH were 5′-CAGCGACACCCACTCCTCCACCTT-3′ and 5′-CATGAGGTCCACCACCCTGTTGCT-3′. Amplified PCR products were separated on an 8% polyacrylamide gel in 0.5× Tris-borate-EDTA buffer and visualized by autoradiography. Amplification yielded amplified fragments of the predicted sizes (cyclin A, 354 bp; GAPDH, 124 bp).
The indicated amounts of purified recombinant GST-E4F Δ60-358 were incubated with 32P-labeled DNA in 20 μl of binding buffer containing 20 mM HEPES (pH 7.9), 50 mM KCl, 1 mM MgCl2, 0.1 mM EDTA, 5 mM dithiothreitol, 0.4 mM ZnCl2, 0.25 mg of bovine serum albumin per ml, and 4% glycerol. For competition binding reactions, the unlabeled competitors were included in the reaction mixtures at the indicated molar excess. When assayed, high-pressure liquid chromatography-purified recombinant HMGA2wt and BSA proteins were included in the indicated amounts. After incubation for 20 min at room temperature, samples were loaded onto a native 6% polyacrylamide gel in 0.5× Tris-borate-EDTA buffer, electrophoresed at 4°C at 15 V/cm, dried, and exposed to X-ray film (Hyperfilm MP; Amersham Biosciences). The following oligonucleotides (only the upper strand is shown) were used in electrophoretic mobility shift assay (EMSA) experiments: CycAwt, 5′-TGTCGCCTTGAATGACGTCAAGGCCGCGA-3′; CycAmut, 5′-TGTCGCCTTGACTCGAGTCAAGGCCGCGA-3′.
Chromatin immunoprecipitations were performed essentially as described previously (52). NIH 3T3 cells, either serum starved for 60 h or restimulated for the indicated time with 10% FCS, were incubated for 10 min with 1% formaldehyde. After addition of 0.1 M glycine, the cross-linked material was broken with a Dounce homogenizer and sonicated until the DNA was fragmented to 500- to 800-bp fragments. Immunoprecipitation was performed with ProtG-Sepharose (KPL), using 3 to 5 μg of the following antibodies: anti-HMGA2 (36), anti-NF-YB (9), anti-human p120E4F (46), anti-mouse p120E4F (23), and anti-leader binding protein 1 (anti-LBP1) (9). The chromatin solution was precleared by treatment with ProtG-Sepharose for 2 h at 4°C, aliquoted, and incubated with the antibodies overnight at 4°C. Before use, ProtG-Sepharose was blocked twice at 4°C with 1 μg of salmon sperm DNA (sheared at 500 bp) per μl and 1 μg of BSA per μl for 2 h and overnight. PCRs were performed with the following primers: cyclin A, 5′-CTGTAAGATTCCCGTCGGGCCTTCG-3′ and 5′-GTAGAGCCCAGGAGCCGCGAG-3′; α-globin, 5′-GGGCAACTGATAAGGATTCC-3′ and 5′-AGCACCATGGCCACCAATCT-3′.
To identify nuclear proteins interacting with HMGA2, a pull-down assay using a GST-HMGA2 fusion protein with a number of different transcription factors known to be involved in growth control and apoptosis was employed. One of the HMGA2-interacting factors was the E1A-regulated transcription factor p120E4F (21). p120E4F is a GLI-Krüppel-related transcription factor containing six C2H2 zinc finger motifs which are clustered in two separate regions; two motifs are contained within an N-terminal domain and mediate the DNA-binding activity, while the remaining four are grouped within a central region. As shown in Fig. Fig.1A,1A, in vitro-translated 35S-labeled p120E4F was retained by GST-HMGA2 but not by GST alone. Moreover, a bona fide interaction between HMGA2 and p120E4F was observed in the presence of high concentrations of ethidium bromide (Fig. (Fig.1A,1A, lanes 5 to 7), which has been shown to disrupt DNA-dependent protein-protein contact (30).
To demonstrate that HMGA2 and p120E4F can also associate in vivo, vectors expressing HMGA2 and an HA-tagged p120E4F were cotransfected in 293 cells. Cell lysates were immunoprecipitated with anti-HMGA2 antibody, and subsequently the immunocomplexes were analyzed by Western blotting with anti-HA antibody. As shown in Fig. Fig.1B,1B, p120E4F associates with HMGA2 (lane 4). Complex formation was not detected in the absence of HMGA2 overexpression in the cells (lane 3). As control, cell lysates used in the experiment were tested in Western blots for the expression of p120E4F and HMGA2 proteins (Fig. (Fig.1B,1B, lanes 1 and 2). Taken together, these results clearly demonstrate that HMGA2 and p120E4F physically associate both in vitro and in vivo.
Next we assessed the functional consequences of the p120E4F-HMGA2 interaction. p120E4F behaves as a transcriptional repressor, and it recently has been demonstrated that inhibition of cyclin A is an event required for p120E4F-dependent cell cycle arrest in G1 and that the cyclin A promoter activity can be repressed by the binding of p120E4F to the CRE site (16). Therefore, we asked whether HMGA2, by associating with p120E4F, could be involved in controlling the cyclin A promoter. The activity of a reporter construct containing the promoter region of the human cyclin A gene cloned upstream of the luciferase gene (CycAwt) (Fig. (Fig.2A)2A) was therefore tested in transient-transfection assays. Cotransfections of CHO cells with the reporter construct and the p120E4F expression vector result, in agreement with previous data (16), in a twofold repression of the luciferase activity (Fig. (Fig.2B,2B, bar 2). Strikingly, this repression could be completely relieved by cotransfecting increasing amounts of the HMGA2-expressing vector (bars 3 and 4). Moreover, the transfection of the HMGA2 expression vector alone, in the absence of p120E4F expression vector, induced the activity of the reporter vector (bars 5 and 6). To confirm that the observed promoter activity was linked to an increase in the amount of the expressed proteins, aliquots of the lysates were subjected to Western blot analysis (Fig. (Fig.2B,2B, lower panels).
The effect on the cyclin A promoter is specific, since the activity of a reporter construct containing a cyclin A promoter mutated within the CRE (CycAmut) (Fig. (Fig.2A)2A) was not modified upon cotransfection of expression vectors for p120E4F, HMGA2, or a combination of both (Fig. (Fig.2C).2C). All of these results suggest that HMGA2 plays an important role in the regulation of the cyclin A promoter through the CRE.
To further strengthen these observations, we carried out transfection experiments with SaOS-2 cells, which express very low levels of HMGA proteins (our unpublished data), and an HMGA2-expressing vector and analyzed the expression of endogenous cyclin A by a semiquantitative RT-PCR. RNA extracted from cells transfected with the HMGA2-expressing vector or with the empty vector was amplified by using primers for cyclin A and for GAPDH as a control. As can be seen in Fig. Fig.3A,3A, the expression of cyclin A in HMGA2-transfected cells (lane 3) was higher than that in control cells (lane 2), while the expression of GAPDH was constant.
Finally, SaOS-2 cells were transfected with the HMGA2-expressing vector and immunostained for overexpressed HMGA2 and endogenous cyclin A. Figure Figure3B3B (left panel) shows a typical result of this experiment. As can be seen in Fig. Fig.3B3B (right panel), the percentage of cyclin A-expressing cells which stained negative for HMGA2 was about 40%, while it increased to about 70% in HMGA2-positive cells. Statistical analysis with Student's t distribution revealed that the increase in the percentage of cyclin A-expressing cells is significant to a confidence level of greater than 99%. No significant difference was detected in the percentage of S-phase cells, as detected by bromodeoxyuridine incorporation, in HMGA2-positive and -negative cells under these experimental conditions (data not shown).
Together, these results demonstrate an HMGA2-dependent induction of the endogenous cyclin A gene.
As a first step to gain insight into the molecular mechanism responsible for this effect, we mapped the regions of p120E4F that are involved in the interaction with HMGA2. To this end, various deletion mutants of p120E4F cloned in the pcDNA3HA vector were in vitro translated and tested for their ability to bind recombinant GST-HMGA2 in pull-down experiments (Fig. (Fig.4A).4A). As summarized in Fig. Fig.4B,4B, progressive deletions from the p120E4F C-terminal end (constructs E4F 694, E4F 551, E4F 427, E4F Δ60-358, and E4F 262; lanes 1 to 13) removing residues 262 to 783 had no effect on HMGA2 binding even though the four-zinc-finger cluster, which is involved in the binding to p53 (46) and p14ARF (43), was completely deleted. A more extended deletion completely abolished the interaction (construct E4F 162, lanes 14 and 15), suggesting that the region containing the first two zinc fingers is involved in the binding with HMGA2. This result was further confirmed by using the construct E4F Δ153-299, which contains the two N-terminal zinc fingers and shows a strong and specific interaction with HMGA2 (lanes 20 and 21). Deletions from the N-terminal end, E4F Δ60-783 and E4F Δ350-783, confirm this result. The weak binding detected with E4F Δ350-783 is very likely due to the presence of four zinc fingers that, although they have a function different from those at the N-terminal region, share structural similarities.
To map the interaction domains also in vivo, vectors expressing wild-type p120E4F and different p120E4F deletions in fusion with an HA tag were cotransfected with wild-type HMGA2 in 293 cells. As shown in Fig. Fig.55 (upper panel), complex formation was detected with E4F 262 (lane 3) and E4F Δ60-358 (lane 5) but not with E4F 162 (lane 4). As a control, the cell lysates used in the experiment were tested by Western blotting for the expression of p120E4F and HMGA2 proteins (Fig. (Fig.5,5, middle and lower panels, respectively). These results are consistent with those obtained in vitro and clearly demonstrate that binding in vivo requires the region containing the two N-terminal zinc fingers of p120E4F.
To identify the domains in HMGA2 that interact with p120E4F, we generated deletions of the HMGA2 protein and tested for their ability to interact with in vitro-translated p120E4F in a far-Western assay. Data obtained from the experiment shown in Fig. Fig.6A6A (upper panel) demonstrate that the interaction with p120E4F is maintained when the acidic C-terminal tail is removed (HMGA294, HMGA283, and HMGA2MAD). Similarly, the deletion of the region containing the third AT hook (HMGA273) is not effective, while further deletions (HMGA254 and HMGA243) completely abolish the binding. As a control, a Western blot analysis was performed in parallel with an antibody raised against a peptide, derived from the N-terminal sequence of HMGA2, common to all the recombinant proteins (Fig. (Fig.6A,6A, lower panel).
These data, summarized in Fig. Fig.6B,6B, indicate that the region between the second and third AT hooks, from amino acid 54 to 73 of HMGA2, is critical for the interaction with p120E4F. Two peptide sequences spanning the region presumably involved in the interaction were expressed within the active site of the E. coli thioredoxin (TrxA) to assess which region was sufficient for binding with p120E4F. TrxA was chosen since its scaffold provides conformational constraint and stability to short peptide sequences (12). Figure Figure6C6C shows a GST pull-down experiment in which peptides from amino acid 44 to 63 and from amino acid 53 to 72 of the HMGA2 sequence were expressed within the TrxA protein fused to GST and incubated with in vitro-translated p120E4F. GST-TRX44/63 (lane 4) was able to bind p120E4F, whereas GST-TRX53/72 (lane 5) behaved as the negative control GST-TRXA (lane 3), demonstrating that the region from amino acid 44 to 63 within HMGA2 is necessary and sufficient for p120E4F binding. These data indicate that, in the context of HMGA2, the minimal region required for specific interaction with p120E4F corresponds to the second AT hook flanked by the spacer region between the second and third AT hooks, revealing, in this respect, a behavior similar to that shown for HMGA2 and NF-κB (36).
We then investigated at the molecular level how the effect of HMGA2 on the cyclin A promoter could be explained. To this end, we tested by EMSA the effect of a purified HMGA2wt recombinant protein on E4F binding to the cyclin A CRE site (Fig. (Fig.7).7). An affinity-purified GST-E4F Δ60-358 fusion protein gave rise to a protein-DNA complex in a dose-dependent fashion (lanes 2 to 5). This complex is specific, since it was competed by an excess of the self cold probe CycAwt (lane 6) but not by the mutant CycAmut probe (lane 7). The E4F-DNA complex was then tested with increasing amounts of HMGA2wt recombinant protein (lanes 8 to 11). As can be seen, increasing amounts of HMGA2wt progressively inhibited the formation of the E4F-DNA complex, which disappeared with the highest concentration of HMGA2wt added (lane 11) while in parallel a fast-migrating complex appeared. This complex is due to the association of HMGA2wt with the DNA, since it also appears in the absence of GST-E4F Δ60-358 fusion protein (lanes 17 to 19), and is specific, since mutations that replace the AT bases present within and 5′ to the CRE in GC abolish the binding of HMGA2wt to the DNA (data not shown). It is interesting that there is a supershifted complex (lane 10), derived from the association of HMGA2wt with the E4F-DNA complex, as an intermediate before the complete disappearance of the E4F-DNA complex. The HMGA2wt-mediated effect is specific, since the addition of BSA had no effect on E4F-DNA complex formation (lanes 12 to 15).
To further demonstrate that HMGA2 could bind to the cyclin A promoter, we performed chromatin immunoprecipitation experiments with anti-HMGA2 affinity purified antibodies (Fig. (Fig.8A).8A). We used mouse NIH 3T3 fibroblasts, which can be arrested in G0 by incubation with 0.5% FCS and restimulated upon serum addition to reenter in the cell cycle. We took samples at different time points (6, 12, 18, and 24 h) poststimulation, corresponding to the G1, G1/S, S, and G2/M phases of the cell cycle, as previously shown and checked by fluorescence-activated cell sorter analysis (reference 9 and data not shown). The immunoprecipitated DNA was used to amplify the cyclin A promoter, as well as an α-globin promoter that is not expressed in these cells. As shown in Fig. Fig.8A,8A, compared to the control antibody, binding of HMGA2 to cyclin A is negligible in G0 and G1 cells but becomes apparent at 18 h, corresponding to S-phase cells, while at 24 h the binding is no longer visible. We could not observe any association of HMGA2 with the control α-globin promoter at the different time points.
If HMGA2 competes with p120E4F for the binding to the cyclin A promoter, as suggested by the EMSA experiments, then the association of these two proteins with the promoter could be mutually exclusive. To test this hypothesis, antibodies against human and mouse p120E4F were used to immunoprecipitate chromatin from growth-arrested and serum-stimulated NIH 3T3 fibroblasts, and the immunoprecipitated DNA was used to amplify the cyclin A promoter. As shown in Fig. Fig.8B,8B, both human and mouse antibodies failed to detect p120E4F associated with the promoter before 24 h, when essentially all cells have passed S phase and are in G2/M or have reentered G1 (reference 9 and data not shown). In particular, p120E4F is not bound at 18 h, when the promoter is active and HMGA2 is bound. As a control for cell cycle progression, we used anti-NF-YB antibodies; the protein was found associated in G1/S, S, and G2/M as previously described (9). These data document the specific, phase-dependent, and mutually exclusive association of HMGA2 and p120E4F with the cyclin A promoter and show the binding of HMGA2 under conditions in which the promoter is transcriptionally active (26).
Growth is one of the fundamental aspects of the development of an organism. This process is tightly controlled by the coordination of proliferation, differentiation, and apoptosis, with disruption of these processes commonly resulting in tumor formation.
In this study we report that HMGA2, an architectural transcription factor which is directly involved in cell proliferation and neoplastic transformation, associates with the transcriptional repressor p120E4F, resulting in the induction of the cyclin A gene. Cyclin A is a key factor controlling cell cycle progression both at the S phase entry and at the G2/M transition by binding to cdk2 and cdc2 kinases, respectively (37, 45).
Although several reports indicate that HMGA proteins are architectural factors, able to enhance the DNA binding of DNA-interacting proteins promoting the formation of enhanceosomes, data have been published showing an inhibitory effect of HMGA on DNA binding of other factors, usually by sterically blocking the functional binding of transcription factors to their recognition sites in gene promoters (42). Would it thus be possible to extend the mechanism that we found for the cyclin A gene to other p120E4F target genes? Although the cyclin A gene is the only cellular target thus far demonstrated for p120E4F, it is likely that other genes involved in the control of cell cycle progression are regulated by p120E4F, as ectopic expression of p120E4F can induce cell cycle arrest at the G1/S and G2/M transitions even when cyclin A mRNA and protein levels are not reduced (22, 44). The association of HMGA2 with p120E4F through the region containing the DNA-binding domain could therefore constitute a switch that displaces p120E4F-containing complexes from cyclin A and other, yet-to-be-defined, p120E4F target genes. According to this hypothesis, the expression of HMGA2 in human tumors could inactivate the repressing effect of p120E4F on cell growth.
Another important consideration refers to the region of HMGA2 involved in p120E4F protein-protein interaction. This is a very short region comprising and flanking the second AT hook, which we have previously shown to be involved in the interaction with NF-κB (36). Moreover, the corresponding region in HMGA1 is involved in the interaction with other transcription factors as well. Two lysine residues within this region were shown to be acetylated by CBP/p300 and P/CAF and to be critical for the disassembly and assembly of the enhanceosome at the beta interferon gene (34). This short domain, which can be targeted by posttranslational modifications, plays an important role in HMGA proteins in establishing connections with nuclear factors and is therefore crucial for the architectural role played by these factors.
The cyclin A promoter is regulated during G0 and G1 by contiguous cis-acting elements, the CDE-CHR bipartite element (8, 27, 38). Different transcription factors, among which are the pocket proteins (pRb and p107), have been shown to be involved in the binding to these elements, but the mechanism remains to be characterized and is still a matter of controversial discussion (27, 38). The cyclin A promoter region also contains a CRE site. Several reports show that the CRE site plays an important role in transcriptional activation of the cyclin A gene (5, 13, 35, 48, 49). Most studies indicate that CRE is constitutively occupied in G1-phase cells, and it has been proposed that the CDE-CHR could act in concert with CRE to confer proper cell cycle regulation to cyclin A gene expression (49). The CRE is recognized by CREB and ATF family members that belong to the family of basic leucine zipper proteins, which are able to form homo- and heterodimers and cross-family heterodimers with members of the AP1 family. As a consequence, different factors binding to CRE have been identified in different cell types. It has been shown that the different compositions of the subunits, as well as their posttranslational modifications, are involved in the response to different stimuli (e.g., serum induction or cyclic AMP) acting through the CRE to regulate the cyclin A gene (5, 13, 35, 48, 49). The CRE site is also recognized by p120E4F, raising the question of how the access of p120E4F to this site and subsequent transcriptional regulation are controlled (16). Interestingly, it has been hypothesized that auxiliary proteins, such as HMGA, could be possible regulators (16). Indeed, HMGA proteins can modulate the access of transcription factors to DNA and are regarded as highly connected nodes of protein-DNA and protein-protein interactions that influence a diverse array of normal and pathological processes, participating in a wide variety of nuclear processes, ranging from chromosome and chromatinic mechanics to architectural transcription factors (40). In this report we show that HMGA2 activates the cyclin A gene through the CRE and that this effect also occurs in the absence of ectopic expression of p120E4F. This could be explained by assuming a competition with the endogenous p120E4F, but it could also be due to an effect of HMGA2 on other transcription factors binding to CRE. Indeed, it has been demonstrated that HMGA1a could enhance the binding of basic leucine zipper-containing factors, increasing the stability of dimers (14). In addition, we demonstrated by EMSA experiments that HMGA2 can enhance the binding of ATF-2 to the CRE (data not shown). Therefore, we propose that HMGA2 regulation of the cyclin A gene occurs by modulating the access of transcription factors binding to the CRE. This occurs through protein-protein interactions, but since we demonstrated that HMGA2 interacts with DNA in a region containing the CRE, it is also possible that structural modifications of DNA induced by HMGA could play a role in this effect. In this scenario, HMGA2 could therefore function by shifting the balance between p120E4F-mediated repression and CREB/ATF activators binding to the CRE, favoring the latter (Fig. (Fig.9).9). Importantly, in this view HMGA2 is not constitutively bound to the cyclin A promoter throughout the cell cycle. Chromatin immunoprecipitation experiments in fact show HMGA2 bound to the promoter only at 18 h after serum induction, which corresponds to the time when the cyclin A gene is activated. In fibroblasts stimulated by serum, HMGA2 is a delayed-early-responsive gene with peak expression at 7.5 to 20 h (31). The expression of HMGA2 is therefore modulated during the progression through the cell cycle, and its kinetic is compatible with the activation of the cyclin A gene. Moreover, the timely regulated association of p120E4F with the promoter is consistent with the proposed model, since p120E4F binds in G2/M but not during S phase, when the promoter is occupied by HMGA2 and the gene is active. Interestingly, p120E4F is not detected in G0, suggesting that is not implicated in cyclin A repression in growth-arrested cells.
Repression of the cyclin A promoter in quiescent cells was found to be associated with recruitment of the E2F-4 transcriptional repressor (reference 39 and references therein). Our finding that another repressor, p120E4F, associates with the cyclin A promoter defines a second level of repressive regulation for this gene.
Our finding that cyclin A is a direct target gene of HMGA2 is consistent with the role proposed for HMGA2 in cell transformation and cell growth during embryogenesis. The HMGA2 protein has been implicated in the pathogenesis of a variety of benign solid human tumors (2, 20, 47, 50), and its role has been demonstrated by using transgenic mice overexpressing wild-type HMGA2 or HMGA2 mutants (1, 3, 6, 18, 19). Moreover, it was reported that the pygmy phenotype in mice was caused by the disruption of both Hmga2 alleles (55). These mice show a small size and a drastic reduction of body fat content, and, interestingly, Hmga2−/− embryonic fibroblasts have a decreased rate of cell proliferation. This suggested an involvement of HMGA2 in the control of critical steps of the progression through the cell cycle, and our results involve cyclin A as a possible effector of HMGA2 in this phenotype.
In conclusion, this work establishes cyclin A as a mediator of HMGA2-dependent cell cycle regulation. The characterization of other molecular partners of HMGA2 will allow definition of the role played by this factor in cell proliferation and tumorigenesis.
We thank A. Skabar for his contribution in the initial part of this work and P. Sandy for critical reading of the manuscript.
This work was supported by grants from Associazione Italiana per la Ricerca sul Cancro (AIRC), Milan, Italy, to R.M., G.D.S, and V.G.; by ASI grant I/R/129/00 to V.G.; by Università degli Studi di Trieste, Italy, to G.D.S., G.M., and V.G.; and by a MIUR Cofin grant to R.M. and V.G. M.G., R.S., C.I., and G.C. are supported by a FIRC fellowship.