Hepatocyte nuclear factor 1α and β control terminal differentiation and cell fate commitment in the gut epithelium

The intestinal epithelium is a complex and dynamic system characterized by a remarkably high degree of cell renewal that requires a tightly controlled coordination of proliferation, migration, differentiation and cell death. In fact, the epithelial cells that are lost from the tip of the villi need to be replaced at a constant and equal rate by cell proliferation in the crypts of Lieberkühn. This leads to a continuous vectorial cell migration from the stem cell compartment at the base of the crypt to the tip of the villus, which is composed of differentiated cells. Villi are populated by three differentiated cell types comprising enterocyte, enteroendocrine and goblet cells, whereas a fourth cell type, the Paneth cell, resides at the bottom of the crypts in the small intestine (reviewed by Sancho et al., 2003; van der Flier and Clevers, 2009).

Several studies have indicated that the Wnt pathway drives the proliferation and maintenance of pluripotent intestinal progenitor cells that, in turn, employ the Notch pathway to make the binary choice between absorptive or secretory cell lineage differentiations. If the Notch pathway is activated in a progenitor cell, it will express Hes1 and differentiate into enterocytes. Conversely, if a precursor cell expresses Atoh1 instead, it will differentiate into one of the three secretory cell lineages – enteroendocrine, goblet or Paneth cells (Jensen et al., 2000; Yang et al., 2001). Cell-type-specific transcription factors might play a crucial role in this dynamic system by modulating the expression of effectors of signalling pathways that in turn are specifically activated or repressed in a cell-type-specific manner. In addition, they could also direct the expression of cell-specific genes that contribute to the function of terminally differentiated cells. In this context, the study of the function of transcription factors offers a unique perspective to understand these complex developmental processes.

Hepatocyte nuclear factor 1α and β (Hnf1a and Hnf1b, formerly known as Tcf1 and Tcf2) genes encode for two closely related dimeric divergent homeobox proteins. Initially discovered in liver, Hnf1 proteins were subsequently shown to be expressed in the epithelia of several organs including pancreas, kidney and intestine. During mouse development, Hnf1β is first expressed in the extra-embryonic visceral endoderm and the primitive gut, whereas Hnf1α is expressed later during hepatic, pancreatic, renal and intestinal organogenesis in combination with Hnf1β (Cereghini, 1996). The study of the phenotype of mice carrying mutations in these genes has shown that Hnf1α and β play important roles in the function and/or differentiation of epithelia. In fact, Hnf1a-deficient mice (Hnf1a^−/−) are born normally, but suffer from hepatic, pancreatic and renal functional defects (Pontoglio et al., 1996; Pontoglio et al., 1997; Pontoglio et al., 1998; Lee et al., 1998; Pontoglio et al., 2001). By contrast, germ-line Hnf1b-deficiency is embryonic lethal because of defective differentiation of extra-embryonic visceral endoderm (Barbacci et al., 1999; Coffinier et al., 1999). Inactivation of Hnf1b demonstrated that this factor plays a crucial role in the differentiation/morphogenesis of the pancreas (Haumaitre et al., 2005), liver (Lokmane et al., 2008) and bile ducts (Coffinier et al., 2002), and in the maintenance of renal tubular structures (Gresh et al., 2004). In humans, patients carrying mono-allelic mutations in HNF1A or HNF1B suffer from maturity onset diabetes of the young (MODY type 3 and 5, respectively) and renal dysfunctions. Furthermore, bi-allelic inactivation of Hnf1a is frequently observed in hepatic adenomas (Bluteau et al., 2002). All of these observations indicate that HNF1α and β are involved in the differentiation program of epithelial structures in several organs including liver, kidney and pancreas. Interestingly, these two proteins are also expressed in the intestinal epithelium but little is known about their function in this organ. To date, several studies have demonstrated that these transcription factors activate the expression of a number of intestinal genes including Fabp1, LPH, Cftr and Glc-6-Pase, which are expressed in differentiated enterocytes (Mouchel et al., 2004; Gautier-Stein et al., 2006; Bosse et al., 2007). An in vitro study demonstrated that Hnf1α and β control the transcription of Dpp4, another marker of differentiated enterocytes (Erickson et al., 2000). Hnf1α and β bind DNA with the same sequence specificity and can form homo- or heterodimers. The gene duplication event at the origin of the Hnf1 family had been selected during evolution with the emergence of the first vertebrates. The fact that these two genes share the same DNA binding specificity and overlapping expression pattern, together with their ability to form heterodimers, suggests that these proteins could play complementary roles. The phenotypic characterization of single- and double-gene inactivation provides a very helpful strategy to understand the dual nature of their genetic program.

In the present study, we investigated the role of Hnf1α and β in the intestinal epithelium. Although the single Hnf1b mutant does not present any severe dysfunction, mice with the intestinal inactivation of Hnf1b in a context deficient for Hnf1a (double mutant Hnf1a^−/−; Hnf1b^Δintestine) survive only few weeks after deletion. We demonstrate here that Hnf1α and β participate in both cell fate commitment and differentiation programs in the intestine. In fact, we observed an increase in the number of goblet cells, a decrease in mRNA levels of several terminal differentiated markers of both enterocytes and secretory cell types and a lack of mature Paneth cells. We further demonstrate that Hnf1α and β control the expression of Jag1, Atoh1 and Cdx2 genes involved in cell commitment and differentiation.

Finally, we demonstrate that the double-mutant mice show a defect in intestinal water absorption that might be responsible for their death. We show that Hnf1α and β directly control the expression of Slc26a3, a gene whose mutations are responsible for a life-threatening form of congenital chloride diarrhoea in human patients. These studies identify crucial direct target genes of the Hnf1 transcription factors and elucidate their function in the gut epithelium.

The generation of Hnf1b^flox/flox and Vil-CreER^T2 mice was previously described (Coffinier et al., 2002; el Marjou et al., 2004). All experiments involving animals were carried out in accordance with French government regulations. Cre recombinase was activated by a single injection of tamoxifen (TAM) solution (50 mg/kg; Inalco). We crossed the Hnf1a^+/+; Hnf1b^flox/flox and Vil-CreER^T2 mouse lines in order to obtain the single Hnf1b mutant mouse model following tamoxifen injection. The Hnf1a^+/+; Hnf1b^flox/flox animals without the Cre recombinase transgene were used as control animals. Then, we crossed Hnf1a^+/−; Hnf1b^flox/flox and Vil-CreER^T2 in order to obtain the concomitant loss of Hnf1α and β in the intestinal epithelium (Hnf1a^−/−; Hnf1b^Δintestine). The Hnf1a^−/−; Hnf1b^flox/flox animals without the Cre recombinase transgene were our single Hnf1a mutant animals. We intraperitoneally injected mice between 4 and 5 weeks after birth with a single dose of tamoxifen and we analyzed the intestinal phenotype of double-mutant mice 2 weeks after injection. All of the mouse strains indicated were treated with the same protocol. Mice were sacrificed by cervical dislocation and the intestine collected.

Intestinal segments from animals were dissected and washed with ice-cold PBS. For small intestine samples, segments (2 cm) from each of the intestinal regions (duodenum, jejunum and ileum) were mixed and homogenized in Trizol. Total RNA was isolated with Trizol reagent (Life Technologies) followed by purification on RNeasy columns (Qiagen). Complementary DNA (cDNA) synthesis was performed according to manufacturer's instructions (SuperScript Kit; Invitrogen). For RT-PCR, cDNA and primers were mixed with SYBR Green RT-PCR Master Mix (Applied Biosystems) and then assayed in an ABI Prism 7000 Sequence Detection System according to manufacturer's instructions. The relative level of each mRNA was calculated using the standard curve method and normalized to the corresponding Gapdh RNA levels. Five independent RNA samples were used for each group (e.g. one group was Hnf1a^−/−; Hnf1b^Δintestine mice) and triplicate reactions of each sample were used to derive the normalized expression level for each gene. The average normalized expression levels were used to determine the average expression level within a group and for statistical comparisons between groups (n=5 for each group). ANOVA and t-tests were performed to measure variations in gene expression between groups.

The primers for each gene analyzed were designed with Primer Express software (Applied Biosystems) and are listed in Table S1 in the supplementary material.

The intestinal tract was dissected as a whole and gently washed with cold PBS to remove any faecal content. The small intestine was rolled up as a ‘Swiss roll’ and fixed in paraformaldehyde 4% for 1 hour at room temperature. Specimens were then washed in cold PBS and included in agarose 4%. They were sectioned with vibratome (100 μm). For the detection of Hnf1α, we used also tissues embedded in Optimum Cutting Temperature compound (Tissue-Teck), frozen and sectioned on a cryostat at 8 μm. For the detection of Hnf1β, immunohistochemistry was performed on a paraffin wax-embedded rolled intestine. Antigen retrieval was performed by incubating the tissue sections in 10 mM sodium citrate, pH 6.0, for 15 minutes at 95°C. For immunohistochemistry, sections were sequentially incubated with biotinylated secondary antibodies (Jackson ImmunoResearch), peroxidase-conjugated streptavidin and DAB substrate (both Vector Labs). The following antibodies were used: rabbit anti-Hnf1α-TC284 (1:500) (Chouard, 1997), monoclonal anti-Hnf1β (1:100; clone 3.12) (Chouard et al., 1997), rabbit anti-ezrin (a gift from D. Louvard, Institut Curie, Paris), rabbit anti-chromogranin A+B (1:50; Progen), UEA1-FITC (1:50; Sigma), rabbit anti-lysozyme (1:200; DakoCytomation).

In situ hybridization (ISH) was carried out on 7 μm slices of the paraffin-embedded specimens. Partial Atoh1 cDNA was obtained by RT-PCR using the primers Atoh1F_IS (CCGTCAAAGTACGGGAACAG) and Atoh1R_IS (GAGTAACCCCCAGAGGAAGC), and then subcloned into TOPO cloning vector (Invitrogen). Digoxigenin-labelled RNA probe was prepared by in vitro transcription with the Digoxigenin RNA Labeling Kit (Roche, France) using T7 or Sp6 RNA polymerases. Sections were incubated overnight at 68°C in prehybridization buffer containing 200 ng/ml of digoxigenin-labelled RNA probe. Immunodetection of the hybridized probe was carried out using an anti-digoxigenin antibody (1:2000, Roche).

Nuclei were prepared from small intestine of control and Hnf1a^−/−; Hnf1b^Δintestine mice. Chromatin preparation and ChIP were performed as previously described (Gresh et al., 2004). ChIP experiments were performed with rabbit anti-Hnf1α-TC284 (1:500, Chouard, 1997) or rabbit anti-Hnf1β (1:200, clone H-85, Santa Cruz Biotechnology) antibodies. Primers were designed using PrimerExpress 2.0 software. For primer sequences, see Table S1 in the supplementary material. Quantification of precipitated DNA fragments was carried out on an ABI Prism 7000 Sequence Detection System using SYBR Green in triplicates. Relative fold-enrichment of DNA fragments was calculated using the following formula: (ChIP_target/ChIP_normalizer)/(input_target/input_normalizer). As normalizer, we used a DNA fragment lacking any Hnf1 site, located in the first intron of the aortic smooth muscle alpha actin 2 (Acta2) gene.

To study the function played by Hnf1β in adult intestinal epithelium, we crossed mice carrying a floxed allele of Hnf1b (Coffinier et al., 2002) with a strain carrying a tamoxifen-inducible Cre recombinase specifically expressed in intestinal epithelium (Vil-CreER^T2 mice) (el Marjou et al., 2004). In this strain, Cre recombinase activity can be induced in all the cell types, including intestinal stem cells, by intraperitoneal injections of tamoxifen, leading to a permanent inactivation of floxed alleles in all intestinal epithelial cells even several weeks after deletion (el Marjou et al., 2004). Our results showed that, after recombination, the loss of Hnf1β expression in the intestine (Hnf1β^Δintestine) did not elicit any overt phenotype, at least over a period of 20 weeks, in spite of a drastically decreased expression of Hnf1β. Conversely, we have previously demonstrated that the loss of Hnf1α (owing to a germ-line-null inactivation) results in liver, kidney and pancreatic dysfunction, whereas the intestine did not present with any particularly severe phenotype (Pontoglio et al., 1996; Shih et al., 2001). The lack of any prominent intestinal phenotype in any of the single-mutant mice raised the possibility that Hnf1α and β could compensate for each other when inactivated independently. As both transcription factors bind to the same DNA sequences, they could play complementary roles during intestinal cell commitment and differentiation. To address the extent of the possible redundancy between these two transcription factors, we analyzed the intestinal phenotype of Hnf1a^−/−; Hnf1b^Δintestine mice (double-mutant mice). In our present study, we principally focused our attention on the small intestine because our Cre inactivation system was particularly efficient in deleting Hnf1b in this compartment (el Marjou et al., 2004).

In both the single- and the double-mutant mice, Hnf1b expression levels were dramatically compromised (90% reduction, Fig. 1A). Conversely, the lack of Hnf1a did not affect Hnf1b mRNA levels in the small intestine of single Hnf1α mutant mice (Fig. 1A). Similarly, the lack of Hnf1b did not affect Hnf1a mRNA levels in the small intestine of Hnf1b^Δintestine mice (see Fig. S1A in the supplementary material). Previous expression studies have revealed that both Hnf1α and β are expressed along the crypt-villus axis in the gut epithelium. These studies were performed both at the protein and mRNA level for Hnf1α (Boudreau et al., 2002; Fang et al., 2006; Serfas and Tyner, 1993) but only at the mRNA level for Hnf1β (Fang et al., 2006; Serfas and Tyner, 1993). Moreover, it was reported that Hnf1α protein levels increased after the suckling weaning transition (Boudreau et al., 2002), whereas in another study, its mRNA decreased during this transition phase (Fang et al., 2006). Thus, owing to the differences observed in expression and to the lack of Hnf1β protein profile, we decided to investigate the Hnf1α and β protein expression patterns in detail at the histological level in 4- to 5-week-old mice. This temporal phase corresponded to the window of time in which we induced Hnf1β inactivation. Immunofluorescence with antibodies specific for Hnf1α revealed that this protein was expressed in all of the nuclei along the crypt-villus axis in the small intestine (Fig. 1B). Depending on the experimental conditions employed, the cells at the very bottom of the crypts gave signals of different intensity. In particular, on vibratome sections (100 μm; Fig. 1B) this signal was much paler than that on sections obtained with a cryostat microtome (see Fig. S1B in the supplementary material). Similarly, immunohistochemistry analysis revealed that Hnf1β was expressed along the crypt-villus axis in the small intestine (Fig. 1D). The specificity of the staining was confirmed by the absence of any signal in mutant mice (Fig. 1C,E; see also Fig. S1C in the supplementary material). This demonstrated that our double-mutant mouse model is completely inactivated for both Hnf1 transcription factors in the gut epithelium. To assess the extent of the tissue-specific inactivation of Hnf1b, we performed PCR to amplify the recombined allele as previously described (Coffinier et al., 2002). We analyzed several tissues and recombination was observed only in the intestine (see Fig. S1D in the supplementary material). Our results showed that the loss of Hnf1β after a single injection of tamoxifen in Hnf1a^−/−; Hnf1b^flox/flox mice resulted in a rapid demise of the animals. More than 50% of the double-mutant animals died within 2 weeks after the deletion, whereas only a few of them (less than 10%) survived to 8 weeks (see Fig. S1E in the supplementary material). Thus, we tried to unravel the cause for the rapid death of double-mutant animals.

To assess the function of Hnf1α and β in the gut epithelium we monitored cell fate commitment and differentiation programs. Upon histological analysis, small intestine of double-mutant animals appeared with a normal crypto-villar architecture (Fig. 1C,E; Fig. 2C). Immunofluorescence with anti-ezrin antibodies (specific for enterocytes) showed a normal signal uniformly localized at the brush border membranes of enterocytes all along the villus both in control (Fig. 2A) and single- or double-mutant mice (Fig. 2B,C). In order to assess the degree of differentiation of the intestinal epithelium, we decided to investigate the expression of markers of enterocyte differentiation. Dpp4, an aminopeptidase localized at the brush border of enterocytes, was shown to be activated by Hnf1α and β in transfection studies (Erickson et al., 2000). When we analyzed the expression levels of Dpp4, we observed that the mRNA level of the enzyme was strongly downregulated in double-mutant animals (80% reduction) and significantly downregulated in single Hnf1α mutant mice (60% reduction; Fig. 2D), indicating that this gene is an in vivo target gene of Hnf1 and that the differentiation program of enterocytes could be altered in our mouse mutant. In addition, Cdx2, a transcription factor involved in intestinal development and enterocyte maturation (Suh and Traber, 1996; Mutoh et al., 2005; Grainger et al., 2010), was downregulated in double-mutant mice (37% reduction) and slightly downregulated in single Hnf1α mutant mice (16% reduction; Fig. 2D). To investigate the relationship between Cdx2 and Hnf1 transcription factors, we took advantage of an in silico approach based on a previously described position-weight matrix (Tronche et al., 1997) (M.P. and S.G., unpublished). This approach revealed that Cdx2 contained a putative binding site for Hnf1 that is well conserved in several vertebrate species. This site, located more than 10 kb upstream of the transcription site, is indeed bound by Hnf1α by chromatin immunoprecipitation (ChIP; Fig. 2E). These results confirm that Hnf1 transcription factors do have a role in the control of enterocyte differentiation and identify a new in vivo target gene.

Normally, goblet cells are present in a rostrocaudal gradient all along the length of the small intestine with the highest level in the ileum. Histological sections stained with UEA1, a lectin specific for mucin-expressing cells, revealed an increased number of goblet cells all along the length of the small intestine in double-mutant mice. We observed an increased number of goblet cells both in the ileum (Fig. 3F compared with 3D) and in the duodenum (Fig. 3C compared with 3A and 3G). By contrast, no significant difference was observed in the single Hnf1α (Fig. 3B compared with A; Fig. 3E compared with 3D) or Hnf1β intestinal single mutant (data not shown). To determine whether the goblet lineage abnormalities in mice were due to abnormal proliferation, we performed double immunostaining with proliferative anti-Histone-H3 phosphorylated on Ser10 and cell-specific UEA1 markers. However, our results did not reveal any significant difference in the extent of proliferation of this cell type (data not shown). The expression levels of Klf4, a transcription factor involved in differentiation of goblet cells, and of Muc3, a goblet-specific gene, were moderately upregulated in double-mutant mice (1.2- and 2.0-fold, respectively; Fig. 3H). This paralleled the significant increase of goblet cells observed in the epithelium. On the contrary, other specific markers for goblet cells such as Muc2, gob-4 (Agr2 – Mouse Genome Informatics) and gob-5 (Clca3 – Mouse Genome Informatics) were strongly downregulated in Hnf1a^−/−; Hnf1b^Δintestine mice (75%, 64% and 81% reduction, respectively; Fig. 3H). This indicated that goblet cells, although present in excess, failed to completely activate their cell-type specific genetic program or they basically reduced gene transcription to be able to normalize with the increased cell number. These alterations in cell fate commitment and/or differentiation were observed only when both factors were inactivated. Moreover, the expression of Elf3, a transcription factor involved in both enterocyte and goblet cell differentiation (Ng et al., 2002), was not affected (Fig. 3H). Hnf4α, a transcription factor essential for development of the colon, regulates goblet-cell maturation (Garrison et al., 2006). Interestingly, Hnf1α and Hnf4α have been described to regulate each other (Odom et al., 2004). To assess the possible contribution of Hnf4α to the phenotype observed, we measured the expression levels of Hnf4a in our mutant animals. Both the single Hnf1a mutant and the double-mutant mouse showed no differences in the expression of Hnf4α (Fig. 3H). In our mutants, enteroendocrine cell differentiation appeared to occur normally as no difference in the number of chromogranin A-positive cells was observed in the small intestine of the double mutant compared with the control animals (data not shown). Conversely, a drastic defect in Paneth cell differentiation was observed as no apparent mature cells were visible at the bottom of the crypts of Hnf1a^−/−; Hnf1b^Δintestine mice (Fig. 4C versus 4A). No significant difference in mature Paneth cells was observed in single Hnf1α or β mutant mice (Fig. 4B versus A; data not shown). Furthermore, markers specific for Paneth cells (Cry1 and Lyz2) were strongly downregulated (respectively 82% and 87% reduction) in the small intestine of Hnf1a^−/−; Hnf1b^Δintestine mice (Fig. 4D). This correlated with the observed drastic disappearance of immunoreactive positive cells at the bottom of the crypts of double-mutant mice. Interestingly, the transcription factor Gfi1, required for intestinal goblet and Paneth cell differentiation (Shroyer et al., 2005), was significantly reduced in double-mutant animals (67% reduction). On the contrary, Sox9 and Msi1, genes required for Paneth cell differentiation in the intestinal epithelium (Bastide et al., 2007; Mori-Akiyama et al., 2007; Murayama et al., 2009), were not affected in Hnf1a^−/−; Hnf1b^Δintestine mice (Fig. 4D). It is known that the Paneth cell maturation program is controlled by the Wnt pathway in the crypt (Andreu et al., 2008; van Es et al., 2005a). However, when we verified the expression levels of c-Myc and Cycd1 (two important Wnt target genes), we did not detect differences in their levels (Fig. 4D), suggesting that this pathway is normally activated and that other genes not necessarily involved in the Wnt pathway might play a key role in the terminal differentiation process of Paneth cells. Paneth cell lifespan is regulated by autophagy, a mechanism in which several genes are involved (Cadwell et al., 2008; Cadwell et al., 2009). To exclude defects in autophagy, we analyzed the expression of two key genes for this process. Our results showed that Atg16l1 and Atg7 were not affected in the double-mutant mice (Fig. 4D), suggesting that the disappearance of Paneth cells in our model is probably due to a differentiation defect.

These results indicate that Hnf1α and β play a crucial role both in cell fate commitment represented by the increased number of goblet cells and in secretory cell-type-specific differentiation represented by the incomplete activation of cell-type-specific gene expression.

Cell fate commitment during the differentiation of intestinal cells occurs in the crypts where both Hnf1α and β are expressed. The number of goblet cells was increased without an increased proliferation rate of goblet precursors suggesting that the two transcription factors might be mechanistically directly involved in cell fate choice. To address this point, we investigated the consequences of Hnf1α and β inactivation in the intestinal epithelium on the intricate transcriptional and signalling network governing the early cell fate decision of all four epithelial lineages found in the small intestine. We focused our attention on the expression of crucial regulators of cell fate commitment and particularly on the effectors of the Notch pathway, a pathway that has been demonstrated to control absorptive/secretory cell fate choice. In this respect, we considered those effectors that are normally expressed in intestine and whose genomic sequences showed a significant enrichment in evolutionarily conserved Hnf1 binding sites. Our results revealed that Jag1, Atoh1 and Notch2 genes were particularly enriched in putative Hnf1 binding sites that are well conserved in several vertebrate species (Fig. 5B). Indeed, expression of both Notch2 and Jag1 was downregulated (47% and 58% reduction, respectively) in Hnf1a^−/−; Hnf1b^Δintestine mice (Fig. 5A). In a similar way, Atoh1 was highly downregulated (77% reduction) in Hnf1a^−/−; Hnf1b^Δintestine mice and slightly downregulated (16% reduction) in Hnf1a^−/− mice. To assess the possible defective activation of Notch signalling, we tested the expression of the typical targets represented by the Hes genes. Our results showed that Hes1, Hes5 and Hes6 expression was downregulated (50%, 63% and 28% reduction, respectively; Fig. 5A).

To verify the possible direct involvement of Hnf1α and β in these transcriptional defects, we performed ChIP with anti-Hnf1α or anti-Hnf1β antibodies. In this way, we assessed the in vivo binding of Hnf1 proteins on several putative Hnf1 binding sites that we selected for their degree of conservation, their genomic position and their statistical Hidden Markov Model (HMM) score. In particular, we selected putative binding sites with HMM ≥5.7 and conserved in at least three species out of fourteen analyzed. Our results showed that sites in Jag1 and Atoh1 genes were bound in vivo in chromatin prepared from small intestine of wild-type animals (Fig. 5B). The specificity of our ChIP experiments was supported by the fact that no significant enrichment for all DNA fragments tested was observed in double-mutant animals (data not shown). In particular, our results showed that both Hnf1α and β could bind the site that we identified in the eleventh intron of the Jag1 genomic sequence (Fig. 5B). Conversely, only Hnf1α (and not β) could bind a site located in Atoh1 (30 Kb downstream to the transcriptional start site in the 3′ region of the gene; Fig. 5B). The putative Hnf1 binding site of Notch2 that we tested was not bound in vivo (Fig. 5B); it is possible that other putative binding sites with a lower HMM score are bound in vivo by Hnf1α and/or β. Evolutionary conserved sites not bound in the intestine are probably involved in the control of these genes in other tissues.

It is known that Hes1 negatively regulates Atoh1 (Jensen et al., 2000; Yang et al., 2001). Hes1 is expressed only in the crypt compartment, whereas Atoh1 is expressed in progenitors of the secretory lineage, as well as in their corresponding fully differentiated cell types (Jenny et al., 2002; Pinto et al., 2003; van Es et al., 2005b; Yang et al., 2001). The analysis of Atoh1 expression by in situ hybridization showed that it was dramatically decreased in the mature secretory cells of the villi in the double-mutant animals (see Fig. S2 in the supplementary material).

The observed cell fate allocation and differentiation defects could not easily explain the rapid death observed in double-mutant animals. Interestingly, we observed that double-mutant mice presented with a sudden and drastic loss of body weight (20%) during the first week after deletion (Fig. 6A). This correlated with the appearance of signs of severe dehydration in double mutants compared with the single-mutant mice (Hnf1a^−/−) or control littermates (see Fig. S3 in the supplementary material). Remarkably, double-mutant animals presented with a significant distension of the intestine compared with wild-type mice. Loops of intestine appeared fluid-filled, with increased volume of the bowel contents (Fig. 6B). Hnf1α single-mutant mice presented with a slightly distended gut, with a milder phenotype compared with the double-mutant mice (Fig. 6B). Interestingly, a closer inspection of villi in the double mutants (Fig. 2C) showed that their structure appeared less turgid compared with that of control animals (Fig. 2A), suggesting that mutant villi might have defective water absorption capacity. Intestinal water transport is linked to solute transports. In this respect, the currently accepted model indicates that active ion transport sets up local osmotic gradients that, in turn, drive water absorption by local osmosis (reviewed by Loo et al., 2002). Hnf1α and β could directly activate the transcription of specific transporter/exchanger genes involved in intestinal water absorption. Among these genes, we focused our attention on Slc26a3, a gene whose mutations are associated to congenital chloride diarrhoea (OMIM 214700). This disease is manifested by enhanced chloride and water loss in stools (Hoglund et al., 1996). The encoded protein is a Cl⁻/HCO ⁻₃ exchanger driving NaCl absorption. The murine model of Slc26a3 knockout recapitulates the human disease as inactivation of the gene resulted in chloride diarrhoea (Schweinfest et al., 2006). On the basis of these observations, we examined expression of Slc26a3 in control, Hnf1a^−/− and Hnf1a^−/−; Hnf1b^Δintestine mice. Interestingly, Slc26a3 expression was completely abolished both in the small intestine and the colon of Hnf1a^−/−; Hnf1b^Δintestine mice (Fig. 6C,D). In single-mutant Hnf1a^−/− mice, Slc26a3 expression was reduced (41% reduction) in the small intestine (but not in the colon) with respect to control animals (Fig. 6C,D). We next tested if this gene could be directly controlled by Hnf1 proteins. Our in silico approach predicted the presence of two Hnf1 putative binding sites, particularly well conserved in several species, in the proximal promoter and in the eighth intron of Slc26a3 genomic sequence (Fig. 6E). Our results showed that both Hnf1α and β did indeed bind those sites, indicating that these factors play, in concert, a direct and crucial role for Slc26a3 gene expression.

The differentiation of the four distinct cell types that populate the villi is controlled by the interplay between signalling pathways and cell-type-specific transcription factors. In the present study, we demonstrate that Hnf1α and β contribute to both cell fate decision and terminal differentiation in the gut epithelium.

Hnf1α and β bind to identical DNA sequences indicating that they could play complementary roles in the organism. Our study shows that single Hnf1a^−/− or Hnf1b^Δintestine animals have no major intestinal phenotypes. However, the lack of both Hnf1α and β in the gut epithelium significantly affects this organ and causes a lethal phenotype.

Hnf1a^−/−; Hnf1b^Δintestine mice die from severe dehydration in a matter of few days after inactivation. This dysfunction is linked to the complete suppression of the expression of Slc26a3, a gene encoding for an anion exchanger involved in intestinal water absorption. A murine model of Slc26a3 gene inactivation showed that newborn pups cannot easily survive after birth (Schweinfest et al., 2006), similar to what happens with newborn children that carry null mutations in this gene (OMIM 214700). We have shown that Hnf1α and β directly control the expression of the Slc26a3 gene, indicating that Hnf1α and β expression in the intestinal epithelium is essential for survival. Previous studies have demonstrated that the Hnf1 transcription factors activate, in vivo, the expression of a number of intestinal genes including Fabp1, LPH, Cftr and Glc-6-Pase, which are expressed in differentiated enterocytes (Bosse et al., 2007; Gautier-Stein et al., 2006; Mouchel et al., 2004). We have confirmed the previous in vitro studies on Dpp4 as a target gene of Hnf1α and β, demonstrating that its levels in vivo were severely affected in our Hnf1 mouse models (Erickson et al., 2000). In addition, Cdx2, a transcription factor involved in enterocyte differentiation (Mutoh et al., 2005; Suh and Traber, 1996), is known to control the expression of Muc2, a marker of goblet cells (Yamamoto, 2003). In our studies, we showed that Cdx2, a gene harbouring an Hnf1 binding site conserved throughout evolution, is defectively expressed in double-mutant mice and is a direct target of Hnf1α. A complex regulatory mechanism was demonstrated for Cdx2 expression (Benahmed et al., 2008; Boyd et al., 2009). Recently, it was demonstrated that Cdx2 controls Hnf1α during embryonic development (Gao et al., 2009). Our data demonstrate that Hnf1 transcription factors partially contribute to Cdx2 expression, suggesting that there is a reciprocal control between Cdx2 and Hnf1α in the intestinal epithelium (Fig. 7).

In addition to the already known role of Hnf1s in controlling the transcription of enterocyte-specific genes, we demonstrate for the first time that they are also involved in production and differentiation of secretory cell lines. Inactivation of both Hnf1α and β gave rise to supernumerary goblet cells not associated with an increase in the proliferation rate of goblet cell precursors. Several studies have demonstrated that the major pathway involved in intestinal cell fate commitment is the Notch pathway (reviewed by Crosnier et al., 2006; Radtke and Clevers, 2005). Intestinal progenitors are committed through a binary cell fate choice. Activation of this pathway leads the intestinal progenitors to differentiate into enterocytes. On the contrary, in the absence of the activation of the Notch pathway, intestinal progenitors express Atoh1 and differentiate into one of the three secretory cell types (Jensen et al., 2000; Yang et al., 2001). Inactivation of Notch signalling resulted in the complete loss of proliferating crypt progenitors and their conversion into postmitotic goblet cells (van Es et al., 2005b). The molecular mechanisms of this conversion were recently elucidated. Notch-mediated Hes1 expression contributes to the maintenance of the proliferative crypt compartment of the small intestine by inhibiting the transcription of two cyclin-dependent kinase inhibitors (Riccio et al., 2008). Gain-of-function studies also demonstrated that Notch activity is required for the maintenance of proliferating crypt cells in the intestinal epithelium (Fre et al., 2005). Interestingly, we showed that Hnf1α and β can activate directly the expression of Jag1, a gene that encodes for a ligand of the Notch pathway and whose genomic sequence contains Hnf1 binding sites conserved throughout evolution that are bound by both Hnf1α and Hnf1β. In line with this observation, the expression of several effectors directly activated by Notch signalling, including Hes1, Hes5 and Hes6, was downregulated in the double mutants. This should explain why our double-mutant mice had a slight, but significant, increased number of goblet cells, which could be ascribed to the specific downregulation of Notch signalling activity. One of the key genes controlling secretory cell lineage commitment is Atoh1. This gene is known to be under the direct repression of Hes1. When Notch signalling is inactive, Atoh1 expression is increased. In spite of a partial inactivation of the Notch pathway, we unexpectedly observed a downregulation of Atoh1 in our double-mutant mice. Interestingly, several Hnf1 binding sites were conserved throughout evolution in this gene and we showed that at least Hnf1α can bind in vivo a distal (30 Kb) downstream transcriptional control element of this gene. Previous studies have shown that a null mutation of Atoh1 leads to a complete depletion of the secretory cell lineage in the intestine, indicating that this factor is required for specification of enteroendocrine, goblet and Paneth cells (Shroyer et al., 2007; Yang et al., 2001). Surprisingly, in our double mutants, the reduction in Atoh1 expression did not prevent secretory cell commitment. In fact, the proportion and differentiation of enteroendocrine cells was not affected and, conversely, the number of goblet cells was paradoxically increased in Hnf1a^−/−; Hnf1b^Δintestine mice. This paradox could be explained by the fact that the expression of Atoh1 is still maintained in the crypts. The residual expression of Atoh1 in our model could be sufficient for the commitment of goblet and enteroendocrine cell fates. Notably, Atoh1 is expressed in progenitors of the secretory lineage as well as in their corresponding fully differentiated cell types (Jenny et al., 2002; Pinto et al., 2003; van Es et al., 2005a; van Es et al., 2005b; Yang et al., 2001), suggesting that Atoh1 could also play a role in postmitotic secretory cells. In our model, Atoh1 appears to be completely abolished in the secretory cells of the villi but not in the crypts supporting the role of Hnf1 in the control of terminal differentiation (Fig. 7).

In our double-mutant mouse model, we detected the lack of mature Paneth cells at the bottom of the crypts. Some transcription factor inactivation models, including Gfi1^−/− or Sox9^−/−, are characterized by a drastic defective differentiation of Paneth cells (Bastide et al., 2007; Shroyer et al., 2005). Interestingly, this defect is normally accompanied by colonization of the bottom of the crypts by proliferative cells. In our double-mutant animals, the apparent lack of Paneth cells did not result in this colonization. This would indicate that committed postmitotic Paneth cell precursors probably occupy their correct position at the bottom of the crypts but simply failed to complete their differentiation program. The expression levels of c-Myc, cyclin D1 (the key effectors of Wnt pathway) and Sox9 (a mediator of the Wnt-dependent program) were not affected, suggesting that the defect of the differentiation program might be Wnt-independent. However, we did not find any difference in the expression levels of Msi1, a gene important for the Paneth cell Wnt-independent differentiation program, suggesting that other genes might be involved in this differentiation program.

Finally, we observed the defective activation of the Notch signalling linked to the decreased expression of Jag1 (Fig. 7). This leads to an increased goblet cell fate commitment. Conversely, we observed defective differentiation of both goblet and Paneth cells as a considerable number of specific secretory markers were downregulated. We checked for the presence on Hnf1 binding sites in all the affected genes involved in terminal differentiation of goblet and Paneth cells. These genes did not present any overtly conserved Hnf1 binding sites. Thus, Hnf1 transcription factors appear to indirectly regulate terminal differentiation markers of secretory cell lineages, possibly through Atoh1. In addition, both Hnf1 transcription factors directly control Slc26a3 a gene involved in intestinal water absorption. Thus, Hnf1α and β both directly and indirectly control terminal differentiation of several cell lineages in the gut epithelium and cell fate commitment through the direct regulation of intestinal key actors (Fig. 7).

In conclusion, Hnf1α and β transcription factors are crucial both for enterocytes and secretory cell lines in the gut epithelium.

This work was supported by Institut National du Cancer (INCA) grant PL 043, the Fondation pour la Recherche Médicale and the Fondation Bettencourt-Schueller (Prix Coup d'Elan). We are grateful to Moshe Yaniv, Jacqueline Barra, Fabiola Terzi and Jonathan Weitzman for helpful discussions and for critical reading of the manuscript. We thank B. Mateescu for antibody anti-Histone-H3 phosphorylated in Ser10. We thank all the members of Pontoglio Laboratory for technical assistance and advice.