Regulation of CCAAT/Enhancer-binding Protein (C/EBP) Activator Proteins by Heterodimerization with C/EBPγ (Ig/EBP)*

The CCAAT/enhancer-binding proteins (C/EBPs) are basic leucine zipper transcription factors that play important roles in regulating cell growth and differentiation. C/EBP proteins form leucine zipper-mediated homodimers but are also capable of heterodimerizing with other C/EBPs in vitro. Here we show that C/EBPβ occurs predominantly as a heterodimer that displays rapid mobility in gel shift assays. Biochemical fractionation and antibody supershift assays demonstrate that the C/EBPβ heterodimeric partner is C/EBPγ (Ig/EBP), a C/EBP protein that has been implicated as an inhibitor of other family members. Although most cell types express C/EBPβ·C/EBPγ heterodimers, macrophages contain a C/EBPβ partner that is serologically distinct from C/EBPγ. We found that C/EBPγ blocked the ability of C/EBPβ and C/EBPγ to activate a reporter gene in L cell fibroblasts but did not inhibit a chimeric C/EBPβ protein containing the GCN4 leucine zipper. Repression by C/EBPγ occurs at the level of transactivation and requires heterodimerization with the C/EBP partner. C/EBPγ was an ineffective repressor in HepG2 hepatoma cells despite forming C/EBP heterodimers, and C/EBPα was not effectively inhibited in either L or HepG2 cells. Our findings demonstrate that C/EBPγ modulates C/EBP activity in a cell- and isoform-specific manner.

Eukaryotic transcription factors commonly occur in families whose members share similar DNA binding specificities and other functional properties. Many transcription factors are dimeric and can form homodimers as well as heterodimers with other family members. The capacity to heterodimerize provides a means of enhancing regulatory diversity, as the various dimeric species within a protein family may exhibit distinct functional properties (1)(2)(3)(4). Thus, it is important to elucidate the dimerization status of transcription factors in vivo to understand the full range of their biological activities and regulation.
In the present study, we have examined the dimerization properties of CCAAT/enhancer-binding proteins (C/EBPs) 1 in cells. C/EBPs are a family of basic leucine zipper (bZIP) DNAbinding proteins (5) consisting of five core members: C/EBP␣, C/EBP␤, C/EBP␦, C/EBP⑀, and C/EBP␥ (Ig/EBP). C/EBP proteins bind to DNA as dimers and display highly related DNA binding and dimerization specificities (reviewed in Refs. 6 -8). C/EBPs are involved in the regulation of many cellular processes. C/EBP␣, C/EBP␤, and C/EBP␦ mediate responses to stress and inflammatory signals, including regulation of acute phase response genes in hepatocytes and expression of proinflammatory cytokine genes in monocytic cells (9 -13). Overexpression experiments and analysis of knockout mice demonstrate that C/EBP proteins also control cell growth and differentiation (6 -8, 14). For example, forced expression of C/EBP␣ and/or C/EBP␤ in precursor cells of the adipocyte, granulocyte, and keratinocyte lineages causes growth arrest and induces cellular differentiation (15)(16)(17). In other contexts, C/EBP␤ has been reported to stimulate cell growth (18,19). Thus, C/EBPs regulate a variety of cellular phenotypes in a wide range of cell types.
In addition to forming homodimers, C/EBP proteins are capable of heterodimerizing with the other family members in vitro (20 -22). Heterodimerization could potentially alter several functional activities of C/EBP proteins, including DNA binding, transactivation potential, responsiveness to signaling pathways, and the ability to cooperate with other transcription factors. It has been assumed that heterodimers between C/EBP family members occur in vivo and possess regulatory activities that are distinct from the homodimeric forms. However, there is only limited evidence for such heterodimers in vivo. An association between C/EBP␣ and C/EBP␤ was observed in transient overexpression experiments (20), and evidence has been reported for C/EBP␣⅐C/EBP␤ heterodimers in liver nuclear extracts (23) and monocytic cells (24). In addition, C/EBPs appear to heterodimerize with proteins from other bZIP subfamilies, including Fos/Jun (25) and ATF/CREB (26 -28).
Here we report that C/EBP proteins in cell and tissue extracts are found predominantly as heterodimers with C/EBP␥. C/EBP␥ is a ubiquitously expressed member of the C/EBP family that was first identified by its affinity for cis-regulatory sites in the Ig heavy chain promoter and enhancer (29). C/EBP␥ contains a C/EBP-like bZIP region but lacks an aminoterminal transactivation domain (30,31) and can inhibit transcriptional activation by C/EBP␣ or C/EBP␤ (31). We show that C/EBP␥ can repress C/EBP␤-and C/EBP␦-mediated transactivation of a reporter gene in fibroblasts in a leucine zipper-dependent manner, indicating that the repression by C/EBP␥ involves heterodimerization with its partner. Interestingly, C/EBP␥ did not repress transactivation by C/EBP␤ or C/EBP␦ in HepG2 hepatoma cells, nor did it inhibit C/EBP␣ activity in either cell type. Thus, the ability of C/EBP␥ to inhibit C/EBP activators is cell-specific and differs for the various C/EBP family members.
Plasmid Constructs-The C/EBP␥ coding sequence was amplified by PCR from a plasmid containing a C/EBP␥ clone of murine origin (29). Two oligonucleotide primers were used, one overlapping the initiation codon with an introduced NcoI restriction site and the second spanning the termination codon with an introduced HindIII restriction site. The PCR product was digested with NcoI (partial) and HindIII (complete) and inserted into the pMEX expression vector to generate pMEX-C/ EBP␥. A bacterial expression construct, pJL6-C/EBP␥, was generated by inserting the digested PCR product into the expression vector pJL6 as described (38). The GAL4-C/EBP␤ hybrid construct was described previously (38).
GAL4-C/EBP␤ transfection assays were conducted using 30 -40% confluent monolayers in 60-mm wells using FuGENE TM (Roche Molecular Biochemicals). One g of (G) 5 E1B-luc reporter plasmid (38), 5 ng of GAL4-C/EBP␤ vector, and 25 ng of Ha-Ras(12V) vector were transfected with varying quantities of the pcDNA3.1 C/EBP␥ expression vector (13.3 ng to 1.25 g). The Renilla luciferase vector, pRL-TK (Promega), was co-transfected as an internal standard for transfection efficiency. Sixteen hours prior to harvesting the cells, the medium was removed and replaced with serum-free media. Cells were collected 48 h after transfection, lysed, and analyzed using the Dual-Luciferase ® assay system (Promega).
Nuclear Extracts-Nuclear extracts from cell lines and transfected cells were prepared by a detergent lysis method. Transfected cells were washed once with phosphate-buffered saline, scraped, and then divided. 20% of the cells from 10-cm dishes were used for luciferase assays by resuspension in detergent lysis solution (100 mM potassium phosphate (pH 7.8), 0.2% Triton X-100, 1 mM dithiothreitol (DTT); CLONTECH Laboratories, Inc.), incubation at room temperature for 5 min and centrifugation. The remaining 80% of the cells were used to make nuclear extracts by resuspension in lysis buffer (20 mM HEPES (pH 7.9), 1 mM EDTA, 10 mM NaCl, 1 mM DTT, 0.1% (v/v) Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 g/ml leupeptin, 5 g/ml aprotinin, 5 g/ml antipain ) and incubation on ice for 10 min. Nuclei were pelleted by centrifugation at 3,500 rpm for 10 min. Proteins were extracted from nuclei by incubation in high salt buffer (25 mM HEPES (pH 7.9), 0.2 mM EDTA, 0.42 M NaCl, 0.2 mM DTT, 25% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 10 g/ml aprotinin, 5 g/ml antipain) at 4°C for 20 min with vigorous shaking. Nuclear debris was pelleted by centrifugation at 14,000 rpm for 5 min, and the supernatant was collected and stored at Ϫ70°C. Nuclear extracts from mouse tissues were prepared by an Nonidet P-40 lysis procedure as described previously (39).
Electrophoretic Mobility Shift Assay (EMSA)-The following doublestranded oligonucleotides containing the wild-type consensus C/EBP site (bold) or a mutant C/EBP site (bold) were used as probes or competitor DNAs.
Wild-type: 5Ј-GATCCATATCCCTGATTGCGCAATAGGCTCAAAA GTATAGGGACTAACGCGTTATCCGAGTTTTCTAG-5Ј Mutant: 5Ј-GATCCATATCCCTGAGGGCGCCCTAGGCTCAAAA GTATAGGGACTCCCGCGGGATCCGAGTTTTCTAG-5Ј The probe was end-labeled using [ 32 P]dCTP and Klenow polymerase. DNA-binding assays were carried out in a 25-l reaction containing 20 mM HEPES (pH 7.9), 200 mM NaCl, 5% (w/v) Ficoll, 5% (v/v) glycerol, 1 mM EDTA, 50 mM DTT, 0.01% Nonidet P-40, 0.06% bromphenol blue, 1.75 g of poly(dI-dC), and 7.5 ϫ 10 4 cpm probe. After incubation for 20 min at room temperature, 10 l of the binding reaction was loaded onto a 6% polyacrylamide gel in 1ϫ TBE (90 mM Tris base, 90 mM boric acid, 0.5 mM EDTA) and electrophoresed at 160 V for 2 h. The gel was dried before autoradiography. Supershift assays were carried out by preincubating the nuclear extract with 2 l of rabbit antiserum at 4°C for 30 min before addition of the binding reaction mixture. Where appropriate, DNA-protein complexes were quantitated using a PhosphorImager and ImageQuant software (Molecular Dynamics). Mixing experiments to form C/EBP heterodimers were incubated at 45°C for 20 min in the presence of 95 mM DTT.
Immunoprecipitation-Approximately 5.8 g of purified His-tagged C/EBP␤ was incubated with 5.8 g of His-tagged C/EBP␥ or truncated C/EBP␤-(191-276) at 45°C for 20 min in the following buffer: 0.2 M Tris (pH 7.5), 0.2 M NaCl, 5 mM EDTA, 10 mM DTT, 0.1% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 g/ml leupeptin, 5 g/ml aprotinin, 5 g/ml antipain. 3 l of NH 2 -terminal C/EBP␤ antiserum and 10 l of protein A-Sepharose beads were then added. As a control, Histagged C/EBP␥ or truncated C/EBP␤-(191-276) were incubated alone in the same manner. After a 3-h incubation at 4°C, the beads were collected by centrifugation and washed three times with buffer. 1ϫ Laemmli sample buffer was added to the beads, and the samples were boiled for 5 min and centrifuged. The eluted proteins were resolved by SDS-PAGE and examined by Western blotting using Super Signal ® West HisProbe TM kit (Pierce).
Immunoprecipitation experiments using nuclear extracts from transfected L or HepG2 cells were performed using the SeizeX immunoprecipitation kit (Pierce). Cells were transfected with 5 g of control plasmid, 5 g of tagged C/EBP␤ vector, or 5 g of tagged C/EBP␤ vector with 5 g of C/EBP␥ vector, and nuclear extracts were prepared. Seventy-five g of N.E. was added to protein A-Sepharose beads crosslinked to the NH 2 -terminal GCN4 antibody and incubated for 3 h at 4°C. The beads were washed three times with buffer and eluted with protein sample buffer and the eluted proteins analyzed by Western blotting.
Western Blotting-Nuclear extracts were mixed with an equal volume of 2ϫ sample buffer (40), heated at 95°C for 5 min, and loaded on precast 12% or 16% SDS-PAGE gels (Novex). Gels were transferred to Immobilon-P membranes (Millipore) and blocked with 5% dry milk in Tris-buffered saline (pH 7.6). Blots were developed using the enhanced chemiluminescence (ECL) detection system (Pierce).

C/EBP␤ Occurs in Cells Predominantly as a Rapidly
Migrating EMSA Complex-To examine the dimeric state of C/EBP proteins in cells, we performed EMSA on a series of nuclear extracts using a consensus C/EBP binding site oligonucleotide as the probe. We initially focused our analysis on C/EBP␤, because this protein is widely expressed in cell lines and tissues. We examined nuclear extracts from cell lines representing glioma (C6-2B), mammary epithelia (EMT6), and macrophages (IC-21) (Fig. 1A). The cell line extracts contained multiple species that bound specifically to the C/EBP motif, as determined by the ability of the unlabeled wild-type probe, but not a mutated oligonucleotide, to compete for binding ( Fig. 1B and data not shown). To identify complexes containing C/EBP␤, we performed supershift analysis using a C/EBP␤ antibody. As shown in Fig. 1A, C/EBP␤ binding activities were observed in all extracts, and the C/EBP␤ EMSA complexes occurred in two forms. One complex exhibited the same mobility as a bacterially expressed C/EBP␤ homodimer (lane 7), whereas a second species (indicated by asterisks) displayed considerably faster mobility in the gel. In all cases this faster migrating form was the predominant C/EBP␤ species. We also analyzed extracts from P388-C␤ cells, which express C/EBP␤ from a retroviral vector (34,35). P388 is a lymphoblastic tumor cell that contains very low levels of endogenous C/EBP proteins except C/EBP␥ (Ref. 41; see below). P388-C␤ nuclear extracts FIG. 1. C/EBP proteins in cells occur primarily as rapidly migrating EMSA complexes. A, EMSA of nuclear extracts from cell lines. A consensus C/EBP binding site was used as the probe. Extracts (3-12 g) were assayed in the presence or absence of the NH 2 -terminal C/EBP␤ antibody. Positions of the rapidly migrating species are indicated by asterisks. Bacterially expressed C/EBP␤ was analyzed in parallel (lane 7) to show the mobility of the C/EBP␤ homodimer. B, specificity of C/EBP⅐DNA complexes. Increasing amounts of unlabeled C/EBP competitor probe (lanes 7-12) or a mutant oligonucleotide with a disrupted C/EBP site (lanes 1-6) were added to binding reactions containing nuclear extracts from L cells. C, EMSA of nuclear extracts from stably transfected P388 lymphoblasts. Extracts from P388 cells and a P388 transfectant expressing C/EBP␤ (P388-C␤) were analyzed as described in panel A.
contained almost exclusively the rapidly migrating form of C/EBP␤ (Fig. 1C, lane 2). Thus, ectopic C/EBP␤ in P388 cells is detected predominantly as a high mobility EMSA species, similar to the results observed for endogenous C/EBP␤ in the other cell lines (Fig. 1A).
The Rapidly Migrating C/EBP Complexes Are Heterodimers-The fact that the C/EBP␤ antibody used for the supershift experiments recognizes the amino terminus of the protein indicates that the rapidly migrating complex contains an intact C/EBP␤ subunit and is not a truncated isoform, such as LIP, that contains the COOH-terminal DNA-binding domain (42,43). The absence of truncated forms of C/EBP␤ in the extracts was confirmed by Western blotting using a COOHterminal C/EBP␤ antibody (data not shown). To test whether the rapidly migrating C/EBP␤ complex is a heterodimer with another cellular protein, we performed a mixing experiment in which purified recombinant C/EBP␤ was combined with nuclear extract from P388 cells and assayed by EMSA (Fig. 2). This procedure generated a rapidly migrating complex that co-migrated with the endogenous C/EBP␤ complex from P388-C␤ cells (compare lanes 1 and 4). Similar results were obtained using recombinant C/EBP␦ and C/EBP␣ (data not shown). In contrast, a chimeric C/EBP␤ protein (C/EBP␤-G LZ ) containing a heterologous leucine zipper from the yeast GCN4 protein did not form the rapidly migrating complex with P388 extract (lanes 5 and 6). These findings show that the C/EBP␤ leucine zipper is necessary to produce the rapidly migrating C/EBP␤ complex and indicate that this species consists of a heterodimer between C/EBP␤ and another nuclear protein.
Identification of C/EBP␥ as the C/EBP␤ Heterodimeric Partner-The rapid mobility of C/EBP␤ heterodimers and their presence in all cell types examined indicate that the dimeric partner is a small, ubiquitously expressed protein. These features suggested that the partner might be C/EBP␥ (Ig/EBP), a 16.4-kDa protein that can dimerize with other C/EBP proteins and is expressed at the mRNA level in many cell types (29,31). A C/EBP␥ polyclonal antiserum raised against recombinant C/EBP␥ (31) did not supershift the C/EBP␤ heterodimer, although it shifted a more rapidly migrating complex that corresponds to a C/EBP␥ homodimer (data not shown). However, biochemical purification of the rapidly migrating complex to near homogeneity suggested that the heterodimerizing factor may indeed be C/EBP␥. 2 Therefore, we generated additional antisera using synthetic peptides corresponding to the C/EBP␥ amino and carboxyl termini and tested the antibodies in EMSA supershift experiments. Both C/EBP␥ antisera supershifted a partially purified, rapidly migrating C/EBP␦ complex isolated from P388-C/EBP␦ cells, as well as a putative C/EBP␥ homodimer also present in this preparation (Fig. 3A). Thus, these two peptide antibodies recognize C/EBP complexes that appear to contain C/EBP␥.
To determine whether the rapidly migrating C/EBP␤ complexes in cell extracts are C/EBP␥ heterodimers, we tested a panel of cell extracts in antibody supershift assays using the COOH-terminal C/EBP␥ antibody. Fig. 3B (upper panel) shows that rapidly migrating C/EBP␤ complexes from glioma cells (lanes 1 and 2) and mammary epithelial cells (lanes 3 and 4), as well as from P388-C␤ cells (lanes 7 and 8), were supershifted by the C/EBP␥ antibody. Similarly, a C/EBP␦ heterodimeric complex from P388-C␦-C1 cells (lanes 9 and 10) reacted with the C/EBP␥ antibody. Interestingly, the rapidly migrating C/EBP␤ complex from IC-21 macrophages was only weakly supershifted by the COOH-terminal C/EBP␥ antibody (lanes 5 and 6) and by the NH 2 -terminal C/EBP␥ antibody (data not shown), despite the fact that this complex migrates identically with C/EBP␤⅐C/ EBP␥ heterodimers from other cells. C/EBP␤ complexes from two other monocyte/macrophage cell lines also did not react appreciably with C/EBP␥ antibodies (data not shown). These findings indicate that a distinct, but functionally related, protein heterodimerizes with C/EBP␤ in monocytic cells. Western blot experiments using the same panel of cell extracts (Fig. 3B, lower panel) confirmed that a protein of ϳ18 kDa, identical in size to ectopically expressed C/EBP␥ (lane 9), was detected in all cells examined including macrophages.
Because our analysis of C/EBP dimerization thus far used transformed or immortalized cell lines, we next wished to determine whether C/EBP␤ heterodimerizes with C/EBP␥ in normal tissues. We prepared nuclear extracts from mouse liver, brain, ovary, and spleen and analyzed them by EMSA and antibody supershift experiments. As shown in Fig. 3C, each extract contained a rapidly migrating C/EBP␤ complex (denoted by an asterisk) that could be supershifted by the C/EBP␤ and C/EBP␥ antibodies. Nuclear extracts from L cells transfected with C/EBP␥ (lane 14) or C/EBP␤ plus C/EBP␥ (lane 15) were analyzed on the same gel to confirm the identities of the homo-and heterodimeric C/EBP␤ and C/EBP␥ EMSA species. In summary, the experiments of Fig. 3 show that C/EBP␤ occurs mainly as a heterodimer with C/EBP␥ in cell lines as well as animal tissues.
C/EBP␥ Causes Cell-specific Repression of C/EBP-mediated Transcription-We next investigated whether heterodimerization with C/EBP␥ affects the transcriptional activity of C/EBP␤. Initially, we examined the effect of C/EBP␥ 2 S. Parkin and P. F. Johnson, unpublished results.

FIG. 2.
Identification of a C/EBP heterodimerizing factor using an in vitro mixing assay. Recombinant C/EBP␤ was assayed by EMSA either alone or after mixing with P388 cell nuclear extract. Nuclear extracts from P388-C␤ cells or P388 cells are shown in lanes 1 and 2, respectively. Recombinant C/EBP␤-G LZ , which contains the GCN4 leucine zipper, did not form a rapidly migrating complex when mixed with P388 extract (lanes 5 and 6). The consensus C/EBP site oligonucleotide was used as the probe. on C/EBP␤-mediated transactivation using a C/EBP-dependent promoter-reporter construct ((DEI) 4 -alb-Luc; Ref. 38) in HepG2 hepatoma cells (Fig. 4A, left panel). C/EBP␤ alone increased reporter expression by ϳ15-fold. Co-transfecting increasing amounts of a C/EBP␥ expression vector did not significantly diminish reporter gene expression, even at the highest dose of C/EBP␥ (8.5 g), which is an 11.3-fold excess of C/EBP␥ over the C/EBP␤ vector. This result was unexpected, because C/EBP␥ was previously found to inhibit the transactivation function of C/EBP␣ and C/EBP␤ in B lymphoma cells, 3T3 fibroblasts, and promonocytic cells (31). Therefore, we performed a similar C/EBP␥ titration experiment in L fibroblastic cells (Fig. 4A, right panel). In these cells C/EBP␥ clearly inhibited C/EBP␤-mediated transactivation of the (DEI) 4 -alb-Luc reporter in a dose-dependent manner. Luciferase activity decreased linearly with the amount of C/EBP␥ vector added and reached 22% of the control level at the maximal dose of C/EBP␥. Similar results were obtained using a reporter construct containing two copies of a consensus C/EBP binding site (data not shown). Thus, C/EBP␥ is capable of inhibiting C/EBP␤ activity in L cells but not in HepG2 hepatoma cells.
To assess the levels of homo-and heterodimeric C/EBP complexes in the transfected cells, we prepared nuclear extracts and subjected them to EMSA analysis (Fig. 4B). C/EBP␤ homodimer and heterodimer levels were increased in the transfected cells. Heterodimers were observed in cells transfected with C/EBP␤ alone, resulting from dimerization with endogenous C/EBP␥ (lane 2). The amount of heterodimeric complex increased with the addition of C/EBP␥ vector, as did the levels of C/EBP␥ homodimer. The EMSA complexes were quantitated by phosphorimaging, and C/EBP␤ homodimer levels were calculated either as the percentage of total C/EBP binding activity (heterodimer plus the two homodimeric species) or as the frac-tion of C/EBP␤ binding activity (C/EBP␤ homodimer plus heterodimer) (Fig. 4A). The proportion of C/EBP␤ homodimer decreased with added C/EBP␥, and the dimerization curves were similar in HepG2 cells (no transcriptional repression by C/EBP␥) and L cells (repression). Western blotting showed that C/EBP␥ did not alter C/EBP␤ levels in the nuclear extracts (Fig. 4C). These experiments show that C/EBP␤⅐C/EBP␥ heterodimers are formed in both cell types, suggesting that heterodimers are transcriptionally active in HepG2 cells but not in L cells.
To further examine whether inhibition of C/EBP␤ by C/EBP␥ requires heterodimerization, we used the zipper-swap mutant, C/EBP␤-G LZ , which is unable to dimerize with C/EBP␥. Fig. 5A shows that C/EBP␥ did not repress C/EBP␤-G LZ activity in either HepG2 or L cells, whereas in HepG2 cells transactivation was actually enhanced at the highest dose of C/EBP␥. EMSA (Fig. 5B) verified that C/EBP␤-G LZ ⅐C/EBP␥ heterodimers were not formed in the transfected cells, although homodimers were observed. Thus, the ability of C/EBP␥ to inhibit C/EBP␤ activity requires heterodimerization between the two proteins and is not the result of competitive binding of transcriptionally inactive C/EBP␥ homodimers to the promoter.
Because the DNA binding activity of C/EBP␤ was not repressed by heterodimerization, it seemed likely that C/EBP␥ inhibits transactivation. To investigate this possibility, we used a GAL4-C/EBP␤ fusion protein whose ability to activate a UAS-dependent reporter gene (G 5 E1b-luc) depends on the transactivation domain (TAD) of C/EBP␤ (38). Normally the activity of full-length C/EBP␤ fused to GAL4 is very low because of strong repression of the TAD by inhibitory sequences located in COOH-terminal regions of the molecule, including the bZIP domain (38,44). However, GAL4-C/EBP␤ can be FIG. 3. The rapidly migrating C/EBP heterodimers contain C/EBP␥. A, the C/EBP␦ heterodimer reacts with two peptide antisera specific for C/EBP␥. A partially purified fraction containing the C/EBP␦ heterodimer isolated from P388-C/EBP␦ cells (data not shown) was analyzed by EMSA supershift using antisera raised against peptides corresponding to the amino or carboxyl termini of C/EBP␥ (lanes 2 and 3, respectively). B, upper panel, nuclear extracts from the cell lines used in Fig. 1 were assayed by EMSA supershift using the COOH-terminal C/EBP␥ antibody. activated by coexpression of oncogenic Ha-Ras. 3 Therefore, we transfected GAL4-C/EBP␤ with a Ha-Ras(12V) vector and the G 5 E1b-luc reporter into L cells along with increasing amounts of the C/EBP␥ vector. As shown in Fig. 6, C/EBP␥ potently inhibited GAL4-C/EBP␤ activity in a dose-responsive manner. This result demonstrates that heterodimerization with C/EBP␥ suppresses the ability of the C/EBP␤ TAD to stimulate tran-scription, even when C/EBP␤ is tethered to DNA through a heterologous DNA-binding domain. C/EBP␥ did not inhibit GAL4-C/EBP␤ activity in HepG2 cells (data not shown), further supporting the observation that C/EBP␥ repression is cell-specific.
We next asked whether C/EBP␥ could repress transactivation by other C/EBP family members (Fig. 7). Neither C/EBP␣ nor C/EBP␦ was inhibited by C/EBP␥ in HepG2 cells, and, in fact, C/EBP␦ activity was stimulated nearly 2-fold at the max- imal C/EBP␥ dose. This enhanced activity was associated with increased expression of C/EBP␦ (data not shown), the mechanism of which is unknown. Similar to C/EBP␤, C/EBP␦ transactivation was repressed by C/EBP␥ in L cells. In contrast, C/EBP␣ activity was unaffected by C/EBP␥ in L cells (Fig. 7) despite the fact that the two proteins formed heterodimers (data not shown). Collectively, our data indicate that C/EBP␣⅐C/EBP␥ heterodimers are active in L cells, whereas C/EBP␤⅐C/EBP␥ and C/EBP␦⅐C/EBP␥ dimers are repressed.

DNA-independent Association of C/EBP␤ and C/EBP␥ in Vitro and in Vivo-
The observation that C/EBP proteins in cells occur predominantly as heterodimers with C/EBP␥ suggested that they might preferentially dimerize with C/EBP␥. Using recombinant His-tagged C/EBP␤ and C/EBP␥ proteins, we performed coimmunoprecipitation assays to compare the ability of C/EBP␤ to self-dimerize and to heterodimerize with C/EBP␥ in vitro (Fig. 8A). Self-dimerization was examined by mixing full-length (p34) C/EBP␤ with a truncated C/EBP␤ protein (C/EBP␤- (192-276)) containing only the bZIP portion of the molecule. Mixtures of full-length C/EBP␤ and either C/EBP␥ or C/EBP␤-(192-276) were immunoprecipitated with an antibody directed against the amino terminus of C/EBP␤. The immunoprecipitates were analyzed by Western blotting using a reagent that detects the polyhistidine tag. In the absence of full-length C/EBP␤, neither of the other proteins was immunoprecipitated (lanes 1 and 2), as expected. When fulllength C/EBP␤ was added to the mixtures, C/EBP␥ and C/EBP␤-(192-276) were detected in the precipitated fraction (lanes 3 and 4). Both proteins were immunoprecipitated with similar efficiency, suggesting that the C/EBP␤ leucine zipper has comparable affinity for itself and for C/EBP␥. Thus, the predominance of heterodimers in vivo may be the result of a molar excess of C/EBP␥ in cells, or, alternatively, dimerization might be regulated by a cellular mechanism such as phosphorylation.
We next examined the association between C/EBP␤ and C/EBP␥ in transfected cells using a co-immunoprecipitation assay. Epitope-tagged C/EBP␤, which contains the NH 2 -terminal 13 amino acids of GCN4 fused to its amino terminus, was expressed in L or HepG2 cells, either alone or with C/EBP␥. Nuclear extracts were prepared and subjected to immunoprecipitation with an antibody recognizing the GCN4 tag (36), FIG. 5. Repression of C/EBP␤ activity by C/EBP␥ requires the C/EBP␤ leucine zipper. A, co-transfection experiments were performed in HepG2 and L cells as described in Fig. 4, except that pMEX-C/EBP␤-G LZ was used instead of wild-type C/EBP␤. Data are the average (Ϯ S.E.) of three experiments. B, EMSA of nuclear extracts from the transfected cells. Positions of C/EBP␤-G LZ homodimers and C/EBP␥ homodimers are indicated.
followed by Western blotting for C/EBP␥. As shown in Fig. 8B (bottom panel), ectopic C/EBP␥ co-immunoprecipitated with C/EBP␤ in L cells and in HepG2 cells (lanes 3 and 6). Thus, in both cell types C/EBP␤ and C/EBP␥ are associated in the absence of DNA. These findings further support the conclusion that impaired heterodimerization does not explain the inability of C/EBP␥ to repress transcription in HepG2 cells. DISCUSSION Our studies demonstrate that C/EBP activator proteins exist predominantly as heterodimers with C/EBP␥ in vivo. By comparing EMSA complexes generated with recombinant C/EBP␤ with those from nuclear extracts, we observed that C/EBP␤ in cell lines and tissues occurs mainly as a rapidly migrating heterodimer. Antibodies specific for the NH 2 and COOH termini of C/EBP␥ supershifted the rapidly migrating C/EBP species, confirming that the heterodimers contain C/EBP␥. Our characterization of C/EBP heterodimers in this study has focused on C/EBP␤, because this isoform is expressed in many  5-7. B, association of C/EBP␤ and C/EBP␥ in transfected cells. Nuclear extracts (20 g) from L and HepG2 cells transfected with the indicated expression vectors were analyzed for expression of C/EBP␥ and epitope-tagged C/EBP␤ by Western blotting (top two panels). The same nuclear extracts (75 g) were subjected to immunoprecipitation using a GCN4 NH 2 -terminal antibody (bottom panel). The immunoprecipitates were analyzed for C/EBP␥ by Western blotting using the COOH-terminal C/EBP␥ antiserum. cell lines. However, C/EBP␦ and C/EBP␣ also occurred as rapidly migrating heterodimers in P388 transfectants that stably express these proteins ( Fig. 3B and data not shown), as well as in transiently transfected L cells (data not shown). Thus, it seems likely that all of the C/EBP activators heterodimerize with C/EBP␥ in vivo.
Heterodimerization with C/EBP␥ did not detectably alter the DNA-binding specificity of its C/EBP partner, because homodimers and heterodimers bound efficiently to a consensus C/EBP element. The major effect of dimerization with C/EBP␥ was to repress C/EBP transactivation function. In L cells, coexpression of C/EBP␥ inhibited the ability of C/EBP␤ and C/EBP␦ to activate transcription from a C/EBP-dependent promoter. A C/EBP␤ chimera containing the GCN4 leucine zipper that cannot heterodimerize with C/EBP␥ was resistant to repression. C/EBP␥ also suppressed transactivation by a GAL4-C/EBP␤ fusion protein. Collectively, these results indicate that heterodimerization with C/EBP␥ inhibits the transcriptional activity of C/EBP␤. At present it is unclear how heterodimerization with C/EBP␥ suppresses transactivation. C/EBP␥ lacks a TAD and by itself neither activates nor represses transcription of target genes (31). 2 It is possible that, because C/EBP heterodimers contain only one activating subunit, they cannot efficiently stimulate transcription. Alternatively, heterodimerization with C/EBP␥ might block access to a coactivator protein for which association with the C/EBP activator involves sequences in the leucine zipper and/or basic region. This possibility is currently under investigation.
The fact that C/EBP␥ did not repress transactivation by any of the C/EBPs in HepG2 hepatoma cells is noteworthy. Analysis of the C/EBP␤ dimeric species expressed in transfected cells showed that heterodimers were produced and their levels increased in proportion to the amount of transfected C/EBP␥ vector. A homodimeric C/EBP␤ complex was observed in both L and HepG2 cells, and this complex did not appreciably diminish with increased C/EBP␥ expression (Fig. 4B). It is possible that a pool of homodimers exists that is resistant to heterodimerization, perhaps because of a specific post-translational modification. However, because the occurrence of these persistent homodimers did not differ in the two cell types, their presence cannot account for the differential repression by C/EBP␥.
Because there was no difference in heterodimer formation in HepG2 and L cells, at least as assessed by EMSA and coimmunoprecipitation experiments, we postulate that C/EBP⅐C/ EBP␥ heterodimers are transcriptionally active in HepG2 cells but not in L cells. There are several potential explanations for this difference in activity. Heterodimers could be the target of activating kinases in HepG2 cells but not in L cells, whereas homodimers might be effective substrates in both cells. Such modifications could occur on either the C/EBP activator protein or the C/EBP␥ subunit. It is also conceivable that protein: protein interactions mediated by the bZIP region are necessary for transcriptional activation and that heterodimeric bZIP domains are differentially active for this function in the two cell types. Irrespective of the mechanism, the ability of C/EBP␥ to affect the transactivation potential of C/EBP activators in a cell-specific manner represents a novel means of controlling C/EBP activity.
A mouse strain carrying a null mutation at the C/EBP␥ locus has been developed (45). Homozygous mutant animals show grossly normal embryonic development and are initially viable after birth. However, the majority of mutant mice die within 48 h of postnatal development. Although the cause of mortality was not determined, the lethal phenotype shows that C/EBP␥ has an essential function in newborn animals and presumably also in adult mice. It remains to be determined whether the lethality of C/EBP␥-deficient mice results from the absence of C/EBP heterodimers, leading to formation of homodimers with altered regulatory activities. Considering the involvement of C/EBP proteins in many biological processes and the predominance of C/EBP⅐C/EBP␥ heterodimers in cells, it is not surprising that deletion of C/EBP␥ would have severe phenotypic consequences. An additional function for C/EBP␥ in lymphoid cells was revealed by analysis of bone marrow chimeras generated from C/EBP␥ null donor cells (45). Natural killer cells derived from mutant donors display reduced cytolytic activity and impaired production of interferon-␥ in response to interleukin-12 or interleukin-18 stimulation. Nevertheless, the molecular basis for defective interferon-␥ gene expression in C/EBP␥-deficient natural killer cells has not been elucidated.
In another study examining C/EBP␥ function in vivo, Zafarana et al. (46) created transgenic mice overexpressing C/EBP␥ in erythroid cells. Animals heterozygous for the C/EBP␥ transgene displayed increased fetal ␥-globin gene expression compared with adult ␤-globin expression, indicating that C/EBP␥ positively regulates ␥-globin transcription. However, when C/EBP␥ expression was increased further by making the transgenic allele homozygous, fetal erythropoiesis was eliminated and the embryos did not survive beyond embryonic day 14.5. These results demonstrate that C/EBP␥ stoichiometry critically affects development of the erythroid lineage. We suggest that the developmental defects associated with high ectopic C/EBP␥ expression may result from decreased levels of C/EBP homodimers in erythroid precursor cells.
In addition to regulating transcription, C/EBP proteins can induce cell growth arrest (47)(48)(49). In proliferating cell lines, C/EBP proteins occur primarily as heterodimers, raising the possibility that heterodimerization with C/EBP␥ mitigates the growth arrest activity of these proteins. In experiments to create P388 cell lines expressing the zipper swap mutant, C/EBP␤-G LZ , only minimal amounts of the mutant protein were detected in stable transfectants whereas the wild type protein could be expressed at much higher levels (41). This result is consistent with the idea that C/EBP␤ must heterodimerize with C/EBP␥ for its expression to be tolerated in proliferating cells. Furthermore, HepG2 hepatoma cells express significantly lower levels of C/EBP␣ and C/EBP␤ than are found in normal, terminally differentiated hepatocytes (50). We speculate that C/EBP␥ may be unable to suppress C/EBPmediated growth arrest in hepatoma cells, similar to its inability to inhibit C/EBP-dependent transcription in these cells. Thus, conversion of hepatocytes to proliferating hepatoma cells might require strong down-regulation of C/EBP␣ and C/EBP␤ expression. In future studies it will be informative to examine the ability of C/EBP␥ to modulate C/EBP-mediated cell growth arrest in various cellular contexts.
The observation that C/EBP activity can be inhibited by heterodimerization with C/EBP␥ suggests that C/EBP dimerization might be regulated to control gene transcription. In this regard, calcium-regulated phosphorylation of a serine residue in the leucine zipper of C/EBP␤ has been linked to its increased transcriptional activity (51). Although the molecular mechanism underlying this activation event has not been elucidated, the authors raised the possibility that phosphorylation of the C/EBP␤ zipper might control dimerization. Our studies indicate that C/EBP␥ could be involved in this putative regulatory mechanism. Although our experiments thus far have focused on artificial promoters, future studies will address potential differences between C/EBP homodimers and heterodimers in activating authentic promoters, in addition to the possibility that C/EBP␥ heterodimerization is regulated by developmental cues or other physiological signals.