|Home | About | Journals | Submit | Contact Us | Français|
The transcription factor CCAAT/enhancer-binding protein α (C/EBPα) coordinates proliferation arrest and the differentiation of myeloid progenitors, adipocytes, hepatocytes, keratinocytes, and cells of the lung and placenta. C/EBPα transactivates lineage-specific differentiation genes and inhibits proliferation by repressing E2F-regulated genes. The myeloproliferative C/EBPα BRM2 mutant serves as a paradigm for recurrent human C-terminal bZIP C/EBPα mutations that are involved in acute myeloid leukemogenesis. BRM2 fails to repress E2F and to induce adipogenesis and granulopoiesis. The data presented here show that, independently of pocket proteins, C/EBPα interacts with the dimerization partner (DP) of E2F and that C/EBPα-E2F/DP interaction prevents both binding of C/EBPα to its cognate sites on DNA and transactivation of C/EBP target genes. The BRM2 mutant, in addition, exhibits enhanced interaction with E2F-DP and reduced affinity toward DNA and yet retains transactivation potential and differentiation competence that becomes exposed when E2F/DP levels are low. Our data suggest a tripartite balance between C/EBPα, E2F/DP, and pocket proteins in the control of proliferation, differentiation, and tumorigenesis.
The CCAAT/enhancer-binding protein α (C/EBPα) belongs to a family of bZIP (basic region leucine zipper) transcription factors that are involved in cell cycle arrest and induction of lineage-specific differentiation genes in several cell types, as shown during hepatic, adipogenic, granulocytic, skin, lung, and placenta development (2, 12, 19, 26, 29, 43, 57). C/EBPα knockout mice die perinatally from hypoglycemia due to defective expression of liver-specific enzymes required for glucose homeostasis (56). Furthermore, C/EBPα-deficient mice lack white adipose tissue and granulocytes of the eosinophil and neutrophil lineages (56, 59). The C/EBPα gene may mutate to produce oncogenic protein forms that are defective for cell cycle inhibition and that no longer promote cell differentiation (19, 22, 34, 37).
Several lines of evidence suggest an intricate relationship between C/EBPα and early gene 2 factor (E2F) gene products. The E2F-dimerization partner (DP) family of transcription factors regulate key genes of cell cycle progression, apoptosis, and DNA damage (17, 33, 41, 44). Formation of E2F-DP heterodimeric complexes is required for induction of E2F-regulated genes (1, 15, 25), while association with pocket proteins (the retinoblastoma family, retinoblastoma protein [pRB], p107, and p130) inhibits the transcriptional activity of E2F and thus restricts cell cycle progression and tumor development (35). Proliferation arrest by C/EBPα involves repression of E2F target genes (50). The murine C-terminal basic region mutant 2 (BRM2) of C/EBPα is unable to repress E2F transcription and induces a myeloproliferative disorder in the mouse (42, 43). BRM2 is of particular interest, since it resembles recurrent C/EBPα mutations isolated from human acute myeloid leukemia (AML) (34, 37, 42). E2F interacts with the bZIP domain of C/EBPα, and yet the N-terminal transactivation domain of C/EBPα is also required for the suppression of E2F genes (9, 18). Curiously, mice expressing the transactivation-deficient N-terminal truncated C/EBPα isoform p30 that is unable to repress E2F develop AML (3, 22), suggesting that the transcriptional activity of C/EBPα is important for its tumor-suppressing function. Failure to abrogate E2F-mediated proliferation may therefore only partially explain the BRM2 phenotype, since the transactivation domain of BRM2 remains intact and the transcriptional capacity of BRM2 may persist (9, 21). Along these lines, it has been shown that C/EBPα-mediated proliferation arrest and differentiation capacity can be separated from each other by highly malignant E7 papilloma viral oncoproteins, independently of pRB (32). BRM2 knockin mice display defects in proliferation and in differentiation of adipocytes and neutrophils (43), suggesting that altered interaction with E2F is involved in regulating both, cell cycle progression, and differentiation. Moreover, a fraction of adult BRM2 mice recover granulopoiesis (42), suggesting that functionality of BRM2 may be restored by readjustment of the balance between E2F and C/EBPα.
The data presented here identify DP as a novel C/EBPα interacting protein and E2F-DP complexes as inhibitors of the transcriptional activity of C/EBPα. Both “activator” and “repressor” E2Fs (E2F1, E2F3, E2F4, and E2F5) in conjunction with DP, but independently of pocket proteins, may suppress transactivation of C/EBPα by interfering with its binding to DNA. DNA binding, transactivation, and differentiation potential of the BRM2 mutant was restored upon knockdown of either E2F or DP proteins. These data suggest an intricate interdependence between transcriptional inactive C/EBPα bound to E2F-DP and transcriptional active C/EBPα bound to DNA. Our data suggest that the ratio between E2F-DP complexes and C/EBPα critically determines precursor cell expansion and C/EBPα-mediated differentiation and suggests a therapeutic opportunity in readjustment of this balance.
pCMV-HA-hDP2, pCMV-HA-hDP1, pBabe-ER-E2F1 wild type (WT) and E132 (33), pE2Fx6-TATA-LUC reporter, and the pRB-binding-defective point mutant E2F1 Y411C (15) were provided by Kristian Helin. The pcDNA3 based amino-terminal hemagglutinin (HA)-tagged E2F constructs were obtained from Stefan Gaubatz. The coding regions of E2F1 and E2F4 contained in the BamHI-EcoRI fragments of these constructs were introduced into the pGEX4T2 BamHI-EcoRI site to generate glutathione S-transferase (GST) fusion proteins. All DP1 and DP2 GST fusion proteins were obtained by introducing a PCR product containing a BamHI and a NotI site, respectively, 5′ and 3′ of the coding region. Alternatively, a PCR product with BamHI-BamHI site was used to generate pCMV-HA-DP2Δ83. The cyclin A binding-deficient pcDNA1-E2F1Δ24 construct was provided by Liang Zhu (24). A BamHI fragment containing the E2F1 DNA-binding-deficient mutant E132 was cloned into pcDNA3 and pGEX4T1. The pBabePuro-based retroviral C/EBPα basic region point mutants (BRM2, I294A and R297A; BRM3, D301A and K304A; and BRM5, Y285A) were obtained from Claus Nerlov (43). For transient transfection, EcoRI-BamHI fragments of these mutants were fused to a carboxy-terminal triple FLAG, contained in a pcDNA3 plasmid. For bacterial expression purposes, DNA encoding the bZIP domain of C/EBPα (both WT and BRM2 mutant), comprising the basic region and the leucine zipper, were fused to an N-terminal His7-tagged protein, contained within the pQLinkH vector (47). The C/EBP-responsive −82 cMGF-luciferase reporter has been described previously (51). Small hairpin RNAs (shRNAs) were expressed in psiRNA (Invivogen), harboring a zeocin selection marker fused to green fluorescent protein (GFP). shRNA oligonucleotides against DP1, E2F1, E2F3, or E2F4 were designed by using the InvivoGen's siRNA Wizard program and subjected to BLAST searching to exclude homology to any additional known sequences. Double-stranded DNA oligonucleotides were ligated to the BbsI site of the psiRNA construct. As a control, a nonspecific shRNA was used. The sequences targeted by shRNAs were as follows: control, 5′-GTC CAT CGA ACT CAG TAG CT-3′; DP1, 5′-GCA GCA TCT CCA ATG ACA AAT-3′; E2F1, 5′-GCC AAG AAG TCC AAG AAT CAT-3′; E2F3, 5′-GCT CAC CAA GAA GTT CAT TCA-3′; and E2F4, 5′-CGA GAG TGA AGG TGT CTG T-3′.
293T, 3T3-L1, Phoenix-E ecotropic retroviral packaging cells, mouse embryonic fibroblasts (MEFs) generated from the C/EBPα knockout strain (56), and pRB−/− p107−/− p130−/− MEFs (10) were grown in Dulbecco modified Eagle medium (DMEM) plus GlutaMAX, supplemented with 10% fetal bovine serum (FBS; Gibco). Prior and during adipogenesis DMEM was replaced by alpha-MEM medium plus GlutaMAX. For detection of endogenous proteins, cells were lysed with RIPA buffer (1% NP-40, 0.1% sodium dodecyl sulfate [SDS], 50 mM Tris-HCl [pH 7.5], 50 mM NaCl, and protease inhibitors) while transiently transfected cells were lysed with Triton buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1 mM EDTA, 1% Triton, and protease inhibitors). Immunoblot analyses were conducted with the following Santa Cruz antibodies: rabbit anti-C/EBPα (14AA), mouse anti-E2F1 (KH95), rabbit anti-E2F1 (C-20), rabbit anti-E2F3 (N-20), rabbit anti-E2F4 (C-20), rabbit anti-DP1 (K-20), rabbit anti-DP2 (C-20), and mouse anti-α-tubulin (B-7), as well as goat anti-GST (Amersham), mouse anti-HA 12CA5 (Roche), mouse anti-FLAG M2 (Sigma), and mouse anti-DP1 Ab-6 (NeoMarkers). Antigen-antibody complexes were detected either by chemiluminescence (ECL System; Amersham) using secondary antibodies conjugated to horseradish peroxidase or by the Odyssey infrared imaging system, using secondary antibodies conjugated to IRDye (Biomol). The latter allowed quantification of signals using the Odyssey software.
Coimmunoprecipitation assays were performed with anti-FLAG-M2-agarose (Sigma) according to the manufacturer's specifications. Immunoprecipitated samples and input control (one-fifth of input lysate) were resolved by immunoblotting. For reporter assays, 293T cells and MEFs were transfected with Metafectene (Biontex) and TransIT-LT1 (Mirus), respectively, according to the manufacturer's specifications. Cells were lysed at 48 h posttransfection, and the reporter activities were determined in a Berthold Lumat LB9501. The luciferase values were normalized to protein levels, and protein expression was controlled by immunoblotting.
Cells were treated with trypsin and resuspended in phosphate-buffered saline (PBS), and GFP-positive cells were sorted by using FACSVantage SE.
Total RNA of cells was isolated using the TriPure isolation reagent (Roche) and cDNA was prepared by using random primers and SuperScriptII reverse transcriptase (Invitrogen) according to the manufacturer's specifications. Real-time reverse transcription-PCR was performed on an ABI Prism 7000 (Applied Biosystems) using SYBR green PCR master mix 7000 (Applied Biosystems) according to the manufacturer's instructions. The relative RNA expression levels were calculated by using the comparative threshold cycle (CT) method, and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) expression values were used to normalize the analyzed RNA levels. The sequences of the primer pairs can be obtained upon request.
Cells were cross-linked with 1% formaldehyde for 20 min. After the addition of glycine to a final concentration of 250 mM, the cells were washed twice with ice-cold PBS and swelled in hypotonic buffer (10 mM Tris-HCl [pH 8], 10 mM NaCl, 0.2% NP-40, and protease inhibitors). Nuclei were resuspended in lysis buffer (50 mM Tris-HCl [pH 8], 10 mM EDTA, 1% SDS, and protease inhibitors) and incubated 20 min on ice prior to sonication with Bioruptor (Diagenode) to an average DNA length of 500 bp. Cell debris were removed by centrifugation at 10,000 × g for 20 min. At this point 10% of the input was kept as a control, whereas the rest of the supernatant was diluted with 2 volumes of dilution buffer (1% Triton X-100, 150 mM NaCl, 20 mM Tris-HCl [pH 8], 2 mM EDTA, and protease inhibitors) and incubated overnight with 5 μg of rabbit anti-C/EBPα (14AA; Santa Cruz) or 5 μg of rabbit IgG (Sigma). Immunoprecipitates were collected with protein G-Dynabeads (Invitrogen). Beads were washed four times with Wash I buffer (20 mM Tris-HCl [pH 8], 2 mM EDTA, 50 mM NaCl, 1% Triton X-100, 0.1% SDS, and protease inhibitors), twice with Wash II buffer (10 mM Tris-HCl [pH 8], 1 mM EDTA, 250 mM LiCl, 1% NP-40, 1% deoxycholic acid, and protease inhibitors), and twice with TE buffer (10 mM Tris-HCl [pH 8], 1 mM EDTA, and protease inhibitors). Protein-DNA complexes were eluted with 100 μl of elution buffer (1% SDS and 50 mM NaHCO3). NaCl was added to a final concentration of 200 mM, and cross-linking was reversed by incubation at 67°C overnight. After 30 min RNase A treatment, DNA was purified by using QIAquick PCR purification kit (Qiagen) and quantified with ABI Prism 7000 (Applied Biosystems) using SYBR green PCR master mix 7000 (Applied Biosystems) according to the manufacturer's instructions. ChIP DNA levels were calculated by using the comparative CT method, normalized to input, and expressed as anti-C/EBPα versus IgG. The primers used in the present study have been previously described: peroxisome proliferator-activated receptor gamma 2 (PPARγ2) promoter (52), PPARγ2 distal (2.8 kb upstream of ATG) (39), and apolipoprotein 2 (AP2) promoter (52).
GST and His fusion proteins were expressed in Escherichia coli BL21(DE3) and prepared according to standard procedures. For in vitro binding assays, in vitro-translated 35S-labeled proteins (Promega TNT kit) were incubated with equal amounts of affinity-purified GST fusion proteins coupled to glutathione-Sepharose. GST protein served as a negative control. The beads were washed three times with NP-40 buffer (0.4% NP-40, 50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM EDTA) and once with Tris-buffered saline. The proteins were resolved by SDS-PAGE and visualized by autoradiography. GST fusion proteins were identified by Coomassie blue staining to verify that equal amounts were present in all reactions. Alternatively, GST proteins were incubated with an equal volume of lysates of 293T cells transfected with C/EBPα expression plasmids. Bound proteins were detected by immunoblotting. GST fusion proteins were identified by Ponceau S staining. His-tagged proteins were bound to Ni beads (Qiagen) in a buffer (pH 8) containing 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 0.3% NP-40, and 20 mM imidazole. This buffer was also used for washing the beads.
Retroviral plasmids were transfected into Phoenix E by using the CaPO4 method. Culture supernatant were recollected 48 h after transfection, passed through 0.45-μm-pore-size polyvinylidene difluoride filters (Millipore), supplemented with 5 μg of Polybrene/ml, and used to infect subconfluent layers of C/EBPα−/− MEFs or 3T3-L1. After 10 h of infection, the cells were selected for 3 days in the presence of 2 μg of puromycin (InvivoGen)/ml.
3T3-L1 or C/EBPα−/− MEFs were differentiated with alpha-MEM medium, supplemented with 10% FBS, 0.5 mM IBMX [3-isobutyl-1-methylxanthine]), 10 μg of insulin/ml, and 1 μM dexamethasone (DEX) for 2 days. From day 3 onward, cells were cultured in alpha-MEM containing only 10% FBS and 10 μg of insulin/ml. The medium was refreshed every second day. After 8 days, cellular morphology was documented by using bright-field microscopy, and the cells were lysed with RIPA buffer. For shRNA experiments, 2 × 105 C/EBPα−/− MEFs were seeded in 24-well containing alpha-MEM medium supplemented with 10% FBS and were transfected 12 h later with 50 ng of psiRNA constructs, when indicated in combination with E2F/DP expression constructs. Drug treatment for induction of adipogenesis was started 24 h posttransfection. Eight days after start of the treatment, cells were washed twice in PBS and fixed 10 min with Roti-Histofix 4% (Roth). GFP expression contained in the psiRNA construct permitted the recognition of transfected cells under UV light using an AxioVert 100 (Zeiss) inverted microscope. Adipocytes were determined by cell morphology, since phase-contrast microscopy allowed a clear recognition of accumulated lipid droplets within the cells. GFP-expressing cells were counted as either nonadipocytes or adipocytes (at least 400 cells per duplicate). Finally, cells were stained with Oil-Red-O and analyzed by bright-field microscopy.
Bandshift and competition analyses with oligonucleotides were performed as described previously (51). The C/EBP probes were derived from the cMGF promoter (51). The E2F probes were derived from the dihydrofolate reductase (DHFR) promoter (48). The relative intensity of signals arising from shifted radiolabeled probes was determined by quantification of signals from scanned autoradiography films using the Odyssey software.
His-bZIP-domain of C/EBPα was overexpressed at 20°C in E. coli Rosetta (DE3). The purification procedure comprises an affinity chromatography on a 5-ml HisTrap FF crude column (GE Healthcare), charged with Ni2+, and a size exclusion chromatography on a Superdex 75 prep grade column (26 by 60 cm; GE Healthcare). The His7 tag was cleaved with tobacco etch virus protease prior to the gel filtration step.
Isothermal titration calorimetry (ITC) measurements (58) were performed in 20 mM HEPES (pH 7.5) and 0.15 M KCl at 25°C using a VP-ITC microcalorimeter (MicroCal, LLC, Northampton, IL). In an experiment 5 μl of DNA solution (200 or 250 μM) was injected into the sample cell containing 20 μM protein solution (monomeric WT and BRM2 variant of C/EBPα bZIP domain). The DNA sequences used were as follows: forward, 5′-GTC AGT CAG ATT GCG CAA TAT CGG TCA G-3′; and reverse, 5′-C TGA CCG ATA TTG CGC AAT CTG ACT GAC-3′ (31). A total of 50 injections were performed with a spacing of 240 s and a reference power of 18 μcal/s. Binding isotherms were plotted and analyzed using Origin Software (MicroCal, LLC).
Statistical differences between indicated values were determined by one-way analysis of variance with Dunnett post-test. A P value of <0.05 was considered statistically significant.
Several studies have shown that interaction between C/EBPα and E2F1 results in repression of E2F target genes (9, 18, 43, 50). Whether and how E2F affects gene activation by C/EBPα has not yet been addressed. The expression of E2F1 or DP1 (or DP2) weakly suppressed C/EBPα, whereas the coexpression of E2F1/DP1 (or E2F1/DP2) strongly repressed the transcriptional activity of C/EBPα (Fig. (Fig.1A1A and data not shown). On the basis of homology and function, E2F proteins can be subgrouped into family members primarily involved in gene activation (“activator” E2F1-E2F3) or in gene repression (“repressor” E2F4 and E2F5). Figure Figure1B1B shows that in conjunction with DP all E2F proteins examined suppressed transactivation by C/EBPα.
Coimmunoprecipitation and in vitro binding studies showed that both DP1 and DP2 bind to C/EBPα and interact with the bZIP domain of C/EBPα (Fig. 2A and B and data not shown). The structural requirements for repression of C/EBPα activity by DP and by E2F1 were therefore examined. First, amino- and carboxy-terminal DP deletions were tested for interaction with C/EBPα. As shown in Fig. 2C, C/EBPα binding depends on conserved N-terminal regions contained in both DPs (DP1, amino acids [aa] 105 to 127; DP2, aa 60 to 82) that partially overlap with the DNA-binding domain (DBD) but not with the E2F-DP dimerization domain (DIM). Our data showed that a C/EBPα-binding-deficient DP2 mutant lacking 83 N-terminal amino acids (equivalent to mutant aa 83 to 385 of Fig. Fig.2A,2A, referred from now on as Δ83) fails to cooperate with E2F1 in repressing the transcriptional activity of C/EBPα (Fig. (Fig.2D).2D). Interaction assays and gene reporter studies showed that mutations within the E2F1 DBD (E2F1 E132) also abrogate binding to C/EBPα (Fig. (Fig.3A)3A) and fail to repress C/EBPα activity (Fig. (Fig.3B).3B). These results are in agreement with previous findings, showing that the E2F1 E132 mutant failed to repress C/EBPα-induced adipogenesis (43). Altogether, these data indicate that the physical interaction between E2F-DP and C/EBPα are required to suppress the transcriptional activity of C/EBPα.
The connection between pRB- and C/EBPα-mediated proliferation arrest and gene repression remains under debate (16, 18, 48). pRB belongs to the family of pocket proteins, including p107 and p130, that regulate the cell cycle and cell differentiation by inhibition of E2F (7). As shown in Fig. Fig.3B,3B, E2F1 mutants that are defective for pRB (Y411C) or for cyclin A-CDK2 binding (Δ24) still repress the transcriptional activity of C/EBPα. MEFs that lack all three pocket proteins (triple-knockout [TKO] cells) (10) were used to further examine the suppression of C/EBPα by E2F-DP. As shown in Fig. Fig.3C,3C, E2F-DP suppressed C/EBPα-mediated transcription also in the absence of pocket proteins. Next, we explored whether repression of E2F by C/EBPα may depend on pRB. Figure Figure3D3D shows that transcriptional activity of E2F1-DP1 was repressed by either pRB or by C/EBPα. Moreover, the pRB-binding-deficient E2F1 mutant (Y411C) that escapes pRB repression (15) was still repressed by C/EBPα. Altogether, these data suggest that in addition to the well-studied repression of E2F by pRB, the repression of E2F by C/EBPα or the repression of C/EBPα by E2F may also occur independently of pRB.
It is thought that basic region mutants of C/EBPα (Fig. (Fig.4A),4A), such as BRM2 and BRM5, fail to abrogate cell proliferation due to defective repression of E2F-regulated S-phase genes (43). Both mutants evoked a myeloproliferative phenotype (42) and also failed to support adipogenic and granulocytic differentiation (43). Although these studies showed that the BRM mutants are deficient for inhibition of proliferation, their transactivation potential remained equivocal (9, 21, 30). Therefore, we compared the transactivation potential of the BRM mutants depicted in Fig. Fig.4A4A with that of WT C/EBPα and BRM3, a mutant that functions similarly as WT (43). WT C/EBPα and BRM mutants were tested at different concentrations in gene reporter assays. As shown in Fig. Fig.4B,4B, all C/EBPα proteins displayed similar transactivation potential at a high C/EBPα concentration (100 ng of expression plasmid, which is within the dosage range frequently used for reporter assays). At a low concentration (1 ng of C/EBPα expression plasmid, which is close to the endogenous level of C/EBPα [data not shown]), however, transactivation by BRM2 and BRM5 was severely diminished compared to the WT or BRM3. This was not due to decreased protein stability or concentration-dependent expression effects, since both BRM2 and BRM5 proteins were expressed at similar amounts compared to the WT or BRM3 (see the expression controls under the graphs in Fig. Fig.4B).4B). Next, we examined the possibility that E2F-DP becomes rate-limiting at a high C/EBPα concentration. As shown in Fig. Fig.4C,4C, BRM2 and BRM5 activities were strongly repressed by E2F-DP compared to WT C/EBPα or BRM3. Binding experiments showed that DP1 and DP2 displayed much stronger interactions with BRM2 and BRM5, compared to WT or BRM3, whereas binding between WT or C/EBPα mutants and E2F1 or E2F4 was indistinguishable (Fig. (Fig.4D4D and data not shown).
Structural and binding analysis of C/EBPα had suggested a crucial role for Tyr285 (mutated in BRM5) in C/EBPα-DNA complexes (30). The crystal structure of C/EBPα may also suggest that Arg297 (mutated in BRM2) interacts with the DNA backbone. In vitro-translated BRM2, BRM5 or WT C/EBPα proteins were subjected to gel retardation analysis using a radiolabeled C/EBP DNA binding site and the same unlabeled DNA probe as a competitor. As shown in Fig. Fig.5A,5A, BRM5 displayed defective DNA binding, whereas BRM2 retained DNA binding capacity. Determination of the DNA binding constants of WT C/EBPα and BRM2 by ITC revealed biphasic binding to C/EBP sites for both WT and BRM2, with 5- to 7-fold reduced binding affinity for BRM2 (Fig. (Fig.5B5B).
E2F may repress the transcriptional activity of C/EBPα either when bound to DNA or by abrogating the binding of C/EBPα to DNA. To distinguish between these possibilities, DNA binding site interaction analysis was performed by gel shift assay. As shown in Fig. Fig.6A,6A, the E2F1-DP1 complex did not bind to the C/EBP probe, neither in the absence nor in the presence of C/EBPα (left panel). Control shifts and antibody super shifts confirmed that the C/EBP site was bound by C/EBPα and that E2F1-DP1 complexes associated with E2F sites only (Fig. (Fig.6A,6A, right panel). Thus, E2F-DP complexes did not detectably associate with C/EBP DNA-binding sites nor with C/EBPα proteins bound to DNA.
Next, we investigated whether E2F-DP complexes interfere with the binding of C/EBPα to DNA. As shown in Fig. Fig.6B,6B, addition of E2F1-DP1 complex diminished the association of C/EBPα with DNA, while individual addition of DP1 or E2F1 did only barely affect C/EBPα binding to DNA. To examine whether contaminating DNA might have affected interaction between C/EBPα and E2F proteins, binding assay samples were treated with DNase (Fig. (Fig.6C,6C, left panel). The removal of residual DNA by DNase treatment slightly increased the interaction between C/EBPα and E2F1, and the addition of the C/EBP DNA binding site (as used in Fig. 6A and B) reduced the interaction between C/EBPα and E2F1 (right panel). These data suggest that binding of C/EBPα to the E2F1 or to C/EBP DNA binding sites is competitive and exclusive.
A physiological function of E2F-mediated disruption of C/EBPα-DNA complexes was addressed using the preadipogenic 3T3L1 cell line that may undergo C/EBPα-dependent differentiation into fat cells (27). 3T3L1 cells were transduced with a conditional estrogen receptor E2F1 fusion construct (ER-E2F1) that may inhibit adipogenic differentiation upon addition of tamoxifen (4OH) (43). As a negative control, the DNA- and C/EBPα-binding-deficient E2F1 mutant E132 was used, which does not repress adipogenesis (43). 3T3L1 expressing ER-E2F1 WT or E132 were hormonally stimulated to differentiate into adipocytes either in the presence or absence of 4OH (Fig. (Fig.7A).7A). A strong inhibition of adipogenesis was observed with the ER-E2F1 WT compared to ER-E132 after 4OH treatment. Some inhibition was also discernible with WT ER-E2F1 in the absence of 4OH, suggesting some leakiness of the ER-fusion proteins to the nucleus. ChIP assay revealed that the association of C/EBPα with adipocytic genes, including apolipoprotein 2 (AP2) and peroxisome proliferator-activated receptor γ2 (PPARγ2), was strongly diminished after activation of WT E2F1 but not of E132 (Fig. (Fig.7B,7B, left panel). Dissociation of C/EBPα from adipocytic genes coincided with decreased levels of adipogenic gene transcripts (Fig. (Fig.7B,7B, right panel). The data suggested that the interaction between E2F-DP and C/EBPα prevented the binding of C/EBPα to promoters of differentiation genes.
MEFs from C/EBPα-deficient mice (αKO MEFs) do not undergo adipocytic differentiation upon hormonal stimulation (56). The induction of adipocytic genes and differentiation into adipocytes was restored by retroviral expression of WT C/EBPα but not by the expression of BRM2 in αKO MEFs (Fig. (Fig.7C,7C, left panel, and Fig. Fig.8C).8C). In accordance with this, WT C/EBPα but not BRM2 associated with promoters of adipocytic genes (Fig. (Fig.7C,7C, right panel). Importantly, knockdown of DP1 with shRNA constructs, however, restored the binding of BRM2 to adipogenic target genes and induced their activation (Fig. (Fig.7D).7D). Adipogenic differentiation by BRM2 was also restored by knocking down either E2F1, E2F3, E2F4, or DP1 proteins (Fig. 8B and C and data not shown) and was overruled by reconstitution of E2F3 or DP1 knockdowns with E2F1 or DP2, respectively, whereas C/EBPα-binding-deficient mutants E2F1 E132 or DP2 Δ83 did not inhibit the function of BRM2 (Fig. 8D and E). Consistently, the adipogenic activity of WT C/EBPα was enhanced by the E2F-DP knockdown constructs and inhibited by WT E2F1 or DP2 but not by the mutants E2F1 E132 or DP2 Δ83, lending further support to a mechanism of E2F-C/EBP cross-inhibition. No adipogenic differentiation was observed with shRNAs in the αKO MEFs, excluding off-target effects of the shRNA constructs (Fig. (Fig.8C).8C). These data show that C/EBPα-mediated adipogenesis is intrinsically repressed by E2F-DP complexes, that BRM2 is more susceptible to E2F-DP-mediated suppression, and that the differentiation potential of BRM2 can be recovered despite its attenuated DNA binding capacity.
C/EBPα-mediated gene regulation entails the activation of cell lineage-specific terminal differentiation genes and the repression of E2F-regulated S-phase genes (43, 50); thus, C/EBPα coordinates both proliferation arrest and differentiation. The C/EBPα mutant BRM2, similar to mutations found in human AML, fails to suppress E2F-regulated S-phase genes or to support adipogenic and granulocytic differentiation (43) and evokes a myeloproliferative disease in mice (42). BRM2 therefore represents a paradigm to examine C/EBPα-regulated proliferation and differentiation processes. Although failure to repress E2F had been documented, it remained to be explored why BRM2 fails to induce differentiation, since it retains an intact transactivation domain and may also bind to cis-regulatory sites (9, 21, 30).
We show here that C/EBPα interacts with DP, the essential dimerization partner of E2F, and that the BRM2 mutant displays enhanced E2F-DP interaction. The data presented show that E2F-DP counteracts induction of differentiation by C/EBPα or by BRM2 and that diminishing E2F-DP complexes restores differentiation by BRM2. This conclusion is also supported by published data, showing that (i) E2F1 downregulation occurs during adipogenesis (11), (ii) forced upregulation of E2F1 inhibits adipogenesis, and (iii) an E2F1 mutant that is unable to repress transcriptional activity of C/EBPα also fails to repress adipogenesis (43). Taken together, our data uncover a novel regulatory function of E2F-DP and suggest reciprocal regulation of the transcriptional activities of C/EBPα and E2F-DP as an important step in proliferation and differentiation control, as summarized in the model shown in Fig. Fig.99.
DP is required for induction of E2F-regulated genes (1, 15, 25). Although many proteins have been identified that interact with E2F, DP binding partners are rare. We show that C/EBPα is a binding partner of DP1 or DP2 and identify a sequence within DP that is required for interaction with C/EBPα but not for interaction with E2F. Although DP and E2F may also bind to C/EBPα independently from each other, efficient repression of the transcriptional activity of C/EBPα requires their combination, strongly suggesting E2F-DP complexes as suppressors of C/EBPα functions.
E2F-DP complexes repress C/EBPα-mediated transactivation by inhibition of binding of C/EBPα to DNA. The BRM2 mutations apparently display two effects, decreasing the binding to DNA, and augmenting binding to DP, suggesting that a balancing mechanism may exist that switches C/EBPα from a DNA-bound transcriptional active to a DP-bound transcriptional inactive state. BRM2 was found to be more susceptibility to inactivation, and yet it retains transactivation and differentiation potential. These data also provide an explanation as to why some studies observed, while others did not, a transcriptional activity of BRM2 (9, 21) and why BRM2 mice may recover granulopoietic potential (42). Different circumstances may affect the equilibrium between BRM2 and E2F-DP proteins, including conformational changes within C/EBPα or posttranslational modifications of C/EBPα.
Other studies proposed that protein-protein interaction between C/EBPα and E2F at E2F cis-regulatory sites are important for E2F repression (50). This model was supported by a study showing that BRM2 failed to interact with E2F4 and to repress E2F-regulated genes (9). However, another study did not confirm differences in binding affinities between BRM2 or WT C/EBPα and E2F1 (21). In agreement with the latter study, our data suggest that BRM2 displays even increased binding to DP. This may seem counterintuitive at first, if one assumes that C/EBPα represses transcription of E2F target genes by binding via DNA-bound E2F complexes only. A recent report, however, has pointed out that repression of E2F by C/EBPβ may require cryptic C/EBP sites in the proximity of consensus E2F sites, such as those contained in the DHFR promoter (16, 48). Our data confirmed that mutation of the cryptic C/EBP site in the DHFR promoter abolishes binding of C/EBPα proteins but does not affect binding of E2F-DP complexes (data not shown). Thus, the data presented here support the notion that C/EBPα represses transcription of E2F-regulated genes through cryptic cis-regulatory C/EBP binding sites (48). This interpretation may also help to explain why the ability of C/EBPα to repress E2F depends on the DNA-binding function of C/EBPα and why BRM2 and BRM5 mutants, both with compromised DNA binding, fail to repress E2F genes.
It is widely accepted that E2F proteins function as repressors by association with pocket proteins. Our data show that E2Fs may repress C/EBPα-target genes independently of pocket proteins and that C/EBPα may repress E2Fs independently of pRB. Thus, pocket proteins are not strictly required for cross-transcriptional repression of E2F and C/EBPα. In agreement with this interpretation are results showing that C/EBPα represses E2F-mediated transcription in Saos cells that are deficient for both pRB and p53 (18). These findings are of major interest, since they suggest pocket protein-independent repressive functions of E2F complexes (23, 33). Although repression of E2F by pRB does not require the presence of cryptic C/EBP binding elements, the antiproliferative function of C/EBPα appears to depend on the presence of pocket proteins (48). This raises the possibility that the collaborative action of C/EBPα and pocket proteins during proliferation arrest may not depend on repression of E2F but may affect another key cell cycle regulator. How exactly the interaction between E2F and C/EBPα at compound sites such as at the DHFR promoter converts both transactivators into repressors still needs to be explored. In any case, our data suggest a mechanism of E2F-DP-mediated disruption of binding of C/EBPα to its differentiation genes that may account for the block of cell differentiation.
The fact that under experimental conditions transcriptional repression of E2F or C/EBPα occurs in the absence of pocket proteins does not, however, exclude a role for pocket proteins in affecting the balance of transcriptional activities between C/EBPα and E2F. In fact, pRB is a good candidate to regulate this balance since pRB displays opposing functions on E2F and on C/EBPα: while pRB represses E2F function, it activates the transcription of C/EBPα genes (8). Both C/EBPα and pRB are required for adipogenesis (5), whereas E2F4 counteracts adipogenesis (11). However, no correlation is observed between pRB-C/EBPα interaction and adipogenesis (43). These data may be explained by our model that predicts activation of C/EBPα-regulated differentiation genes by pRB-mediated neutralization of E2F and block of differentiation by dissociation of pRB, increasing the availability of E2Fs (Fig. (Fig.9).9). When pRB is hypophosphorylated it associates with E2F, reducing the amount of available E2F that may bind to and inhibit C/EBPα. Proliferating cells contain high levels of pRB in phosphorylated, E2F-nonbinding form (4). As cells progress toward terminal differentiation the level of phosphorylated pRB decreases (49), resulting in enhanced pRB-E2F complex formation and release of C/EBPα from E2F-C/EBPα complexes. This would result in concomitant elevation of free C/EBPα proteins that support differentiation, e.g., toward the adipocytic lineage. Our data also show that the negative effect of E2F on adipogenesis depends on C/EBPα, since the knockdown of E2F proteins in the absence of C/EBPα does not restore adipogenesis. Thus, we favor a scenario wherein pRB is a regulator of available E2F-DP complexes that in turn titrates C/EBPα activities in proliferation and differentiation control.
Interestingly, C/EBPα and C/EBPβ appear to be functionally distinct, as Sebastian and coworkers have shown that C/EBPβ fails to repress E2F in an equivalent TKO cell line (46, 48). It is known that despite their extended homology and partial redundancy (2, 6, 20), C/EBPα and C/EBPβ may also have antagonistic functions, e.g., in the case of skin tumorigenesis, where the lack of C/EBPα promotes tumorigenesis (28), whereas the lack of C/EBPβ protects against chemical-induced skin carcinogenesis (60). Curiously, pRB deficiency also protects from tumorigenesis in a murine skin carcinogenesis model (45). These results may point to a general difference in pocket protein requirements of C/EBPα- versus C/EBPβ-mediated E2F functions. Alternatively, the different TKO cell lines used in the Johnson lab studies and in our study may have acquired additional mutations that obscure distinct aspects of C/EBP biology. Nevertheless, how pocket proteins modify the functional interactions between E2F and C/EBP family members requires further investigations.
In addition to the adipogenic phenotype, BRM2 mice display impaired myelopoiesis and serve as a model for C-terminal AML-C/EBPα mutations (42). We show that knockdown of E2F restores the adipogenic potential of BRM2 and one could therefore envisage that interfering with the interaction between E2F-DP and C/EBPα would also restore granulocytic differentiation and revert the tumorigenic phenotype.
The data presented here suggest that E2F represses the transactivation of C/EBPα target genes by disrupting the binding of C/EBPα to its cis-regulatory sites. Interestingly, disturbed DNA binding of C/EBPα is frequently observed in AML patients that carry C-terminal point mutations or small insertions in the basic region of C/EBPα (13, 14, 37). This often occurs in conjunction with N-terminal frameshift mutations that cause expression of the truncated p30 C/EBPα isoform that is defective for E2F repression and for transactivation (18, 22, 37). Modeling human leukemogenesis in the mouse has shown that the combinations of these two types of mutations accelerate the malignancy of AML-type leukemias compared to homozygous mutant mice (3). Abrogation of C/EBPα functions has also been reported in AML-ETO and BCR-ABL leukemias (36, 38), and ectopic expression of C/EBPα induces the differentiation of such leukemic cells (53, 54). Moreover, the loss of C/EBPα expression in the epidermis increases susceptibility to Ras-induced skin tumorigenesis (28), similarly to the overexpression of E2F1 or DP1 transgenes in the mouse (40, 55). To further delineate the contribution of failure to activate terminal differentiation genes in tumorigenesis, it would be interesting to identify a DP mutant that retains E2F coregulation but is defective for C/EBPα interaction. Pharmacological interference of C/EBPα-E2F-DP interaction and/or restoration of DNA-binding functions of C/EBPα might also provide a therapeutic opportunity in various proliferative diseases.
Katrin Zaragoza was partially supported by a fellowship of the MDC-HU International Ph.D. Program. The Protein Sample Production Facility at the Max Delbrueck Center is funded by the Helmholtz Association of German Research Centres.
We are grateful to Elisabeth Kowenz-Leutz, Hein te Riele, Kristian Helin, Claus Nerlov, Stefan Gaubatz, and Liang Zhu for kindly providing cells and plasmids. We thank Hans-Peter Rahn for fluorescence-activated cell sorting; Jeske Smink for critical comments on the manuscript; and Janett Tischer, Silke Kurths, and Ingrid Berger for excellent technical assistance.
Published ahead of print on 22 February 2010.
†The authors have paid a fee to allow immediate free access to this article.