cAMP Response Element-binding Protein Interacts with the Homeodomain Protein Cdx2 and Enhances Transcriptional Activity*

Cdx2 encodes for a homeodomain protein that is expressed in intestinal epithelial cells. The Cdx2 protein triggers intestinal differentiation in cell lines and is necessary for maintenance of the intestinal phenotype in mice. CBP (cAMP response element-bindingprotein) is a transcriptional co-activator that interacts with many transcription factors and components of the basal transcriptional machinery. In this study, we demonstrate that CBP is markedly induced upon differentiation of the Caco-2 intestinal cell line and augments Cdx2-dependent transcriptional activity. Cdx2 interacts with the amino-terminal domain of CBP, and the two proteins coexist in vivo within the same nuclear protein complex. Moreover, expression of the CBP domain that interacts with Cdx2 acts as a dominant-negative inhibitor of transcriptional activation by Cdx2. These findings demonstrate a direct interaction between an intestinal homeodomain protein and CBP and suggest that CBP participates in the network of transcriptional proteins that direct intestinal differentiation.


Cdx2 encodes for a homeodomain protein that is expressed in intestinal epithelial cells. The Cdx2 protein triggers intestinal differentiation in cell lines and is necessary for maintenance of the intestinal phenotype in mice. CBP (CREB-binding protein) is a transcriptional co-activator that interacts with many transcription factors and components of the basal transcriptional machinery. In this study, we demonstrate that CBP is markedly induced upon differentiation of the Caco-2 intestinal cell line and augments Cdx2-dependent transcriptional activity. Cdx2 interacts with the amino-terminal domain of CBP, and the two proteins coexist in vivo within the same nuclear protein complex. Moreover, expression of the CBP domain that interacts with Cdx2 acts as a dominant-negative inhibitor of transcriptional activation by Cdx2. These findings demonstrate a direct interaction between an intestinal homeodomain protein and CBP and suggest that CBP participates in the network of transcriptional proteins that direct intestinal differentiation.
In mammals, visceral endoderm gives rise to the gastrointestinal epithelium. Several lines of evidence suggest that the caudal-type homeobox gene Cdx2 is involved in the regulation of intestinal epithelial cell development and maintenance of the differentiated phenotype. Cdx2 encodes a transcription factor that is expressed in differentiated enterocytes (1) and binds to cis-elements present in the gene promoters of enterocytespecific genes including sucrase-isomaltase (SI) 1 (2) and intestinal phospholipase A/lysophospholipase (3). Expression of Cdx2 in intestinal cell lines and in an epithelial-mesenchymal co-culture system has shown that it is involved in the control of intestinal cell proliferation and differentiation (4 -6). Finally, colon tumors develop upon loss of Cdx2 expression in colonocytes of mice that are heterozygous for the null allele of Cdx2 (7).
The molecular mechanisms of how Cdx2 regulates transcriptional initiation of enterocyte genes have not been well defined.
We hypothesized that co-factors that link DNA-binding proteins to the basal transcriptional apparatus may be involved in regulation of Cdx2-dependent transcription. CBP (CREB-binding protein) was first characterized as a co-factor for the CREBbinding protein that potentiates transcriptional activity (8). Since then, CBP, and its family member p300, have been shown to bind to interact with many transcription factors in a number of different families in addition to CREB, including helix-loop-helix proteins and nuclear receptors (9 -11). CBP functions as an adaptor protein for complex transcriptional regulatory elements by enhancing the interaction between transcription factors and components of the basal transcriptional machinery. Additionally, inactivation of CBP is critical for the adenoviral protein E1A to induce oncogenic transformation and to inhibit differentiation suggesting that CBP plays a role in cell growth and differentiation (12,13). Alterations of the human CBP gene have been implicated in hematological malignancies as well as in congenital malformation and mental retardation (14). Recently, it has been shown that inhibition of CBP in Caenorhabditis elegans leads to developmental arrest, with most embryos showing a lack of endodermal cells (15).
In this study, we show that CBP interacts with Cdx2 and is functionally important for Cdx2-dependent transactivation. These findings demonstrate that CBP interacts with intestinal homeodomain transcription factors and suggest that CBP may modulate the function of Cdx2 in the regulation of enterocyte proliferation and differentiation.
The DNA sequence encoding for the amino acids 181 to 269 of Cdx2 was generated by polymerase chain reaction. The oligonucleotides were designed in manner to introduce a SalI and a XbaI at the ends of the polymerase chain reaction products. The polymerase chain reaction product was cloned into SalI/XbaI restriction sites of pSG424 vector encoding for the Gal4 DNA binding domain (16). The resulting HindIII/ XbaI fragment was subcloned into pRc/CMV generating the pRc/CMV-Gal4-Cdx2(181-269) plasmid. The pRc/CMV-Gal4 plasmid was obtained by subcloning the HindIII/XbaI fragment of pSG424 into pRc/ CMV HindIII/XbaI restriction sites.
Cell Culture and Transfection-Caco-2 and NIH-3T3 cells were maintained in Dulbecco's modified Eagle's medium with high glucose, 10% fetal calf serum, and 1% penicillin/streptomycin. The cells were placed in 6-well plates 24 h before transfection with Lipofectin (Life Technologies, Inc.) following the manufacturer's recommendations. Luciferase activity was determined 48 h after the transfection using the * This work was supported by National Institutes of Health Grant RO1-DK46704 and the Molecular Biology Core of the Center for Molecular Studies in Digestive Diseases at the University of Pennsylvania (P30-DK50306) (to P. G. T.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by a fellowship from the Fonds de la recherche en Sauté du Québec.
In Vitro GST-Protein Interaction Assay-GST-protein interaction assays were performed as described elsewhere (19) with slight modifications. Briefly, bacterial expression of GST fusion proteins was induced in medium containing 0.1 M isopropyl-1-thio-␤-D-galactopyranoside for 3 h. The bacterial pellet was resuspended in 500 l of PBS and sonicated to disrupt the bacteria. The bacterial remnants were removed by centrifugation, and the supernatant was incubated with 75 l of GST beads (Amersham Pharmacia Biotech) on a rotating wheel for 30 min at 4°C. The beads were washed 4 times with PBS and once with HND buffer (10 mg/ml bovine serum albumin, 20 mM Hepes, 50 mM NaCl, 0.1% Nonidet P-40, 5 mM dithiothreitol).
Proteins were labeled with [ 35 S]methionine by coupled in vitro transcription and translation using the TNT reticulocyte lysate system (Promega) according to the manufacturer's instructions. The GST beads with bound fusion proteins were incubated with labeled proteins in 200 l of HND buffer at 4°C. The beads were then washed with MTPBS (PBS, 0.1% Nonidet P-40) buffer and proteins eluted from the beads in SDS sample buffer. The eluted proteins were separated on a 10% acrylamide SDS-polyacrylamide gel electrophoresis gel.
Co-immunoprecipitation and Immunoblots-Co-immunoprecipitations were performed using an immunoprecipitation kit (Protein G) (Boehringer-Mannheim) following the manufacturer's recommendations. Immunoprecipitation of Flag-tagged proteins were performed using 3 g of anti-Flag M2 antibodies (Kodak). The immunoprecipitates were solubilized in SDS sample buffer and loaded on a 4% acrylamide SDS-polyacrylamide gel electrophoresis gel for detection of CBP and a 12% gel for Cdx2 and Flag-tagged proteins. Proteins were transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA). The membrane was blocked for 2 h in PBS (150 mM NaCl, 7 mM Na 2 HPO 4 , 2 mM KCl, 800 M KH 2 PO 4 , pH 7.4) containing 0.1% Tween 20 and 10% skim milk. Cdx2 was detected by using rabbit anti-Cdx2 polyclonal antibody. 2 CBP was detected using rabbit polyclonal A-22 antibody (Santa Cruz Biotechnology, Santa Cruz, CA). E-Cadherin was detected using mouse monoclonal 36 antibody (Transduction Laboratories, Lexington, KY). Proteins were detected on the membranes using ECL-plus (Amersham Pharmacia Biotech) in combination with goat anti-rabbit antibodies for CBP and Cdx2 (Amersham Pharmacia Biotech) and in combination with anti-mouse antibodies for E-Cadherin (Santa Cruz Biotechnology).

RESULTS AND DISCUSSION
Cdx2 is an important transcription factor for activating the SI promoter in intestinal cell lines (2). Caco-2 cells, derived from a human colonic adenocarcinoma, spontaneously differentiate following confluence in culture, including a marked increase in the expression of SI mRNA and protein (6,20,21). Therefore, we examined the expression of Cdx2 protein during Caco-2 cell differentiation to determine whether an increase in this transcription factor may be involved in SI induction. The amount of Cdx2 protein was not different in pre-and postconfluent cells in comparison to E-Cadherin, a protein that is unchanged during Caco-2 differentiation (22) (Fig. 1). Because CBP is able to act as a co-activator for many transcription factors, we also examined its expression. CBP was not detectable in preconfluent Caco-2 cells and only detectable at confluency after long exposure of the immunoblot (Fig. 1). However, following confluency there was marked induction of CBP protein (Fig. 1). This observation suggested that CBP may be an important co-factor for mediating the function of Cdx2 in the activation of SI gene transcription.
Based on these findings in Caco-2 cells, we explored the functional interaction of Cdx2 and CBP. The adenovirus E1A oncoprotein inhibits the ability of CBP to co-activate transcription in a variety of cell lines (23,24). We examined the functional consequence of expression of E1A on transactivation of the SI promoter by Cdx2 in Caco-2 cells, which have been shown to express endogenous Cdx2 and support SI promoter transcription via interaction of Cdx2 with the SIF1 promoter element (2). Expression of wild-type E1A resulted in marked reduction of SI promoter activity in Caco-2 cells and abolished superactivation of the SI promoter when Cdx2 was over expressed ( Fig. 2A). Moreover, an E1A mutant that lacks the ability to interact with retinoblastonia protein but does bind to CBP (E1ACXdl) also inhibited Cdx2 activation of the promoter. In contrast, expression of a mutant form of E1A (E1A(⌬2-36)) that is unable to bind CBP, had no effect on SI promoter activation by Cdx2 (Fig. 2A). We next examined the ability of E1A to affect Cdx2-dependent activation of the SI promoter in 2 D. Silberg and P. G. Traber, unpublished results. the nonintestinal fibroblast cell line NIH-3T3. Similar to results in Caco-2 cells, E1A and the E1ACXdl mutant abolished Cdx2 activation of the SI promoter, whereas E1A(⌬2-36) had no effect (Fig. 2B).
These results suggested that CBP was involved in the control of SI promoter activation. However, the region from nucleotides Ϫ183 to ϩ54 that constitutes the SI promoter contains a number of DNA regulatory binding sites (25), each of which bind proteins that might be affected by E1A-CBP interactions. Because the presence of the Cdx binding site is absolutely required for promoter activity (2), experiments with mutations in this site could not alone answer the question of whether other SI promoter elements are involved in mediating the effect of E1A on Cdx2-dependent activation. Thus, to determine whether transactivation by Cdx2 is specifically inhibited by E1A, the same experiments were performed in Caco-2 cells using a reporter construct that linked Cdx binding sites upstream of a minimal thymidine kinase promoter (pTK-SIF1(ϩ4)-luc) (18). As with the SI promoter, Cdx2-induced transactivation of this construct was abolished by wild-type E1A, but not by the E1A mutant that is unable to bind to CBP (Fig. 2C). These modifications in luciferase activity were directly related to the presence of the SIF1 elements, because no luciferase activity was unchanged when transfections were done using pTK-luc (Fig. 2C). These results suggest that CBP may be directly involved in the ability of Cdx2 to activate gene transcription.
The ability of CBP to directly augment transactivation by Cdx2 was examined by co-transfection of expression vectors for Cdx2 and CBP with the Cdx2-responsive reporter plasmid pTK-SIF1(ϩ4) in Caco-2 (Fig. 3A). CBP overexpression in Caco-2 cells resulted in a dose-dependent increase in Cdx2-dependent transactivation with a maximum of a 3-fold increase over that with Cdx2 alone (Fig. 3A). The induction of transcription by CBP was greater in the presence of higher levels of Cdx2, suggesting that Cdx2 may be limiting in Caco-2 cells. Addition of the wild-type E1A expression vector resulted in loss of reporter gene expression whereas E1A(⌬2-36) had no effect on the ability of CBP to augment Cdx2 transactivation (Fig.  3A). Transient overexpression of CBP or Cdx2 in Caco-2 cells did not change the endogenous expression of Cdx2 and CBP, respectively (Fig. 3B). Furthermore, transient overexpression of E1A or E1A(⌬2-36) in Caco-2 cells did not interfere with Cdx2 and CBP expression (Fig. 3B). These results demonstrate that in Caco-2 cells, CBP is able to increase Cdx2 transactivation activity when SIF1 elements are located immediately upstream of a heterologous promoter.
In previous studies, we have shown that Cdx2 is able to activate transcription when its DNA binding elements are placed in an enhancer position on heterologous promoters (18). These studies further showed that the cellular context is important for Cdx2 to activate transcription from an enhancer context, because Caco-2 cells supported Cdx2-dependent enhancer activation, whereas NIH-3T3 cells did not. These results led to the conclusion that there may be cell-specific adaptor proteins that are important for Cdx2 function on certain transcriptional elements (18). We found that CBP was not able to augment Cdx2 transactivation when Cdx binding elements were placed in an enhancer context (data not shown). Thus, consistent with the widespread expression of CBP, it does not appear to be responsible for the previously observed cell-specific function of Cdx2 on enhancer elements.
The mechanism by which CBP augments Cdx2 transcriptional activity could be via direct protein-protein interaction, by common interaction with a third protein, or by indirect effects on other components of the transcriptional machinery. To determine whether there is direct interaction between Cdx2 and CBP, in vitro protein interaction assays were performed. The bacterially expressed fusion protein GST-CBP(1-450) was able to interact with in vitro labeled Cdx2 (Fig. 4A). In contrast, there was very weak interaction with GST-CBP(451-682) and no interaction with GST-CBP(1000 -1500). These data raised the question of whether the homeodomain of Cdx2 interacted with GST-CBP(1-450), similar to a recently described POUhomeodomain protein Pit-1 (26). A peptide containing the homeodomain of Cdx2 and excluding the majority of the aminoterminal and carboxyl-terminal domains was found to bind specifically to GST-CBP(1-450) (Fig. 4B).
We next examined whether CBP and Cdx2 interact within the cell. Co-immunoprecipitation was performed to determine whether Cdx2 and CBP coexist within the same protein complex. Protein extracts of Caco-2 cells transfected with expression vectors for Flag-CBP and/or Flag-Cdx2 were subjected to immunoprecipitation followed by immunoblot analysis. Results of these experiments showed that Cdx2 protein was co-immunoprecipitated with Flag-CBP and that endogenous CBP was immunoprecipitated with Flag-Cdx2 (Fig. 4C). Nonimmune antibodies were unable to immunoprecipitate either Flag-CBP or Flag-Cdx2 and anti-Flag antibodies were unable to immunoprecipitate CBP or Cdx2 (Fig. 4C). Thus, CBP and Cdx2 are associated in an immunoprecipitable protein complex in Caco-2 cells.
The interaction of Cdx2 with a specific domain of CBP provided an avenue to test the functional specificity of the interaction. It was previously shown that short regions of the CBP protein have a dominant-negative effect on the functional interaction between CBP and nuclear hormone receptors (17). Therefore, we expressed HA-tagged domains of CBP with Cdx2 and CBP in Caco-2 cells to assess the ability of different domains to interfere with transcriptional activation. Expression of HA-CBP(1-450) inhibited CBP-dependent augmentation of Cdx2-induced transcription in a dose-dependent manner, whereas HA-CBP(451-682) had little effect (Fig. 5). This evidence substantiates the hypothesis that Cdx2 interacts with the amino-terminal domain of CBP in the cell nucleus.
Taken together, these results demonstrate that the aminoterminal domain of CBP is able to interact with Cdx2 and that this interaction results in increased Cdx2-dependent transactivation. Recently, it has been shown that CBP interacts with the POU-homeodomain protein Pit-1 and that this interaction is functionally important for transcriptional activation (26). Our findings represent the second description of an interaction between homeodomain proteins and CBP and extends the regulatory implications beyond the subfamily of POU-homeodomain proteins. Cdx2 has been shown to be a critical component of the transcriptional machinery of intestinal epithelial cells and, as a result, plays a critical role in intestinal development and differentiation. Based on these studies, we hypothesize that CBP may modulate the function of Cdx2 on complex promoters and thus may influence intestinal differentiation and development.