CCAAT/enhancer-binding protein family members recruit the coactivator CREB-binding protein and trigger its phosphorylation.

CCAAT/enhancer-binding protein (C/EBP) family members are transcription factors involved in important physiological processes, such as cellular proliferation and differentiation, regulation of energy homeostasis, inflammation, and hematopoiesis. Transcriptional activation by C/EBPalpha and C/EBPbeta involves the coactivators CREB-binding protein (CBP) and p300, which promote transcription by acetylating histones and recruiting basal transcription factors. In this study, we show that C/EBPdelta is also using CBP as a coactivator. Based on sequence homology with C/EBPalpha and -beta, we identify in C/EBPdelta two conserved amino acid segments that are necessary for the physical interaction with CBP. Using reporter gene assays, we demonstrate that mutation of these residues prevents CBP recruitment and diminishes the transactivating potential of C/EBPdelta. In addition, our results indicate that C/EBP family members not only recruit CBP but specifically induce its phosphorylation. We provide evidence that CBP phosphorylation depends on its interaction with C/EBPdelta and define point mutations within one of the two conserved amino acid segments of C/EBPdelta that abolish CBP phosphorylation as well as transcriptional activation, suggesting that this new mechanism could be important for C/EBP-mediated transcription.

The CCAAT/enhancer-binding protein (C/EBP) 1 family is composed of pleiotropic transcription factors involved in tissuespecific metabolic gene transcription, in signal transduction activated by several cytokines, and in cell differentiation (for a review, see Refs. [1][2][3][4][5][6][7]. Six members of the family have been described so far: C/EBP␣, C/EBP␤, C/EBP␦, C/EBP␥, C/EBP⑀, and C/EBP (8). C/EBP isoforms bind to their cognate DNA element through a bipartite domain called bZIP. This domain consists of a basic region, contacting DNA, and a homo-or heterodimer-forming region called the leucine zipper (9). Because of the high conservation in the bZIP domain, C/EBP family members are able to form homo-or heterodimers, and all, except C/EBP, bind to the same cis-regulatory elements.
Knock-out mice were generated for different C/EBP isoforms (reviewed in Refs. 7, 8, and 10). These C/EBP-deficient mice display various phenotypes extending from perinatal lethality (for C/EBP␣) to subtle abnormalities. These different phenotypes suggest that C/EBP family members are not functionally redundant, which, to a certain extent, was confirmed by studies in cell cultures (reviewed in Ref. 7). Because the DNA binding domain of the C/EBP isoforms is highly similar, these functional differences must be mostly due to specific properties of their transactivation domain.
Relatively little is known about the way C/EBP family members activate transcription. C/EBP␣ interacts with TBP and TFIIB, two major components of the general transcription machinery (25). Moreover, C/EBP␣ and C/EBP␤ were shown to recruit the chromatin remodeling complex SWI/SNF (26,27). Modification of chromatin is an important step in the activation of gene transcription. Unlike ATPase/helicase-type remodeling complexes such as SWI/SNF, other large complexes contain proteins with histone acetyltransferase (HAT) enzymatic activity (28). These coactivators acetylate the histone tails of nucleosomes, thus favoring chromatin remodeling and activation of transcription. CREB-binding protein (CBP) and p300 belong to this class of coactivators. CBP was first identified through its ability to bind to the transcription factor CREB (29). Since then, several other transcription factors were shown to require CBP for efficient transactivation (for a review, see Refs. 30 -33). CBP functions as an adaptor between the tissue-and sequence-specific transcription factors and the general transcriptional machinery. Therefore, CBP is believed to activate gene transcription by recruiting basal transcription factors (TFIIB, TBP, and RNA polymerase II holoenzyme), by modifying chromatin structure via histone acetylation, and finally, by recruiting other histone acetyltransferases, such as SRC-1, ACTR, or P/CAF. CBP and p300 were shown to participate in C/EBP␣-and C/EBP␤-mediated gene transcription (34 -40). A direct interaction of C/EBP␤ with the E1A-binding domain of p300 was identified (35), suggesting that these C/EBP isoforms activate transcription, at least in part, by recruiting CBP/p300.
In this study, we provide evidence that interaction of C/EBP family members with CBP triggers its phosphorylation. We show for the first time that C/EBP␦ binds to CBP, and we identify two amino acid segments in C/EBP␦ that are necessary for this interaction. Mutations in one of these segments abolish C/EBP␦-mediated transactivation as well as CBP phosphorylation. Together, these results demonstrate that CBP is phosphorylated when it interacts with C/EBP family members, suggesting that CBP phosphorylation could play a key role in C/EBP-mediated gene transcription.
Cell Culture and Transfection Assays-HEK 293T cells were cultured on gelatin-coated plates in high glucose GLUTAMAX™ Dulbecco's modified Eagle's medium (Invitrogen) containing 10% heat-inactivated newborn calf serum (Invitrogen), 100 units/ml penicillin G, and 100 g/ml streptomycin sulfate (Invitrogen). Calcium phosphate transfection assays were performed according to an improved protocol described by Jordan et al. (44). Briefly, 293T cells were seeded at 8 ϫ 10 5 cells/60-mm plate and, 24 h later, were transiently transfected with 8 g of total DNA. The cell culture medium was changed 8 h later, and 24 h after transfection, cellular lysates were prepared for luciferase assays or immunoblot analysis.
Luciferase Assays-Transfected cells were washed with PBS, lysed in 250 l of 1ϫ cell culture lysis buffer (Promega), and centrifuged in a microcentrifuge for 2 min at 4°C. To test the samples for luciferase activity, 20 l of the 250ϫ diluted supernatant was assayed in a Turner-Designs TD-20/20 luminometer using 100 l of luciferase assay reagent (Promega). Luciferase activity was normalized to total cellular protein (Bio-Rad protein assay). All experiments were performed in triplicate.
Preparation of Cellular Extracts-PC12 cell nuclear extract was prepared according to Dignam et al. (45). To perform experiments with recombinant proteins expressed in 293T cells, whole cell extracts were prepared by lysing the cells in 50 mM Hepes, pH 7.6, 250 mM NaCl, 0.2 mM EDTA, 0.5% Nonidet P-40, 10 M NaF, and 10 M Na 3 VO 4 for 30 min at 4°C. Then the extracts were cleared by centrifugation at 16,000 ϫ g for 10 min. To prevent proteolysis, a protease inhibitor mixture for mammalian tissue (Sigma) and calpain inhibitor were included in the lysis buffer.
Immunoblot Analysis-After separation on SDS-polyacrylamide gels (SDS-PAGE), the proteins were transferred to polyvinylidene difluoride membranes with a semidry blotting system (Bio-Rad) for 45 min at 20 V as in Cardinaux et al. (46). Blots were blocked overnight at 4°C in TBST containing 10 mM Tris-HCl, 150 mM NaCl, pH 7.4, and 0.05% Tween 20, supplemented with 10% skim milk powder and 1% bovine serum albumin. Blots were subsequently incubated with a primary antibody in TBST plus 1% skim milk powder for 2 h at room temperature. Full-length CBP was detected with an anti-CBP 451-682 polyclonal antibody as in Cardinaux et al. (47), FLAG-tagged CBP fragments were detected with anti-FLAG M2 monoclonal antibody (Sigma), and HAtagged C/EBP isoforms were detected with anti-HA rat monoclonal antibody (Roche Applied Science). Finally, polyvinylidene difluoride membranes were incubated with horseradish peroxidase-conjugated secondary antibodies and developed using chemiluminescence detection kits (ECL (Amersham Biosciences) or SuperSignal® West Femto (Pierce)).
GST Pull-down Assays-GST fusion proteins were expressed in Escherichia coli (BL21-Codon Plus®, Stratagene) and purified over glutathione-Sepharose beads (Sigma). A 5 M concentration of the appropriate GST-C/EBP fusion proteins were incubated with 10 l of glutathione-Sepharose beads (Amersham Biosciences) in 150 l of buffer A (50 mM Hepes, pH 7.6, 1 M NaCl, 0.2% Nonidet P-40, 0.1 mM EDTA, 1 mM dithiothreitol) for 1 h at 4°C. The beads were washed twice with 1 ml of buffer A and twice with 1 ml of HEG100 (20 mM Hepes, pH 7.6, 10% glycerol, 100 mM KCl, 0.2 mM EDTA, 1 mM dithiothreitol). The GST-C/EBP bound beads were incubated for 2 h at 4°C with 450 g of proteins from PC12 nuclear extract. ϳ100 g of proteins from transfected 293T whole cell extract, or 20 l of in vitro translated CBP 1680 -1892 (using the TNT® T7 coupled reticulocyte lysate system from Promega) in 300 l of HEG100 supplemented with a protease inhibitor mixture for mammalian tissue (Sigma). The beads were washed three times with 1 ml of HEGN300 (20 mM Hepes, pH 7.6, 10% glycerol, 300 mM KCl, 0.1 mM EDTA, 0.1% Nonidet P-40, 1 mM dithiothreitol), resuspended in 25 l of SDS-PAGE loading buffer, and heated to 95°C for 5 min. For immunoblot analysis, 20 l were electrophoresed on a 6% SDS-PAGE for full-length CBP or on a 12% SDS-PAGE for CBP 1680 -1892 , and for Coomassie staining, 5 l were electrophoresed on a 12% SDS-PAGE.
Dephosphorylation of CBP 1680 -2441 -FLAG-tagged CBP 1680 -2441 was immunoprecipitated from whole cell extracts diluted in 50 mM NaCl, 10 mM Tris-HCl, pH 7.8, using 10 l of anti-FLAG M2-agarose beads (Sigma). For dephosphorylation assays, beads were washed three times in the same buffer, and for control reactions, the washing buffer was completed with 10 M NaF and 10 M Na 3 VO 4 . CBP fragments were incubated for 20 min at 37°C in 50 l containing 50 mM Tris-HCl, pH 8.5, 5 mM MgCl 2 , and 3 units of shrimp alkaline phosphatase (Promega). Control reactions were performed at 37°C in the same buffer without shrimp alkaline phosphatase, supplemented with 10 M NaF and 10 M Na 3 VO 4 to inhibit endogenous cellular phosphatases. Reactions were stopped by adding SDS-PAGE loading buffer, and FLAGtagged CBP 1680 -2441 was detected by immunoblot analysis. For each condition, the input is also shown (i.e. CBP fragments left on ice in 50 mM NaCl, 10 mM Tris-HCl, pH 7.8, and then eluted with SDS-PAGE loading buffer).

CBP Coactivates C/EBP␣, C/EBP␤, and C/EBP␦-mediated
Transcription-p300 and CBP are transcriptional coactivators for C/EBP␣ and C/EBP␤ (34 -40). To investigate whether C/EBP␦ is activating transcription through the same mechanisms, we performed a series of reporter gene assays using a synthetic C/EBP-responsive luciferase reporter gene (pGL3-5ϫC/EBP), which contained five copies of a consensus C/EBPbinding site in front of the SV40 promoter and the luciferase gene. 293T cells were transiently transfected with this reporter and the same amount of expression vectors encoding HAtagged C/EBP␣, C/EBP␤, or C/EBP␦ (Fig. 1). C/EBP␣ and C/EBP␤ increased luciferase gene expression by ϳ2-fold, whereas C/EBP␦-mediated transcription was 12-fold higher than basal luciferase activity. Regardless of these differences in transcriptional activity, overexpression of CBP increased C/EBP-mediated transcription by 2-3-fold for each C/EBP fam-ily member. The levels of the C/EBP isoforms were not affected in the presence of CBP (data not shown). Taken together, these data indicate that CBP is a rate-limiting coactivator involved in C/EBP␣-, C/EBP␤-, and C/EBP␦-mediated transcription and suggest that these C/EBP isoforms activate transcription at least partly by recruiting CBP.
C/EBP␤ and C/EBP␦ Interact with the E1A Binding Domain of CBP-The amino terminus of C/EBP␤ interacts with the E1A binding domain of p300 (35). However, interactions of C/EBP␤ or C/EBP␦ with CBP have not yet been shown. To bridge this gap, we performed a series of GST pull-down experiments with various purified proteins containing the transactivation domain of C/EBP␤ or C/EBP␦ fused to GST (Fig. 2). First, we showed that the amino acids 1-196 of C/EBP␦ were interacting with CBP from a PC12 cells nuclear extract ( Fig.  2A), suggesting that C/EBP␦ might recruit CBP in vivo. It was shown that C/EBP␤ recruits p300 through its E1A binding domain (35). To test whether C/EBP␤ or C/EBP␦ interact with the same region of CBP, we then expressed in HEK 293T cells a FLAG-tagged protein containing CBP amino acids 1680 -1892. A whole cell extract was prepared and incubated with GST-C/EBP␦  or various deletions of the C/EBP␤ transactivation domain fused to GST (Fig. 2B). A similar interaction with CBP 1680 -1892 was observed for C/EBP␤ 22-227 , C/EBP␤  , and C/EBP␦  , whereas further deleting C/EBP␤ to amino acid 103 reduced significantly its interaction with the E1A binding domain of CBP. Together, these experiments strongly suggest that C/EBP␤ as well as C/EBP␦ recruit the coactivator CBP through its E1A binding domain.
To confirm that C/EBP␤ and C/EBP␦ are able to interact with the E1A binding domain of CBP directly, without the help of accessory nuclear factors, CBP 1680 -1892 was translated in vitro and labeled with [ 35 S]methionine (Fig. 2C). GST-C/EBP␤  and GST-C/EBP␦  both pulled down the CBP fragment, thus demonstrating a direct interaction between C/EBP␤ or C/EBP␦ and the E1A binding domain of CBP.
C/EBP␣, C/EBP␤, and C/EBP␦ Modify the Electrophoretic Mobility of CBP-While we were studying C/EBP-CBP interactions, we noticed that the migration of CBP in a SDS-polyacrylamide gel was affected when it was coexpressed with C/EBP␣, C/EBP␤, or C/EBP␦ (Fig. 3). To better characterize this change in mobility, we next tried to find a CBP fragment displaying a similar shift in the presence of C/EBP␦. We first tested a fragment encoding amino acids 1680 -1892 of CBP, because we had shown in Fig. 2 that it was interacting with C/EBP␦. However, no mobility shift was observed with this CBP fragment (Fig. 4). Then we tested several CBP constructs, and, interestingly, we found that the electrophoretic mobility of CBP 1680 -2441 , containing the E1A binding domain and the C terminus of CBP, was modified when it was coexpressed with C/EBP␣, C/EBP␤, or C/EBP␦ (Fig. 4B). CBP 1893-2441 was not shifted in the presence of C/EBP␦, demonstrating that amino acids 1680 -1892 were required to induce a mobility shift of CBP 1680 -2441 . On the whole, these data suggest that by interacting with the E1A binding domain of CBP, C/EBP family FIG. 1. CBP potentiates C/EBP-mediated transcription. HEK 293T cells were transfected as described under "Experimental Procedures" with 2 g of pGL3-5ϫC/EBP and 0.25 g of pcDNA3-HA-C/EBP␣, -␤, or -␦, in the absence (Ϫ) or in the presence (ϩ) of 4 g of pcDNA3-CBP-2ϫFLAG. pcDNA3 was used to set the total amount of DNA to 8 g for each condition. Results are displayed as the mean Ϯ S.E. (n ϭ 3) relative luciferase activity. Values are normalized for protein levels. Note that these results were obtained in the same experiment but that a different scale was used for C/EBP␦ for presentation purposes.

FIG. 2. CBP interacts with C/EBP␤ and C/EBP␦ through its E1A binding domain.
A, binding of full-length CBP (CBP FL ) from PC12 nuclear extract to GST-C/EBP␦  . The specificity of the interaction was confirmed with a negative control consisting of GST alone. B, binding of the E1A binding domain of CBP (CBP 1680 -1892 ) expressed in HEK 293T cells to various GST-C/EBP␤ constructs or to GST-C/EBP␦   members induce covalent modifications in the C-terminal third of CBP, which in turn modify its migrating properties.
The Mobility Shift of CBP 1680 -2441 Is Due to Phosphorylation-The most common covalent modification of proteins that induce a mobility shift in SDS-polyacrylamide gel electrophoresis is phosphorylation. To determine whether the slow migrating forms of CBP 1680 -2441 were actually generated by phosphorylation, we tested whether the mobility shift observed in the presence of the C/EBP family members was affected by phosphatase treatment. As shown in Fig. 5, shrimp alkaline phosphatase totally abolished the slow migrating forms of CBP 1680 -2441 observed when C/EBP␣, C/EBP␤, or C/EBP␦ was coexpressed with this CBP fragment. This experiment thus demonstrated that the C/EBP-induced mobility shift is due to CBP phosphorylation.
To exclude a nonspecific effect of protein overexpression, we then transfected a series of constructs expressing different proteins together with CBP 1680 -2441 (Fig. 6). First, we showed that the truncated form of C/EBP␤, called LIP (48), which does not contain a transactivation domain, did not trigger a mobility shift in CBP 1680 -2441 . Moreover, CREB and ATF-1 that interact with the CREB-binding domain of CBP did not affect the mobility of the CBP fragment. These data thus confirmed that CBP phosphorylation was specifically triggered by the coexpression of transcriptionally active C/EBP family members.
The transcriptional activation of a protein kinase by the C/EBP isoforms could explain the appearance of the slow migrating forms of CBP. To test this possibility, we coexpressed CBP 1680 -2441 together with a fusion protein consisting of the Gal4 DNA binding domain and amino acids 1-152 of C/EBP␦ (Fig. 6, last lane). The bZIP DNA binding domain of C/EBP is lacking in this deletion mutant; therefore, it cannot activate any cellular genes. The mobility shift of CBP was still induced by Gal4-C/EBP␦ 1-152 , thus excluding a C/EBP-mediated transcriptional activation of a kinase. This last piece of data suggests that CBP phosphorylation is triggered by its interaction with the transactivation domain of C/EBP␦ even if C/EBP is not bound to DNA. (25) defined two regions highly conserved among C/EBP␣, C/EBP␤, and C/EBP␦ that they called box A and box B. Interestingly, deletions or mutations within these regions strongly decrease C/EBP␣ and C/EBP␤ transcriptional activity (25,36,40,49,50). Therefore, we first asked whether these conserved regions, known to be necessary for C/EBP␣-and C/EBP␤-mediated gene transcription, were equally important for C/EBP␦. As shown in Fig. 7A, in C/EBP␦, box A extends from amino acid 54 to 67, whereas box B consists of the amino acids 81-86. We performed a series of GST pull-down experiments and showed that deletion of box A or box B almost completely abolished the interaction of C/EBP␦ with full-length CBP from PC12 cell nuclear extract and strongly diminished the interaction with CBP 1680 -1892 (Fig. 7B). Furthermore, deletion of box A or box B markedly decreased the transcriptional activity of the corresponding Gal4-C/EBP␦ 1-152 mutants (Fig. 7C).

Two Conserved Amino Acid Segments of C/EBP␦ Are Critical for CBP Recruitment-Nerlov and Ziff
We then introduced some point mutations into C/EBP␦ based on previously described mutations in C/EBP␣ (25). In this C/EBP isoform, Tyr 67 of box A and Phe 77 and Leu 78 of box B were shown to be important for the interaction with TBP and TFIIB as well as for the activation of transcription. Introducing corresponding alanine point mutations into C/EBP␦ by changing Tyr 64 (mut 1), Leu 81 and Phe 82 (mut 2), or all three amino acids (mut 3) strongly diminished binding of full-length CBP as well as the interaction with CBP 1680 -1892 (Fig. 7B). Mutation of Leu 81 and Phe 82 in box B (mut 2) was clearly deleterious for the transcriptional activity of Gal4-C/EBP␦ 1-152 in contrast to the effect of Tyr 64 mutation, which is more difficult to understand (Fig. 7C). Tyr 64 mutation on its own did not impair Gal4-C/EBP␦ 1-152 activity, whereas the incorporation of this mutation into mut 2 further reduced its transcriptional activity (mut 3). Therefore, these data highlighted the importance of Leu 81 and Phe 82 for C/EBP␦-mediated transcription, showing a role for Tyr 64 only in combination with mutations of these two amino acids. Mutating Tyr 64 to alanine strongly reduced CBP binding (Fig. 7B, mut 1), whereas this mutation had no effect on the transcriptional level (Fig. 7C). This apparent discrepancy may be due to the intrinsic properties of the GST pulldown and luciferase assays. In the transfected cells, the interaction of C/EBP␦ and CBP occurred in a context that is difficult to mimic in vitro. In other words, Tyr 64 mutation might only marginally affect CBP recruitment in vivo, despite its effect in the GST pull-down experiments. Taken as a whole, these results indicate that the interaction of C/EBP␦ with CBP depends on two conserved regions that are critical for the transcriptional activity of C/EBP␦. C/EBP␦ Box B Is Required for CBP Phosphorylation-Having defined deletion and alanine mutations that affect CBP recruitment and transcriptional activity, we next tested the effect of these mutations on the ability of C/EBP␦ to induce CBP phosphorylation. Surprisingly, deletion of box A (⌬54 -67), which strongly reduced the interaction with CBP in the GST pull-down experiments, did not abolish the mobility shift of CBP 1680 -2441 (Fig. 8A). In contrast, deletion of box B (⌬81-86) or both box A and box B resulted in a strong inhibition of CBP phosphorylation. These results therefore suggest that CBP phosphorylation mostly rely on the integrity of box B. This was confirmed by testing the effect of Gal4-C/EBP␦ 1-152 alanine point mutations that were coexpressed with CBP 1680 -2441 in HEK 293T cells. As shown in Fig. 8B, replacement of Leu 81 and Phe 82 with alanine strongly reduced the mobility shift of the CBP fragment, whereas Tyr 64 mutation had no effect. Taken together, our data thus suggest that C/EBP␦ is triggering CBP phosphorylation mostly through box B.

DISCUSSION
Numerous studies have highlighted the presence of C/EBP binding sites in the promoter of genes involved in a variety of cellular processes (reviewed in Ref. 7). However, many questions remain about how C/EBP family members activate transcription. In this paper, we provide evidence that C/EBP-mediated transcription involves the recruitment of the coactivator CBP. We show that the interaction of C/EBP␦ with the E1A binding domain of CBP relies on two amino acid segments called box A and box B that are conserved between the activating members of the C/EBP family. Interestingly, C/EBP␣, C/EBP␤, and C/EBP␦ induce slow migrating forms of fulllength CBP, as well as of a C-terminal fragment of CBP, when they are coexpressed in HEK 293T cells. This mobility shift is due to CBP phosphorylation, because slow migrating forms disappear after phosphatase treatment. Finally, we show that deletion of box B or mutation of Leu 81 and Phe 82 , which alters CBP recruitment and transcriptional activity, almost completely abolishes CBP mobility shift as well, suggesting a link between CBP phosphorylation and C/EBP-mediated gene acti- vation. The effects of mutations that affect C/EBP␦ box A are more difficult to interpret. Deletion of box A (⌬54 -67) strongly reduces transcriptional activity and interaction with CBP but does not alter CBP phosphorylation. Moreover, mutation of Tyr 64 within box A markedly decreases CBP binding but neither affects transcriptional activity nor CBP phosphorylation. As mentioned earlier, this apparent lack of data correlation might be due to the properties of the assays used to monitor CBP binding, C/EBP-mediated transcriptional activity, and CBP phosphorylation. A mutation that reduces the interaction with CBP in vitro might not significantly alter CBP phosphorylation in vivo, because the nuclear environment could favor weak and transient interactions that would not be visible in the GST pull-down assays. Alternatively, CBP could be phosphorylated very rapidly even if the C/EBP⅐CBP complex is not stable. Altogether, C/EBP␦ mutations affect similarly CBP phosphorylation and transcriptional activity, except for the box A deletion mutant that strongly reduces transcriptional activity but not CBP phosphorylation. Other factors might interact with box A and participate in C/EBP-mediated transcriptional activation. Deletion of box A would thus impair recruitment of these factors, whereas Tyr 64 mutation would not. These factors could be, for instance, TBP or TFIIB, because they were shown to interact with this region in C/EBP␣ (25). However, the involvement of other coactivators is also possible.
We show that CBP phosphorylation relies both on the E1A binding domain, with which C/EBP isoforms interact, and the C-terminal glutamine-rich domain. This suggests that the interaction of C/EBP with CBP somehow recruits a protein kinase that phosphorylates one or several sites between amino acids 1893 and 2441. Recruitment of this yet uncharacterized kinase could occur in different ways. For instance, this kinase could be steadily associated with the C/EBP isoforms and could phosphorylate CBP when the C/EBP⅐CBP complex is built up. Alternatively, C/EBP binding may regulate the availability of a phosphorylation site by inducing a conformational change in CBP, or instead, the complex formed by C/EBP and CBP could be recognized by the kinase. Solving this issue will require identification of the protein kinase phosphorylating CBP in the presence of C/EBP. We tried to diminish CBP phosphorylation using many protein kinase inhibitors but were unable to determine which protein kinase is involved in this process.
A rather different interpretation of our data would be that C/EBP activates the transcription of a kinase gene whose product would then phosphorylate CBP. However, the following pieces of evidence argue against this possibility. First, our data suggest that CBP phosphorylation site(s) should be located between amino acids 1893 and 2441; nevertheless, CBP phosphorylation occurs only if the domain with which C/EBP interacts is present. If the role of C/EBP were solely to activate the transcription of a CBP kinase, then this domain would probably not be required. Second, deletion of box A (⌬54 -67) drastically reduced C/EBP␦ transcriptional activity. Accordingly, this C/EBP␦ mutant should not activate the kinase gene as efficiently as wild-type C/EBP␦, and thus CBP phosphorylation should be greatly reduced. However, CBP phosphorylation still occurred with this C/EBP␦ mutant, suggesting that the transcriptional activation of a kinase is not involved. Finally, the best evidence so far comes from the data obtained with the Gal4-C/EBP␦ 1-152 construct. The bZIP DNA binding domain of C/EBP␦ is lacking in this deletion mutant, and hence, it cannot bind to DNA or heterodimerize with endogenous C/EBP family members. Therefore, it is very unlikely that it could activate the gene of a CBP kinase. Nevertheless, CBP phosphorylation was induced by Gal4-C/EBP␦ 1-152 , suggesting that CBP phosphorylation is triggered by its interaction with the transacti-vation domain of C/EBP␦ rather than by the transcriptional activation of a protein kinase.
During the preparation of this manuscript, Schwartz et al. (51) showed that C/EBP family members also trigger the phosphorylation of p300 C-terminal part. Conserved posttranslational modifications of CBP and p300 suggest that they are important for gene regulation mediated by C/EBP and potentially other transcription factors as well. Identifying which serines or threonines are phosphorylated in CBP 1680 -2441 is an important issue to determine the functional role of this phosphorylation. In this CBP fragment, two putative phosphorylation sites were previously identified: first, a protein kinase A phosphorylation site on serine 1772 (52) that might increase Pit-1 function in the presence of cAMP (53), although this issue remains controversial (54); second, a protein kinase B phosphorylation site (threonine 1872 in mouse CBP) that modulates the interaction of C/EBP␤ with p300/CBP and mediates the effect of insulin on gene expression (38). We mutated Ser 1772 or Thr 1872 to alanine, but this had no effect on CBP phosphorylation induced by C/EBP␦ (data not shown).
What could be the role of CBP phosphorylation in C/EBPmediated gene transcription? It has been long known that CBP and p300 are phosphoproteins (reviewed in Refs. 31 and 33). For instance, cell cycle-dependent phosphorylation of p300 was first observed by Yaciuk and Moran (55). Likewise, retinoic acid and E1A induce phosphorylation of p300 during the differentiation of F9 cells (56). However, the huge size of p300 and CBP hindered the identification of phosphorylation sites within these coactivators. Consequently, only a few studies could correlate p300/CBP phosphorylation with a functional effect. For example, Ser 301 of CBP was identified as a major target of CaMKIV phosphorylation in hippocampal neurons (57). Mutation of Ser 301 impaired N-methyl-D-aspartate-and CaMKIVstimulated transcription, demonstrating that CaMKIV signaling contributes to CREB/CBP-dependent transcription by phosphorylating CBP at Ser 301 . In addition, Ser 89 of p300 was shown to be phosphorylated in vivo, most probably by protein kinase C␣ (58). Another report identified this site as a target of AMP-activated protein kinase (59). In both cases, Ser 89 phosphorylation is inhibitory, since it either diminishes the p300 intrinsic HAT activity (60) or blocks the interaction of p300 with nuclear receptors (59). Several other studies reported regulation of p300 or CBP by protein kinases such as CaMKIV (61)(62)(63), MAPK (64 -68), or cyclin E-Cdk2 (69 -71), but none of them identified the amino acids targeted by these kinases. Interestingly, however, many of these studies showed that CBP or p300 were more important targets of protein kinases than the transcription factors themselves. For instance, MAPK appears to stimulate Elk-1-mediated gene expression by phosphorylating the C-terminal part of CBP rather than Elk-1 itself (64). Similarly, the activation of c-Jun by calcium does not require c-Jun phosphorylation but rather activation of CBP by calcium/calmodulin-dependent kinases (63). Pit-1-mediated regulation of the growth hormone and the prolactin genes involves the cAMP-protein kinase A pathway. However, mutating all the protein kinase A consensus sites in Pit-1 has no effect on the transcription of the growth hormone gene, suggesting that protein kinase A-mediated phosphorylation of CBP is responsible for cAMP activation (72). The prolactin gene is also inducible by cAMP without involving Pit-1 phosphorylation (68). In this case, cAMP was suggested to induce the MAPK family, which in turn phosphorylates CBP/p300 to stimulate transcription.
Stimulation of HAT activity was shown to be a way of increasing the coactivator function of CBP in a phosphorylationdependent manner. A remarkable increase in histone acetyla-tion was detected when the C-terminal region of CBP was phosphorylated in vitro by the p44 MAPK/extracellular signalregulated kinase 1 (67). Similarly, the cyclin E⅐Cdk2 (to be consistent with C/EBP⅐CBP complex, which may target the same sites as MAPK, phosphorylates CBP in a cell cycle-dependent manner, thus enhancing its HAT activity (71). With this in mind, we measured the HAT activity of CBP coexpressed with C/EBP␣, C/EBP␤, or C/EBP␦ but found no significant difference with the activity of CBP expressed alone (data not shown).
In conclusion, one can speculate that C/EBP-mediated phosphorylation of CBP might have many diverse effects. For instance, it could induce a conformational change in CBP that might favor its interaction with components of the basal transcription machinery (e.g. TBP or TFIIB). On the other hand, phosphorylation of CBP could trigger the recruitment of other chromatin remodeling complexes. Sorting out this issue will require identification of which protein kinase as well as which phosphorylation sites are involved in this process. However, regardless of what will result from these studies, one can already anticipate that CBP phosphorylation is an important modulatory mechanism in C/EBP-mediated gene transcription.