Cloning of the Human Claudin-2 5 (cid:1) -Flanking Region Revealed a TATA-less Promoter with Conserved Binding Sites in Mouse and Human for Caudal-related Homeodomain Proteins and Hepatocyte Nuclear Factor-1 (cid:1) *

Claudin-2 is a structural component of tight junctions in the kidneys, liver, and intestine, but the mechanisms regulating its expression have not been defined. The 5 (cid:1) -flanking region of the claudin-2 gene contains binding sites for intestine-specific Cdx homeodomain proteins and hepatocyte nuclear factor (HNF)-1, which are conserved in human and mouse. Both Cdx1 and Cdx2 activated the claudin-2 promoter in the human intestinal epithelial cell line Caco-2. HNF-1 (cid:1) augmented the Cdx2-induced but not Cdx1-induced transcriptional activation of the human claudin-2 promoter. In mice, HNF-1 (cid:1) was required for claudin-2 expression in the villus epithelium of the ileum and within the liver but not in the kidneys, indicating an organ-specific function of HNF-1 (cid:1) in the regulation of claudin-2 gene expression. Tight junction structural components, which determine epithelial Region of the Human Claudin-2 Gene and Cloning of Human Claudin-2— A degenerate primer approach with primers 5 (cid:1) -TGG ATG GA(A/G) TGT GC(A/T/G/C) AC(A/T/G/C) CA(C/ T)-3 (cid:1) and 5 (cid:1) -GA GCA (G/A)GA (A/G)AA GCA (A/T/G/C)AG (A/G/T/ C)AT(G/T/A)AT (A/G/T/C)CC-3 (cid:1) , corresponding to the mouse claudin-2 sequence, used to amplify 407 bp of the open reading frame of human claudin-2. Data base searches with the putative human clau-din-2 sequence identified several human claudin-2 expressed sequence tag clones, which were used to complement the 5 (cid:1) and 3 (cid:1) sequence. PCR with the primers 5 (cid:1) -GCT TCT ACT GAG AGG TCT G-3 (cid:1) and 5 (cid:1) -TTC TTC ACA CAT ACC CTG-3 (cid:1) , and DNA sequencing was utilized to confirm the expression of the full-length human claudin-2 sequence in T-84 cells. DNA and amino acid sequence of human clau-din-2 GenBank TM 5 (cid:1) -flanking of claudin-2 PCR (cid:1)

Claudin-2 is a regulatory component of tight junctions in the liver, the kidneys, and the epithelium of the small and large intestines (1,2). Claudin-2 expression has been demonstrated to be involved in the regulation of the intestinal barrier function by immune modulators (3). Claudins form a family of proteins composed of at least 24 members, which are expressed in an organ-specific manner and regulate the tissue-specific physiological properties of tight junctions (4,5). Tight junctions not only create a primary barrier to prevent paracellular passage of solutes and pathogens, but they also restrict the lateral diffusion of membrane lipids and proteins to maintain cellular polarity (5)(6)(7)(8). Evidence is mounting that claudins are actively involved in the regulation of paracellular transport of ions through tight junctions (9,10). The modulation of selective transport through tight junction may require the coordinated expression of distinct claudins in a particular cell type (1,11). Therefore, the regulation of claudin expression may determine the fundamental ability of the intestinal epithelium to modulate water or ion transport and barrier function. However, the transcriptional events involved in the organ-specific expression of claudins have not been determined.
Cdx1 and Cdx2 are members of the caudal-related homeobox gene family based on their sequence homology to the caudal gene of Drosophila melanogaster (12)(13)(14). In vitro and in vivo studies of Cdx1 and Cdx2 suggest that these transcription factors are important in the early differentiation and maintenance of intestinal epithelial cells. In vitro experiments show significant functional effects of Cdx genes on intestinal differentiation (15,16), proliferation (15,17), and intestine-specific gene transcription (18 -22). Overexpression of Cdx2 in undifferentiated IEC-6 intestinal epithelial cells leads to the development of a differentiated phenotype (15). Cdx1 and Cdx2 have been shown to regulate intestine-specific gene transcription by binding to several intestine-specific promoters (18 -20, 23, 24). In intestinal epithelial Caco-2 cells, Cdx2 expression induces the expression of sucrase isomaltase (SI) 1 and lactase-phlorizin hydrolase (LPH), two markers of intestinal differentiation (25).
In this report, we examined the Cdx-and HNF-1␣-mediated regulation of the claudin-2 promoter in the human intestinal epithelial cell line Caco-2 and determined the claudin-2 mRNA and protein expression in HNF-1␣-deficient mice. These experiments identify claudin-2 as a target of Cdx homeoproteins and HNF-1␣ function in human intestinal epithelial cells. HNF-1␣ regulated the complex pattern of claudin-2 expression along the crypt-villus axis of the mouse ileum and was required for claudin-2 expression in the liver.
Cell Culture-Cells from the human colon cancer-derived cell line Caco-2, the human hepatocellular carcinoma-derived cell line HepG2, and the mouse mesenchymal cell line NIH3T3 were obtained from the American Type Culture Collection (Manassas, VA). These cells were grown in Dulbecco's modified Eagle's medium (Cellgro; Mediatech Inc., Herndon, VA), supplemented with 100 IU/ml penicillin, 100 g/ml streptomycin, and 10% (for HepG2 and NIH3T3) or 20% (for Caco-2) heat-inactivated fetal calf serum (Sigma) in a humidified 5% CO 2 atmosphere at 37°C. The human colon cancer-derived cell line T-84 cells were grown in Dulbecco's modified Eagle's medium with Ham's F-12 medium (1:1) with the antibiotics described above and 10% heat-inactivated fetal calf serum. Data base searches with the putative human claudin-2 sequence identified several human claudin-2 expressed sequence tag clones, which were used to complement the 5Ј and 3Ј sequence. Additional PCR with the primers 5Ј-GCT TCT ACT GAG AGG TCT G-3Ј and 5Ј-TTC TTC ACA CAT ACC CTG-3Ј, and DNA sequencing was utilized to confirm the expression of the full-length human claudin-2 sequence in T-84 cells. DNA and amino acid sequence of human claudin-2 has been submitted to GenBank TM and is available under the accession number AF250558. The GenomeWalker kit (CLONTECH, Palo Alto, CA) was used to isolate the 5Ј-flanking region of the human claudin-2 gene. In brief, the first PCR was performed with gene-specific primer 1 (5Ј-CAA AAG CCC CAG AAG GCC TAG GAT GTA G-3Ј; ϩ30 to ϩ57 relative to the adenosine of the methionine start codon of the human claudin-2 cDNA; GenBank TM accession numbers AF250558 or AF177340) and adaptor primer 1. The second PCR was done with gene-specific primer 2 (5Ј-GGC AGA CCT CTC AGT AGA AGC GTC TTC-3Ј; Ϫ27 to Ϫ1; corresponding to the 493-519 sequence of AF177340) and adaptor primer 2. The longest PCR fragment was purified and subcloned into pCR2.1 vector (Invitrogen). The resulting plasmid was designated as pCR-hCL2p and sequenced.
Deletion Constructs, Mutagenesis, and Reporter Gene Assay-The KpnI/XhoI fragment of pCR-hCL2p was subcloned into the KpnI/XhoI site of the pGL3B vector (Promega, Madison, WI). Various length fragments of the 5Ј-flanking region of the human claudin-2 gene were amplified by PCR and subcloned into pGL3B. To obtain the Ϫ62 construct, the HindIII/XbaI fragment from the Ϫ84 construct was ligated to the EcoRI/HindIII-digested Ϫ84 construct with complementary 38base oligonucleotides (designated as Ϫ62wt, from Ϫ62 to Ϫ31; sense, 5Ј-AAT TCA TAT TTA ATC TGG TTT ATG GAT TTT TTT TAG GT-3Ј; antisense, 5Ј-CTA GAC CTA AAA AAA ATC CAT AAA CCA GAT TAA ATA TG) with 5Ј-EcoRI and 3Ј-XbaI overhangs (underlined). To make mutant claudin-2 promoter constructs, mutated 38-base oligonucleotides were substituted for the wild type sequence. For Mut1, Mut2, and Mut1 ϩ 2, 5Ј-AAT TCA TAT TTA ATC TGG TGG CTG GAT TTT TTT TAG GT-3Ј, 5Ј-AAT TCA TAT TTA ATC TGG TTT ATG GAT TTT TTG GCG GT-3Ј, and 5Ј-AAT TCA TAT TTA ATC TGG TGG CTG GAT TTT TTG GCG GT-3Ј were used, respectively. (Only sense strands are shown; nucleotide substitutions are indicated with bold type.) For reporter assays, a DNA transfection mixture was prepared consisting of 1 g of the reporter construct and 20 ng of pRL-CMV (Promega) as an internal control. The cells were split onto 6-well plates 18 h before transfection. The cell confluency at transfection was 40 -60%. The individual DNA mixtures were transfected with LipofectAMINE Plus (Invitrogen) according to the manufacturer's protocol. For cotransfection experiments, 0.5 g of the expression vector was transfected along with reporter vectors. pcDNA3.1 vector (Invitrogen) was used to equalize the amount of transfected DNA. The cells were harvested 48 h after transfection, and the luciferase activity was measured using the dual-luciferase reporter assay system (Promega) and a luminometer.
Transfection efficiencies were normalized to Renilla luciferase activity of the pRL-CMV vector, and the results are expressed as the mean relative luciferase activity Ϯ S.D. of at least three independent experiments.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear proteins were prepared as previously described (33). Cytosolic fractions obtained during this procedure were separated for Western blot analysis. The double-stranded oligonucleotides, Ϫ62wt, Mut1, Mut2, and Mut1 ϩ 2, were used as probes or cold competitors to analyze the interaction between Cdx protein and DNA. The HNF-1 wild type probe from Ϫ67 to Ϫ51 of the human claudin-2 gene sequence consisted of complementary 29nucleotide oligonucleotides (sense, 5Ј-AAT TCC TGG TCA ATA TTT AAT CTG T-3Ј, and antisense, 5Ј-CTA GAC AGA TTA AAT ATT GAC CAG G-3Ј) with 5Ј-EcoRI and 3Ј-XbaI overhangs (underlined). Mutant HNF complementary oligonucleotides were as follows: sense, 5Ј-AAT TCC TAA TTC AGG TTT AAT CTG T-3Ј, and antisense, 5Ј-CTA GAC AGA TTA AAC CTG AAT TAG G-3Ј (nucleotide substitutions are indicated with bold type).
The probes were labeled with Klenow enzyme by fill-in incorporation with nucleotide triphosphates, including [␣-32 P]dATP. The binding reaction was performed as previously described (34). For a competition assay, a 100-fold excess of unlabeled oligonucleotide was added to the reaction. To perform supershift assay, the binding mixtures were incubated for 10 min at room temperature in the presence of 1 l of antibodies. The samples were fractionated on 4% nondenaturing polyacrylamide gel in 0.5ϫ TBE buffer. The resultant DNA-protein complexes were detected by autoradiography.
Western Blot Analysis-The protein concentration of each sample was quantified by the Bradford method. The samples were electrophoresed through a 4 -20% gradient SDS-polyacrylamide gel and transferred onto polyvinylidene difluoride membranes (Millipore, Bedford, MA). The blots were blocked overnight at 4°C with 10% dry milk in PBS containing 0.1% Tween 20 (PBS-T), followed by the incubation for 3 h at room temperature with primary antibodies diluted in blocking buffer at 1:1000. After washing in PBS-T for 30 min, the blots were incubated with secondary antibodies diluted in blocking buffer for 45 min at room temperature. The hybridized bands were detected by an ECL kit (Amersham Biosciences) according to the manufacturer's instructions.
RNA Extraction and Northern Blot Analysis-Total RNA was isolated from tissues using Trizol reagent (Invitrogen). Total RNA (30 g) was electrophoresed in a 1% agarose formaldehyde gel and transferred to a nylon membrane (Magna NT, MicroSeparations Inc., Westbrough, MA) by capillary blotting. The probes were labeled with [␣-32 P]dCTP using a Rediprime random primer labeling kit (Amersham Biosciences). The membranes were hybridized with radiolabeled probes in Quickhyb solution (Stratagene, La Jolla, CA) at 65°C for 1 h. The membranes were washed with 0.1% SDS, 2ϫ sodium chloride sodium citrate buffer at room temperature for 15 min and at 65°C for 10 min. The blots were analyzed by autoradiography. The probes used to detect claudin-1, claudin-2, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were described previously (3). Other probes were: Cdx1, a 0.9-kb Hin-dIII/XbaI fragment of pRc/CMV-Cdx1; Cdx2, a 0.9-kb HindIII fragment of pRc/CMV-Cdx2; and HNF-1␣, a 0.4-kb SmaI fragment of pBJ5mHNF1␣. The HNF-1␣ probe derives from a unique sequence in HNF-1␣ cDNA and does not cross-hybridize with HNF-1␤. Northern blots were densitometrically analyzed, and gene-specific mRNA expression levels were normalized to GAPDH mRNA expression levels in the same samples and expressed as the mean density/area calculated from three independent experiments.
Tissue Preparation and Immunostaining-Mice carrying the HNF-1␣ null allele were obtained from Dr. Frank J. Gonzalez (National Institutes of Health, Bethesda, MD) (35). Homozygous HNF-1␣ null and wild type littermates were obtained by mating heterozygous carriers. All of the animal experiments were performed in accordance with National Institutes of Health guidelines and protocols approved by the Subcommittee on Research Animal Care at our institute. The liver and kidney were removed and washed with ice-cold PBS. Segments of 2 cm from the most proximal jejunum and the most distal ileum were collected. For immunostaining, small tissue blocks were mounted in OCT compound and frozen in dry ice-ethanol. For RNA extraction, small pieces of tissues were snap frozen at Ϫ80°C.
4-m-thick cryosections of frozen tissues were prepared. The sections were air-dried and fixed in methanol at Ϫ20°C for 10 min followed by rehydration in PBS at 4°C for 30 min as previously described (2). The sections were blocked with 5% normal donkey serum in PBS (blocking solution) for 1 h at 20°C and incubated with primary antibodies or normal rabbit serum diluted at 1:100 with blocking solution for 3 h at room temperature. After three washes with PBS, the slides were incubated at room temperature with fluorescein isothiocyanate-labeled anti-rabbit antibody (Vector Laboratories, Burlingame, CA) diluted at 1:500 with blocking solution for 1 h in the dark and analyzed with an AX-70 Olympus fluorescent microscope.

Isolation of the 5Ј-Flanking Region of the Human Claudin-2
Gene-The 5Ј-flanking region of the claudin-2 gene isolated from the human genomic library and T-84 cell-derived claudin-2 cDNA sequence were confirmed by sequence comparison with the human genomic clone AL158821. A BLAST search revealed that the gene encoding human claudin-2 is located on chromosome X, mapping to q22.3-23. The claudin-2 mRNA expressed in T-84 cells contains an open reading frame of 693 bp. Human and mouse claudin-2 have a high sequence identity of 87% on the mRNA level and 93% identity on the amino acid level.
Comparison with the mouse claudin-2 promoter in genomic data bases revealed that the promoters of the human and mouse claudin-2 genes possess a remarkable homology of 84% for the region of Ϫ1 to Ϫ400 (Fig. 1). The mouse claudin-2 cDNA (GenBank TM accession number AK004990) recovered by cap trapping revealed the putative transcriptional start site at 152 bp upstream of the translational start codon. The transcriptional initiation site is located within a consensus initiator element (Inr; NCANNNNN) (36,37).
The promoters of the human and mouse claudin-2 genes have no TATA box near the putative transcriptional initiation site ( Fig. 1). However, a CAAT box is located at Ϫ60 to Ϫ63 bp, and two E boxes (CANNTG) are located at Ϫ198 to Ϫ195 bp and Ϫ67 to Ϫ62 bp ( Fig. 1), suggesting that regulatory elements to initiate gene transcription are present.
To identify the regions involved in regulating claudin-2 gene transcription, sequentially deleted 5Ј-flanking regions (Ϫ1041, Ϫ393, Ϫ84, Ϫ62, and Ϫ31 to ϩ148) were cloned into the reporter plasmid pGL3B. Reporter constructs were transfected into intestinal epithelial cell line Caco-2, hepatic cell line HepG2, or fibroblast cell line NIH3T3. In Caco-2 cells the claudin-2 promoter fragments containing Ϫ1040 to Ϫ62 bp of the 5Ј-flanking region induced a 18 -29-fold increase in relative luciferase activity above that observed after transfection with the control null reporter construct (Fig. 2). In contrast, the same promoter regions achieved only a 7-11-fold increase when transfected into HepG2 and NIH3T3 cells (Fig. 2).
Removal of the putative AP-1 site and the NF-B site decreased the promoter activity slightly. Disruption of the HNF-1-binding site in the Ϫ62-bp construct did not alter the promoter activity significantly in Caco-2 cells. However, removal of the Cdx-binding sites resulted in a loss of promoter activity in Caco-2, HepG2, and NIH3T3 cells (Fig. 2).
Claudin-2 Promoter Activity Is Regulated by Cdx Homeodomain Protein Overexpression in Caco-2 Cells-To examine the function of the two Cdx sites, mutations were introduced into CdxA (Mut1), CdxB (Mut2), or both (Mut1 ϩ 2) (Fig. 3A). As shown in Fig. 3B, mutation in the CdxA (Mut1) or CdxB (Mut2) site decreased the promoter activity to 30 and 61% of that observed with Ϫ62 wild type construct, respectively. When both sites were mutated (Mut1 ϩ 2), promoter activity was decreased to 15% of the wild type construct, comparable with the Ϫ31 construct lacking both Cdx sites.
Next we determined the ability of Cdx1 and Cdx2 to activate the claudin-2 promoter. As shown in Fig. 4A, Cdx2 but not Cdx1 protein was detectable in nuclear proteins from Caco-2 cells. Transient expression with either Cdx1 or Cdx2 alone or in combination resulted in the strong expression of these proteins in the nuclei of Caco-2 cells 48 h after transfection (Fig. 4A). Ectopic expression of Cdx1 did not alter the expression level of Cdx2, nor did Cdx2 overexpression induce Cdx1 protein expression in the nuclei (Fig. 4A).
As shown in Fig. 4B, Cdx1 overexpression resulted in a 3.5-fold increase of the promoter activity driven by the Ϫ62 construct containing both intact Cdx sites in Caco-2 cells (92fold relative to the activity of null pGL3B vector). In contrast, Cdx2 overexpression increased the activity of the same construct up to 6.7-fold (177-fold of null pGL3B vector activity).
Overexpression of Cdx1 and Cdx2 together did not significantly increase the activity above the values achieved by Cdx2 alone (Fig. 4B). Although mutation of either the CdxA (Mut1) or CdxB (Mut2) site retained the ability to respond to Cdx2 overexpression, promoter activities induced by Cdx2 overexpression were less than 25 and 43% of that observed in Ϫ62 construct, respectively. Similarly, Mut1 and Mut2 constructs were less sensitive to Cdx1 overexpression. In the absence of both Cdx sites neither Cdx1 nor Cdx2 overexpression induced a significant induction of reporter gene transcription in Caco-2 cells (Fig. 4B). The ability of Cdx2 to induce a stronger induction of claudin-2 promoter activity in comparison with Cdx1 was specific for Caco-2 cells. As demonstrated in Fig. 4C, Cdx1 and Cdx2 enhanced claudin-2 promoter activity in fibroblasts 2.7-and 2.8-fold, respectively, whereas Cdx1 induced a 2.9-fold and Cdx2 induced a 6.7-fold higher promoter activity in Caco-2 cells.
Cdx-2 Binds to the Cdx-responsive Elements of the Human Claudin-2 Promoter-To further define the interaction between Cdx2 and the two Cdx sites, EMSA and supershifts were performed with nuclear proteins from Caco-2 cells. These experiments were carried out in post-confluent Caco-2 cells because it was shown that specific Cdx-DNA complexes can be obscured by unspecific binding of unknown peptides in nuclear proteins from preconfluent Caco-2 cells (22).
As shown in Fig. 5A, three DNA-protein complexes (A, B, and C) were observed when binding reactions were carried out with radiolabeled wild type oligonucleotide containing both intact Cdx sites (Fig. 5A, lane 1). Unlabeled Mut2 oligonucleotide with an intact CdxA but a mutated CdxB competed with the formation of all three complexes, whereas unlabeled Mut1 oli- gonucleotide with a mutated CdxA but an intact CdxB prevented only the formation of complex B (Fig. 5A, lanes 3 and 4). In supershift assays, anti-Cdx2 antibody shifted only complex A to reveal two distinct Cdx2-containing protein-DNA complexes (Fig. 5A, lane 7). In contrast, anti-Cdx1 antibody did not affect the mobility of the complexes (Fig. 5A, lane 6). The Cdx2-containing complex A was also formed with radiolabeled Mut2 oligonucleotide used as a probe (Fig. 5A, lane 10), suggesting that this complex is preferentially formed with CdxA. Complex B did not form on the Cdx sites because this complex was detected and consequently competed by all three mutated oligonucleotides (Fig. 5A, lanes 3-5, 9, 11, and 13). Mutation in the CdxA site greatly reduced the formation of complex C, suggesting the that formation of complex C is dependent on this site (Fig. 5A, lanes 8 and 12).
Within the SI gene promoter two adjacent Cdx consensus sites may be able to direct the formation of Cdx2 homodimers (18). We therefore further characterized the potential coordi-nation of Cdx2 binding by the two Cdx sites. In these experiments increasing amounts of nuclear proteins of Cdx2-transfected Caco-2 cells were used. Complex A, which was supershifted by the anti-Cdx2 antibody, was observed even in the absence of the CdxA site when more nuclear protein was used (Fig. 5B, lane 7 and 8). However, most of the Cdx2containing complexes formed in the presence of the CdxA site and did not require the CdxB site (Fig. 5B, lanes 2-4 and  10 -12). In addition, when 12 g of nuclear proteins from Cdx2transfected Caco-2 cells were used, an additional complex (complex D) was observed, which was shifted by anti-Cdx2 antibody (Fig. 5B, lanes 3, 4, 11, and 12). Although we could not visualize an additional supershifted band derived from complex D, it may correspond to Cdx2 homodimers, which could not be distinguished in supershifts from monomeric complexes (18).
HNF-1␣ Enhances Cdx2-mediated Activation of Human Claudin-2 Promoter in Caco-2 Cells-Cdx2 has been shown to regulate intestine-specific LPH gene expression in synergy with HNF-1␣ (21). We therefore determined whether the HNF-1 site in the human claudin-2 promoter could contribute to transcriptional regulation. We compared the effect of HNF-1␣ and HNF-1␤ overexpression, because both proteins share highly homologous DNA-binding domains but have distinct activation domains (42). As shown in Fig. 6A, Cdx1 and Cdx2 overexpression resulted in 3-and 5-fold increases of the promoter activity driven by Ϫ84 construct (100-and 170-fold relative to that of null pGL3B vector), respectively. However, transfection of either HNF-1␣ or HNF-1␤ alone was not able to increase promoter activity. In contrast, cotransfection of HNF-1␣ together with Cdx2 but not Cdx1 resulted in a 9-fold increase of the promoter activity (293-fold of the null pGL3B activity) (Fig. 6A). Disruption of the HNF-1 site in the Ϫ84 reporter construct prevented a synergistic cooperation of HNF-1␣ and Cdx2 (Fig. 6A, Ϫ62 construct). As shown in Fig.  6B, transfection of Caco-2 cells with HNF-1␣ expression constructs resulted in an increase of HNF-1␣ expression in both the cytosolic and nuclear protein fractions. In contrast, Cdx2 was exclusively expressed in nuclear protein fractions of Caco-2 cells even after ectopic expression (Fig. 6B).
HNF-1␣ Binds Its Recognition Sequence within the Human Claudin-2 Promoter-To further determine the interaction between HNF-1 proteins and the HNF-1-binding site in the human claudin-2 promoter, EMSA and supershifts were performed with nuclear proteins from Caco-2 cells. As shown in Fig. 7, a single DNA-protein complex was observed when nuclear proteins from mock-transfected Caco-2 cells was used (lane 1). The addition of 100-fold excess of unlabeled wild type but not mutant oligonucleotide prevented the formation of this complex (Fig. 7, lanes 2 and 3). The HNF-1 consensus sequence-protein complex was supershifted efficiently by anti-HNF-1␣ antibody but only to a small extent by anti-HNF-1␤ antibody (Fig. 7, lanes 4 and 5). Transfection with either HNF-1␣ or Cdx2 alone did not alter the formation of this complex (Fig. 7, lanes 6 and 7). In contrast, cotransfection with Cdx2 and HNF-1␣ together resulted in the increased formation of the complex, which was supershifted by the anti-HNF-1␣ antibody (Fig. 7, lanes 8 and 9).
HNF-1␣ Is an Organ-specific Regulator of Claudin-2 Expression-The in vitro experiments identified HNF-1␣ as a potential regulator of claudin-2 expression. Cdx1 and Cdx2 expression is restricted to the intestine, whereas HNF-1␣ is also a regulator of gene expression in the liver and kidneys, organs in which claudin-2 is expressed (1,5,42). In contrast to Cdx2deficient animals (43), HNF-1␣-deficient mice are viable and survive to adulthood (35,44). We utilized these mice to determine the potential contribution of HNF-1␣ in the expression of claudin-2 in different organs. Analysis of the claudin-2 mRNA and protein expression in these animals revealed that HNF-1␣ was required for expression of claudin-2 in the liver (Fig. 8A). Claudin-2 mRNA and protein expression was absent in the liver of HNF-1␣-deficient animals, whereas claudin-1 mRNA expression was unaffected (Fig. 8). In contrast, HNF-1␣ was not required for claudin-2 mRNA and protein expression in the kidneys (Fig. 8).
We next analyzed the expression of claudin-1 and claudin-2 along the cephalo-caudal and crypt-villus axes in wild type and HNF-1␣ deficient mice, because the in vitro experiments sug-gest the ability of HNF-1␣ to regulate claudin-2 expression in the presence of Cdx homeodomain proteins in intestinal epithelial cells. Densitometric analysis of Northern blots after normalization to GAPDH mRNA expression demonstrated that claudin-2 mRNA was differentially expressed along the cephalo-caudal axis. In wild type mice, claudin-2 mRNA was expressed at 17.2 Ϯ 1.7-fold higher levels in the ileum than in the jejunum (Fig. 9, A and B), in good agreement with the recent analysis of claudin-2 protein expression in the rat intestine (2). In contrast, claudin-1 mRNA was expressed at a 2.4 Ϯ 0.4-fold higher level in the jejunum than in the ileum (Fig. 9, A  and B). The claudin-2 mRNA expression pattern correlated with Cdx1 mRNA expression in the same intestinal segments, which increased 5 Ϯ 0.5-fold from the jejunum to the ileum (Fig. 9A). However, Cdx2 mRNA expression levels were similar in the jejunum and ileum (Fig. 9A). In the absence of HNF-1␣, claudin-2 expression decreased by 55 Ϯ 10% in the ileum (Fig.  9, A and B). This regulation was specific for claudin-2, because claudin-1 mRNA expression was not altered in the absence of HNF-1␣ in the mouse jejunum and ileum (Fig. 9, A and B).
The reduction of claudin-2 mRNA expression could be due to an overall reduction of claudin-2 gene transcription or a reduced expression in specific intestinal epithelial cell subsets. We therefore determined the expression and subcellular distribution of claudin-1 and claudin-2 proteins in the ileum by immunostaining along the crypt-villus axis in wild type and HNF-1␣-deficient mice (Fig. 9C). The claudin-2 protein was expressed in tight junctions of the crypt and villus epithelium of the ileum in wild type mice (Fig. 9C, panels B and G). In the absence of HNF-1␣, claudin-2 expression was restricted to the tight junctions of the crypt epithelium (Fig. 9C, panels A and F). Claudin-1 expression was not altered in the absence of HNF-1␣ and was observed in tight junctions of the crypt and villus epithelium of the ileum in both wild type and HNF-1␣deficient mice (Fig. 9C, panels C, D, H, and I). Incubation with rabbit control serum did not result in detectable immunoreactivity (Fig. 9C, panels E and J). DISCUSSION The diverse claudin family of tight junction-associated proteins has the potential of directing the variability of paracellular transport and barrier functions within gastro-intestinal organs (5). Recent evidence demonstrated that claudins are not only involved in the induction of tight junction formation but are also able to regulate water-and ion-specific paracellular transport mechanisms (9,10). Loss of claudin-16 results in the inability to absorb magnesium in the thick ascending limb of Henle (10). Claudin-4 expression resulted in the specific decrease in absolute sodium permeability, whereas claudin-2 appeared to increase paracellular conductance in kidney epithelial cells without changing the paracellular transport of inert compounds (9). The molecular mechanisms orchestrating the organ-specific expression of claudin-2 are unknown. In this report we provide the first insights into the transcriptional activation events, which regulate the complex expression pattern of claudin-2.
We demonstrate that the mouse and human claudin-2 promoters contain conserved binding sequences for Cdx homeodomain proteins and for the POU homeodomain family member HNF-1␣. Cdx1 and Cdx2, intestine-specific homeobox proteins, play an important role in the transcription of the intestinespecific expression of several genes such as SI (15), LPH (21), and guanylyl cyclase C (45). HNF-1␣ and HNF-1␤ were first identified as liver-enriched transcription factors involved in the expression of several plasma proteins, including albumin and clotting factors (49) and can act either as homodimers or heterodimers (26). There is increasing evidence that HNF-1␣ is crucial for the transcription of the intestine-specific genes such as LPH (21) and SI (29,50).
Our experiments provide the first demonstration that Cdx homeodomain proteins can initiate transcriptional activation of a TATA-less promoter. In contrast HNF family members have been demonstrated to activate tissue type-specific expression of Ksp cadherin (cadherin-16), which lacks TATA boxes (48).
Similar to the SI and LPH genes, the claudin-2 promoter has two putative Cdx-binding sites. The Cdx-binding site containing the region of the claudin-2 promoter mediated basal transcriptional activation in intestinal epithelial cells but also had activity in fibroblasts and HepG2 cells. Similar to our results, the consensus Cdx-binding site-containing promoter have been demonstrated to induce transcriptional activation in fibroblasts without Cdx proteins by undetermined mechanisms (47). This promoter region may comprise a core promoter that contains transcriptional elements sensitive to activation by factors in nonepithelial cells in the absence of Cdx and HNF-1␣. In addition to tissue-specifically expressed transcription factors like Cdx1 and Cdx2 silencer, binding upstream of the investigated promoter region may be necessary to direct tissue typespecific expression of claudin-2.
Both Cdx1 and Cdx2 can interact with the Cdx consensus sites within the claudin-2 promoter, although Cdx2 is the more potent activator of the claudin-2 gene transcription in Caco-2 cells. The stronger induction of claudin-2 promoter activity by Cdx2 in comparison with Cdx1 was specific for Caco-2 cells, suggesting that in these cells Cdx2 may cooperate with other factors enhancing its transcriptional activity. Our experiments identified HNF-1␣ as a potential candidate, because it was able to enhance Cdx2-but not Cdx1-induced claudin-2 promoter activity in Caco-2 cells.
Our results are consistent with the previous observation that Cdx2 is more effective than Cdx1 in transcriptional activation of the clusterin gene promoter (46). Although both Cdx-binding sites were required for full transcriptional activity of the human claudin-2 promoter in Caco-2 cells, Cdx2 binding occurred primarily at the CdxA site. The second CdxB site may serve primarily to support Cdx2 homodimer or oligomer formation, as has been proposed for the two Cdx sites in the SI promoter (18). Alternatively, additional transcription factors may require CdxB to bind and enhance Cdx2-mediated transcription. The involvement of additional transcriptional activators may be particularly necessary in the activation of the mouse claudin-2 promoter, in which the second Cdx-binding site present in the human promoter is not completely preserved.
In Caco-2 cells, HNF-1␣ was able to enhanced claudin-2 promoter activity only in the presence of overexpressed Cdx2. HNF-1␣ has been demonstrated to synergize with Cdx2 to induce LPH gene transcription (21). However, whereas the LPH promoter can be activated by the expression of HNF-1␣ alone, activation of the claudin-2 promoter by HNF-1␣ in Caco-2 cells was dependent on the recruitment of overexpressed Cdx2 to its binding site. This cooperation was specific for HNF-1␣ because HNF-1␤ failed to enhance Cdx2-mediated activation of the claudin-2 promoter. These results are similar to the previous observations that HNF-1␤ was less potent than HNF-1␣ as a transactivator of LPH (51), SI (29), and ␣ 1 -antitrypsin (52) genes. Collectively, the promoter analysis revealed the ability of Cdx homeodomain proteins and HNF-1␣ to bind to their recognition sequences in the claudin-2 promoter and to regulate the activation of this promoter in Caco-2 cells.
We analyzed wild type and HNF-1␣-deficient mice to assess the role of HNF-1␣ in the regulation of claudin-2 expression. These experiments indicate that HNF-1␣ can regulate claudin-2 expression in an organ-specific manner. HNF-1␣ was required for claudin-2 expression in the liver. HNF-1␣-deficient mice have enlarged fatty livers and dysregulated fatty acid homeostasis, which have been traced in part to a reduced expression of liver fatty acid-binding protein (53). It is currently not clear whether the lack of claudin-2 contributes to the disturbed liver function in HNF-1␣-deficient animals.
In contrast, claudin-2 mRNA and protein expression in proximal tubules of the kidneys were not altered in the absence of HNF-1␣. In the kidney HNF-3 may compensate for the lack of HNF-1␣ in the activation of the claudin-2 promoter. HNF-3 has recently been shown to mediate the kidney-specific expression of Ksp cadherin through a motif similar to that of the HNF-3-CAAT box-containing sequence found in the claudin-2 promoter partially overlapping with the HNF-1 consensus sequence (48).
In the absence of HNF-1␣, claudin-2 was still expressed in the small intestine, although its expression was restricted to the crypt epithelium. The loss of claudin-2 expression in intestinal villi epithelium may be due to the lack of HNF-1␣, which has been demonstrated to be predominantly expressed in the intestinal epithelial cells of the small intestinal villi (28).
HNF-1␣, Cdx1, and Cdx2 are differential expressed along the crypt-villus axis of the small intestine (28,31). Cdx1 expression has been demonstrated to localize to intestinal crypts, whereas Cdx2 expression was observed to extend into small intestinal villi (31). However, recent experiments with antibodies recognizing phosphorylated Cdx2 demonstrated activated Cdx2 in small intestinal crypts (54). If the regulation of claudin-2 expression in mice corresponds to its regulation in Caco-2 cells, HNF-1␣ may be required to enhance Cdx2-mediated claudin-2 expression in the intestinal villi, whereas Cdx1 and/or Cdx2 may drive the remaining expression of claudin-2 in the crypt epithelium of HNF-1␣-deficient mice.
The function of HNF-1␣ in the transcriptional regulation of claudin-2 expression was specific because claudin-1 expression was not regulated in the absence of HNF-1␣ in the jejunum or ileum. The different transcriptional regulation of claudin-1 was further apparent in the distinct expression pattern along the cephalo-caudal axis and the unaltered expression along the in crypt-villus axis in the absence of HNF-1␣. The impact of HNF-1␣ gene disruption on the gut has not been examined in detail. The loss of claudin-2 expression in the ileal villi and the liver may contribute to the severe phenotype of the HNF-1␣deficient mice. HNF-1␣ gene disruption in mice leads to dwarfism because of reduced insulin-like growth factor-1 synthesis and an early onset form of type 2 diabetes mellitus because of impaired glycolytic signaling (35,44,55). However, impaired intestine-and liver-specific secretive or absorptive function may relate to these phenotypes. Further analysis of the HNF-1␣-deficient mice should prove valuable to uncover additional roles of claudin-2 in the regulation of organ-specific functions.
Our studies suggest that the expression of claudin-2 is under FIG. 9. Claudin-1 and claudin-2 expression in the small intestine of HNF-1␣-deficient and wild type mice. A, Northern blot analysis of claudin-1, claudin-2, Cdx1, Cdx2, and HNF-1␣ mRNA expression in the jejunum and ileum of HNF-1␣-deficient (Ϫ/Ϫ) and wild type (ϩ/ϩ) mice. B, densitometric analysis of claudin-1 (open bars) and claudin-2 (black bars) mRNA expression in the presence or absence of HNF-1␣ gene. Expression levels of claudin-1 and claudin-2 mRNAs were normalized for GAPDH mRNA levels in the same RNA isolations and expressed as relative density/area (mean Ϯ S.D., n ϭ 3). C, immunostaining of claudin-2 and claudin-1 protein in the ileum of HNF-1␣-deficient (Ϫ/Ϫ) and wild type (ϩ/ϩ) mice. Frozen sections were stained with either anticlaudin-2 (panels A, B, F, and G) or anti-claudin-1 (panels C, D, H, and I) antibody and fluorescein isothiocyanate-labeled anti-rabbit secondary antibody. Panels E and J, control staining with rabbit serum and secondary antibody. The arrows indicate stainings of claudins in tight junctions. Original magnifications were 40ϫ in panels A-E and 100ϫ in panels F-J. the regulatory control of HNF-1␣ in the liver and small intestinal villi in mice. Whereas in the liver HNF-1␣ is required for claudin-2 expression, in the intestine HNF-1␣ may cooperate with additional factors to extend claudin-2 expression from the crypt into the functionally distinct villus intestinal epithelial cell compartment. It remains to be determined whether the augmentation of claudin-2 gene expression by HNF-1␣ in this compartment is dependent on Cdx2 as observed in Caco-2 cells. Together our experiments support a model in which claudin-2 expression is governed by distinct organ-specific transcriptional mechanisms involving homeodomain proteins.