Cdk2-dependent Phosphorylation of Homeobox Transcription Factor CDX2 Regulates Its Nuclear Translocation and Proteasome-mediated Degradation in Human Intestinal Epithelial Cells*

By having demonstrated previously that p27 Kip1 , a potent inhibitor of G 1 cyclin-cyclin-dependent kinases complexes, increases markedly during intestinal epithelial cell differentiation, we examined the effect of p27 Kip1 on the activity of the transcription factor CDX2. The present results revealed the following. 1) p27 Kip1 with the CDX2 transcription factor. 2) In contrast to CDX2 mRNA levels, CDX2 protein expression levels significantly increased as soon as Caco-2/15 cells reached confluence, slowed their proliferation, and began their differentiation. The mechanism of CDX2 regulation is primarily related to protein stability, because inhibition of proteasome activity increased CDX2 levels. CDX2 in proliferative epithelial cells. Cdk2 CDX2 and phosphorylated CDX2, glutathione S -transferase and immunoprecipitation with proliferating Caco-2/15 Caco-2/15 to glutathione-Sepharose. beads was washed resuspended in sample buffer, and was transferred onto nitrocellulose after Immunological detections were performed with antibodies recognizing Cdk2. The other half of the beads was washed four times with lysis buffer followed by three times with ice-cold kinase buffer prior to performing the kinase assay. Similar results were obtained in three different experiments. B, HEK293 cells were transfected with CDX3 or CDX3(RNL/ANA) expression vectors. Thirty six hours after transfection, cells were treated with cycloheximide (25 (cid:2) g/ml) for 8 h followed by cell lysis at the end of the treatment. CDX3 and CDX3(RNL/ANA) protein levels were determined by Western blotting. Results are repre- sentative of three independent experiments. cells were transfected with 0.1 (cid:2) g of sucrase-isomaltase ( SI )-lucif- erase reporter vector and 0.025 (cid:2) g CDX3 or CDX3(RNL/ANA) expression vectors or the pBAT vector (EV). Two days after transfec- tion, cells were lysed, and luciferase activity was measured. The increase in luciferase activity was calculated empty level at Results

Although cell differentiation and growth arrest are two distinct processes occurring independently, they are generally intimately linked because cell cycle exit is a prerequisite for terminal differentiation. Mechanisms operating at the interface between growth arrest and differentiation programs play critical roles in establishing the differentiated states of various cell types. Such mechanisms are also likely to be important in maintaining the irreversibility of terminal differentiation, with their disruption having an important role in cell immortalization and tumorigenesis (1). The intestinal epithelium represents an attractive system in which to study the relationship between cell cycle and cytodifferentiation. Proliferating, differentiating, and functional cells are organized into well defined compartments in this polarized tissue, from which the entire sequence of developmental events is on display at any given moment in time (2,3). The crypt compartment houses the proliferative cells that give birth to the four types of differentiated enterocytes (absorptive enterocytes, goblet, enteroendocrine, and Paneth cells) that, with the exception of Paneth cells, populate the villus compartment. The decision to differentiate is abruptly taken by committed crypt cells while in their most rapid state of proliferation (4). These newly differentiated cells acquire their distinctive ultrastructural features and cell surface markers upon leaving the proliferative cell cycle, at the top of the crypts or the base of the villus (3,(5)(6)(7). A better understanding of the interaction between G 1 phase regulators and transcription activators of intestinal differentiation is required to delineate the molecular mechanisms that coordinate cell growth arrest and differentiation in the intestinal epithelium.
Several cell cycle-associated proteins have been directly implicated in the growth arrest that precedes terminal differentiation of intestinal epithelial cells, including the retinoblastoma proteins pRb and p130, p21 Cip/Waf , p27 Kip1 , and p57 Kip2 (8 -11). Although these proteins are clearly important for cell cycle arrest, little is currently known about their potential functions during cell differentiation. By exploiting the use of a p27 Kip1 antisense cDNA to specifically deplete the intestinal epithelial cell line Caco-2/15 of this particular inhibitor, we previously demonstrated that p27 Kip1 is required for complete intestinal epithelial differentiation. Indeed, stable expression of a p27 Kip1 antisense cDNA is sufficient to markedly block expression of several differentiation markers, namely sucraseisomaltase, lactase-phlorizin hydrolase, and alkaline phosphatase (11). Accordingly, Tian and Quaroni (9) have reported that forced overexpression of p27 Kip1 in the human intestinal epithelial crypt cell line HIEC, by infection with recombinant Adp27 adenovirus, resulted in induction of cell differentiation 6 days after infection. These authors suggested that p27 Kip1 is more directly related to expression of differentiated traits in intestinal epithelial cells than p21 Cip or other cell cycle inhib-itors. However, the mechanisms by which p27 Kip1 promotes intestinal epithelial cell differentiation remain to be elucidated. The demonstration that p27 Kip1 is involved in intestinal epithelial cell differentiation also has potential important implications for development of cancer in this tissue. Indeed, several studies in humans have demonstrated abnormally low levels of p27 Kip1 protein in carcinomas, correlating well with both histological aggressiveness and patient mortality (12).
Several lines of evidence demonstrate that CDX2, a homeodomain protein related to the Drosophila caudal gene, is important both for pattern formation in the developing embryo and activation of intestine-specific genes, including sucraseisomaltase and lactase-phlorizin hydrolase, among others (13)(14)(15). Additionally, CDX2 expression can negatively regulate the proliferation of intestinal epithelial and colon cancer cells in culture while promoting the acquisition of a mature enterocyte phenotype consisting of a polarized, columnar shape with apical microvilli and tight junctions (14,16). Supporting in vitro observations, CDX2ϩ/Ϫ heterozygous mice develop hamartomas or polyps containing heterotopic intestinal tissue in the colon (17). CDX2 expression drops consistently in colon cancers in relation with the severity of dysplasia (18 -20).
Because CDX2 and p27 Kip1 are both required for intestinal epithelial cell differentiation and could play an important role in colorectal cancer, we analyzed their potential molecular relationship in human intestinal epithelial cells. The present results demonstrate that CDX2 is associated with p27 Kip1 and the cyclin-dependent kinase, Cdk2, and is a substrate for Cdk2. In addition, results indicate that phosphorylation of CDX2 by Cdk2-associated complexes targets CDX2 for rapid degradation by the ubiquitin/proteasome pathway. The targeted degradation of CDX2 following its phosphorylation by Cdk2 identifies a new mechanism through which CDX2 activity can be regulated in coordination with cell cycle machinery.

Materials
[␥-32 P]ATP and [␥-32 P]orthophosphate were obtained from Amersham Biosciences. Monoclonal antibody HSI-14 against sucrase-isomaltase was kindly provided by Dr. A. Quaroni (Cornell University, Ithaca, NY). Rabbit polyclonal antibody against GST 1 (B14), Cdk2 (M2), CRM1 (H-300), and p27 Kip1 (C- 19) were from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse monoclonal antibody against CDX2 was from Biogenex, San Ramon, CA. The monoclonal HA (F-7) antibody raised against a peptide from hemagglutinin HA1 protein was purchased from Santa Cruz Biotechnology. Antibody-recognizing actin was purchased from Roche Applied Science. Recombinant p27 Kip1 protein was from Santa Cruz Biotechnology. Recombinant active cyclin E-Cdk2 complex was from Upstate Biotechnology, Inc. (Lake Placid, NY). The specific inhibitor of protein synthesis (cycloheximide) was purchased from Calbiochem. All other materials were obtained from Sigma unless stated otherwise.

Cell Culture
The Caco-2/15 cell line was obtained from A. Quaroni (Cornell University, Ithaca, NY) and was cultured as described previously (11). The human embryonic kidney 293 cells (ATCC; American Type Culture Collection, Manassas, VA) were cultured in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% FBS. The human colon cancer cell line DLD-1 (ATCC) was cultured in RPMI medium (Invitrogen) containing 10% FBS.

Expression Vectors and Constructs
The full-length mouse p27 Kip1 (␥EXlog(ϩ), Novagen), provided by Dr. J. Massague (Memorial Sloan-Kettering Cancer Center, New York), was subcloned into the expression vector pECE in-frame with the hemagglutinin epitope (HA) (Dr. J. Pouysségur, University of Nice, Nice, France) (21). The GST-p27 Kip1 construct was generated using a p27 Kip1 /pECE as the template (21), and the sense and antisense oligonucleotides 5Ј-CCG CGT GGA TTC TCA AAC GTG AGA GTG AGA  GTG TCT-3Ј and 5Ј-CCG GAA TTC TTA CGT CTG GCG TCG AAG  GCC-3Ј, respectively. The resulting DNA fragment was subcloned into the BamHI-EcoRI sites of the pGEX4T-2 vector. The p57 Kip2 expression vector was a kind gift from Dr. Stephen Elledge (Department of Biochemistry, Howard Hughes Medical Institute, Houston, TX). Expression vectors for HA-tagged wild-type Cdk2 (HA-Cdk2) and a dominantnegative form of Cdk2 (HA-Cdk2DN) were obtained from Dr. James M. Roberts (Division of Basic Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA). The hamster CDX3 expression vector was kindly provided by Dr. M. S. German (University of California, San Francisco) (22). The full-length CDX3 cDNA was subcloned into the expression vector pBAT in-frame with the HA epitope. The HA-CDX3 was constructed by adding the HA epitope by PCR using an oligonucleotide containing the sequence encoding the HA epitope: primer sense, 5Ј TCA CTA GCC TAG GAC ACC ATG TAT GAT GTT CCT GAT TAT GCT AGC CTC CCG ATG TAC GTG AGC TAC CTC CTA 3Ј; primer antisense, 5Ј CCG GAA TTC TCT GCC GCC GCT 3Ј. The PCR product was inserted into the pBAT vector (Invitrogen) into the AvrII-EcoRI sites. Double R162A/L164A mutations were introduced into the pGEX4T2-CDX3 vectors by using a QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The pCDNA3 vector encoding HA6-tagged ubiquitin was kindly provided by S. Meloche (Institut de Recherche en Immunologie et Cancérologie, Université de Montréal, Quebec, Canada) and described elsewhere (23). The sucrase-isomaltase reporter construct used for luciferase assays contained the human sucrase-isomaltase promoter from residues Ϫ183 to ϩ54 cloned upstream of the luciferase gene of the pGL2 reporter construct (25). Authenticity of all DNA fragments generated by PCR was confirmed by DNA sequence analysis.

Immunoprecipitation-Electrophoretic Mobility Shift Assays in HEK293 and Caco-2/15 Cells
HEK293 cells were seeded in 100-mm dishes and transfected by calcium phosphate precipitation with 1 g of p27 Kip1 or 1 g of p57 Kip2 or the relevant expression vectors and 1 g of CDX3 or HA-CDX3 expression vectors per dish. Two days after transfection, immunoprecipitation-electrophoretic mobility shift assays (IP-EMSA) were performed as described previously (24).

GST Fusion Protein Purification
Hamster CDX3 wild type and mutated forms of CDX3 were ligated downstream of the glutathione S-transferase sequence in a pGEX plasmid (Amersham Biosciences). The recombinant plasmid was introduced into Escherichia coli BL21 DE3, and the fusion protein was produced by growing 500 ml of bacterial culture to an optical density between 0.9 and 1.1 and treating the culture with 0.5 mM isopropyl-1-thio-␤-Dgalactopyranoside for 3 h. Cells were recovered and resuspended in SB buffer (16 mM sodium phosphate, pH 7.4, 150 mM NaCl, 15% glycerol, 0.02% Triton X-100, 1 mM dithiothreitol, 10 g/ml leupeptin, 1 g/ml pepstatin A, 10 g/ml aprotinin) and sonicated. Triton X-100 was added to the lysates to a final concentration of 1%. The bacterial lysates were incubated on ice for 15 min and centrifuged at 12,000 rpm for 15 min. The supernatants were recovered and mixed with 55 l of a 1:1 suspension of glutathione-Sepharose 4B beads (Amersham Biosciences), and the resulting mixture was rotated at 4°C for 1 h. The beads were then washed extensively in SB buffer and used for in vitro binding assays as described previously (26).

GST Pull-down Assays with Caco-2/15 Cell Lysates
Cells were grown in 60-mm dishes in Dulbecco's modified Eagle's medium supplemented with 10% FBS. The cells were lysed in 1 ml of lysis buffer (150 mM NaCl, 1 mM EDTA, 40 mM Tris, pH 7.6, 1% Triton X-100) supplemented with protease inhibitors (0.1 mM PMSF, 10 g/ml leupeptin, 1 g/ml pepstatin A, 10 g/ml aprotinin) and phosphatase inhibitors (0.1 mM orthovanadate, 20 mM para-nitrophenyl phosphate, 40 mM ␤-glycerophosphate). Lysates were centrifuged at 13,000 ϫ g for 10 min. 7.5 g of GST-CDX3 fusion or GST proteins were coupled to 50 l of glutathione-Sepharose. The Caco-2/15 lysates were incubated for 2 h at 4°C with the immobilized fusion proteins by end-over-end rotation. The beads were washed four times with the lysis buffer after which Laemmli buffer was added and the mixture boiled for 5 min. Bound proteins were visualized by SDS-PAGE (12% acrylamide gels) and immunoblotting as described below. In other experiments, the beads were washed four times with lysis buffer followed by washing three times with ice-cold kinase buffer (20 mM MOPS, pH 7.0, 30 mM MgCl 2 , 1 mM dithiothreitol) before performing the kinase assay in the presence or absence of (R)-roscovitine.
GST Pull-down Assays with p27 Kip1 and Cyclin E/Cdk2 Complex 7.5 g of GST-CDX3 fusion or GST proteins were coupled to 50 l of glutathione-Sepharose. In some cases, p27 Kip1 (150 ng) was first incubated for 1 h at 4°C with the immobilized fusion proteins by end-overend rotation. The beads were washed five times with the lysis buffer after which the cyclin E-Cdk2 complex (150 ng) was incubated for 1 h at 4°C with the immobilized fusion proteins. Again the beads were washed five times with lysis buffer. In other cases, the cyclin E-Cdk2 complex was first incubated with the immobilized fusion proteins, and p27 Kip1 was added thereafter. Laemmli buffer was added to the beads and the ensuing mixture boiled for 5 min. Bound proteins were visualized by SDS-PAGE (12% acrylamide gels) and immunoblotting as described below.

Preparation of Polyclonal Antiserum against CDX2/3 (CDX2/3-NR)
Studies were conducted in agreement with the principles and procedures outlined in the Canadian Guidelines for Care and Use of Experimental Animals. Antibodies were raised in rabbits (New Zealand, female) by multisite injections of CDX3-GST protein, as described previously (27). The antibodies were purified by affinity chromatography using CDX3-GST immobilized to an agarose gel. The column coupling CDX3-GST was performed using the AminoLink Plus kit (Pierce) according to the manufacturer's instructions. Antibodies were tested on intestinal mucosae, Caco-2/15 cells, and HEK293 cells transfected with the pcDNAneo3/CDX3 construct for their specificity in Western blot and immunofluorescence by comparison with nonimmune sera and mouse monoclonal antibody against CDX2 provided by Biogenex (CDX2-Biogenex). Results obtained with CDX2/3-NR antibody were quite similar to those obtained with CDX2-Biogenex antibody. The best working concentration for the CDX2/ 3-NR antibody was established at 1:20,000 for immunoblot experiments and at 1:1000 for immunofluorescence experiments.

Protein Expression and Immunoblotting
Cells were lysed in SDS sample buffer (62.5 mM Tris-HCl, pH 6.8, 2.3% SDS, 10% glycerol, 5% ␤-mercaptoethanol, 0.005% bromphenol blue, and 1 mM PMSF). Proteins (20 -40 g for large gels or 5-10 g for mini-gels) from whole-cell lysates were separated by SDS-PAGE in 12.5% acrylamide gels. Proteins were detected immunologically after transfer onto nitrocellulose membranes (Amersham Biosciences). Protein and molecular weight markers (Bio-Rad) were localized by staining with Ponceau Red. Membranes were blocked for 1 h at 25°C in PBScontaining powdered milk followed by overnight incubation with the first antibody in blocking solution and then washed and incubated with horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit (1: 1000) IgG for 1 h, after determination of the best concentrations for each antibody. Blots were visualized by the ECL system (Amersham Biosciences). Protein concentrations were measured using a modified Lowry procedure with bovine serum albumin as standard (28).

Immunofluorescence Microscopy of Cultured Cells
Caco-2/15 cells grown on sterile glass coverslips were washed twice with ice-cold PBS. Cultures were then fixed with 30% methanol and 70% acetone for 15 min at Ϫ20°C, permeabilized with a 0.1% Triton X-100 solution in PBS for 10 min, and blocked with PBS and 2% bovine serum albumin for 20 min at room temperature. Cells were subsequently immunostained for 1 h with the primary antibody (CDX2, p27 Kip1 , or CRM1) and for 30 min with the secondary antibody (fluorescein isothiocyanate-labeled goat anti-rabbit). In some experiments, nuclei were stained with 4,6-diamidino-2-phenylindole. Negative controls (no primary antibody) were included in all experiments.

GST-CDX3 Phosphorylation Assays
Kinase assays were performed using the Cdk2 kinase buffer supplied by the manufacturer. Briefly, Cdk2 buffer was supplemented with 500 M ATP, 1 Ci of [␥-32 P]ATP, 5 g of GST-CDX3, and 1-40 ng of activated cyclin E-Cdk2 complexes and incubated at 30°C for 30 min. Reactions were stopped by addition of Laemmli buffer. Radiolabeled GST-CDX3 was separated by SDS-PAGE and processed for autoradiography.

In Vivo Phospholabeling
Cdk2, HA-CDX3, p27 Kip1 , or their relevant expression vectors were transfected in HEK293 cells. Twenty four hours after transfection, cells were labeled for 2 h in phosphate-depleted medium with 100 Ci/ml of [ 32 P]orthophosphate (Amersham Biosciences). After labeling, the cells were lysed in lysis buffer, and 32 P-phosphate-labeled CDX3 was isolated by immunoprecipitation with HA antibody. Incorporation of [ 32 P]orthophosphate was detected by autoradiography.

Northern Blot
Total cellular RNAs were prepared from Caco-2/15 cells at subconfluence (day Ϫ2), confluence (day 0), and 6 and 15 days post-confluence by the guanidine isothiocyanate/phenol method (TRIzol, Invitrogen), as described previously (29). RNAs were subjected to agarose gel electrophoresis with formaldehyde and transferred onto nylon membranes (Nytran, Schleicher & Schuell). Equal RNA loading was confirmed by hybridization to a ribosomal L32 gene probe. Hybridizations were performed with a random-primed 32 P-labeled probe (Amersham Biosciences) of a PCR-amplified human CDX2 fragment from nucleotides 1102 to 1706.

Animals and Human Specimens
CD-1 mice (10 -12 weeks) were purchased from The Jackson Laboratory (Bar Harbor, ME), fed Purina chow ad libitum, and kept in a controlled temperature and light cycle environment (20°C; 12 h light and 12 h darkness). All studies were conducted in agreement with the principles and procedures outlined in the Canadian Guidelines for Care and Use of Experimental Animals. Tissues from human fetuses varying in age from 18 to 20 weeks of gestation (post-fertilization fetal ages were estimated according to Streeter (30)) were obtained from normal elective pregnancy terminations. No tissue was collected from cases associated with a known fetal abnormality or fetal death. Studies were approved by the Institutional Human Subject Review Board.

Indirect Immunofluorescence in Human and Mouse Intestine
Segments of human and mouse jejunum and colon were rinsed with 0.15 M NaCl, cut into small fragments, embedded in optimum cutting temperature compound (Canemco Supplies, St-Laurent, Quebec, Canada), and quickly frozen in liquid nitrogen. Frozen sections 2-3-m thick were spread on silane-coated glass slides and air-dried for 1 h at room temperature before storage at Ϫ80°C (11). For indirect immunofluorescence, sections were fixed with 2% formaldehyde in PBS (pH 7.4; 45 min, 4°C) prior to detection of 4,6-diamidino-2-phenylindole (Vector Laboratories) and CDX2 (CDX2/3NR) (1:1000, 2 h, room temperature). Secondary antibodies consisted of goat anti-rabbit immunoglobulin Gfluorescein isothiocyanate from Roche Applied Sciences. Negative controls (no primary antibody) were included in all experiments. Observations and acquisition of images were performed with a Leica DM-RXA fluorescence microscope (Leica Microsystems, Wetlar, Germany) equipped with a cooled CCD camera (Micromax-5MHz-1300Y, Princeton Instruments Inc., Trenton, NJ). Quantifications were performed on native images using the Metamorph Imaging System version 6.1 (Universal Imaging Corp., West Chester, PA). Student's t test was used for statistical analysis.

Cell Fractionation Along the Crypt-Villus Axis
Segments of mouse jejunum were inverted onto polyethylene tubing, ligated at both extremities, and washed extensively with KRP buffer, pH 7.5, as described previously (31). Segments were then incubated under agitation with ice-cold isolation buffer (2.5 mM EDTA and 0.25 mM NaCl) for 2 min. After each interval, cell suspensions were centri-p27 Kip1 and Cdk2 Regulate CDX2 Stability fuged at 400 ϫ g for 5 min. Pellets were then washed with ice-cold KRB buffer and lysed in chilled Triton lysis buffer (150 mM NaCl, 1 mM EDTA, 40 mM Tris, pH 7.6, 1% Triton X-100, 0.1 mM PMSF, 10 g/ml leupeptin, 1 g/ml pepstatin, 10 g/ml aprotinin, 0.1 mM orthovanadate, and 40 mM ␤-glycerophosphate). Crypt or villus origin of the various cell fractions was determined by evaluation of the activities of several brush border enzymes as described previously (11,31).

In Situ Hybridization
For riboprobe production and RNA, T3 sense (GGTCAAATTAACC-CTCACTAAAGGGAGACTGGTTTCAGAACCGCAGAGC) and T7 antisense (GGTCACCTAATACGACTCACTATAGGGAGCCAGCCTGGAA-TTGCTCTGCC) oligonucleotides were used to amplify 223-bp fragments of human CDX2. RNA was synthesized in vitro using T3 and T7 polymerase (New England Biolabs) according to specifications of the manufacturer. In situ hybridization was performed according to the Wilkinson procedure (32) modified by Lantz et al. (33). Signal detection was achieved using the digoxigenin system following the instructions of the manufacturer (Roche Applied Sciences). Briefly, jejunum from human fetuses varying in age from 18 to 20 weeks of gestation were dissected and fixed for 4 h in 4% paraformaldehyde at 4°C. Tissues were embedded in paraffin from which 5-m sections were generated and spread onto glass slides. The sections were rehydrated, deparaffinized, and treated with 10 mg/ml proteinase K for 4 min, post-fixed in 4% paraformaldehyde for 5 min, acetylated in 1 mM triethanolamide, 0.25% acetic anhydride for 10 min, and hybridized overnight at 55°C in hybridization solution containing 54% formamide, 5ϫ SSC, 5ϫ Denhardt's solution, 0.25 mg/ml yeast tRNA, 0.5 mg/ml herring sperm DNA, and 300 ng/ml digoxigenin-labeled riboprobe. Subsequent to hybridization, the first wash was performed at 55°C in 0.2ϫ SSC for 1 h and 5 min at room temperature. Sections were blocked for 1 h in blocking buffer (1% blocking solution; Roche Applied Sciences) in 0.1 M maleic acid, 0.2 M NaCl, pH 7.5, followed by incubation with antidigoxigenin 1:1500 (Roche Applied Sciences) for 3 h. Slides were exposed to 5-bromo-4-chloro-3-indolyl-phosphate and 4-nitro blue tetrazolium chloride (Roche Applied Sciences) until color reactivity was fully developed.

In Vivo Ubiquitination Assay
HEK293 cells were transfected by calcium phosphate precipitation with expression vectors encoding CDX3 and HA6-tagged ubiquitin. Thirty six hours after transfection, cells were treated with MG132 for the indicated times and lysed in Triton X-100 lysis buffer supplemented with 10 mM N-ethylmaleimide. Lysates were cleared of cellular debris by centrifugation (13,000 rpm, 10 min, 4°C). Lysates were further incubated for 10 min with 2.5 mM dithiothreitol to quench any free N-ethylmaleimide. CDX3 was immunoprecipitated as described above, and ubiquitin-containing conjugates were detected by immunoblotting with a specific antibody against the HA epitope.

Transient Transfections and Luciferase Assays
Subconfluent Caco-2/15 cells were seeded in 24-well plates and transfected by lipofection (Invitrogen) as described before (11) with 0.1 g of sucrase-isomaltase-luciferase reporter and 0.025 g of the relevant expression vector (pBAT) containing wild-type CDX3 or CDX3(RNL/ ANA) mutant per well. Two days after transfection, cells were lysed, and luciferase activity was measured. The pRL-SV40 Renilla luciferase vector (Promega, Madison, WI) was used as a control for transfection efficiency. Two days after transfection, luciferase activity was measured according to the Promega protocol.

Data Presentation
Assays were performed in either duplicate or triplicate. Typical Western blots shown are representative of three independent experiments. Densitometric analyses were performed using the Scion Image 4.02 software package (Scion Corp., Frederick, MA). Representative results of in situ indirect immunofluorescence from three independent experiments are shown.

RESULTS
p27 Kip1 Specifically Interacts with CDX2/3-In an earlier study (11), we demonstrated that stable expression of a p27 Kip1 antisense cDNA was sufficient to markedly block sucrase-isomaltase gene and protein expression in differentiating Caco-2/15 cells. The sucrase-isomaltase promoter was also shown to contain several cis-acting elements important for transcriptional regulation (34). One of these elements, the SIF1 element, interacts with CDX transcription factors (35). The CDX2/3 homeoproteins (the protein designated CDX3 in the hamster and CDX2 in the mouse and humans) stimulate differentiation and expression of sucrase-isomaltase in intestinal epithelial cells (14). Finally, in muscle cells, it has been proposed recently (36) that p57 Kip2 induces cell differentiation by direct interaction with MyoD. Hence, we hypothesized that a similar mechanism could be involved in the cell differentiation effect of p27 Kip1 in intestinal epithelial cells. Total cell extracts from HEK293 cells overexpressing HA-p27 Kip1 , CDX3, or HA-p27 Kip1 with CDX3 were immunoprecipitated with the HA antibody. The complexes were dissociated with deoxycholate and used for IP-EMSA assays (24) with the sucrase-isomaltase CDX-binding site (SIF1). The IP-EMSA revealed a specific CDX3-DNA complex in p27 Kip1 immunoprecipitates of cell extracts overexpressing both CDX3 and p27 Kip1 (Fig. 1A, lane 2). To confirm the physical interaction between p27 Kip1 and CDX3, the IP-EMSA method was used together with cell extracts from differentiated Caco-2/15 cells immunoprecipitated with the p27 Kip1 antibody. Fig. 1B (lane  3) shows that the p27 Kip1 antibody did indeed immunoprecipitate the sucrase-isomaltase CDX DNA complex. This interaction was specific because CDX-containing complexes were not immunoprecipitated with nonrelated antibody (Fig.  1B, lane 1) or p57 Kip2 antibody (Fig. 1B, lane 2). The interaction between CDX2/3 and p27 Kip1 was also assessed in a pull-down assay with the use of the GST-p27 Kip1 fusion protein. Fig. 1C shows that the GST-p27 Kip1 protein was able to pull down CDX2 protein from Caco-2/15 cell lysates, as opposed to the GST protein alone. In addition, in newly postconfluent cells, both p27 Kip1 and CDX2 were localized into the nucleus, with superimposition of both staining clearly showing co-localization of these proteins (Fig. 1D). These data strongly indicate that p27 Kip1 interacts with CDX2/3 transcription factors.
Association of Cdk2 and Phosphorylation of CDX2 by Cdk2 in HEK293 and Caco-2/15 Cells-During intestinal cell differentiation, we have previously reported an increased association of p27 Kip1 with Cdk2 correlating with a strong decline in Cdk2-associated complex activity (11). To determine whether CDX2/3 can also associate with Cdk2, pull-down assays were performed using the GST-CDX3 fusion protein to absorb naive Caco-2/15 cell lysates. The absorbed material was analyzed by Western blot with the Cdk2 antibody. Cdk2 from Caco-2/15 total cell extracts was used as a positive control ( Fig. 2A, lanes 1 and 2). As shown in Fig. 2A (lanes  4 -8), there was a significant amount of Cdk2 bound to the GST-CDX3 fusion protein. GST protein alone did not pull down the Cdk2 protein ( Fig. 2A, lane 3). In order to explore the possible physical interactions between p27 Kip1 , Cdk2 and CDX2/3, an in vitro GST pull-down assay was performed with the GST-CDX3 protein, the recombinant cyclin E-Cdk2 complex, and p27 Kip1 protein (Fig. 2B). The GST-CDX3 recombinant protein was also able to pull down the p27 Kip1 protein (Fig. 2B, lanes 2, 3, and 5) as opposed to the GST protein (lane 6) or beads alone (lane 1). Furthermore, GST-CDX3 interacted with cyclin E-Cdk2 complex (Fig. 2B, lanes 3-5). The interaction still occurred when both p27 Kip1 and cyclin E/Cdk2 proteins were included with GST-CDX3 (Fig. 2B,  lanes 3 and 5). Most interestingly, the association of p27 Kip1 was markedly increased when cyclin E/Cdk2 was first incubated with GST-CDX3 (Fig. 2B, lane 5). These results indicate that CDX2, p27 Kip1 , and cyclin E/Cdk2 proteins can coexist in the same complex. To determine the functional relevance of this CDX3-Cdk2 association, the capacity of pulled down Cdk2 to phosphorylate the GST-CDX3-fusion protein was evaluated in a kinase assay. As shown in Fig. 2C, pulled down Cdk2 from subconfluent Caco-2/15 cells efficiently phosphorylated the GST-CDX3 protein, as confirmed by the potent inhibitory effect of the specific Cdk inhibitor (R)-roscovitine. (R)-Roscovitine, although not monospecific, is a Cdk inhibitor currently undergoing clinical testing and shown to cause a dose-dependent decrease in Cdk2 kinase activity in vitro (37). However, in post-confluent Caco-2/15 cells, kinase(s) other than Cdk2 appeared to associate and phosphorylate GST-CDX3 because (R)-roscovitine only modestly reduced CDX3 phosphorylation. In a previous study, we demonstrated a direct phosphorylation of CDX3 by p38 mitogen-activated protein kinase in post-confluent Caco-2/15 cells (38). Here, an in vitro kinase assay using bacterially expressed GST-CDX3 protein as substrate and recombinant active cyclin E-Cdk2 complex confirmed that Cdk2 was able to potently phosphorylate GST-CDX3 to a significant level (Fig. 2D). In addition, immunoprecipitated Cdk2 from growing subconfluent Caco-2/15 cells, but not post-confluent cells, potently phosphorylated GST-CDX3 (results not shown). We next examined whether CDX3 was phosphorylated by Cdk2 in vivo. HEK293 cells were transfected with either CDX3, HA-Cdk2, p27 Kip1 , or their respective empty vectors and subjected to 32 P phospholabeling. 32 P phospholabeling of immunoprecipitated CDX3 protein showed strong incorporation of 32 P into CDX3 in cells transfected with Cdk2 (Fig. 2E, lane 1  versus lane 4). There was much less incorporation of 32 P observed in cells in which p27 Kip1 was co-expressed (Fig. 2E,  lane 2 versus lane 1). Taken together, these results demonstrate that CDX2/3 may indeed be a specific target for Cdk2associated complexes.
Intestinal Epithelial Cell Differentiation Correlates with Increased Protein Expression of CDX2, p27 Kip1 , and Inhibition of Cdk2 Kinase Activity-Caco-2/15 cells, which spontaneously differentiate into an enterocyte phenotype after confluence (39), were harvested at 60% (subconfluent) and at 100% confluence (day 0 -1), and at 7 and 14 days post-confluence and were analyzed by Western blotting to confirm the timing of p27 Kip1 , CDX2, and sucrase-isomaltase protein ex-FIG. 1. p27 Kip1 specifically interacts with CDX2. A, co-transfection of CDX3 and HA-p27 Kip1 expression vectors was performed in HEK293T cells. 36 h after transfection, cells were lysed, followed by immunoprecipitation (IP) of HA-p27 Kip1 using antibody against HA. The supernatants and immunoprecipitates were incubated with the SIF1 DNA-binding sitelabeled probe after which DNA-protein complexes were separated from the free probe on a native polyacrylamide gel for electrophoretic mobility shift assays. B, nuclear protein extracts from 7-day postconfluent Caco-2/15 cells were immunoprecipitated (Ip) with antibodies against p57 Kip2 or p27 Kip1 . Supernatants and pellets were incubated with the SIF1 DNAbinding site-labeled probe, and DNA-protein complexes were separated from the free probe on a native polyacrylamide gel for electrophoretic mobility shift assay. C, lysates from post-confluent differentiated Caco-2/15 cells were prepared and incubated with 5 g of GST alone or with purified GST-p27 Kip1 bound to glutathione-Sepharose (pull-down GST-p27 Kip1 ). The beads were washed and resuspended in SDS sample buffer, and the bound material was transferred onto nitrocellulose after SDS-PAGE. Immunological detections were performed using antibodies recognizing CDX2 and GST proteins. D, newly confluent Caco-2/15 were fixed with methanol/acetone for immunofluorescence study and co-stained for p27 Kip1 (red) and CDX2 (green) proteins. Bars, 20 m. p27 Kip1 and Cdk2 Regulate CDX2 Stability pression in parallel with Cdk2 kinase activity (all performed in the same cell lysates). Consistent with previous observations (11,38), decreased activity of Cdk2-associated complexes (phosphorylated histone H1) became apparent and significant after confluency, concomitantly with the induction of p27 Kip1 protein (5.3-fold) and CDX2 (3.9-fold) protein at day 7 post-confluency, in comparison with levels observed in subconfluent cells (Fig. 3A). Notably, sucrase-isomaltase protein levels were progressively increased during Caco-2/15 differentiation. Thereafter, both p27 Kip1 and CDX2 protein levels were similarly increased by 4.7-and 2.9-fold, respectively, correlating with the almost complete inhibition (by 80%) of Cdk2 kinase activity (Fig. 3A). Identical results were obtained in Western blot analysis performed with two different CDX2/3 antibodies, CDX2/3-NR and CDX2-Biogenex antibodies (data not shown, see "Experimental Procedures"). By contrast, Northern blot (Fig. 3B) analysis revealed that CDX2 mRNA levels did not differ significantly between subconfluent and post-confluent differentiating Caco-2/15 cells. Hence, these data indicate that there is a clear correlation between protein levels of expression of p27 Kip1 and CDX2 and the decreased kinase activity of Cdk2-associated complexes.
Distribution of CDX2 in Human and Mouse Intestinal Epithelium-The pattern of CDX2 protein expression was further analyzed in normal epithelial cells along the crypt-villus axis of the mouse adult jejunum as well as in the mid-gestational

FIG. 2. Association of Cdk2 and phosphorylation of CDX2 by Cdk2 in HEK293 and Caco-2/15 cells.
A, lysates from subconfluent and post-confluent differentiated Caco-2/15 cells were prepared and incubated with 5 g of GST alone or with purified GST-CDX3 bound to glutathione-Sepharose (pull-down GST-CDX3). The beads were washed and resuspended in SDS sample buffer, and the bound material was transferred onto nitrocellulose after SDS-PAGE. Immunological detections were performed using antibodies recognizing Cdk2. B, GST pull-down assay was performed with GST-CDX3 or GST (as a control of specificity) attached to Sepharose beads and incubated with recombinant p27 Kip1 protein (lanes 2, 3, 5, and 6) and cyclin E-Cdk2 complex (lanes 3-6). Lane 3, p27 Kip1 protein was first incubated with beads for 1 h, and after several washings, the cyclin E-Cdk2 complex was added for 1 h; lane 5, the cyclin E-Cdk2 complex was first incubated with beads, and p27 Kip1 was added 1 h later. After repeated washing procedures, the precipitated proteins were subjected to SDS-PAGE and detected by Western blot. C, lysates from subconfluent and 14-day postconfluent Caco-2/15 cells were prepared and incubated with 5 g of GST alone (not shown) or with purified GST-CDX3 bound to glutathione-Sepharose. The beads were washed four times with lysis buffer followed by three times with ice-cold kinase buffer prior to performing the kinase assay in the presence or absence of 20 M (R)-roscovitine. Kinase activity is demonstrated by the phosphorylation of GST-CDX3. Similar results were obtained in three different experiments. D, kinase assays were performed by incubating active recombinant cyclin E-Cdk2 complex for 30 min with 5 g of GST-CDX3 fusion protein as described under "Experimental Procedures." The reaction was stopped by addition of Laemmli buffer and radiolabeled GST-CDX3 fusion protein separated on SDS-PAGE and autoradiographed. E, co-transfections of HA-CDX3 and p27 Kip1 or Cdk2 expression vectors were performed in HEK293T cells. 36 h after transfection, cells were labeled with [ 32 P]phosphate for 2 h. The cells were subsequently lysed, and [ 32 P]phosphate-labeled CDX3 was isolated by immunoprecipitation with CDX2/3 antibody and analyzed by Western blot and autoradiography. The results are representative of three independent experiments. p27 Kip1 and Cdk2 Regulate CDX2 Stability human fetal jejunum, because at this stage, the crypt-villus axis is morphologically and functionally similar to its adult counterpart (40). Immunofluorescence staining revealed that CDX2 was detected in the nuclei of all enterocytes along the human (Fig. 4A) and mouse crypt-villus axis (data not shown). However, fluorescence intensity was consistently weaker in the lower portion of the crypt than in the remaining epithelium (Fig. 4, B and C). Paneth cells are differentiated cells that exist in the crypt. However, CDX2 protein has been reported to be weakly expressed in adult mouse Paneth cells (41), and a negligible Paneth cell compartment is present in the mid-gestational human fetal jejunum (42). Determination of the net fluorescence intensity for 26 crypts from 8 fetal human jejuna confirmed that cells located in the lowest portion of the crypt expressed significantly (p Ͻ 0.05) lower levels of CDX2 that those located in the upper region of the crypt. Expression of CDX2, p27 Kip1 , Cdk2, and cyclin E proteins was also analyzed by Western blot in crypt and villus cell populations isolated from adult mouse jejunum according to a modified Weiser procedure (31). Biochemical analysis of sucrase-isomaltase, alkaline phosphatase (not shown), and trehalase activities in recovered intestinal epithelial cell preparations revealed relatively pure preparations of poorly differentiated crypt to fully differentiated villus cell populations (Fig. 4D). Western blot analysis of CDX2 protein expression well confirmed the expression patterns observed in immunofluorescence experiments. Of note, CDX2 immunostaining was also increased along the co-lonic gland axis (data not shown). By contrast, in situ hybridization revealed that CDX2 mRNAs were homogeneously expressed along the entire crypt-villus axis of the human intestine (Fig. 4E). Hence, these results indicated that the gradient of CDX2 protein expression along the crypt-villus axis of the intestine may be due to an increase in the stability of CDX2 protein rather to an increase in mRNA levels.
CDX2 Stability Is Increased during Caco-2/15 Cell Differentiation-Because, in contrast to mRNA levels, CDX2 protein levels increased during Caco-2/15 cell differentiation, the stability of CDX2 was compared in subconfluent, confluent, and 7-day post-confluent cells, following cycloheximide treatment. As shown in Fig. 5A, the rate of degradation of CDX2 was clearly higher in subconfluent growing cells than in post-confluent cells. Quantitation of the data revealed that the half-life of the protein was around 3 h in subconfluent cells, compared with that of confluent (5 h) or 7-day post-confluent cells (Ͼ8 h) (Fig. 5B). Expression levels of actin, a stable protein, was not modified until 8 h after cycloheximide addition (data not shown). These results indicate that stability of CDX2 is increased during intestinal epithelial cell differentiation.
CDX2 and p27 Kip1 Proteins Are Degraded through the Ubiquitin-Proteasome Pathway-Because many short lived proteins are degraded by a ubiquitin-proteasome-dependent mechanism (43), the effect of the peptide-aldehyde MG132, a potent inhibitor of the 20 S proteasome (44), was tested on the stability of CDX2 protein in subconfluent proliferating Caco-2/15 cells. p27 Kip1 , which is degraded through the ubiquitin-proteasome system in proliferative cells (45), was used as a positive control. Treatment of subconfluent proliferating Caco-2/15 cells with 25 M MG132 led to an accumulation of p27 Kip1 protein after 4 h (Fig. 6A, middle panel). In parallel, CDX2 protein levels increased by 3-4-fold 4 h after addition of the proteasome inhibitor (Fig. 6A, upper panel). Addition of MG132 did not significantly alter CDX2 or p27 Kip1 protein levels in post-confluent nonproliferating Caco-2/15 cells (data not shown). Of note, CDX2 protein expression was also enhanced in the colon cancer cell line DLD-1 following MG132 treatment (Fig. 6B). These observations are consistent with the view that CDX2 is degraded by the proteasome in proliferative cancer intestinal epithelial cells.
Most substrates of the proteasome are conjugated to multiple molecules of ubiquitin prior to their degradation (46). Treatment of cells with MG132 blocks the degradation of ubiquitinated proteins but does not interfere with ubiquitin conjugation to protein substrates, thereby resulting in transient accumulation of ubiquitin-protein conjugates. To demonstrate that CDX2 is ubiquitinated in vivo, HEK293 cells were cotransfected with CDX2 and HA6-tagged ubiquitin followed by treatment with MG132. Cellular extracts were then subjected to immunoprecipitation with anti-CDX2 and analyzed by anti-HA immunoblotting. High molecular weight species corresponding to multiubiquitinated forms of CDX2 were detected only in cells treated with MG132 (Fig. 6C, see arrow and arrowheads), demonstrating that CDX2 proteolysis in vivo is dependent on ubiquitin conjugation.
The Cytoplasmic Ubiquitin-Proteasome System Degrades CDX2 in Intestinal Epithelial Cells-To determine whether the degradation of CDX2 occurs in the nucleus or in the cytoplasm, degradation of CDX2 was examined in subconfluent proliferating Caco-2/15 cells in the presence of either MG132 or leptomycin B in combination with cycloheximide. Leptomycin B is known to inhibit the CRM1-dependent nuclear export pathway through a covalent interaction at Cys 529 of CRM-1 (47) and has been used to determine whether nuclear export is required in the degradation of nuclear proteins. As shown in Figs. 5 and   FIG. 3. Intestinal epithelial cell differentiation correlates with increased protein expression of CDX2, p27 Kip1 and inhibition of Cdk2 kinase activity. A, Caco-2/15 cells were harvested at 70% (subconfluence (subc.)), 100% confluence (day 0 -1), and 7 and 14 days postconfluence. Cell extracts were separated by 10% SDS-PAGE and proteins analyzed by Western blotting for expression of sucrase-isomaltase, p27 Kip1 , Cdk2, and CDX2. Cell extracts (400 g) were immunoprecipitated with a specific antibody to Cdk2, and the kinase activity of Cdk2 was demonstrated by the phosphorylation of histone H1. B, total RNA was extracted from Caco-2/15 cells at different stages of confluence (Ϫ2, 0, 6, and 15 days post-confluence). Northern hybridization was performed using a 32 P-labeled human CDX2 probe. An L32 probe was used to evaluate the relative amounts of mRNA transferred to the membrane. The results are representative of three independent experiments. p27 Kip1 and Cdk2 Regulate CDX2 Stability 7A, the half-life of CDX2 protein in subconfluent proliferating Caco-2/15 cells was found to be about 3-4 h and was strongly extended when proteasome activity was blocked with MG132. Most interestingly, CDX2 was significantly stabilized by the presence of leptomycin B, indicating that CDX2 is degraded in the cytoplasm. To ensure that leptomycin was active in the inhibition of nuclear export, a series of immunofluorescence experiments were carried out in proliferative Caco-2/15 cells. As seen in Fig. 7B (panel 1), CDX2 was mostly localized to the nucleus in control conditions. Treatment of Caco-2/15 cells with leptomycin B enhanced the nuclear staining of CDX2. In addition, a sharp staining at the nuclear envelope was also observed (Fig. 7B, panel 5, see arrowheads). In similar fashion, following leptomycin B treatment, CRM1 staining accumulated at the nuclear envelope (Fig. 7B, panel 7 versus panel 3, see  arrowheads). Overall, these results indicate that nuclear export is required for the degradation of CDX2 and that CRM1 directs CDX2 nuclear export in intestinal epithelial cells.
Cdk2 Activity Promotes CDX2 Degradation via the Ubiquitin-Proteasome Pathway-Earlier studies (48) demonstrated that p27 Kip1 must be phosphorylated by Cdk2 in order to be degraded by the proteasome. In Fig. 2, we demonstrated that CDX2 is a substrate for phosphorylation by Cdk2. To test the hypothesis that Cdk2-dependent phosphorylation of CDX2 promotes its degradation, the effect of the Cdk2 inhibitor (R)-roscovitine was first examined on CDX2 and p27 Kip1 expression. Treatment of proliferating Caco-2/15 cells with (R)-roscovitine potently blocked cyclin E-Cdk2 activity (data not shown) and led to an increase in expression levels of CDX2 and p27 Kip1 proteins (Fig. 8A) and comparatively to nontreated Caco-2/15 cells at the same time points (data not shown). Next, the influence of ectopic expression of Cdk2 on the levels of co-expressed CDX2 was examined in transiently transfected Caco-2/15 cells treated with cycloheximide. As shown in Fig.  8B, CDX2 exhibited a significant shortened half-life when coexpressed with Cdk2 in subconfluent proliferating Caco-2/15 cells. This suggests that the increased levels of CDX2 observed following (R)-roscovitine treatment is due to the stabilization of the protein. Immunofluorescence experiments shown in Fig. 8C revealed significant cytoplasmic and nuclear staining for CDX2 after 16 h of MG132 treatment (Fig. 8C, panel 2). By contrast, (R)-roscovitine induced CDX2 accumulation into the nucleus only (Fig. 8C, panel 3). Of note, when (R)-roscovitine and MG132 were added to subconfluent proliferating Caco-2/15 p27 Kip1 and Cdk2 Regulate CDX2 Stability cells, CDX2 protein dramatically accumulated into the nucleus and was absent in the cytoplasm (Fig. 8C, panel 4). These results suggest that nuclear export and degradation of CDX2 are both dependent on Cdk2 activity.
Mutational Analysis of 162 RNL 164 Residues on CDX3 Abolishes Binding of Cdk2-Inhibitors, activators, and substrates of Cdks utilize a cyclin-binding sequence, known as a Cy or RXL motif, to bind directly to the cyclin subunit (49). We previously found a docking domain for Cdk2 homologous to the general consensus sequence RXL and localized between amino acids at positions 162 and 164 of the transactivation domain of CDX. Each residue in the core Cy motif from CDX3 ( 162 RNL 164 ) was mutated to alanine. To determine whether CDX3(RNL/ ANA) can also directly associate with Cdk2, pull-down assays were performed using the wild-type GST-CDX3 and GST-CDX3(RNL/ANA) fusion proteins to absorb naive Caco-2/15 cell lysates. The absorbed material was analyzed by Western blot with the Cdk2 antibody. As shown in Fig. 9A, GST-CDX3 was able to pull down Cdk2 in contrast to GST-CDX3(RNL/ANA). Accordingly, GST-CDX3(RNL/ANA) was much less phosphorylated in comparison to GST-CDX3, following an in vitro kinase assay. The stability of the CDX3(RNL/ANA) mutant was compared with the stability of wild-type CDX3 in transfected HEK293 cells following cycloheximide treatment. As shown in Fig. 9B, expression levels of CDX3(RNL/ANA) mutant was not modified 8 h after cycloheximide addition, in contrast to wildtype CDX3 that was markedly degraded 8 h after cycloheximide treatment. In addition, we also compared the transcriptional activity of wild-type CDX3 and CDX3(RNL/ANA) mutant on sucrase-isomaltase transcription. As shown in Fig.  9C, the mutation of Ala 162 -Ala 164 significantly improves CDX3 transcriptional activity on sucrase-isomaltase promoter activity, when compared with wild-type CDX3. Overall, these results clearly indicate that the mutation of Ala 162 -Ala 164 on CDX3, which diminishes CDX3/Cdk2 association and Cdk2-mediated phosphorylation, improves CDX3 protein stability and transcriptional activity. DISCUSSION CDX transcription factors play multiple functions in the establishment and maintenance of the intestinal epithelium. One critical feature of the organization of the intestinal epithelium is the compartmentalization of proliferating cells in crypts and the continuous renewal of differentiated cells on the villus from this proliferating compartment. This apparent antagonism between proliferation and terminal differentiation of intestinal epithelial cells implies that signaling pathways driving proliferation must be suppressed to allow differentiation. Indeed, recent studies have demonstrated the role of G 1 cyclins and their partners (Cdk and Ckis) in cell cycle arrest and enterocyte differentiation program (7)(8)(9)(10)(11). Because cell cycle exit and induction of intestinal epithelial cell differentiation only take place in G 1 (11, 50 -52), CDX2 expression, which is already detected in proliferating intestinal epithelial cells, must be precisely controlled at this stage of the cell cycle. Here our study provides the first evidence that the cell cycle inhibitor p27 Kip1 and the cyclin-associated kinase Cdk2 physically interact with CDX2 protein and that Cdk2 activity promotes proteasome-dependent degradation of CDX2 in intestinal epithelial growing cells.
Little is known about the regulation of CDX2 activity. Phosphorylation of CDX2 regulates its transcriptional activity and may modulate its interactions with other transcription factors and cofactors (35,38,53). However, the mechanisms regulating CDX2 gene expression during cell differentiation and during the progression of colonic adenocarcinomas remain elusive. Indeed, the low rate of genomic alteration reported at the CDX2 locus in human colorectal tumors indicates that decreased CDX2 expression as opposed to structural mutation might be responsible for its down-regulation in intestinal cancers (54). Although prior p27 Kip1 and Cdk2 Regulate CDX2 Stability work had hinted that PTEN and phosphatidylinositol 3-kinase might regulate CDX2 expression in some cellular backgrounds (55), others (56) failed to obtain evidence to support a role for endogenous phosphatidylinositol 3-kinase in the repression of CDX2 expression in colon cancer cells.
In the present study, we were able to show that CDX2 expression levels significantly increased as soon as Caco-2/15 cells reached confluence, slowed their proliferation and began their differentiation. While this work was in progress, another study (57) reported such up-regulation of CDX2 protein expression during Caco-2 cell differentiation. Our results also demonstrate that the mechanism of CDX2 regulation is primarily related to protein stability, because inhibition of proteasome activity increased ubiquitin-CDX2 conjugates and CDX2 protein levels, whereas very little change in CDX2 mRNA levels was found during intestinal differentiation. In addition, experiments performed with cycloheximide revealed that the halflife of CDX2 protein was significantly enhanced in differentiated versus undifferentiated proliferative Caco-2/15 cells. Therefore, it appears that CDX2 is subject to ubiquitin-dependent degradation in proliferative intestinal epithelial cells. These data are supported by immunofluorescence experiments demonstrating a gradient of CDX2 protein expression along the colonic axis and along the small intestinal crypt-villus axis with higher levels being detected in differentiated epithelial cells. Such a pattern of CDX2 protein expression in the intestine has also been reported by James et al. (13), Silberg et al. (41), and Kim et al. (55). However, others have reported that CDX2 protein was equally distributed along the total length of the colonic crypt and crypt-villus axis of the small intestine (53,58). These conflicting observations could result from the use of different antibodies or of different concentrations of antibodies. Indeed, we detected the CDX2 gradient after a serial dilution of CDX2 antibodies to establish optimal staining concentration.
Nuclear export is required for the ubiquitin-dependent degradation of other proteins such as p27 Kip1 (59) Immunological detections were performed with antibodies recognizing Cdk2. The other half of the beads was washed four times with lysis buffer followed by three times with ice-cold kinase buffer prior to performing the kinase assay. Similar results were obtained in three different experiments. B, HEK293 cells were transfected with CDX3 or CDX3(RNL/ANA) expression vectors. Thirty six hours after transfection, cells were treated with cycloheximide (25 g/ml) for 8 h followed by cell lysis at the end of the treatment. CDX3 and CDX3(RNL/ANA) protein levels were determined by Western blotting. Results are representative of three independent experiments. C, subconfluent Caco-2/15 cells were transfected with 0.1 g of the sucrase-isomaltase (SI)-luciferase reporter vector and 0.025 g of the CDX3 or CDX3(RNL/ANA) expression vectors or the pBAT vector (EV). Two days after transfection, cells were lysed, and luciferase activity was measured. The increase in luciferase activity was calculated relative to the empty vector level (pBAT) of sucrase-isomaltase-luciferase, which was set at 1. Results are the mean Ϯ S.E. of at least three separate experiments. *, significantly different from control at p Ͻ 0.05 (Student's t test). p27 Kip1 and Cdk2 Regulate CDX2 Stability and p53 (61) as well as IB␣ (62). Hence, these results, which couple nuclear export with degradation of the substrate, suggest that export to the cytoplasmic ubiquitin-proteasome machinery may represent a common route for the degradation of nuclear substrates. However, it has been demonstrated recently that MyoD is ubiquitinated and degraded in the nucleus (63). Our data indicate that MG132 enhances the amount of CDX2 into the cytoplasm, indicating that CDX2 is degraded through the cytoplasmic ubiquitin-proteasome system. This cytoplasmic degradation of CDX2 is dependent in large part on its active nuclear export by the CRM1 receptor. CRM1 possesses the broadest range of substrates among the exportin receptors and mediates the nuclear export of numerous signaling molecules and transcription factors (64). Our proposed hypothesis of CDX2 cytoplasmic degradation is further supported by the following evidence. First, treatment of Caco-2/15 cells with leptomycin, a potent and selective inhibitor of CRM1 function, causes nuclear accumulation of CDX2. Second, CDX2 and CRM1 accumulate at the nuclear envelope of intestinal epithelial cells following leptomycin treatment. Third, CDX2 and CRM1 co-precipitate in pull-down assays (preliminary results). CRM1 mediates nuclear export by binding to a leucinerich nuclear export signal motif in the export substrate (64). The first nuclear export signal identified was the human immunodeficiency virus, type 1, Rev protein, and a classical nuclear export signal consensus sequence (LX (2-3) LNL) has been identified based on conserved clustering of leucine residues (65). We have found similar motifs localized between amino acids 218 and 224 of the CDX2 and CDX3 proteins. However, a functional relationship remains to be determined.
Our data suggest that CDX2 degradation may be triggered by direct Cdk2-dependent phosphorylation of CDX2, in a manner similar to Cdk2-phosphorylation/degradation of p27 Kip1 (48) and MyoD (66). Indeed, the steady-state level of CDX2 protein is increased in the presence of MG132, a proteasome inhibitor, and (R)-roscovitine, a small molecule that specifically targets the ATP-binding site of Cdks, and especially Cdk2, indicating that both inhibition of proteasome function or Cdk2 activity stabilizes CDX2. The observation that treatment with (R)-roscovitine leads to accumulation of CDX2 into the nucleus suggests that phosphorylation of CDX2 by Cdk2 may be a prerequisite for CRM1 binding, nuclear export, and cytoplasmic degradation. Cdk2-mediated phosphorylation is often involved in proteolytic degradation of other cell cycle and differentiation regulators such as p27 Kip1 (48), MyoD (66), Ctd1 (67), ␤-catenin (68), and cyclin E (69). Cdk2-associated complexes could be targeted to the transcription factor CDX2/3 by a docking domain that is distinct from the phosphoacceptor motifs. In fact, we observed a docking domain for Cdk2 homologous to the general consensus sequence RXL (49) and localized between amino acids 162 and 164 of the transactivation domain of CDX2 and CDX3 proteins. Site-directed mutagenesis of this sequence impaired the association of Cdk2 to GST-CDX3. Hence, our data support the model that CDX2 undergoes CRM1-dependent nuclear export and cytoplasmic degradation in cells in which Cdk2 is highly activated, such as proliferative intestinal epithelial undifferentiated crypt cells and colon cancer cells.
The Cdk2-dependent decrease in CDX2 expression reported here in proliferative Caco-2/15 is fully consistent with the decline of expression of this tumor suppressor in colorectal cancer. Indeed, Cdk2-associated complexes play a key role in intestinal carcinogenesis. Cyclin E amplification has been identified in colon cancer (70), and cdk2 amplification is associated with concurrent cyclin E gene amplification in colorectal carcinomas (71). Furthermore, colorectal tumors have higher levels of Cdk2-associated cyclin E (72) and cyclin A (73). How-ever, loss of CDX2 expression in some colorectal cancer cells may also result from defects in trans-acting pathways regulating CDX2 transcription, as recently demonstrated by Hinoi et al. (56), indicating that the abundance of CDX2 protein could be regulated by both transcriptional and post-translational (proteolysis) pathways.
Finally, regulation of CDX2 through Cdk2-associated complexes may account, at least in part, for p27 Kip1 -dependent promotion of intestinal epithelial cell differentiation (9,11). Indeed, an increased association of p27 Kip1 with Cdk2, correlating with the inhibition of Cdk2 activity, has been observed during intestinal epithelial cell differentiation. Accordingly, expression of p27 Kip1 in vivo was barely detected in the nuclei of most crypt cells; however, its nuclear expression began to increase significantly in cells of the upper region of the crypt and toward the villus (11). We have also demonstrated previously that whereas Cdk2 protein expression is similar between undifferentiated and differentiated intestinal epithelial cells (Fig. 4D), Cdk2 kinase activity is negligible in primary cultures of differentiated villus cells compared with that found in proliferating intestinal crypt cells (11). Hence, we believe that the increased CDX2 protein stability observed during intestinal epithelial cell differentiation is likely caused by the loss of kinase activity of Cdk2-associated complexes (and not expression) due to increased p27 Kip1 protein expression.
It has been reported that p27 Kip1 homozygous null mice are free from severe developmental and proliferative intestinal disorders (74 -76). However, p27 Kip1 is the major cell cycle inhibitor synthesized in differentiating Caco-2/15 cells and in primary cultures of human differentiated enterocytes, and inhibition of this induction interferes with differentiation (11). An attractive possibility is that some p27 Kip1 functions may be redundant and may be compensated for by the other members of the Cip/Kip family, namely p21 Cip and p57 Kip2 . In this regard, we noted a significant increase of expression of p57 Kip2 in p27 Kip1 antisense-transfected intestinal cells (11). Whereas p27 Kip1 -deficient mice do not spontaneously develop intestinal tumors, they show markedly increased predisposition to adenoma and adenocarcinoma development in the small intestine and colon in response to four diverse carcinogens: ENU, ␥-radiation (75), DMBA (77), and DMH (78). In addition, reduction of p27 Kip1 greatly accelerated the rate of development of intestinal tumors with pre-existing mutations in Apc (78), and loss of p27 Kip1 , an important inhibitor of Cdk2-associated complexes, is associated with aggressive behavior in colorectal adenocarcinomas (12). Most interestingly, a greater percentage of colon tumors from DMH-treated p27 Kip1 -deficient mice were adenocarcinomas, with features that included complete invasion through the muscularis into the serosal space, and lymphatic vessel penetration. These more aggressive phenotypes were not observed in tumors from p27 Kip1 wild-type littermates (78). This suggests that an additional tumor-suppressing function of p27 Kip1 may be to control tumor cell differentiation, migration, and/or invasion.
In conclusion, Cdk2-associated complexes, by targeting CDX2 for ubiquitination and consequent degradation, might counteract the ability of CDX2 to block cell cycle progression (14,16,79). In a similar manner, cyclin E-Cdk2 down-regulates the ability of p27 Kip1 (48) and MyoD (66) to block cell cycle progression. The purpose of CDX2 degradation in cycling undifferentiated cells is an interesting query; CDX2, like p27 Kip1 , may participate in the establishment of a restriction point in enterocytes that controls the decision to proliferate or to exit from the cell cycle and differentiate.