CBP Alleviates the Intramolecular Inhibition of ATF-2 Function*

The transcription factor ATF-2 (also called CRE-BP1), whose DNA-binding domain consists of a basic amino acid cluster and a leucine zipper (b-ZIP) region, binds to the cAMP response element as a homodimer or as a heterodimer with c-Jun. The amino-terminal region of ATF-2 containing the transcriptional activation domain is phosphorylated by stress-activated kinases, which leads to activation of ATF-2. We report here that CBP, which was originally identified as a co-activator of CREB, directly binds to the b-ZIP region of ATF-2 via a Cys/His-rich region termed C/H2, and potentiates trans-activation by ATF-2. The b-ZIP region of ATF-2 was previously shown to interact with the amino-terminal region intramolecularly and to inhibit trans-activating capacity. The binding of CBP to the b-ZIP region abrogates this intramolecular interaction. The adenovirus 13S E1A protein which binds to the b-ZIP region of ATF-2 also inhibited this intramolecular interaction, suggesting that both CBP and 13S E1A share a similar function as positive regulators of ATF-2. We found that the b-ZIP regions of c-Jun and CREB also interact with the C/H2 domain of CBP, suggesting that CBP acts as a regulator for a group of b-ZIP-containing proteins. These results shed light on a novel aspect of CBP function as a regulator for a group of b-ZIP-containing proteins.

So far, a number of transcription factors of the ATF/CREB family have been identified. All members of this family contain a DNA-binding domain consisting of a cluster of basic amino acids and a leucine zipper region, the so-called b-ZIP. They form homodimers or heterodimers through the leucine zipper and bind to the cAMP response element (CRE) 1 (1). Among many of the transcription factors of the ATF/CREB family, three factors, ATF-2 (also called CRE-BP1), ATF-a, and CRE-BPa, form a subgroup (2)(3)(4)(5). A common characteristic of this group of factors is the presence of a transcriptional activation domain containing the metal finger structure located in the amino-terminal region (5,6). These factors are capable of form-ing homodimers or heterodimers with c-Jun and bind to CRE (1,5,6). Among these three factors, ATF-2 has been more extensively studied and shown to be ubiquitously expressed with the highest level of expression being observed in the brain (7). The mutant mice generated by gene targeting exhibited decreased postnatal viability and growth with a defect in endochondrial ossification and a decreased number of cerebellar Purkinje cells (8). The stress-activated protein kinases (SAPK) such as Jun amino-terminal kinase and p38 phosphorylate this group of factors at sites close to the amino-terminal transcriptional activation domain, and stimulate their trans-activating capacity (9 -11). Since a group of factors of the ATF/CREB family including CREB are activated via direct phosphorylation by cAMP-dependent protein kinase (PKA) (12), these two groups of factors, CREB and ATF-2, are linked to the distinct signaling cascades involving the PKA and SAPK pathways. The adenovirus 13S E1A activates CRE-dependent transcription, and this transcriptional activation is mediated by ATF-2 (13,14). E1A binds to the b-ZIP region of ATF-2 (15). Recently, it was reported that the b-ZIP region of ATF-2 interacts with the amino-terminal region intramolecularly (16), and this interaction appears to inhibit the trans-activating capacity of ATF-2. However, the co-activator that binds to the aminoterminal activation domain remains unidentified, and the mechanism of transcriptional activation by ATF-2 needs to be clarified.
The transcriptional co-activator CBP was originally identified as a protein that binds to the PKA-phosphorylated form of CREB (17). CBP also binds to multiple components of the basal transcriptional machinery, including TFIIB (18) and the RNA polymerase II holoenzyme complex (19), suggesting that CBP serves as a molecular bridge between CREB and the basal transcriptional machinery. In addition to CREB, many other transcription factors including c-Jun (20), c-Fos (21), c-Myb (22), nuclear hormone receptors (23,24), Stat2 (25), and MyoD (26) were recently demonstrated to bind to CBP (for review, see Ref. 27). CBP contributes to the transcriptional activation mediated by each of these factors. Although multiple transcription factors bind to CBP, there is a striking difference in the role of CBP depending on the transcriptional activator. For instance, CBP binds to the transcriptional activation domains of CREB and c-Myb (17,22). In the case of nuclear hormone receptors, however, other co-activators bind to the transcriptional activation domain, and CBP binds to a different domain, indicating that CBP functions as a integrator for nuclear hormone receptors (23,24).
The amount of CBP in mammalian cells appears to be limiting, as a 50% reduction in the amount of CBP causes abnormal pattern formation in human (known as Rubinstein-Taybi syndrome) (28) and mouse (29). Recent genetic analyses using Drosophila CBP mutants indicated that CBP is required for transcriptional activation by Cubitus interruptus (Ci) and Dorsal (Dl), which are homologs of mammalian factors glioblastoma (GLI) and NF-B, respectively (30,31). These results suggest that the decreased expression level of target genes of these transcription factors such as Bmp and Twist lead to the deficiency in pattern formation.
In addition to the finding that CBP itself has histone acetyltransferase (HAT) activity (32,33), CBP forms a complex with multiple HATs such as P/CAF, ACTR, and SRC-1 (34 -36), suggesting that the CBP complex contributes to transcriptional activation by disrupting the repressive chromatin structure. The cbp gene family contains at least one other member, p300, that was originally identified through its ability to bind to the adenovirus E1A protein (37), and E1A binds to both CBP and p300 (38,39). Binding of E1A to CBP inhibits transcriptional activation mediated by CBP. Since E1A and HAT P/CAF bind to the same region of CBP, the mechanism of inhibition of CBP activity by E1A was postulated to be due to blocking of P/CAF binding to CBP (34).
To investigate the possibility that CBP is also involved in transcriptional activation by ATF-2, we have examined for a direct interaction between CBP and ATF-2. Our results indicate that CBP functions as a regulator of ATF-2 by binding to its b-ZIP region.

EXPERIMENTAL PROCEDURES
Plasmid Construction-To express various forms of GST-CBP fusion proteins in Escherichia coli, the plasmids pGEX-KIX, pGEX-Bromo, pGEX-C/H2, and pGEX-C/H3 were made by inserting the appropriate fragment encoding the 265-(amino acids 454 -718), 104-(amino acids 1087-1190), 437-(amino acids 1190 -1626), and 257-amino acid (amino acids 1621-1877) regions of mouse CBP, respectively, into the appropriate site of the pGEX vector (see Ref. 40; Amersham Pharmacia Biotech). The plasmid to express the GST fusion protein containing the amino-terminal 253 amino acids of ATF-2 was constructed using the pGEX-2TK vector. The modified pSP65 vector pSPUTK (Stratagene) was used for in vitro transcription/translation of various forms of ATF-2. A series of mutants of ATF-2 was described previously (6). The plasmid to express CREB, c-Jun, c-Fos, or E1A by in vitro transcription/ translation system was also made using pGEM (Promega), pBluescript, and the pcDNA3 vector (Invitrogen), respectively. The plasmids to express Gal4-CBPC/H2 in which the C/H2 domain of CBP (amino acids 1182-1500) was fused to the DNA-binding domain of Gal4 (amino acids 1-147) were constructed by the polymerase chain reaction-based method using the cytomegalovirus promoter-containing vector, pSTCX556 (41). The plasmid encoding the VP16 fusion protein containing the carboxyl-terminal region of ATF-2 (amino acids 291-505), CREB (amino acids 191-341), or c-Jun (amino acids 201-334) joined to the VP16 activation domain was constructed similarly. The plasmids to express CBP and E1A were described previously (22).
In Vitro Binding Analysis with GST Fusion Proteins-The GST pulldown assay using GST-CBP and in vitro translated ATF-2, CREB, c-Jun, or c-Fos was essentially performed as described (22). The expression of the GST fusion protein or GST alone in E. coli and preparation of the bacterial lysates containing these proteins were done as described (42). Samples of bacterial lysate containing 20 g of GST or GST-CBP were rocked for 2-3 h at 4°C with 100 l of glutathione-Sepharose beads (Amersham Pharmacia Biotech). The beads were washed with 1 ml of PBS 5 times and then with 1 ml of binding buffer (20 mM Hepes, pH 7.7, 75 mM KCl, 0.1 mM EDTA, 2.5 mM MgCl 2 , 1% skim milk, 1 mM dithiothreitol, 0.05% Nonidet P-40). Various forms of ATF-2, CREB, c-Jun, or c-Fos were synthesized with [ 35 S]methionine using an in vitro transcription/translation kit according to the procedures described by the supplier (Promega). Then, a sample from the reaction was mixed with 750 l of binding buffer and the GST or GST-CBP affinity resin. After rocking at 4°C overnight, the resin was washed with 1 ml of binding buffer 5 times and mixed with SDS-sample buffer, and the bound proteins were released by boiling. The proteins were analyzed by SDS-PAGE followed by autoradiography. In the experiments to examine the effects of phosphorylation by PKA, 25 l of lysate containing the in vitro translated CREB was mixed with 175 l of the kinase buffer to give final concentrations of 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 12 mM MgCl 2 , and 2.5 mM ATP, and the mixture was incubated with or without 100 units of the catalytic subunit of bovine PKA (Sigma). The potato acid phosphatase treatment of the in vitro translated CREB was performed as described (22) except for the use of 1 milliunit of phosphatase.
To examine the intramolecular interaction of ATF-2, the GST pulldown assay using the GST fusion protein containing the amino-terminal portion of ATF-2 and the in vitro translated carboxyl-terminal region of ATF-2 was done similarly, except for the use of a low stringency binding buffer (10 mM Hepes, pH 8.0, 50 mM KCl, 2.5 mM MgCl 2 , 50 mM ZnCl 2 , 0.025% Nonidet P-40, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 100 g/ml bovine serum albumin).
To investigate whether the SAPK-phosphorylated activation domain of ATF-2 binds to CBP, GST pull-down assays were performed as follows. HepG2 cells were transfected with 5 g of the ATF-2 expression plasmid, and the transfected cells were treated with sorbitol (0.5 M) or PBS for 30 min before the preparation of cell lysates. Three days after transfection, cell lysates were prepared using the lysis buffer (50 mM Hepes, pH 7.5, 250 mM NaCl, 0.2 mM EDTA, 50 mM NaF, 0.5% Nonidet P-40, 2 mM Na 3 VO 4 , 25 mM ␤-glycerophosphate, and 0.1 nM okadaic acid) containing the protease inhibitor mixture (Boehringer Mannheim) and mixed with the resin containing 20 g of the GST-CBP fusion protein containing various parts of CBP. The bound proteins were eluted by SDS sample buffer and analyzed by Western blotting using the anti-ATF-2 antibody and chemiluminescent detection reagents (New England Biolabs).
Two-hybrid Assay in Mammalian Cells-To investigate the in vivo interaction between the C/H2 domain of CBP and the b-ZIP region of ATF-2, two-hybrid assays were done using HepG2 cells essentially as described (43). A mixture containing 1 g of the luciferase reporter plasmid containing three copies of the Gal4-binding site linked to the TK promoter, 3 g of the Gal4-CBPC/H2 expression plasmid, 3 g of VP16-ATF-2, or VP16 expression plasmid, and 0.5 g of the internal control plasmid pRL-TK (Promega), in which the sea-pansy luciferase gene is linked to the TK promoter, was transfected into HepG2 cells by using the CaPO 4 method. Luciferase assays were performed using the dual-luciferase assay system (Promega).
Co-immunoprecipitation-For co-immunoprecipitation of CBP and ATF-2, a mixture of 5 g of the CBP expression plasmid, pcDNA3-CBP, and 5 g of the plasmid to express wild type ATF-2 was transfected by using the CaPO 4 method into 293T cells which corresponds to the adenovirus type 5-transformed human embryonic kidney 293 cells containing the SV40 large T antigen (44). Two days after transfection, cells were lysed by rocking in the lysis buffer (50 mM Hepes, pH 7.5, 250 mM NaCl, 0.2 mM EDTA, 10 mM NaF, 0.5% Nonidet P-40) at 4°C for 1 h and centrifuged at 13,000 rpm for 20 min. The supernatant was mixed with 1.5 volume of the lysis buffer lacking NaCl to decrease the salt concentration to 100 mM. Lysates were immunoprecipitated using anti-CBP antibodies NT (Upstate Biotechnology Inc.) or anti-␤-galactosidase, and the immune complexes were separated on 10% SDS gels and analyzed by Western blotting using anti-ATF-2 antibodies (7) and ECL detection reagents (Amersham Pharmacia Biotech). For co-immunoprecipitation of Gal4-CBPC/H2 and VP16-CREBDBD or VP16-c-JunDBD, a mixture of 5 g of the Gal4-CBPC/H2 expression plasmid and 5 g of the plasmid to express VP16-CREBDBD or VP16-c-JunDBD was transfected into 293 cells. Immunoprecipitation was performed similarly using anti-VP16 antibody V-20 (Santa Cruz), and the immune complexes were separated on 10% SDS gels and analyzed by Western blotting using anti-Gal4 antibody DBD (Santa Cruz) and chemiluminescent detection reagents (New England Biolabs).
CAT Co-transfection Experiments-The CAT co-transfection assay was done essentially as described (14). A mixture of 4 g of the reporter plasmid DNA pMFcolCAT6MBS-I, in which the CAT gene is linked to the mouse ␣2(I)-collagen promoter and six tandem repeats of the Mybbinding site MBS-I, 5 g of effector plasmid DNA to express c-Myb-ATF-2 fusion protein consisting of the DNA-binding domain of c-Myb and full-length ATF-2, and 1 g of the internal control plasmid pact-␤gal was transfected into Chinese hamster ovary cells (CHO-K1). The plasmid to express c-MybDBD-ATF-2-Gla4, in which the carboxyl-terminal 512 amino acids of c-MybDBA-ATF-2 containing the b-ZIP region was replaced by the DNA-binding domain of Gal4 (amino acids 1-147), was constructed using the polymerase chain reaction-based method and also used for co-transfection assays as the effector plasmid. To examine the effect of E1A and CBP, 5 g of the E1A13S expression plasmid and 6 g of the CBP expression plasmid were added. The total amount of DNA was adjusted to 21 g by adding the control plasmid DNA lacking the cDNA to be expressed. Forty hours after transfection, cell lysates were prepared, and CAT assays were done. The amounts of lysates used for the CAT assay were normalized with the ␤-galactosidase activity expressed from the internal control plasmid pact-␤-gal. For UV stimu-lation, cells were washed with PBS 8 h before preparation of cell lysates and irradiated (45 J/m 2 ) for 15 s. All co-transfection experiments were repeated at least two times, and the difference between each set of experiments was no more than 20%. Typical results are shown in Fig. 4.

RESULTS
In Vitro Binding of CBP to ATF-2-To examine whether CBP directly binds to ATF-2, we first used the GST pull-down assay. The full-length form of mouse CBP synthesized using an in vitro transcription/translation system bound to the GST fusion protein containing full-length ATF-2 (data not shown). To narrow down the specific region in CBP responsible for interaction with ATF-2, a series of GST fusion proteins containing various parts of CBP were made and used for the GST pull-down assays (Fig. 1, A and B). The 35 S-ATF-2 protein was synthesized using an in vitro transcription/translation system and mixed with various GST-CBP resins. Approximately 51% of input ATF-2 bound to the GST fusion protein containing the Cys/His-rich region of CBP, termed the C/H2 domain, whereas other GST fusion proteins containing other regions of CBP and GST alone as a control did not interact with ATF-2 (less than 2% of input) (Fig. 1C).
To examine further which region of ATF-2 interacts with CBP, we used deletion mutants of ATF-2 for the binding assay (Fig. 2, A and B). We previously identified the amino-terminal region between amino acids 19 and 50, which contains a metal finger, as a transcriptional activation domain (6). The NT50 mutant lacking this activation domain still retained almost full capacity to bind to CBP. About 27 and 30% of the input wild type ATF-2 protein and NT50 bound to the GST-CBP resin, respectively. The two amino-truncated mutants, NT253 and NT341, lacking the amino-terminal 253 and 341 amino acids, respectively, also bound to CBP as efficiently as the wild type. These results indicate that CBP binds to the carboxyl-terminal 164-amino acid region containing the b-ZIP region. The two mutants lacking the cluster of basic amino acids of the b-ZIP region, ⌬BR and NT253⌬BR, failed to interact with CBP. In addition, the two mutants in which the third and fourth leucine in the b-ZIP region were changed to valine, L34V and NT253L34V, did not bind to CBP. The CT91 mutant lacking the carboxyl-terminal 91 amino acid region bound to the GST-CBP resin, but its binding efficiency was significantly lower than that of wild type. These results indicate that CBP binds to the b-ZIP region of ATF-2 (amino acids 338 -407) and that the region downstream from the b-ZIP region enhances the interaction with CBP.
In Vivo Interaction between CBP and ATF-2-To confirm the in vivo interaction between ATF-2 and CBP in mammalian cells, co-immunoprecipitation was performed (Fig. 3A). The two plasmids to express ATF-2 and CBP were transfected into 293T cells, and the cell lysates were immunoprecipitated with anti-CBP antibody or control antibody against anti-␤-galactosidase. ATF-2 was co-immunoprecipitated with anti-CBP antibody but not with the anti-␤-galactosidase antibody.
To confirm further the in vivo interaction between ATF-2 and CBP, we performed in vivo two-hybrid assays in mammalian cells (Fig. 3B). Two chimeric proteins were created by fusing the CBP fragment containing the C/H2 region in frame to the DNA-binding domain of Gal4 and by fusing the carboxylterminal region of ATF-2 containing the b-ZIP to the transcriptional activation domain of VP16. Transcriptional activation was then examined in HepG2 cells transfected with a combination of these constructs. The basal activity is the luciferase activity obtained by a combination of Gal4 DNA-binding domain and VP-16. The VP16 protein fused to the carboxylterminal region of ATF-2 stimulated Gal4-CBP activity 7.3fold, whereas VP16 alone stimulated only 2.2-fold.
Furthermore, VP16-ATF-2 enhanced the Gal4 activity only 1.7-fold. These results indicated that the C/H2 domain of CBP interact with the carboxyl-terminal region of ATF-2 containing the b-ZIP structure.
Potentiation of ATF-2-dependent trans-Activation by CBP-To investigate the role of CBP in transcriptional activation by ATF-2, we performed some CAT co-transfection experiments using Chinese hamster ovary cells (Fig. 4). Since ATF-2 was expressed in all of the cell lines examined, it was difficult to analyze the transcriptional activation resulting from the exogenous ATF-2 expressed from the transfected DNA. Therefore, we used the fusion protein consisting of the c-myb gene product (c-Myb) and ATF-2 (14). c-Myb The results of the binding assays shown in C are indicated on the right. The relative binding activities are designated ϩ and Ϫ, which indicate binding of 50 -60% and less than 2% of the input protein, respectively. B, analysis of the GST-CBP fusion proteins. The bacterial lysates containing 3-5 g of various GST-CBP fusion proteins or control GST were mixed with the glutathione-Sepharose resin and washed. The bound proteins were analyzed by 10% SDS-PGE followed by Coomassie Brilliant Blue staining. C, binding of ATF-2 to various types of GST-CBP fusion proteins. In the input lane, 35 S-labeled wild type ATF-2 was synthesized in vitro and analyzed by 10% SDS-PAGE. In other lanes, the 35 S-ATF-2 was mixed with various GST-CBP affinity resins, and the bound proteins were analyzed by 10% SDS-PAGE followed by autoradiography. The binding assays were performed using the affinity resin containing 20 g of each of the fusion proteins or control GST. The amount of ATF-2 in the input lane was 10% that used for the binding assay.
is a sequence-specific DNA-binding protein, which is predominantly expressed in immature hematopoietic cells but not in many other cells. Therefore  1 and 4). Co-transfection of the CBP expression plasmid with the MybDBD-ATF-2 expression plasmids enhanced the level of CAT activity about 2.7fold (cf. lanes 4 and 5). The E1A13S product also potentiated by 2.6-fold the trans-activation by MybDBD-ATF-2 as reported previ-ously (cf. lanes 4 and 6), but additional enhancement of the transactivating capacity of MybDBD-ATF-2 was not observed by coexpression of both CBP and E1A (lane 7). Thus, both E1A and CBP enhance the trans-activating capacity of ATF-2, but they cannot additionally stimulate ATF-2 activity.
To confirm further that CBP and E1A stimulate the ATF-2 activity via binding to the b-ZIP region of ATF-2, we constructed the fusion protein MybDBD-ATF-2-Gal4, in which the b-ZIP region was replaced by the DNA-binding domain of Gal4. This fusion protein enhanced the CAT expression from the pMFcolCAT6MBS-I reporter 3.1-fold. As expected, neither CBP nor E1A enhanced the trans-activating capacity of this fusion protein (Fig. 4, lanes 8 -10). Although we expected that this fusion protein would have the higher trans-activating capacity The results of the binding assays shown in B are indicated on the right. The relative binding activities of the mutants are designated ϩ and Ϫ, which indicate the binding of 25-50% and less than 2% of the input protein, respectively. B, binding of ATF-2 mutants to CBP. In the input lanes, various forms of ATF-2 shown above each lane were synthesized in vitro and analyzed by 10% (upper panel) or 12% (lower panel) SDS-PAGE. In other lanes, the 35 S-labeled proteins shown above each lane were mixed with the GST-CBP affinity resin, which contains the C/H2 region, and the bound proteins were analyzed by 10 or 12% SDS-PAGE followed by autoradiography. In the input lanes, the amount of protein was 10% that used for the binding assay. Less than 0.5% of the input ATF-2 proteins bound to the control GST resin (data not shown).

FIG. 3. In vivo interactions of ATF-2 and CBP.
A, co-immunoprecipitation. Whole-cell lysates were prepared from 293T cells transfected with the two plasmids to express ATF-2 or CBP and immunoprecipitated (IP) with the anti-CBP antibody or the control antibody anti-␤galactosidase (anti-␤-gal). The immune complexes were analyzed by 10% SDS-PAGE, followed by Western blotting using the anti-ATF-2 antibody. In lane 1, an aliquot of the whole-cell lysate was directly used for Western blotting. B, mammalian two-hybrid assay. Structures of the Gal4-CBP and VP16-ATF-2 fusion proteins used for two-hybrid assay are indicated. HepG2 cells were cotransfected with the luciferase reporter plasmid containing the Gal4-binding sites, the expression plasmid for the Gal4-CBP fusion protein or Gal4 alone, the expression plasmid for the VP16-ATF-2 fusion protein or VP16 alone, and the internal control plasmid pRL-TK. The experiments were repeated twice, and the average degree of activation is indicated by a bar graph with S.E., and significant activation is shown by a solid bar. than that of MybDBD-ATF-2 due to a lack of intramolecular association, the trans-activating capacity of this fusion was almost the same as that of MybDBD-ATF-2. This could be due to the lower protein stability of this fusion protein.
UV irradiation is known to lead to activation of SAPK which then phosphorylates ATF-2 at Thr-69, Thr-71, and Ser-90. This phosphorylation enhances the trans-activating capacity of ATF-2, possibly by enhancing the binding to a putative coactivator. By treating the transfected cells with UV, we performed similar co-transfection experiments to those described above (lanes 11-15). CBP and E1A stimulated the trans-acti-vation by MybDBD-ATF-2 5.4-and 9.3-fold, respectively, after treatment of the cells with UV light (cf. Lanes 12-14). These results indicated that SAPK can enhance the ATF-2 activity additionally in the presence of CBP or E1A.
Inhibition of the Intramolecular Interaction of ATF-2 by CBP-The b-ZIP region of ATF-2 interacts with the aminoterminal region intramolecularly (16), and this interaction appears to inhibit the trans-activating capacity of ATF-2. Therefore, we speculated that CBP might block this intramolecular interaction by directly binding to the b-ZIP region of ATF-2. To examine this possibility, we investigated the effect of CBP on the intramolecular interaction between the amino-and carboxyl-terminal regions of ATF-2 (Fig. 5A). The carboxyl-terminal 165-amino acid region of ATF-2 (amino acids 341-505) synthesized in the in vitro translation system bound to the GST fusion To examine the trans-activating capacity of ATF-2, the fusion protein consisting of the DNA-binding domain of c-Myb and the full length of ATF-2 protein was used. Many endogenous CRE-binding proteins affect the assay using the CRE-containing reporter but not using the Myb site containing reporter, because the level of c-Myb in Chinese hamster ovary cells is very low. The structures of the MybDBD-ATF-2 and MybDND-ATF-2-Gal4 fusion proteins used are indicated. The structure of the pMFcolCAT6MBS-I reporter plasmid is also shown. Chinese hamster ovary cells were transfected with a mixture of the CAT reporter plasmid containing Myb-binding sites pMFcolCAT6MBS-I, the MybDBD-ATF-2 or MybDND-ATF-2-Gal4 expression plasmid, the CBP expression plasmid, the 13S E1A expression plasmid, or the control plasmid, and the internal control plasmid pact-␤-gal. CAT assays were performed, and the degree of trans-activation (compared with samples without any effector plasmid) was measured. Experiments were repeated three times, and the average degree of trans-activation is indicated by a bar graph with S.E. In lanes 11-15, the transfected cells were irradiated by UV to induce the phosphorylation of ATF-2 by SAPK. Enhancement of the trans-activating capacity of ATF-2 by CBP or E1A is shown by a solid bar graph.

FIG. 5. Effect of CBP and E1A on the intramolecular interaction of ATF-2.
A, inhibition of the intramolecular interaction of ATF-2 by CBP. GST pull-down assay was performed using the GST-ATF-2 fusion protein, which contains the amino-terminal 253-amino acids region of ATF-2, and the in vitro translated 35 S-ATF-2 mutant, which has the carboxyl-terminal 165-amino acid region containing the b-ZIP domain. To examine the effect of the CBP C/H2 domain on the intramolecular interaction of ATF-2, increasing amounts of in vitro translated CBP C/H2 domain were added. As a control, the in vitro transcription/ translation reaction for CBP C/H2 was performed without T7 RNA polymerase. The relative amounts of carboxyl-terminal 35 S-ATF-2 bound to the GST-ATF-2 resin are plotted below. B, inhibition of the intramolecular interaction of ATF-2 by 13S E1A. Similar assays were performed using the in vitro translated 13S E1A as a competitor. DBD, DNA-binding domain.
protein containing the amino-terminal 253-amino acid region of ATF-2. To examine the effect of CBP C/H2 domain on the intramolecular interaction of ATF-2, increasing amounts of in vitro translated CBP C/H2 domain was added as a competitor. The CBP C/H2 domain inhibited the binding of the carboxylterminal ATF-2 to the GST fusion containing the amino-terminal ATF-2 in a dose-dependent manner. Addition of 15 l of CBP C/H2 inhibited by 66% the interaction between the aminoand carboxyl-terminal regions of ATF-2. As a control, the in vitro transcription/translation reaction for CBP C/H2 protein was performed without T7 RNA polymerase, and lysates were used as a competitor. These lysates did not affect the intramolecular interaction of ATF-2.
We also examined whether the E1A13S product could also inhibit the intramolecular interaction of ATF-2, like in the case of CBP (Fig. 5B). The in vitro translated E1A inhibited the binding of the carboxyl-terminal ATF-2 to the GST fusion containing the amino-terminal ATF-2 in a dose-dependent manner, whereas the control lysates synthesized without T7 RNA polymerase did not. Thus, both CBP and E1A block the intramolecular interaction between the amino-and carboxylproximal regions of ATF-2 by interacting with the b-ZIP region.
Interaction between CBP and Other b-ZIP-containing Proteins-Our results described above indicated that CBP directly binds to the b-ZIP region of ATF-2 via the C/H2 domain of CBP. The b-ZIP domain is a common structure shared by members of the ATF/CREB and Jun/Fos family. Therefore, we examined whether the C/H2 domain of CBP could also interact with the b-ZIP region of other members of the ATF/CREB and Jun/Fos family. The in vitro translated CREB protein bound efficiently to the GST fusion containing the C/H2 domain (46% of the input), whereas in vitro translated c-Jun bound with low efficiency but still significantly (5% of the input) (Fig. 6A). In contrast, c-Fos failed to bind to the C/H2 domain of CBP. The smaller fragment containing the b-ZIP region of CREB or c-Jun efficiently bound to the GST-C/H2 fusion, but the c-Fos b-Zip region did not. These results indicate that the b-ZIP region of some transcription factors binds to the C/H2 domain of CBP (Fig. 6B). To confirm the in vivo interaction between the C/H2 domain and the b-ZIP region of CREB or c-Jun, co-immunoprecipitation was performed (Fig. 6C). The plasmid to express VP16 fused to the b-ZIP region of CREB or c-Jun was cotransfected into 293 cells with the plasmid to express the Gal4 fusion with the C/H2 domain of CBP, and the cell lysates were immunoprecipitated with anti-VP16 antibody. The Gal4-C/H2 fusion was co-immunoprecipitated with VP16-CREB or VP16c-Jun, but not with VP16 alone. To confirm further that CREB binds to CBP via the two domains, b-ZIP and KIX, we examined the effect of PKA treatment on the CREB-CBP interaction. PKA treatment enhanced 3-fold the binding of in vitro translated CREB to the GST-CBP fusion protein containing the KIX domain of CBP, whereas the binding of CREB to the GST fusion containing the C/H2 domain of CBP was not affected by PKA treatment (Fig. 6D, upper panel). In addition, we also examined the effect of potato acid phosphatase, since the in vitro translated CREB would be already phosphorylated by endogenous PKA in reticulocyte lysates. The phosphatase treatment decreased the CREB-KIX interaction 5.2-fold but did not affect the CREB-C/H2 interaction. Thus, CREB binds to the C/H2 and KIX domain of CBP via its b-ZIP and KID domains in a phosphorylation-independent and -dependent manner, respectively. These results suggest that CBP can bind to various transcription factors of the ATF/CREB family through the b-ZIP domain.
CBP Does Not Bind to the SAPK-phosphorylated Activation Domain of ATF-2-CBP binds to the PKA-phosphorylated KID domain of CREB via the KIX domain present in the aminoproximal portion of CBP. In addition, CBP binds to the b-ZIP region of CREB through the C/H2 domain as described above. This raised the possibility that CBP might also bind to the SAPK-phosphorylated transcriptional activation domain in the amino-terminal region of ATF-2. To investigate this possibility, we examined whether SAPK-phosphorylated ATF-2 preferentially binds to CBP (Fig. 7). The HepG2 cells transfected with the ATF-2 expression plasmid were treated by sorbitol which leads to activation of SAPK, and whole cell lysates were prepared. Lysates were incubated with the GST-CBP resins containing the KIX, bromo, or C/H2 domain of CBP, and the bound ATF-2 proteins were detected by anti-ATF-2 antibodies. ATF-2 bound to the C/H2 domain, and the binding efficiency for C/H2 was not affected by phosphorylation by SAPK. In addition, neither the unphosphorylated form nor the phosphorylated forms bound to the bromo or KIX domain. These results suggest that an uncharacterized co-activator other than CBP binds to the SAPK-phosphorylated transcriptional activation domain of ATF-2. DISCUSSION Our results indicate that CBP stimulates transcriptional activation by ATF-2 via binding to the b-ZIP region of ATF-2. CBP abrogates the intramolecular interaction between the amino-and carboxyl-terminal regions of ATF-2 which inhibit the trans-activating capacity of ATF-2. Adenovirus E1A 13S, which is known to interact with the b-ZIP region of ATF-2, also inhibits the intramolecular interaction of ATF-2. In this context, CBP and E1A stimulate ATF-2 activity through a similar mechanism (Fig. 8). This is consistent with our observation that CBP and E1A do not additionally enhance ATF-2 activity (Fig. 4). In the case of CREB, CBP interacts with CREB via two regions; the KIX domain binds to the PKA-phosphorylated KID domain, and the C/H2 domain of CBP interacts with the b-ZIP region of CREB in a phosphorylation-independent manner. Recently, an interaction with CBP through two sites was also reported for NF-B; the two sites of NF-B bind to CBP in a phosphorylation-dependent and -independent manner, respectively (45). However, the transcriptional activation domain in the amino-terminal region of ATF-2 did not directly bind to CBP even after phosphorylation by SAPK. Therefore, an uncharacterized co-activator may bind to this region and mediate transcriptional activation by ATF-2 (Fig. 8). Recently, UTF1 (undifferentiated embryonic cell transcription factor 1), which can binds to the activation domain of ATF-2 and also to the TFIID complex, was identified (46). Although UTF1 is specifically expressed in undifferentiated embryonic cells, the related proteins could be also expressed in many other cell types and may function for trans-activation by ATF-2. Thus, CBP functions as a regulator of ATF-2, like in the case of nuclear hormone receptors. The transcriptional activation domains of nuclear hormone receptors bind to co-activators such as SRC-1, whereas CBP interacts with another region distinct from that of the nuclear hormone receptor molecule (23,24). Since SRC-1 was recently found to directly bind to CBP (23,(47)(48)(49), the uncharacterized co-activator of ATF-2 may also interact with CBP.
Our results indicate that the b-ZIP region of ATF-2, c-Jun, and CREB functions not alone as a DNA-binding domain but also as an interaction domain for CBP. Probably the protein surface of the b-ZIP region which is exposed to the solvent serves for interaction with CBP. This is not surprising as there already exist a number of examples of DNA-binding domains that interact with specific proteins. For instance, E1A interacts with the DNA-binding domain of a number of transcription factors such as the b-ZIP of ATF-2, the metal fingers of Sp1, and the basic helix-loop-helix of upstream stimulatory factor (15). In addition, the DNA-binding domain of c-Myb is known to interact with HSF3 and Cyp40 (43,50). Interestingly, the glu-cocorticoid receptor directly binds to the leucine zipper region of the c-Jun/c-Fos heterodimer and inhibits AP-1 activity (51,52). The direct binding of glucocorticoid receptors to the b-ZIP region of c-Jun may inhibit the interaction between CBP and c-Jun. This mechanism may partly contribute to the inhibition of AP-1 activity by the glucocorticoid receptor.
E1A 13S protein stimulates CRE-dependent transcription through binding to the b-ZIP region of ATF-2. Since both the amino-terminal metal finger and the carboxyl-terminal b-ZIP region of ATF-2 are required for E1A-induced trans-activation (53), E1A cooperatively functions with an unidentified co-activator that binds to the amino-terminal transcriptional activation domain of ATF-2. The results of our co-transfection experiments indicate that E1A and CBP do not additionally stimulate transcriptional activation by ATF-2. This suggests that E1A and CBP cannot bind simultaneously to the b-ZIP region of ATF-2. E1A may drive out CBP from the b-ZIP region of ATF-2 and may replace its function as a regulator.
Recently, Liang and Hai (54) reported an interaction between ATF-4 and CBP via multiple domains. Both the aminoterminal and carboxyl-terminal regions of ATF-4 bind to the four regions of CBP as follows: the KIX domain, C/H3 domain, carboxyl-terminal Q-rich region, and the HAT domain. The carboxyl-terminal region of ATF-4 contains the b-ZIP domain and the HAT domain comprising the C/H2 region. Therefore, the interaction between the b-ZIP region and the C/H2 domain of CBP may also occur in the case of ATF-4. During the preparation of this manuscript, a report by Kawasaki et al. (55) showed that the trans-acting complex on the c-jun promoter contains p300 and ATF-2. The relatively broad region between the amino-terminal transcriptional activation domain and the b-ZIP region of ATF-2 (amino acids 112-350) was demonstrated to interact with the region containing both the bromo and C/H2 domains of CBP. However, our results indicate that the N-truncated mutant lacking the amino-terminal 341 amino acid region binds to CBP and that the two mutants of the b-ZIP region do not bind to CBP. In addition, not only ATF-2 but also the b-ZIP regions of other transcription factors including CREB and c-Jun were confirmed to bind to CBP. Furthermore, our FIG. 7. Effect of SAPK-dependent phosphorylation of ATF-2 on binding to CBP. The HepG2 cells transfected with the ATF-2 expression plasmid were treated with sorbitol (ϩ), which leads to activation of SAPK, or PBS(Ϫ) as a control. Whole-cell lysates were prepared and mixed with the GST-CBP resin containing various portions of CBP as indicated above each lane. The bound proteins were eluted, separated on 10% SDS-PAGE, and analyzed by Western blotting using the anti-ATF-2 antibody (upper panel). In the lower panel, whole cell lysates were directly analyzed by Western blotting using the anti-ATF-2 antibody.
FIG. 8. Schematic model for the role of CBP. In the absence of CBP or 13S E1A, the amino-terminal region of ATF-2 containing the activation domain intramolecularly interacts with the carboxyl-terminal b-ZIP region. In this state, the amino-terminal activation domain cannot interact with co-activator molecules. CBP or 13S E1A directly bind to the b-ZIP region of ATF-2 and inhibit the intramolecular interaction. When ATF-2 is bound to CBP or 13S E1A, the amino-terminal activation domain is exposed and able to interact with a co-activator. data indicate that the C/H2 domain of CBP is sufficient for interaction with ATF-2, and that the bromo domain is not required.