Structural and Functional Studies of CCAAT/Enhancer-binding Protein ε*

CCAAT/enhancer-binding protein (C/EBP) ε is a critical transcription factor for differentiation of myeloid cells. Structural and functional relationships of C/EBPε were explored by recombinant protein studies, gene mutation, and transactivation assays. Evidence strongly suggested that C/EBPε does not have disulfide bonds. Transactivation analysis of C/EBPε having mutations of each of three conserved cysteines (C345, C148S, and C280S) indicated that the three mutant proteins had almost the same activity as the wild type. Dimer formation of C/EBPε was not detected using both reducing and non-reducing SDS-polyacrylamide gel electrophoresis with Western blot analysis from either bacterial or mammalian expressed C/EBPε. Furthermore, C/EBPε mutant C280S gave a gel band similar to that for wild type, suggesting that this C-terminal, conserved cysteine is not involved in disulfide bond formation in vivo, even though previous data for C/EBPβ suggested that dimers may formin vitro utilizing this conserved cysteine residue. Mutational studies of conserved residues in the activating domain 1 (ADM1) and ADM2 of the amino region of the gene indicated that negative charge is critical for transactivational activity of C/EBPε. Mutational analyses of hydrophobic amino acids in ADM1 suggested that these residues do not play a key role in transactivational activity. Further mutational studies indicated that, although the N-terminal 32-amino acid peptide of C/EBPε isoform p32 did not greatly influence the transactivation activity compared with p30 isoform, this peptide does modulate transactivation activity. Domain swapping experiments substituting the ADM1 domain of various C/EBPs for C/EBPε showed that the C/EBPα and -δ but not -β ADM1 markedly enhanced the chimeric C/EBPε transcriptional activity. Based on mutational data and possible mRNA structure, we hypothesized about the effect of mRNA structure on translation of the two major C/EBPε isoforms: p32 and p30. The data suggested a very stable 8-base pair double helical structure with one strand sequence including the initial codon for p32 and complementary strand with the initial codon for p30.

The CCAAT/enhancer-binding protein (C/EBP) 1 family consists of six members (1), including the highly related C/EBP␣ (2), C/EBP␤ (3), C/EBP␦ (4) and C/EBP⑀ (5,6), and the less related C/EBP␥ (7) and C/EBP (8). Like other leucine zipper transcription factors, they contain a basic amino acid domain and a leucine zipper region allowing the formation of homoand heterodimers when binding to DNA (9). They share similar DNA binding specificities (9). The C/EBPs serve as critical regulators in the differentiation processes of a variety of mammalian cells including adipocytes, hepatocytes, and myeloid cells (10,11). In the hematopoietic system, early myeloid progenitors have elevated levels of C/EBP␣ that decrease during granulocytic differentiation, whereas expression of C/EBP␤ and C/EBP␦ is low in early myeloid stem cells and increase during granulocytic differentiation (11). Mice with genetic disruption of C/EBP genes have confirmed the importance of these proteins in myelopoiesis. These C/EBPs, together with other factors such as Myb, AML1, or PU.1 can induce the expression of myeloid-specific target genes (12)(13)(14)(15)(16)(17). C/EBP⑀ is expressed almost exclusively in the myeloid lineage of the hematopoietic system and functions during terminal differentiation of neutrophils and to a lesser extent macrophages; it also is involved in the regulation of cytokine gene expression in macrophages and T lymphocytes (5, 6, 18 -20). A neutrophil-specific, secondary granule deficiency syndrome in two individuals with repeated infections with germline mutations of C/EBP⑀ gene has recently been reported (21,22).
Structural/functional studies of the C/EBPs have been only partially explored using molecular genetic methods (23)(24)(25)(26)(27)(28)(29)(30). The conserved basic amino acid-rich region and leucine zipper domain at the carboxyl end probably adopt a helical configuration upon binding with DNA (31,32); the structures of the activation and repression regions are unknown. Several regulatory sites that can be phosphorylated have been identified (30,33,34). The C-terminal conserved cysteine of C/EBP␤ has been demonstrated to form a disulfide bond after dimer formation in vitro (9). In this report, we provide evidence that no disulfide bonds exist in C/EBP⑀ and conjecture that the same probably pertains to C/EBP␣ and C/EBP␦. Additionally, no disulfide linkage occurs between monomers after dimer formation in vivo. Alanine scanning studies of the activating domain 1 (ADM1) and ADM2 of C/EBP⑀ indicate that negatively charged amino acid residues in these two regions are critical for its transactivational activity. Translation analysis of these mutants suggests that formation of a special secondary structure of C/EBP⑀ mRNA results in the translation of the two major C/EBP⑀ isoforms.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The GST fusion protein vectors were constructed using the Escherichia coli expression vector, pGEX-5X-1 (Amersham Pharmacia Biotech). Full-length human C/EBP⑀ gene was obtained by PCR using pcDNA3 epsilon as template utilizing two primers, JT32N and NDELC (primer sequences listed in Table I). EcoRI and SalI sites were introduced into each C/EBP⑀ gene at the beginning and end of the constructs, respectively. Pfu DNA polymerase (Stratagene, La Jolla, CA) was employed, and the PCR was carried out under standard conditions. The PCR product was digested by EcoRI and SalI and inserted into the corresponding sites of pGEX-5X-1, downstream of the GST gene.
For promoter-reporter assays in mammalian cells, a CMV promoter expression vector, pCMV-␤-gal (GIBCO), was cut by EcoRI and XhoI, and the CMV promoter fragment was purified and ligated with the C/EBP⑀ gene with EcoRI and SalI cohesive sites. The wild-type C/EBP⑀ gene placed in the CMV-C/EBP⑀ plasmid (p32) was prepared from the GST-C/EBP⑀ (GST-⑀) construct. C280S mutant C/EBP⑀ gene (p32C280S) was obtained by PCR with primers JT32N and C280SA (Table II). Both C34S and C148S mutant genes were obtained by two rounds of PCR. Equal amounts of the two purified PCR products from primers JT32N and C34SR, and from primers NDELC and C34SF, were mixed as template for the full-length C34S mutant gene by a second round of PCR using primers JT32N and NDELC. The C148S mutant gene was prepared in the same manner as C34S mutant by using for the first reaction, primers JT32N and C148SR and primers NDELC and C148SF, and then primers JT32N and NDELC were used for the second round of PCR. All three of the PCR mutant genes were cleaved with EcoRI and SalI and inserted into the CMV promoter expression vector at EcoRI and XhoI sites, separately. Other mutant constructs were prepared in a similar manner. The mutant genes p32Y6A, p32Y7A, p32E8A, p32C9A, p32E10A, p32P11A, p32R12A, p32E8/10A, p32Y6/ 7A, p32S2E, p32S2ET5D, ␣p32, ␤p32, and ␦p32 used the wild-type C/EBP⑀ gene as template, and with the respective mutant primer and NDELC primer in the PCR amplification (Table I). The p32Y6/7AR12A was obtained with p32R12A as template. The mutant genes (p30, p30E3A, p30E5A, p30D9A, p30Y13A, ␣p30, ␤p30, ␦p30, and ␤p30R6A) were obtained also with the wild-type C/EBP⑀ gene as template, and with the respective mutant primer and NDELC primer in one PCR. Mutant gene p32M33I was obtained just like p32C148S and p32C34S. Human C/EBP␣ gene (generous gift from Dr. Daniel G. Tenen, Harvard Medical School, Cambridge, MA) was used to subclone the activation domain of C/EBP␣. A second PCR linked this fragment to the Nterminal 32-amino acid truncated C/EBP⑀ isoform with the N-terminal 32 amino acids deleted to get ␣Ap32. All recombinant constructs were confirmed by sequencing the DNA.

Expression and Purification of GST Fusion Proteins-
The GST fusion gene constructs were used to transform E. coli strain BL21. Single colonies of BL21 bearing GST-C/EBP⑀ gene were inoculated into 3 ml of 2ϫ YT medium containing 50 g/ml ampicillin. The cells grew at 37°C overnight with shaking. After 10-fold dilution into fresh 2ϫ YT medium, the cells were incubated with shaking for about 2 h until an A 600 of 0.5-2.0 was reached. IPTG was added to a final concentration of 0.5 mM, and the cells were cultured for another 6 h. Cells were spun down and kept at Ϫ20°C. For purification of the GST fusion proteins, the cells from 20 ml of culture medium were suspended in 1 ml of ice-cold PBS buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3) and sonicated in short bursts on ice. The supernatant was loaded onto a spin column (Centri-Spin-10, Princeton Separations) containing 20 l of PBS buffer and glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech). The elute from the column was reloaded 3-5 times to increase the recovery of the GST fusion protein. The beads containing the GST fusion protein were washed with 400 l of PBS buffer to purify partially the GST fusion protein. The yield for the GST-C/EBP⑀ fusion protein was about 1.5 g/ml cell culture, while the yield for GST was about 50 g/ml cell culture. 12% PAGE Ready Gels (Bio-Rad ) were used for SDS-PAGE analysis of the proteins. We tried to use factor Xa (751 units/mg, Amersham Pharmacia Biotech) to cleave C/EBP⑀ from the bound fusion protein. Only GST was released from the bound beads, but a negligible C/EBP⑀ band was detected in the fraction that was subjected to factor Xa cleavage, probably as a result of C/EBP⑀ degradation by contaminating proteases (data not shown).
Analysis of GST-C/EBP⑀-Unlike many purified GST proteins, the isolated GST-C/EBP⑀ contained a high content of nucleic acid probably as a result of its DNA binding capacity. For thiol group determination of the C/EBP⑀ protein, the Ellman's method was used with a minor modification (35). Protein samples were placed in 6 M guanidine HCl and PBS buffer. 5,5Ј-Dithio-bis-2-nitrobenzonic acid (Sigma) was prepared in the same buffer to 2 mM. DTT was used as a thiol group standard. 20 l of sample was mixed with 50 l of 5,5Ј-dithio-bis-2nitrobenzonic acid solution at room temperature. A 412 of the mixture was monitored in a microcuvette. Protein concentration was determined by the experimental equation: [protein mg/ml] ϭ 1.5 ϫ A 280 Ϫ 0.75 ϫ A 260 (36). Fresh GST-C/EBPC/EBP⑀ or GST protein samples were washed from the beads by 6 M guanidine HCl in PBS buffer and used for thiol group determination. For reduced preparation of the GST proteins, the protein was washed off of the beads with 6 M guanidine HCl in PBS buffer, and 100 mM DTT was added to a final concentration of 10 mM. The reduction reaction was allowed to proceed at 37°C for 40 min (37,38). Spin column with gel filtration medium was used to remove excess reductant and denaturant. The thiol group content of the protein was immediately measured.

TABLE I
Oligonucleotides used in the construction of the C/EBP mutations Western Blot of C/EBP⑀-Either reduced (0.1 M DTT) or non-reduced samples were employed for detection of dimers, which were possibly linked by disulfide bond(s), using 12% SDS-PAGE. For GST-C/EBP⑀ analysis, the proteins on the gel were transferred to a nitrocellulose membrane overnight in a cold room in Towbin transfer buffer (25 mM Tris, 192 mM glycine, 20% methanol, 0.1% SDS, pH 8.4) at a voltage of 22 V. The transferred membrane was exposed to 5% milk powder in TPBS buffer (10 mM sodium phosphate, 0.9% NaCl, 0.1% Tween 20) to block nonspecific protein binding sites. Rabbit polyclonal IgG against the rat C/EBP⑀ C-terminal epitope (C-22, LRNLFRQIPEAASLIKGVG-GCS, Santa Cruz Biotechnology, 100 g/ml) was used as the first antibody (1 g of antibody/ml) in the TPBS buffer containing 5% milk powder for 1.5 h. After washing twice with TPBS buffer, 15 l of second antibody, anti-rabbit IgG (horseradish peroxidase-linked whole antibody from donkey, Amersham Pharmacia Biotech), was added to 8 ml of 5% milk powder TPBS buffer and incubated for 1 h. After washing twice with TPBS, the membrane was subjected to a 1.5-ml mixture of equal volume of luminal/enhancer solution and stable peroxide solution (Pierce) for 5 min, then exposed to a x-ray film for several seconds.
Western blot analysis of C/EBP⑀ expressed from mammalian cells was performed in a similar manner as described above for bacterial expressed GST-C/EBP⑀. Much more sensitive rabbit polyclonal antibodies raised against the N-terminal C/EBP⑀ peptide were utilized as first antibody for this assay using a protein concentration of 1 g/ml (6).
Mammalian Cell Transfections and Protein Activity Analysis-For promoter-reporter assays in Jurkat cells, 3 ϫ 10 6 cells in 0.8 ml of serum-free RPMI 1640 medium were transfected with ϳ3 g of DNA (0.4 g of pCMV-C/EBP⑀ or mutant vectors, 2.5 g of pMim-luc, and 0.05 g of pRLSV40) in 20 l of Lipofectin and 200 l of Opti-MEM I reduced serum medium in 35-mm tissue culture plates as described by the manufacturer (Life Technologies, Inc.). The plasmid pRLSV40 (Promega) was included as a monitor of transfection efficiency and pMimluc (a gift from Dr. Achim Leutz, Max Delbruck-Centrum for Molekulare Medizin, Berlin, Germany) containing a C/EBP consensus DNA sequences as a reporter of transactivation. After 6 h of exposure to the DNA/liposome complexes, 4 ml of RPMI 1640 containing 10% fetal bovine serum was added. The cells were activated with 10 ng/ml 12-Otetradecanoylphorbol-13-acetate and 125 ng/ml calcium ionophore (A23187, Sigma) at 24 h after transfection. At 60 h after transfection, 1.5 ml of cell culture was centrifuged, and the cells were washed with 200 l of PBS buffer. Dual-luciferase assay kit (Promega) was used for analysis of activity of both firefly and Renilla luciferases. The washed cells were lysed in 30 l of passive lysis buffer at room temperature for 15 min., and 20 l of supernatant was used for assay of enzymatic activity. For experiments utilizing Myb, 0.4 g of plasmid containing c-myb gene under the control of the CMV promoter was mixed with the above 3 g of DNA. Either reduced or non-reduced total proteins from 100 l of cell culture were loaded onto a 12% SDS-polyacryamide gel for Western blot analysis to examine for disulfide bond formation.
C/EBP⑀ mRNA Secondary Structure Analysis-Possible mRNA structure of C/EBP⑀ was simulated with a RNA folding program (mfold version 3.0; Ref. 51). The folding temperature was fixed at 37°C. No other constraint information was entered during our simulation. Nterminal C/EBP⑀ mRNA sequences were used for this simulation, and structures having the lowest free energy were obtained.

C/EBP⑀ Contains No Disulfide
Bonds-Analysis of DNA coded protein sequences showed that C/EBP⑀ contained five cysteine residues (Fig. 1) (5). Analysis to determine if this protein has disulfide bonds was pursued to understand its structure. As the intact C/EBP⑀ protein is not easily available, the GST-C/EBP⑀ protein was used. The GST contains four free cysteines without a single disulfide bond and can serve as a control (39,40). The GST-C/EBP⑀ protein was removed from the glutathione beads by 6 M guanidine HCl in PBS buffer, and used directly for thiol group determination. Both freshly prepared GST-C/EBP⑀ and the reduced form gave ϳ10 free thiol FIG. 1. Protein sequence analysis and chimeric construction of C/EBP⑀. Panel A, conserved cysteine containing regions in human C/EBP⑀ (5, 6). Conserved cysteines are underlined, and amino acid residue numbers are shown below the sequence. Homologies of other human C/EBPs (2)(3)(4) in the basic/leucine zipper domain to this region of C/EBP⑀ are indicated as a percentage, which is shown in brackets. Panel B, ADM1 (amino acids 1-12) and ADM2 (amino acids 33-48) sequences of C/EBP⑀ are aligned with those of C/EBP␣, C/EBP␤, and C/EBP␦. The identical amino acids are represented by vertical lines, and conservative substitutions with colons. Panel C, schematic representation of the C/EBP⑀ chimeric constructs. The black bars show the ADM1 replacements and gray bars the ADM2 replacements. The names of the established constructs are on their right. ␣Ap32 represents the ADM1 of C/EBP␣ and the connecting sequence between ADM1 and ADM2 of C/EBP␣ linked to the ADM2 and the remaining carboxyl sequences of C/EBP⑀. groups, which was very close to the expected number of 9 (Table II). Both fresh GST and reduced GST gave ϳ4 thiol groups. RNase A was employed as another control. The data in Table II suggest that the active C/EBP⑀ protein contains no disulfide bonds. The partially purified GST or GST-C/EBP⑀ protein showed simultaneously oxidation at neutral pH (pH 7.3). In the presence of 6 M guanidine HCl, 70% of the thiol groups in the GST protein were observed to be oxidized in 1 week at 4°C, neutral pH. During this long period of storage of the GST-C/EBP⑀ protein on beads at 4°C, 0.1 mM DTT was added to the PBS buffer to prevent oxidation of the protein.
Nevertheless, the content of thiol groups dropped from 10.0 to 8.9 in 6 days under these conditions.
Effects of Mutations of Conserved Cysteines of C/EBP⑀ on Its Transactivation Activity-Among five cysteines in C/EBP⑀, three of them are relatively conserved within the C/EBP family members: Cys-34, Cys-148, and Cys-280 (Fig. 1A). In order to understand the role of these conserved cysteines in transactivation by C/EBP⑀, site-directed DNA mutagenesis was employed to change the corresponding cysteines to serines. The mutant genes were cloned into a eukaryotic expression vector and transfected into Jurkat cells (human T lymphocytes). No distinguishable changes in ability to transactivate were observed between wild-type and single site cysteine to serine mutants (Fig. 2). The expression levels of these mutants were almost the same as that of the wild-type C/EBP⑀ as evidenced by Western blot assay (Fig. 2). These results indicate that the three conserved cysteines are not critical for the transactivation of C/EBP⑀ and may not be involved in disulfide bond formation between two C/EBP⑀ monomers. Data have suggested previously that C/EBP⑀ may interact with Myb (41), therefore, the transactivation experiments were repeated with a c-myb expression vector included in the assays. Again, no notable changes in activity of the mutants as compared with the wild-type C/EBP⑀ were noted (data not shown).
Reducing and Non-reducing Gel Analysis of C/EBP⑀ Proteins-In order to know whether a disulfide bond was involved in dimer formation, C/EBP⑀ proteins expressed from either bacterial or mammalian cells were subjected to either reducing or non-reducing SDS-polyacrylamide gel analysis. Analysis of total bacterial expressed proteins in both the presence and absence of DTT showed that the vast majority of GST-C/EBP⑀ protein was in its monomer form (58 kDa; Fig. 3A, lanes 3, 4, 7,  and 8). No obvious dimer band (about 120 kDa) could be seen from the corresponding non-reducing lanes (lanes 4 and 8). Small amounts of high molecular weight proteins were detected because of the absence of DTT, but not those bands that would correspond to the dimers. A 50-kDa band below the GST-C/EBP⑀ most likely represents the N-terminal region of the C/EBP⑀-GST fusion product. A putative basic/leucine zipper domain was observed in lane 7. The proteins expressed from mammalian cells were also studied (Fig. 3B). The C280S mutant was included in the analysis, as Cys-280 was proposed to be involved in disulfide bond formation between C/EBP␤ monomers in vitro (9). Under both reducing and non-reducing conditions, the expressed C/EBP⑀ proteins were in their monomer form with no obvious dimer bands.
Mutational Studies of the ADM1 of C/EBP⑀-The activation and repression domains of C/EBP⑀ have been roughly defined by reporter gene analysis of various segments of C/EBP⑀ (23)(24)(25)(26)(27)(28)(29)(30). The amino acid sequences of the ADM1 and ADM2 of the C/EBP⑀ activation domain have been aligned with those of C/EBP␣, -␤, and -␦ (Fig. 1B). Full-length C/EBP⑀ contains both the ADM1 and ADM2 regions, while an N-terminal truncated FIG. 2. Transactivation activity of C/EBP⑀ bearing mutations of the conserved cysteines. Promoter-reporter assays were performed in Jurkat cells. These cells were transiently transfected with either pCMV-C/EBP⑀ or the cysteine mutants of C/EBP⑀, as well as the pMimluc as reporter and the pRLSV40 in order to measure transfection efficiency. Dualluciferase assay was used to measure activity of both firefly and Renilla luciferases. The level of transactivation was quantitated and normalized to the activity mediated by wild-type C/EBP⑀, which was set at 100%; values represent mean Ϯ S.D. of three independent transfections. A representative Western blot of expressed C/EBP⑀ and its mutant proteins harvested from Jurkat cells are shown on the top. Affinity column purified C/EBP⑀ polyclonal antibodies raised in rabbits were employed as first antibody. Two C/EBP⑀ isoforms are indicated by arrows. The band above the C/EBP⑀ protein represents a nonspecific band. C/EBP⑀ isoform (p30, with the first 32 amino acids deleted) retains only ADM2 (17,29). In order to get a clear understanding of the structure/function relationship of the activation domain of C/EBP⑀, both alanine scanning mutagenesis and domain swapping with the ADM1 of other C/EBP genes were used to construct a series of mutants in either the ADM1 or ADM2 region. If the transactivation activity of p32 transfected cells was taken as 100%, the activity for p30 was about 30%; but the ratio for expressed p32 compared with p30 isoform proteins was about 3 to 1 (Fig. 4A), indicating that the p32 and p30 isoforms gave nearly comparable transactivation activity in our assay system when adjusting for protein expression. Mutation of two tyrosines, p32Y6A and p32Y7A, produced a C/EBP⑀ that showed a slightly higher transactivation activity than the wildtype molecule. Combining both mutations (p32Y6/7A) gave the same activity as either alone (Fig. 4A). Further mutations explored the importance of the negative charge at the carboxyl end of the molecule. For mutant p32E8A and p32E10A, each gave slight activity, while p32R12A demonstrated a higher activity than the wild-type C/EBP⑀. The negative to positive charge change at both p32E8/10R nearly extinguished transactivational activity (Fig. 4A). Sequence analysis of the ADM1 regions of C/EBP-␣, -␤, and -␦ indicates that, unlike C/EBP⑀, each has a glutamic acid at position 2 (Figs. 1B). A glutamic acid at codon 2 of C/EBP⑀ was substituted for its serine (p32S2E), which resulted in an 1.8fold enhanced activity of p32 (Fig. 4A). This emphasizes that the negative charge of glutamic acid is important for transactivation activity. We also constructed a double mutant of C/EBP⑀ (p32S2ET5D) in order to mimic C/EBP␣ at these two positions; these changes almost doubled the transcriptional activity compared with wild-type C/EBP⑀ (Fig. 4A).
We substituted the entire ADM1 (codons 1-12) of C/EBP␣ (␣p32), C/EBP␤ (␤p32), and C/EBP␦ (␦p32) for the existing ADM1 of C/EBP⑀32, and examined the ability of these chimerics to transactivate a myeloid promoter (Figs. 1C and 4A). Both the ␣p32 and ␦p32 enhanced activity nearly 2-fold as compared with the wild-type C/EBP⑀. In contrast, the chimeric containing the ADM1 of C/EBP␤ conferred about the same activity as C/EBP⑀. In further experiments, the ADM1 and the connecting sequences between ADM1 and ADM2 of C/EBP␣ were linked to the rest of the C/EBP⑀, beginning at ADM2. This chimeric tripled the transactivating activity as compared with the wildtype p32 isoform of C/EBP⑀ (Fig. 5A).
Sequence analysis of ADM1 also indicates that C/EBP␤ has one more tyrosine residue between positions 7 and 8 as compared with the other C/EBPs (Fig. 1B). This may account for its lower activity, as mutant p32Y6A and p32Y7A both gave higher transactivating activity (Fig. 4A). A deletion mutant ␤p32⌬Y7 was constructed, and demonstrated nearly a 3-fold greater activity compared with either p32 or ␤p32. In summary, mutational studies showed that several negative charges of the first 32 amino acid peptides enhanced transcriptional activity of C/EPB⑀, while hydrophobic replacements in its ADM1 destroyed its activity. Furthermore, substitution of ADM1 of either C/EBP␣ or C/EBP␦ for the ADM1 of C/EBP⑀ enhanced the activity of the molecule.
Mutational Studies of the ADM2 of C/EBP⑀-Both alanine scanning mutagenesis and C/EBP fragment swapping were also used to explore transactivational activity of the ADM2 of C/EBP⑀. Negative charge appears also to be important for its activity. Mutants p30E3A, p30E5A, and p30D9A (p30 sequence numbering of ADM2, Fig. 1B) all gave lower activities as compared with p30 C/EBP⑀ (Fig. 4B). ADM2 domain swapping showed that both ␣p30 and ␦p30 gave an equivalent activity as p30 C/EBP⑀, while ␤p30 showed lower activity (Fig. 4B). Sequence analysis of the ADM2 region indicates that unlike C/EBP-␣, -␦, and -⑀, C/EBP-␤ has a basic arginine residue at position 6 ( Fig. 1B). This probably results in the lower activity of ␤p30, because a positive to neutral charge mutant ␤p30R6A resulted in a chimeric with higher activity (Fig. 4B). Unlike ADM1, the hydrophobic environment is not favorable for transactivation activity as the mutant p30Y13A gave a little lower activity than p30 (Fig. 4B). Mobility differences of the C/EBP⑀ mutants as related to charge were noted on Western blot analysis. Those proteins with a more positive charge (p32E8/10R and ␤p32) or a less negative charge (p32E8A, p32E10A, p30E3A, p30E5A, p30D9A) migrated more quickly on the gel (Fig. 4, A and B).
Effect of mRNA Structure on Translation of Two C/EBP⑀ Isoforms-The possible mRNA structure of C/EBP⑀ was simulated with a RNA folding program (mfold, version 3.0; Ref. 51). The sequences within 500 bp of the start of transcription were input and analyzed. A stable RNA structure was always identified at the 5Ј terminus of C/EBP⑀ after various sizes of the transcript were analyzed (Fig. 5). The coding amino acid for full-length C/EBP⑀ (p32) was marked in bold and italic numbers. This sequence includes the initial codon for full-length C/EBP⑀ and the initial codon for the truncated p30 C/EBP⑀ missing the N-terminal 32 amino acids. An 8-base pair doublehelical structure could be observed, with one strand sequence including the initial codon for the full-length C/EBP⑀ and the complementary strand representing the initial codon for the p30 truncated C/EBP⑀. The very close contact of these two initiation codons perhaps allows ribosomal co-translation of these two C/EBP⑀ isoforms. Mutational destruction of this putative structure resulted in a marked diminution of translation of the p30 isoform as shown in Fig. 4A. The mutants, p32S2E,  p32S2ET5D, ␣p32, ␤p32, ␦p32, ␣Ap32, and ␤p32 ⌬Y7, showed almost no expression of the p30 isoform as a result of destruction of this helical structure. Furthermore, the mutation of the initial codon of p30 C/EBP⑀ (p32M33I) resulted in p32 expression activity nearly equivalent to the wild-type p32, and no p30 expression occurred (Fig. 4A). In contrast, mutations at other positions (amino acids 6 -12; p32Y6A, p32Y7A, p32E8A, p32C9A, p32E10A, p32P11A, p32R12A, p32E8/10R, p32Y6/7A, and p32Y6/7AR12A), which did not intrude structurally on the 8-base pair helical structure, gave the expected 3 to 1 ratio of p32 and p30 C/EBP⑀ expression (Fig. 4A). DISCUSSION The majority of our knowledge about the structure and function of the C/EBPs derives largely from analysis of reporter systems defining activation and repression regions of this protein family (23)(24)(25)(26)(27)(28)(29)(30). Structural knowledge about conserved amino acid residues in the activation regions of the C/EBPs is limited. The C/EBP␤ protein during preparation of nuclear extract has been reported to be degraded rapidly by an endogenous protease with the basic/leucine zipper domain being the FIG. 4. Effect of ADM1 and ADM2 mutations on transactivation activity of C/EBP⑀. Promoter-reporter assays were employed for these mutational studies. Jurkat cells were transiently transfected with either pCMV-C/EBP⑀ (p32) or its mutants, together with the pMim-luc as reporter and the pRLSV40 as a measurement of transfection efficiency. The level of transactivation was quantitated and normalized to the activity of the wildtype C/EBP⑀, which was set at 100%. Values represent the mean Ϯ S.D. of three independent transfections. A representative Western blot of expressed C/EBP⑀ and its mutant proteins harvested from Jurkat cells is shown on the top panel. C/EBP⑀ affinity-purified, rabbit polyclonal antibodies were employed as first antibody. Two C/EBP⑀ isoforms are indicated by arrows as p32 and p30. The band above the C/EBP⑀ protein represents a nonspecific band. Panel A shows the data for ADM1 studies. Panel B displays the data for ADM2.
major remaining product C/EBP␤ (41). Likewise, during preparation of C/EBP⑀ protein by factor Xa cleavage of the GST-C/ EBP⑀ fusion protein, we found that the C/EBP⑀ protein is very unstable, making the recovery of intact C/EBP⑀ protein difficult. As GST contains four free thiol groups, we could gleam the status of the thiol groups of C/EBP⑀ by studying the GST-C/ EBP⑀ fusion protein. With partially purified GST-C/EBP⑀ protein ( Fig. 1), we determined that no disulfide bonds were present in C/EBP⑀ (Table II). Other C/EBPs, such as C/EBP␣ and -␦, may also possibly contain no disulfide bonds, as judged from their protein sequence homologies to that of the C/EBP⑀ (Fig.  1B). The transactivation analysis of C/EBP⑀ after mutation of each of the three conserved cysteines is also consistent with our hypothesis. Each of these mutants gave almost the same potency of transactivation as the wild-type protein, indicating that those three cysteines may not be involved in disulfide bond formation (Fig. 2). Additionally, these cysteines probably are not involved in binding to metal ions, as metal binding cysteines often play a key role in protein function. We have shown previously that Myb interacts and enhances the transactivation by C/EBP⑀ (17). In the presence of Myb, the three cysteine mutations of C/EBP⑀ showed no distinguishable change of activity as compared with co-transfection of c-Myb and the wildtype C/EBP⑀, suggesting that these cysteines may not participate in the interaction with Myb (data not shown).
The C/EBP␣, -␤, -␦, and -⑀ proteins as well as several other leucine zipper proteins such as Fra-1, Jun-B, and Fos-B (42-44) contain a conserved cysteine at their C-terminal region (Fig. 1A). A study has reported that this conserved cysteine could allow efficient disulfide cross-linking between paired leucine zipper helices, and all pairwise combination of dimer interactions among those family members were possible (9). Using non-reducing SDS-polyacrylamide gel analysis, we tested newly expressed C/EBP⑀ proteins within cells of both bacterial and mammalian origins. In both situations, we could not detect a notable dimer band on the gel, nearly all of the protein was in monomer form (Fig. 3). The mutant C280S, which would result in the loss of a putative covalent linkage between two monomers, gave a nearly identical gel band to that of the wild-type C/EBP⑀ (Fig. 3B). Meanwhile, the mutant C280S protein displayed the same transactivation activity as that of the wild-type C/EBP⑀ (Fig. 2). This strongly suggests that this conserved cysteine is not involved in disulfide bond formation in vivo. Studies have shown that oxidized old protein of a mutant C/EBP(L12V), which should be incapable of forming a dimer, could indeed form a dimer with wild-type C/EBP (9). Two cysteines can form a disulfide bond in vitro by random oxidation. We also observed the oxidation of both GST and GST-C/EBP⑀ proteins in vitro. Another study demonstrated that redox changes affect the in vitro DNA binding capacity of some leucine zipper proteins (45). The bacterially expressed DNA binding domains of Fos, Jun, and BZLF1 were unable to bind DNA under non-reducing conditions, whereas the binding of the C/EBP␣ DNA binding region was unaffected. Sensitivity to redox state is due to the presence of a conserved cysteine residue in the basic DNA binding motif of the Fos, Jun, and BZLF1 proteins but not C/EBP␣. Under non-reducing conditions, an intermolecular disulfide bridge was formed between the cysteine residues of each basic motif within a dimer, which prevented binding to DNA. C/EBP␣ could bind DNA in either the absence or presence of DTT. Furthermore, the nuclear extracts contained a moderately heat-stable factor, other than reduced glutathione, that could activate the DNA binding ability of F B G Z (F B ϭ Fos basic motif; G Z ϭ GCN4 leucine zipper) (44). This suggests that stronger reducing conditions exist in the nucleus than in vitro.
We propose that formation of a dimer disulfide bond at the conserved C-terminal cysteine of the C/EBP proteins may not exist in vivo. This cysteine in other homologous proteins such as Fra-1, Jun-B, and Fos-B may also not be involved in disulfide bond formation. In fact, many other leucine zipper proteins, including GCN4, can form dimers without the corresponding cysteine (32). Furthermore, a chimeric protein, in which the leucine zipper of C/EBP␣ was replaced by the analogous region of GCN4, showed similar DNA binding and transactivation activities as that of C/EBP␣ itself (22).
The DNA binding domain of C/EBPs is an ␣-helical structure. The configuration of the activation and repression regions of this family of proteins is still unknown, probably because of their flexible conformation. The negative charged residues in the poorly structured activation regions has been proposed to be responsible for stimulating the formation and activity of transcriptional preinitiation complexes, and these regions have been called "acidic blobs" (46 -48). Hydrophobic amino acid residues are also important for the potency of some transcriptional activators (49,50). In order to obtain a clearer understanding of the importance of the conserved residues in the activation region of C/EBP⑀, alanine scanning mutagenesis of the ADM1 and ADM2 regions of C/EBP⑀ was performed; furthermore, interchange of the ADM1 and ADM2 domains of C/EBP⑀ for those of other C/EBP family members (domain swapping) was examined.
Negative charge was critical for transactivational activity in both of the ADM1 and ADM2 regions. The negative to neutral charge mutants (p32E8A, p32E10A, p30E3A, p30E5A, and p30D9A) resulted in each case in lower activity compared with wild-type C/EBP⑀ (Fig. 4, A and B). Additionally, the negative to positive charge mutations (p32E8/10R) markedly diminished activity. The result of hydrophobic residue mutation (p30Y13A) in ADM2 agreed with the established theory that hydrophobic amino acid residues are important for the activity of some transcriptional activators (49,50).
The domain swapping analyses indicated that when ADM 1 FIG. 5. Effect of partial C/EBP⑀ mRNA secondary structure on its translation. The possible mRNA structure of C/EBP⑀ was simulated with a RNA folding program (mfold, version 3.0; Ref. 51). The folding temperature was fixed at 37°C. No other constraint information was entered during the simulation. Structure having lowest free energy was obtained. A 152-nucleotide sequence is shown above. This sequence includes the initial codon for full-length C/EBP⑀ and the initial codon for the truncated p30 C/EBP⑀ as marked by arrows. An 8-base pair doublehelical structure was also indicated, with one strand sequence including the initial codon for the full-length C/EBP⑀ and the complementary strand representing the initial codon for the p30 truncated C/EBP⑀.