|Home | About | Journals | Submit | Contact Us | Français|
Hepatocyte nuclear factor 4α (HNF-4α) is a transcription factor which is highly expressed in the intestinal epithelium from duodenum to colon and from crypt to villus. The homeostasis of this constantly renewing epithelium relies on an integrated control of proliferation, differentiation, and apoptosis, as well as on the functional architecture of the epithelial cells. In order to determine the consequences of HNF-4α loss in the adult intestinal epithelium, we used a tamoxifen-inducible Cre-loxP system to inactivate the Hnf-4a gene. In the intestines of adult mice, loss of HNF-4α led to an increased proliferation in crypts and to an increased expression of several genes controlled by the Wnt/β-catenin system. This control of the Wnt/β-catenin signaling pathway by HNF-4α was confirmed in vitro. Cell lineage was affected, as indicated by an increased number of goblet cells and an impairment of enterocyte and enteroendocrine cell maturation. In the absence of HNF-4α, cell-cell junctions were destabilized and paracellular intestinal permeability increased. Our results showed that HNF-4α modulates Wnt/β-catenin signaling and controls intestinal epithelium homeostasis, cell function, and cell architecture. This study indicates that HNF-4α regulates the intestinal balance between proliferation and differentiation, and we hypothesize that it might act as a tumor suppressor.
Hepatocyte nuclear factor 4 (HNF-4) belongs to the superfamily of nuclear receptors (8). In mammals, it is represented by two paralogs: HNF-4α, which is expressed in the liver, kidneys, pancreas, and intestine, and HNF-4γ, which is expressed in the same organs except the liver (14, 42, 50). HNF-4γ has not been extensively studied. However, mice lacking expression of the Hnf-4g gene do not present an overt phenotype (19). HNF-4α has been essentially studied in liver and hepatocyte cell lines. Mutations in the human Hnf-4a gene are associated with maturity onset diabetes of the young (MODY1), a disorder that is characterized by the early onset of type 2 diabetes (20). Disruption of Hnf-4a gene expression leads to embryonic death at the gastrulation stage (embryonic day 6.5) (15). HNF-4α is dispensable for hepatic specification but necessary for hepatoblast differentiation into hepatocytes (29). Inactivation of the Hnf-4a gene in fetal liver revealed that HNF-4α also controls epithelial morphology (41), and transcriptome analysis of Hnf-4a knockout in fetal liver showed that genes encoding proteins of all categories of cell junctions (adherens, desmosomes, tight) are targets of HNF-4α (5). Expression of HNF-4α in dedifferentiated hepatocytes provokes the reexpression of epithelial markers, such as E-cadherin (48). Its ectopic expression in mesenchymal cells induces the expression and the proper localization of tight and adherens junction proteins (10, 41), as well as the formation of microvilli (12), suggesting that HNF-4α is a central regulator of epithelial morphogenesis. In adult mouse liver, conditional disruption of the Hnf-4a gene revealed a key role for HNF-4α in the control of genes involved in amino acid, glucose, and lipid metabolism (21). All together, these studies show that HNF-4α is at the crossroads between liver morphogenesis and function (43). By contrast, the role of HNF-4α in the intestine is less well understood.
The intestinal epithelium establishes a physical barrier between the external and internal compartments and plays a key role in the absorption and transfer of nutrients. These functions, in constantly renewing tissue, rely on the functional architecture of the epithelial cells, as well as on the complex and precise control of proliferation, differentiation, and apoptosis to ensure the homeostasis of the intestinal epithelium (46). The permanent renewal of epithelial cells is accomplished by the stem cells located at the bottom of the crypts that generate four differentiated cell types in strictly controlled proportions: Paneth cells, goblet cells, enteroendocrine cells, and the absorptive enterocytes which represent 90% of the intestinal epithelial cells. The last three cell types differentiate during their migration upward from crypts toward the tips of the villi, where they die by apoptosis and are shed into the lumen (17, 33), the entire process being completed within 3 to 5 days in mice for enterocytes. The fourth cell type, i.e., Paneth cells, is located in the bottom of the crypts. The precise molecular mechanisms sustaining these proliferation/differentiation/apoptosis events are not well understood. Besides cell-cell contacts and epithelium-mesenchyme interactions, different transcription factors have been implicated in these events, among which are Cdx1, Cdx2, and GATA4, -5, and -6. Results from our group (2, 9, 45) and from studies based on transcriptome, metabolome, and bioinformatic analyses have suggested an important role for HNF-4α in the regulation of the enterocyte phenotype (30, 35, 49). Recently, it has been shown that HNF-4α is essential for the normal embryonic development of the mouse colon (18).
The aim of the present study was to determine the role of HNF-4α in the adult mouse small intestine. Using an inducible and tissue-specific Cre-loxP system to circumvent the early embryonic lethality of HNF-4α knockout mice, we generated adult mice lacking HNF-4α in the intestinal epithelium and demonstrated that HNF-4α plays a pivotal role in the homeostasis of the intestinal epithelium, in the epithelial cell architecture, and in the barrier function of the intestine.
Hnf-4aloxP/loxP mice harboring an Hnf-4a allele in which exons 4 and 5 were flanked by loxP sites (Fig. (Fig.1A)1A) have been described previously (21). The tamoxifen-dependent 9-kb-villin-Cre-ERT2 recombinase mouse transgenic line (16) was used to produce inducible and intestinal epithelium-specific inactivation of the Hnf-4a gene. villin-Cre-ERT2 mice were mated with Hnf-4aloxP/+ mice to obtain Hnf-4aloxP/loxP; villin-Cre-ERT2 mice. Mice were genotyped by PCR of tail biopsy DNA with previously described primers (16, 21). The Cre recombinase activation in intestinal epithelial cells was induced by daily intraperitoneal injection of 600 μg/15 μl (ethanol/phosphate-buffered saline [PBS], vol/vol) tamoxifen (Sigma) for 5 days as previously described (53). Mice were euthanized on day 8 following the first injection. Two strains were utilized in this study, Hnf-4aloxP/loxP as control mice and Hnf-4aloxP/loxP; villin-Cre-ERT2 as Hnf-4aintΔ mice (where int stands for intestinal and Δ stands for deletion), both receiving tamoxifen treatment. The recombined allele was detected with the R primer described in reference 21 and the Δbis primer 5′-TGCTCCGTAGGAAGTCACAGG-3′ (Fig. (Fig.1A)1A) after rough scraping of the intestine to prepare the epithelial sample for genotyping. Experimental mice were 2-month-old males fed a standard diet ad libitum. The animal care and experimental procedures used in this study conform to the French guidelines for animal studies.
Mice were euthanized, and their intestines were removed and flushed gently with PBS. The small intestine was taken out after the duodenum and cut into three equal parts named i1, i2, and i3 (Fig. (Fig.1B).1B). For histological analyses, pieces of the proximal jejunum (1 cm) (Fig. (Fig.1B)1B) were immediately embedded in tissue-tek or fixed overnight at 4°C in alcohol-formalin-acetic acid before being embedded in paraffin. Immunostaining with anti-chromogranine A, anti-bromodeoxyuridine (anti-BrdU), and anti-caspase 3 antibodies was performed with 5-μm paraffin sections. Histological colorations and other immunostainings were performed with 6-μm cryosections. Periodic acid-Schiff (PAS) (32) staining was performed by using a standard histological protocol (28). Immunostaining was performed as previously described (38). The primary antibodies used were goat anti-HNF-4α (1/250), goat anti-HNF-4γ (1/5,000) (C-19 and C-18, respectively; Santa Cruz Biotechnology), rabbit anti-lysozyme (1/200) (A0099; Dako), rabbit anti-active caspase 3 (1/100) (557035; BD Pharmingen), rabbit anti-chromogranin A (1/200), rabbit anti-Ki67 (1/100) (ab45138 and ab15580, respectively; Abcam), rat anti-BrdU (1/50) (ab6326; Abcam), rabbit anti-EBP50 (1/100) (PA1-090; Affinity Bioreagents), and mouse anti-E-cadherin (1/250) (ECCD-2; Zymed Laboratories). The secondary antibodies were Alexa-fluor-conjugated donkey anti-goat 546, goat anti-rat 488, and donkey anti-rabbit 488 (Molecular Probes). Nuclear counterstaining was performed with 4′,6-diamidino-2-phenylindole (DAPI). Immunostaining was examined by epifluorescence microscopy (Axiophot microscope connected to an Axiocam camera using the Axiovision 4.5 software; Carl Zeiss). The β-catenin staining was performed according to standard procedures described previously (44), with mouse anti-β-catenin at 1/50 (610154; Transduction Laboratories). A horseradish peroxidase-labeled secondary antibody (Amersham Biosciences) and 3,3′-diaminobenzidine were employed for revelation.
Fragments of the proximal jejunum (1 cm) were prepared as previously described (38). Observations were made with a JEOL CX100 electron microscope equipped with a Gatan digital camera (3.11.0), and the micrographs were processed with Gatan software.
Eight days after the first tamoxifen injection, mice were injected intraperitoneally with 120 mg BrdU (Sigma) per kg of body weight and sacrificed 90 min later. Paraffin sections of alcohol-formalin-acetic acid-fixed jejunum were incubated for 30 min in 2.5 N HCl before processing for immunostaining with anti-BrdU antibody (Abcam). Antigen retrieval was performed by boiling slides in 10 mM citrate buffer (pH 6) for 10 min. After washes in PBS, immunostaining was done as described above, with a 1/50 dilution of a BrdU-directed monoclonal antibody.
The epithelium from villi and crypts was isolated as previously described (2, 45), with Cell Recovery Solution (BD Biosciences) and a chelating buffer, respectively. The enrichment of epithelial cells of villi versus the epithelial cells of crypts, and reciprocally, was measured by the mRNA levels of apoA4, a gene specific for enterocytes and those of Lgr5, a gene specifically expressed in the stem cells of crypts. The ratio of the mRNA levels in the two compartments expresses the enrichment factor. The level of apoA4 mRNA is 7-fold higher in the villus fraction than in the crypt fraction, and that of Lgr5 mRNA is, reciprocally, 7.36-fold higher in the crypt fraction than in the villus fraction, indicating a 7-fold enrichment in each fraction. Normalization was done with cyclophilin mRNA.
Total RNA from villus or crypt epithelial cells from the mouse jejunum (i1) was isolated by using RNAplus reagent (qBiogene) according to the manufacturer's instructions. Reverse transcription (RT) was performed with 1 μg of RNA in a 20-μl reaction mixture. Semiquantitative real-time PCR was performed with the LightCycler system by using SYBR green according to the manufacturer's procedures (Roche Molecular Biochemicals). The sequences of the primers used are reported in Table Table1.1. After tamoxifen injection, the activated Cre recombinase deleted exons 4 and 5 of the Hnf-4a gene. The loss of normal HNF-4α mRNA was quantified with primers located in exons 3 and 5. Results are expressed as the ratio of the mRNA of interest to cyclophilin mRNA.
Total proteins from the villus epithelial cells from mouse jejunum (i1) were isolated with a lysis buffer (20 mM Tris, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 0.5% sodium deoxycholate) containing a protease inhibitor cocktail (Sigma). Nuclear and cytoplasmic proteins from villus or crypt epithelial cells of the jejunum were isolated with the NE-PER nuclear and cytoplasmic extraction reagent kit (Pierce). Protein concentrations were measured with the Bio-Rad DC protein assay (Bio-Rad). For sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis, samples with equal amounts of protein were boiled for 10 min in SDS-reducing buffer and separated on a 0.1% SDS-containing polyacrylamide gel. Proteins were transferred onto nitrocellulose membranes and probed with primary antibodies. The primary antibodies used were goat anti-HNF-4α (1/5,000), rabbit anti-p21 (1/1,000) (C-19 and H-164, respectively; Santa Cruz Biotechnology), rabbit anti-claudin 7 (1/2,000), rabbit anti-claudin 2 (1/2,000), rabbit anti-ZO-1 (1/100) (34-9100 32-5600 and 61-7300, respectively; Zymed Laboratories), rabbit anti-EBP50 (1/2,000) (PA1-090; Affinity Bioreagents), rat anti-E-cadherin (1/2500) (ECCD-2; Takara), mouse anti-β-catenin (1/2,000) (610154; BD Transduction Laboratories), mouse anti-actin (1/2,000) (MAB1501R; Chemincon), rabbit anti-LDH (lactate dehydrogenase) (1/20,000) (ab52488; Abcam), and rabbit anti-SP1 (1/1,000) (PEP-2; Santa Cruz Biotechnology). A horseradish peroxidase-labeled secondary antibody (Amersham Biosciences) was used and detected by the chemiluminescence method (ECL; Amersham Biosciences). The quantitative analyses were performed with a high-performance calibrated imaging densitometer (Bio-Rad GS-800) by using PD Quest and ImageQuant 5.2 software. Results are expressed as the ratio of the protein of interest to Sp1 for nuclear proteins or actin for cytoplasmic proteins.
In vivo paracellular permeability was measured as previously described after mice were given an oral bolus of fluorescein isothiocyanate-labeled dextran 4 kDa (FITC-dextran; Sigma) of 60 mg/100 g body weight (40). The paracellular permeability of the ileal mucosa was assessed in 0.196-cm2 Ussing chambers by measuring the mucosal-to-serosal flux of FITC-dextran as previously described (4). FITC-dextran was assayed by fluorimetry (excitation wavelength, 490 nm; emission wavelength, 520 nm). Mucosal-to-serosal flux was expressed as picomoles per hour per square centimeter of ileal mucosa.
The vectors encoding rat HNF-4α2 (pMT2-HNF-4α) and β-galactosidase (pRSV-β-Gal) were as previously described (2). The T-cell factor (TCF)-responsive TOPflash vector (Millipore) expressing luciferase driven by multiple TCF-responsive elements was utilized to evaluate the activity of β-catenin. TCF-responsive elements were mutated into the FOPflash vector (Millipore) used as a negative control. HCT116 human colorectal cancer cells were plated into 12-well plates at 105 per well and grown in Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum (Invitrogen) maintained at 37°C in a humidified, 5% CO2-containing atmosphere. Transfection was performed 72 h after plating by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were transfected with 200 ng of TOP or FOP vector along with 300 ng RSV-βGal plasmid and increasing amounts of PMT2-HNF-4α. Cells were then harvested at 48 h posttransfection. β-Galactosidase and luciferase assays were performed with a luminometer and a microplate reader. Luciferase activity was normalized to β-galactosidase activity.
The plasmid encoding Flag-Tcf4 was a generous gift from M. Idogawa (23, 24). Caco-2/TC7 cells were cultured in Dulbecco modified Eagle medium supplemented with 20% heat-inactivated fetal bovine serum and antibiotics. HCT116 cells were cultivated and transfected with plasmid encoding HNF-4α and Flag-Tcf4 as described previously. Cells were lysed on ice in a mixture of 50 mM Tris HCl (pH 8.0), 150 mM NaCl, 5 mM EGTA (pH 8.0), 50 mM NaF (pH 8.0), 10% glycerol, 1.5 mM MgCl2, and 1% Triton X-100 containing freshly added protease inhibitor cocktail (Sigma). Lysates were clarified by centrifugation at 4°C; protein concentrations were determined with a bicinchoninic acid protein quantification kit (Interchim). Immunoprecipitations were performed overnight at 4°C with rabbit, mouse, and goat control immunoglobulins G (IgG; Santa Cruz Biotechnology), mouse anti-β-catenin (610154; BD Transduction Laboratories), rabbit anti-HNF-4 (H-171; Santa Cruz Biotechnology), mouse anti-Flag (M2; Sigma), or rabbit anti-Tcf4 (H-125; Santa Cruz Biotechnology) antibodies. Five micrograms of each antibody or control IgG was used for 1 mg of transfected HCT116 and CaCo-2/TC7 cell extracts. Immunocomplexes were collected on protein A-Sepharose beads and protein G-Sepharose beads (GE Healthcare) during the last 45 min. The beads were pelleted by centrifugation, washed three times with a mixture of Tris HCl (pH 8.0), 150 mM NaCl, 5 mM EGTA, and 1% NP-40 and then boiled for 10 min in SDS sample buffer. Western blots were probed with goat anti-HNF-4α (1/5,000) (C-19; Santa Cruz Biotechnology), goat anti-Tcf4 (1/2,000) (N-20; Santa Cruz Biotechnology), and mouse anti-β-catenin (1/2,000).
Results were expressed as means ± the standard errors of the means. The statistical significance of differences was determined by an unpaired Student t test by using Excel software.
To examine the role of HNF-4α in the intestinal epithelium of adult mice, floxed Hnf-4a mutant mice (21) containing two loxP sites flanking exons 4 and 5 of the Hnf-4a gene (Hnf-4aloxP/loxP mice) (Fig. (Fig.1A,1A, top) were crossed with the villin-Cre-ERT2 line (16) to generate Hnf-4aloxP/loxP; villin-Cre-ERT2 mice. The expression of HNF-4α was quantified by semiquantitative RT-PCR on mRNA isolated from 2-month-old wild-type mice, Hnf-4aloxP/loxP mice, and Hnf-4aloxP/loxP; villin-Cre-ERT2 mice without CRE recombinase activation. Whatever the genotype, the levels of Hnf-4a mRNA were identical and we demonstrated an absence of hypomorphism for Hnf-4a gene expression. There was no detectable background recombination in the presence of the villin-Cre-ERT2 transgene and no phenotypic differences between these strains. The different genotypes were equally viable and were obtained at the expected ratios in all crosses (data not shown).
The CRE recombinase activity was induced by daily injection of tamoxifen on 5 consecutive days in 2-month-old mice. The deletion of exons 4 and 5 by the induction of CRE recombinase activity in Hnf-4aloxP/loxP; villin-Cre-ERT2 mice generated Hnf-4aintΔ mice (Fig. (Fig.1A,1A, top). Hnf-4aloxP/loxP mice, i.e., mice lacking the Cre-ERT2 transgene, were used as controls and injected with tamoxifen under the same conditions as the Hnf-4aintΔ mice. Mice were euthanized on day 8 following the first injection, a time interval corresponding at least to one complete intestinal epithelium renewal cycle. The removal of exons 4 and 5 was specific for all segments of the intestine and was not observed in the other organs expressing HNF-4α, i.e., the liver, kidneys, stomach, or pancreas (Fig. (Fig.1A,1A, bottom). The inactivation was very efficient, as the drop in Hnf-4a mRNA expression reached 90% and 87% in the villus and crypt epithelium of the jejunum, respectively (Fig. (Fig.1C).1C). Similar results were obtained in the ileum and colon (data not shown). Accordingly, the level of HNF-4α protein was decreased by 90% in the jejunum epithelium of Hnf-4a intestine-null mice (Fig. (Fig.1D).1D). As previously shown for wild-type mice (45), HNF-4α was expressed in the nuclei of all epithelial cells along the crypt-to-villus axis in control Hnf-4aloxP/loxP mice (Fig. (Fig.1E).1E). HNF-4α was not detectable in epithelial cells of Hnf-4aintΔ mice (Fig. (Fig.1E)1E) and remained at very low levels up to 30 days after tamoxifen injections (Fig. (Fig.2A).2A). Together, these results indicated that injections of tamoxifen resulted in an intestine-specific recombination of the floxed Hnf-4a gene in Hnf-4aloxP/loxP; villin-Cre-ERT2 mice and that the loss of HNF-4α protein was observed in all epithelial cells along the crypt-to-villus axis.
Histological examination of intestinal sections from Hnf-4aloxP/loxP and Hnf-4aintΔ mice showed that the general morphology of the intestinal epithelium was conserved in Hnf-4aintΔ mice (Fig. (Fig.3A)3A) but that the depths of the crypts were significantly increased in the Hnf-4aintΔ mice (Fig. 3B and C), whereas the sizes of the villi remained similar in the two groups of mice. Within the crypt compartment of Hnf-4aintΔ mice, the size of the cells appeared normal, suggesting that crypt hypertrophy was the consequence of hyperplasia and of aberrant proliferation. To test this hypothesis, Hnf-4aintΔ and Hnf-4aloxP/loxP mice were injected with BrdU and euthanized 90 min later. The average number of BrdU-positive cells per crypt was twofold higher in Hnf-4aintΔ mice than in Hnf-4aloxP/loxP control mice (Fig. 3D and E). In the colon, where the inactivation of HNF-4α was efficient, the number of proliferative cells was also increased (Fig. 4A and B). The proliferative index (i.e., the ratio of the number of proliferative cells to the total number of cells in the crypt) was higher in Hnf-4aintΔ mice (Fig. (Fig.3F),3F), suggesting a loss of intestinal homeostasis. Immunostaining of the Ki67 protein confirmed the increased epithelial proliferation in crypts of Hnf-4aintΔ mice (Fig. (Fig.3G).3G). As the length of the villi was not affected, we measured the frequency of apoptosis by immunostaining of the activated caspase 3 in the intestinal epithelium. Hnf-4aintΔ epithelium presented fourfold more activated-caspase 3-positive cells than that of Hnf-4aloxP/loxP controls (Fig. 3H and I). This observation could explain the normal length of the villi despite the increased number of proliferative cells.
Interestingly, 30 days after tamoxifen injection, the observed effects on crypt size and proliferation were still significant but less pronounced than those observed 8 days after tamoxifen injection (Fig. 2B, C, and D).
The canonical Wnt pathway was shown to play an important role in maintaining the proliferation capacity of the intestinal epithelium (46). We observed an enhancement of intracellular β-catenin staining in crypts of Hnf-4aintΔ mice compared to that in Hnf-4aloxP/loxP control mice (Fig. (Fig.5A).5A). We further analyzed the level of nuclear and cytoplasmic β-catenin in crypt cells. The amount of cytoplasmic β-catenin was unchanged (Fig. (Fig.5B,5B, top), whereas that of nuclear β-catenin was increased by 50% (Fig. (Fig.5B,5B, bottom) in Hnf-4aintΔ mice compared to that in Hnf-4aloxP/loxP control mice. The cytoplasmic and nuclear fraction purity was assessed through LDH (cytoplasmic) and Sp1 (nuclear) immunoblot assays. Contamination of the nuclear fraction by cytoplasm was negligible (Fig. (Fig.5B).5B). The accumulation of cytoplasmic and nuclear β-catenin is the hallmark of active Wnt signaling; we analyzed the expression of several classical target genes of Wnt pathway activation in crypt-enriched epithelial cells. The expression of Tcf-4, axin 2, c-Myc, and Lgr-5 was significantly increased in Hnf-4aintΔ mice (Fig. (Fig.5C).5C). In parallel, the expression of the cell cycle inhibitor p21 was significantly decreased in crypt-enriched (data not shown) and villus-enriched epithelial cells from Hnf-4aintΔ mice at both the mRNA (Fig. (Fig.5C)5C) and protein (Fig. (Fig.5D)5D) levels. No significant change was observed in the expression of p27 (data not shown) or cyclin D1 (Fig. (Fig.5C),5C), which are also regulators of the cell cycle.
To further address the relationships between HNF-4α and the Wnt/β-catenin pathway, we transfected the TOP/FOP luciferase reporter vectors, a TCF/β-catenin-directed transcription system (47), together with increasing amount of an HNF-4α-expressing vector, into the HCT116 colon carcinoma cell line, which does not express endogenous HNF-4α (22). HNF-4α strongly (80%) decreased the β-catenin/TCF-dependent luciferase activity in a dose-dependent manner (Fig. (Fig.5E).5E). An interaction between HNF-4α and Tcf-4 or β-catenin has been investigated. Through coimmunoprecipitation experiments with HCT116 (Fig. (Fig.6A)6A) and Caco-2/TC7 (Fig. (Fig.6B)6B) cells, we showed that HNF-4α interacts with Tcf-4 but not with β-catenin. All together, these results, which were obtained in vivo and in vitro, indicated that HNF-4α interfered with the Wnt/β-catenin pathway to control the proliferation of crypt epithelial cells.
We analyzed the effect of HNF-4α loss on the intestinal epithelial cell lineage. The four differentiated cell types were reported to originate from intestinal stem cells. Activation of the Notch signaling pathway is known to induce Hes-1 gene expression and commitment of epithelial cells to the absorptive lineage, leading to differentiated enterocytes, through repression of Math-1, encoding a transcription factor necessary for cell commitment to the secretory lineage (46). In Hnf-4aintΔ mouse crypts, we observed a 1.8-fold increase in the Math-1 mRNA level and no significant modification of the Hes-1 mRNA level (Table (Table22).
Consistent with this observation, the average number of goblet cells detected by PAS (32) staining was significantly increased (1.3-fold in villi and 1.7-fold in crypts) in Hnf-4aintΔ mice (Fig. 7A and C). The mRNA encoding trefoil factor 3, a marker of differentiated goblet cells, was also increased, as well as that of Muc2, the major gel-forming mucin in goblet cells (Table (Table2).2). Goblet cells were mature, as confirmed by PAS staining and electron microscopy (Fig. 7A and B). The loss of HNF-4α affected neither the number of Paneth cells stained with an anti-lysozyme antibody (Fig. (Fig.7F)7F) nor the level of MMP-7 or Sox-9 mRNA (Table (Table2).2). The cryptdin-1 mRNA level was only moderately affected (Table (Table2).2). We examined and counted the enteroendocrine cells after immunostaining of chromogranin A (Fig. 7D and E). In Hnf-4aintΔ mice, we observed a 1.4-fold decrease in the enteroendocrine cell number in villi and a 1.4-fold increase in crypts, resulting in an unchanged total (crypts plus villi) enteroendocrine cell number. The expression level of genes encoding the gut hormones glucose-dependent insulinotropic peptide, somatostatin, and glucagon was significantly decreased (Table (Table2),2), suggesting that the corresponding endocrine subtypes were reduced. The transcription factors Ngn-3 and NeuroD were reported to be involved in enteroendocrine specification and differentiation (31). The mRNA encoding Ngn-3 was significantly increased in Hnf-4aintΔ mice, whereas that encoding NeuroD remained unchanged (Table (Table22).
Our previous work (2, 9) and a bioinformatic analysis (49) showed that, in the intestine, HNF-4α influences enterocyte function by regulating genes involved in lipid metabolism. Thus, we analyzed the expression of several genes involved in the function of lipid transfer in enterocytes. In Hnf-4aintΔ mice, there was a significant reduction in the mRNAs coding for microsomal triglyceride transfer protein, apolipoproteins B and AIV, and intestinal fatty acid binding protein (Table (Table22).
In villi, the expression of some transcription factors such as HNF-1α and Cdx2 was unaffected by the loss of HNF-4α, whereas that of Klf4, a Krüppel-like factor expressed in terminally differentiated mucosecretory cells, was decreased by 30% (Table (Table2).2). In crypts, HNF-1β (or Tcf2) expression was significantly increased (Table (Table2).2). Notably, the absence of HNF-4α induced a strong expression of HNF-4γ in crypts (Table (Table22 and Fig. Fig.8A)8A) which was still observed 30 days after tamoxifen injection (Fig. (Fig.8B)8B) without modifications of HNF-4γ expression in villi (Table (Table22).
All together, these results showed that intestinal disruption of the Hnf-4a gene perturbed the homeostasis of the intestinal epithelium lineage and the terminal differentiation of enterocytes and of some endocrine subtypes.
HNF-4α has been shown to be essential for morphological differentiation of hepatocytes by regulating the expression of cell-cell junction-associated proteins (41). We thus investigated the impact of intestinal HNF-4α loss on the cell architecture of adult small intestine epithelium and particularly on the organization of the cell-cell junctions. E-cadherin (adherens junctions) mRNA and protein expression was unchanged in Hnf-4aintΔ mice (Fig. 9A and B), but the protein was not properly located to membranes and was abundantly present in the cytoplasm (Fig. (Fig.9C).9C). EBP50 (ERM [ezrin-radixin-meosin] binding protein, also named Na+/H+ exchanger regulatory factor 1 or NHERF1) is a scaffolding protein which organizes these ERM proteins at the apical membrane (37), and the gene that encodes it is also a target of HNF-4α (12). Accordingly, the mRNA and protein levels of EBP50 were significantly decreased in Hnf-4aintΔ mice (Fig. 9D and E).
The expression of some genes encoding junctional proteins was further analyzed. Desmocollin 2 and plakophilin 2 (desmosomes) remained unchanged in Hnf-4aintΔ mice, whereas the mRNA encoding Gjb1 (gap junctions) was 1.7-fold increased in mutant mice (Fig. 10A). Important modifications in the expression of tight junction proteins were observed in Hnf-4aintΔ mice, namely, a significant decrease in claudins 4 and 7 and ZO-1 mRNA levels associated with a marked increase in that of claudin 2 (Fig. 10A). These variations correlated with those observed at the protein levels (Fig. 10B and C). Examination of epithelial cell-cell contacts at the ultrastructural level showed that the intercellular space of tight junctions was 1.7-fold wider in hnf-4αintΔ mice than in control mice (Fig. 11A). This structural change might account for the increased paracellular permeability that we quantified both in vivo (Fig. 11B) and ex vivo in Ussing chambers by measuring the mucosal-to-serosal flux of FITC-dextran (Fig. 11C).
The results presented here show that the nuclear receptor HNF-4α plays a critical role in the homeostasis of intestinal epithelium, in the epithelial cell architecture, and in the barrier function of the intestine. We report that loss of HNF-4α in the adult small intestine affects the Wnt signaling pathway, leads to enhanced proliferation in crypts, and induces an increased number of mucosecretory cells. In absorptive cells, HNF-4α exerts a major role in the activation of genes involved in lipid transfer by fully differentiated enterocytes (6). Moreover, loss of HNF-4α also disturbs the epithelial cell architecture by modifying the expression of proteins involved in cell-cell adhesion complexes, specifically in the tight junctions, leading to increased in vivo paracellular permeability in Hnf-4aintΔ mice. From these results, we hypothesize that HNF-4α plays dual roles in the adult intestinal epithelium: HNF-4α controls the functional differentiation of epithelial cells and finely modulates the homeostasis of the epithelium.
The importance of HNF-4α as a master regulator of development and differentiation has been established in the liver (15, 29, 41) and in the developing embryonic colon (18). In addition, increasing evidence suggests that HNF-4α plays a role in regulating proliferation. Indeed, HNF-4α was shown to inhibit cell proliferation in vitro (11, 22, 32) and to slow the progression of hepatocellular carcinoma in mice (27). We report here that loss of HNF-4α in the adult intestinal epithelium results in increased cell proliferation in the crypt compartment. This is correlated with downregulation of p21 and overexpression of c-Myc. It was recently reported that HNF-4α increases the expression of the p21 gene by competing with c-Myc for the p21 promoter (11, 22), a mechanism that could account for the downregulation of p21 expression observed in Hnf-4aintΔ mice. The canonical Wnt pathway was shown to play an important role in maintaining the proliferation capacity of the intestinal epithelium (46). Tcf-4, whose intestinal inactivation leads to an absence of proliferating cells, is the most prominent effector of Wnt/β-catenin signaling in the gastrointestinal tract (25, 51). The expression of Tcf-4 is upregulated in the absence of HNF-4α, concomitantly with that of β-catenin/Tcf-4 target genes, c-Myc, axin 2, Lgr-5, and claudin 2. Furthermore, we demonstrate that HNF-4α is able to decrease Wnt-Tcf4/β-catenin-driven transcription activity. This modulation of the Wnt signaling pathway could be due to an interaction between HNF-4α and Tcf-4 (7) or β-catenin, as shown for other members of the nuclear receptor family (39). Through immunoprecipitation experiments with Caco-2/TC7 and HCT116 cells, we showed that HNF-4α interacts with Tcf-4 but not with β-catenin. One can hypothesized that HNF-4α acts in the crypt to sequester Tcf-4 and thereby represses its transcriptional activity. All together, these observations imply that HNF-4α could regulate the balance between proliferation and differentiation by interfering directly with the Wnt signaling pathway in the intestinal epithelium.
We show here that disrupting Hnf-4a gene expression in the intestinal epithelium of adult mice is associated with a marked increase in mucus-secreting cells. In the intestinal epithelium, cells commit either to the absorptive lineage when the Notch signal is activated or to the secretory lineage (enteroendocrine, goblet, and Paneth cells) when the Notch signal is off (46). Math-1, which is repressed by the Notch signaling effector Hes-1, is expressed at an early progenitor stage in cells that are committing to the secretory lineage. Hnf-4aintΔ mice show a marked increase in Math-1 expression, which might explain the higher number of mucus-secreting cells, as well as the upregulation of their differentiation markers trefoil factor 3 and Muc2. Inducible inactivation of the Notch signaling pathway in the adult intestinal epithelium leads to increased expression of Math-1 with a massive conversion of proliferative crypt cells into goblet cells without impact on the other secreting cells (52). Consistent with these observations, there is no global change in the number of enteroendocrine and Paneth cells in Hnf-4aintΔ mice. However, the number of enteroendocrine cells along the crypt-to-villus axis reveals a paradox: compared with controls, Hnf-4aintΔ mice exhibit more enteroendocrine cells in crypts and fewer in villi, as well as increased expression of Ngn-3, which is involved in enteroendocrine specification (31), and decreased expression of some differentiation markers such as glucagon, glucose-dependent insulinotropic peptide, and somatostatin. We hypothesize that the higher level of Ngn-3 mRNA in Hnf-4aintΔ mice drives more cells toward the enteroendocrine lineage but that the absence of HNF-4α prevents their full terminal differentiation. Commitment to the absorptive lineage is driven by the Notch signal and the transcriptional factor Hes-1. In Hnf-4aintΔ mice, the expression of Hes-1 and the number of enterocytes, estimated by the size of villi, are unchanged. Thus, the main role of HNF-4α in the absorptive cells is to control genes specifically expressed in fully differentiated enterocytes, notably, those involved in the transfer of dietary lipids.
It is noteworthy that Hnf-4g, the paralog of Hnf-4a, is expressed in intestinal villi (45) but not in the liver (50). Our results suggest that HNF-4γ, whose expression is similar in the villus compartment and induced in the crypt compartment of Hnf-4aintΔ mice, may act to partly compensate for the lack of HNF-4α. This hypothesis is supported by the observation that HNF-4γ is still present in crypts at 30 days after tamoxifen injections. At this time, the size of crypts and the proliferation status are closer to those of control mice. A recent paper from Babeu et al. (3) reported results obtained with mice where the same Hnf-4aloxP/loxP gene was inactivated in the intestine but by using a constitutively active Cre recombinase driven by a 12.5-kbp villin promoter which is active at embryonic days 12.5 to 13.5 (34). The animals were viable and did not present any overt phenotype or intestinal dysfunction. The results we obtained with adults 30 days after tamoxifen injection are similar to those of Babeu et al. (3), i.e., a subtle and statistically significant increase in crypt cell proliferation and of the number of goblet cells. Our observations support the hypothesis of functional compensation by HNF-4γ. It may be hypothesized that in the study of Babeu et al., the absence of an overt phenotype could be due to the presence of HNF-4γ, even if a slight decrease in the expression of Hnf-4g was observed in the whole intestinal epithelium of their mice. We have previously reported that, in vitro, the two isoforms have similar properties, i.e., affinity for their binding site and transactivation of a reporter gene (2). However, mice lacking expression of HNF-4γ are viable and have no evident phenotype, indicating that the two proteins are not fully redundant. Deciphering the different roles of the two isoforms is still a challenging task.
HNF-4α plays an essential role in the architecture of epithelial cells (5, 10, 12, 41, 48). In the adult intestinal epithelium, loss of HNF-4α induces a distension of tight junctions that is associated with changes in the expression of tight junction proteins such as claudins 2, 4, and 7 and ZO-1. These modifications could be due to the increased expression of Tcf-4, which is a negative regulator of claudin 7 expression and a positive regulator of claudin 2 expression (13). Claudins are responsible for the gate function of tight junctions, which control paracellular permeability. Increased claudin 2 expression was reported to be responsible for increased epithelial permeability (54). Accordingly, we observed an increased paracellular permeability of Hnf-4aintΔ mouse intestine in vivo and in vitro. Interestingly, HNF-4α has recently been shown to have a role in protection against inflammatory bowel diseases, a pathology where the epithelial barrier is impaired (1). Furthermore, loss of tight junction structure and function is frequently observed in epithelium-derived cancers and overexpression of claudin 2 can stimulate the invasion and migration activities of cancer cells (36). Loss of HNF-4α in intestinal epithelium also induces a drastic decrease in EBP50 expression. ERM proteins regulate the organization and function of specific cortical structures in polarized epithelial cells, and EBP50 organizes these ERM proteins at the apical membrane (37). The observed downregulation of EBP50 expression in Hnf-4aintΔ mice could explain the concomitant E-cadherin membrane delocalization. Indeed, low expression of EBP50 decreased the interaction between β-catenin and E-cadherin, leading to disorganization of adherens junctions and increased cell proliferation and motility (26).
Our work in vivo demonstrates for the first time that HNF-4α is a transcription factor at the crossroads between intestinal epithelium homeostasis and epithelial cell architecture in the adult mouse intestine by controlling cell proliferation in crypts and functional differentiation in villi. Indeed, HNF-4α interferes with the Wnt/β-catenin signaling pathway and its loss destabilizes adherens and tight junctions. It is currently admitted that the deregulation of the Wnt/β-catenin signaling pathway is an early event in the colorectal cancer progression cascade and that the destabilization of cell-cell contacts takes part in the epithelial-to-mesenchymal transition, a crucial process in tumor progression. Since HNF-4α regulates the balance between proliferation and differentiation, we hypothesize that it might act as a tumor suppressor.
We thank N. Comuce, C. Lasne, and V. Chauffeton for animal care. Electron microscopy analysis was performed by using facilities of the Centre de Recherche des Cordeliers, UMRS 872. We thank M. Pontoglio for helpful scientific discussions and access to unpublished results, J. M. Lacorte and V. Carrière for their help during genotyping and RT-PCR experiments, respectively, and M. Rousset for critical reading of the manuscript.
This work was supported by INSERM, France, and Pierre & Marie Curie University, Paris, France. A.-L. Cattin is the recipient of a doctoral fellowship from the Ministère de l'Enseignement Supérieur et de la Recherche.
Published ahead of print on 5 October 2009.