Glucose Tolerance Is Improved in Mice Invalidated for the Nuclear Receptor HNF-4γ: A Critical Role for Enteroendocrine Cell Lineage

During past decades, changes in dietary habits, including increments in calorie and saturated fatty acid intake, have occurred concomitantly with a rise of metabolic diseases, such as diabetes, obesity, and the metabolic syndrome (1). These metabolic disorders, which are known risk factors for cardiovascular diseases, represent emerging medical and public health concerns. Intestine, which is the first organ facing digestion products and which contributes to energy homeostasis through the transfer of nutrients to the organism, receives little attention about its potential role in the onset of metabolic disorders compared with liver, pancreas, muscle, and adipose tissue.

Several transcription factors (CDX1, CDX2, GATA 4/5/6) are reported to be involved in intestinal differentiation and regulation of gut-specific gene expression (2–5). Approaches based on transcriptome, metabolome, and bioinformatic analyses and data from in vivo experiments indicate that hepatocyte nuclear factor-4α (HNF-4α) is involved in the regulation of enterocyte phenotype (6–8).

HNF-4 belongs to the nuclear receptor superfamily. In mammals, two paralog genes encode the HNF-4α and HNF-4γ forms. HNF-4α is expressed in liver, kidney, pancreas, and intestine (9), and its invalidation is lethal in mice at day 6.5 of embryogenesis (10). In human, heterozygous mutations in the HNF4A gene lead to maturity-onset diabetes of the young type 1 (11) and contribute to susceptibility to type 2 diabetes (12,13). Numerous studies have shown that HNF-4α plays pleiotropic roles in liver functions and is a central transcription factor at the crossroads between epithelial morphogenesis and functions (14,15).

HNF-4γ was first identified in human kidney (16,17) and in mouse pancreas (18). It is also expressed in intestine (19) but almost absent from liver (17,18), and it has been shown to be highly expressed during intestine development (20). We reported previously that HNF-4α is expressed in intestinal crypts and villi, whereas the γ form is restricted to villi (21). These different distributions combined with the absence of compensation by HNF-4γ for some phenotypic traits of intestinal Hnf4a knockout mice, namely impairment of intestinal epithelium homeostasis, cell architecture, and fatty acid uptake (22–24), suggest that the α and γ forms play specific and different roles in intestine.

The physiological role of HNF-4γ has been much less studied than that of HNF-4α. Hnf4g^−/− mice do not exhibit an overt phenotype. Succinctly, they present weight gain but decreased food intake and lower energy expenditure and locomotor night activity (25). We aimed to determine the roles of HNF-4γ in adult mice and to explore the underlying mechanisms. Using the constitutive Hnf4g knockout model, we demonstrate that loss of HNF-4γ leads to an increased enteroendocrine L-type cell number and to an overproduction of GLP-1, thus generating hyperinsulinemia and improving glucose tolerance during an oral glucose tolerance test (OGTT) through the incretin effect of GLP-1.

Total and constitutive Hnf4g gene invalidation was as described in Gerdin et al. (25). Hnf4g^+/− mice obtained from Deltagen, Inc., on a C57BL/6J genetic background, were mated to obtain Hnf4g^−/− mice on the same genetic background. In experiments, we compared Hnf4g^−/− male mice with C57BL/6J wild-type male Hnf4g^+/+ mice matched for age and housed in the same room.

Specific intestinal Hnf4a gene invalidation in adult mice (Hnf4a^Δint) was described previously (22,23,26). For experiments, 6-month-old male control Hnf4a^loxP/loxP and Hnf4a^loxP/loxP/villin-CreERT2 mice received tamoxifen treatment (23), and analyses were performed 10 days later.

All animals were housed in the specific-pathogen free facility of the Centre de Recherche des Cordeliers on a 12-h light/dark cycle and fed a standard diet (A03/R03; SAFE). Experimental procedures were according to French guidelines for animal studies from the Comité National de Réflexion Ethique sur l’Expérimentation Animale Charles Darwin (Ce5/2009/045).

All experiments were performed on conscious mice. After overnight fasting (15 h), mice received a glucose load of 3.6 g/kg for OGTTs or 2 g/kg for intraperitoneal glucose tolerance tests (IPGTTs). The oral glucose challenges were administrated using a gavage needle (20 gauge, 38 mm long, curved, with a 21/4-mm ball end). After overnight fasting, exendin (9-39) (5 μg/mouse in NaCl 9 g/L) (Abcam) or saline (vehicle) was injected intraperitoneally 30 min before the oral glucose load (3.6 g/kg) as previously described (27). For insulin tolerance tests (ITTs), mice fed ad libitum were injected intraperitoneally with 1 unit/kg insulin (Actrapid; Novo Nordisk). For streptozotocin treatment (Sigma-Aldrich, St. Louis, MO), mice received streptozotocin (150 mg/kg) or vehicle (citrate buffer pH 4–4.5) intraperitoneally after overnight fasting. Blood glucose concentrations were measured with a glucometer (Accu-Chek Go; Roche). Blood samples (70 μL at 0, 10, 30, 60, and 180 min after glucose challenge) were collected from the tail into EDTA-precoated microvettes (Sarstedt) with DPP-IV inhibitor (DPP4-010 [Millipore] 4 μL/100 μL blood) to prevent inactivation of plasma GLP-1. Plasma insulin, leptin, peptide YY (PYY), and glucose-dependent insulinotropic polypeptide (GIP) were measured by Luminex technology (multiplex mouse gut hormones kit; Roche). Total GLP-1 was measured with an ELISA kit (Millipore).

Mice were killed. Brown and white inguinal, perirenal, and epididymal adipose tissues were removed and weighed. Pieces of jejunum or colon and pancreas tail were immediately fixed overnight at 4°C in formalin-acetic acid-alcohol before embedding in paraffin. Immunostaining was performed on 5-μm paraffin sections (22). Primary antibodies are listed in Supplementary Table 1. Horseradish peroxidase–labeled secondary antibodies were rabbit anti-guinea pig and goat anti-rabbit IgG (Amersham Biosciences). 3,3′-Diaminobenzidine was used for revelation. Tissue sections were counterstained with hematoxylin. For immunofluorescence, the secondary antibody was Alexa Fluor 546 donkey anti-rabbit (Molecular Probes). Nuclei were stained with DAPI. Immunostaining was examined by epifluorescence microscopy (Axiophot microscope, AxioCam camera, AxioVision 4.5 software; Zeiss). Numbers of chromogranin A–, GLP-1–, or PYY-positive cells per villus or per crypt of jejunum or colon were estimated in three to five animals per genotype.

Morphometric analysis of pancreatic islets in the whole pancreas was as previously described (28). Surfaces occupied by insulin staining and total pancreatic tissue were quantified (10× objective, Leica microscope and camera, Leica QWin software; Leica Microsystems GmbH, Wetzlar, Germany). The ratio of Ki67-positive cell number to islet area was quantified under a light microscope.

After flushing with PBS, the jejunum was cut into small pieces and incubated overnight (4°C) in Cell Recovery Solution (BD Biosciences) (29) containing 2% protease inhibitor cocktail (Sigma-Aldrich). Epithelial cell homogenate was filtered, washed with PBS, and centrifuged to obtain villus epithelial cells homogenized for RNA extraction.

Mouse islets were isolated with collagenase solution (1 mg/mL; Sigma-Aldrich), separated on a Histopaque gradient (Sigma-Aldrich), and handpicked under microscope (Leica Microsystems). Batches of 50 islets were sequentially incubated for 1 h at 37°C with 2.8 mmol/L glucose, 16.7 mmol/L glucose in the presence or absence of GLP-1 (Sigma-Aldrich), and 50 mmol/L KCl in Krebs-Ringer bicarbonate HEPES buffer. Total islet insulin content was extracted in acid ethanol (1.5% volume for volume HCl in 75% volume for volume ethanol). Islet insulin content and insulin secretion were assayed by ELISA kit (Mercodia).

Total RNA was isolated from jejunum or colon epithelial cells by RNA-PLUS solution (Qbiogene) and from pancreatic islets by RNeasy Mini Kit (Qiagen). Reverse transcription was performed in 20 μL reaction mixture with 5 μg or 400 ng RNA. Semiquantitative real-time PCR was performed with SYBR Green (Applied Biosystems) in a Stratagene system. Primer sequences are reported in Supplementary Table 2.

Vectors encoding rat HNF-4α2 (pMT2-HNF-4α), mouse HNF-4γ (pMT2-HNF-4γ), and β-galactosidase (pRSV-β-Gal) were as previously described (29). Vectors encoding the luciferase gene under the control of the −2.4-kb rat proglucagon promoter (pGL4-PPGLuc) (30,31) and Pax6 (pBAt14-mPax6) (32) were gifts from T. Kieffer (University of British Columbia, Vancouver, British Columbia, Canada) and M.S. German (University of California, San Francisco, San Francisco, CA), respectively. One microgram of DNA was transfected using X-tremeGENE HP DNA Transfection Reagent (Roche) into 12-well plates 24 h after plating COS-7 cells (0.75 × 10⁵ cells/well) or in suspended GLUTag cells before plating at high density (2.5 × 10⁵ cells/well) on Matrigel. The GLUTag cell line was kindly provided by D. Drucker (Mount Sinai Hospital, Toronto, Ontario, Canada). Cells were transfected with 200 ng pGL4-PPGLuc vector, 200 ng pRSV-β-Gal plasmid, and increasing amounts of pMT2-HNF-4α, pMT2-HNF-4γ, or pBAt14-mPax6 and analyzed at 48 h posttransfection.

Results are expressed as mean ± SEM. Statistical analyses were performed using GraphPad Prism (GraphPad Software, La Jolla, CA). Two-group comparisons were performed using Mann-Whitney U nonparametric tests. Two-way ANOVA (Bonferroni posttest) or one-way ANOVA (Kruskal-Wallis test, Dunn posttest) test was performed for comparisons of more than two groups. P < 0.05 was considered statistically significant.

Growth curves showed that the Hnf4g^−/− mice weighed more than the Hnf4g^+/+ mice, the difference becoming significant at 4 months (Supplementary Fig. 1A). Although Hnf4g^−/− mice did not eat more than Hnf4g^+/+ mice (Supplementary Fig. 1B), they were slightly larger (4%) (Supplementary Fig. 1C) and had heavier inguinal and perirenal adipose tissues, which represent visceral adipose tissue (Supplementary Fig. 1D), although no difference was found in the weights of brown and epididymal adipose tissues. We thus focused on energy metabolism, exploring glucose homeostasis in 2- and 6-month-old mice (before and after weights became statistically different between Hnf4g^−/− and Hnf4g^+/+ mice).

Although basal blood glucose and insulin levels did not differ significantly between 6-month-old Hnf4g^−/− and Hnf4g^+/+ mice (Table 1), during an OGTT, 2- and 6-month-old Hnf4g^−/− mice had a lower glycemic peak 30 min after the glucose bolus than the Hnf4g^+/+ mice (Fig. 1A). Furthermore, the area under the curve was twofold lower in Hnf4g^−/− than in Hnf4g^+/+ mice. Such an improvement in glucose tolerance could be due to a defect in intestinal glucose absorption, higher insulin sensitivity, or higher insulin secretion by pancreatic β-cells.

Glycemia did not differ between Hnf4g^−/− and Hnf4g^+/+ mice during the first 10 min of OGTT (Fig. 1B), indicating that the enterocyte glucose absorption was not altered by Hnf4g gene invalidation. Although basal insulinemia was similar in the two groups, the level of plasma insulin peak was threefold higher in Hnf4g^−/− mice than in control mice during OGTT (Fig. 1C, left panel). Furthermore, the insulinogenic index, which represents the first-phase insulin response to glucose challenge, was fivefold higher in Hnf4g^−/− than in control mice (Fig. 1C, right panel). Both results indicate that Hnf4g gene invalidation leads to an improvement of oral glucose–stimulated insulin secretion. Finally, measurement of blood glucose during an ITT indicated that 2- or 6-month-old Hnf4g^−/− mice were not more sensitive to insulin than Hnf4g^+/+ mice (Supplementary Fig. 2A and Fig. 1D). Contrary to OGTT, during IPGTT, no difference in glycemia (Supplementary Fig. 2B and Fig. 1E) or plasma insulin (area under the curve 42,157 ± 6,378 and 40,312 ± 6,548 pg/mL/min for Hnf4g^−/− and Hnf4g^+/+, respectively) (data not shown) was observed between Hnf4g^−/− and Hnf4g^+/+ mice (2 and 6 months old). Failure of pancreatic β-cells to secrete more insulin after intraperitoneal glucose challenge versus oral glucose challenge revealed a major role of intestine in glucose-stimulated insulin secretion improvement in Hnf4g^−/− mice.

We hypothesized an increased intestinal secretion of enteropeptides, such as GLP-1 or GIP, in response to oral glucose challenge only. Enteropeptide secretion is accompanied by amplification of insulin secretion through the so-called incretin effect, which is not observed upon intraperitoneal glucose administration.

Plasma concentrations of total GLP-1 and GIP were measured to determine incretin hormone levels. Because of the extensive and extremely rapid degradation of the active 7-36 amide GLP-1 (2–4-min half-life) by DPP-IV in plasma, we determined the total GLP-1 plasma concentration, which reflects the oral glucose-stimulated GLP-1 secretion. In 6-month-old mice, plasma GLP-1 concentration was higher in Hnf4g^−/− than in Hnf4g^+/+ mice both in the fasted state (Table 1) and in response to glucose challenge (Fig. 2A, left panel). By contrast, plasma GIP concentrations were similar in both groups in the fasted state (Table 1) and after an oral glucose challenge (Fig. 2A, right panel), suggesting a key role for GLP-1.

To validate a direct incretin effect of GLP-1 on the exaggerated insulin peak in Hnf4g^−/− mice, the GLP-1 receptor was blocked by its antagonist exendin (9-39), which inhibits GLP-1 signaling and its effect on insulin secretion. An OGTT was performed 30 min after intraperitoneal injection of exendin (9-39) or vehicle. Glucose tolerance was impaired in both groups of mice in the presence of exendin (9-39) compared with the condition without exendin (9-39) (Fig. 2B). Accordingly, plasma insulin decreased in both groups during OGTT in the presence of exendin (9-39) (Fig. 2C). However, the glucose tolerance of Hnf4g^−/− mice in the presence of exendin (9-39) was similar to that of Hnf4g^+/+ mice without exendin (9-39). Ten minutes after glucose challenge in the presence of exendin (9-39) or not, plasma GLP-1 (Fig. 2D, left panel) but not plasma GIP (Fig. 2D, right panel) increased significantly in Hnf4g^−/− compared with Hnf4g^+/+ mice. These results demonstrate a direct effect of GLP-1 on the improved response of Hnf4g^−/− mice to a glucose tolerance test.

The size of crypts and villi was measured in 6-month-old Hnf4g^−/− and Hnf4g^+/+ mice, and no obvious difference between both groups was observed (Supplementary Fig. 3A). The enteroendocrine cells were counted in jejunum of Hnf4g^−/− and Hnf4g^+/+ mice after immunostaining of chromogranin A, a specific marker of secretion granules of endocrine cells. In jejunum of Hnf4g^−/− mice, the enteroendocrine cell number doubled in villi and remained unchanged in crypts compared with Hnf4g^+/+ mice (Fig. 3A). Quantification of the GLP-1–positive cells showed that Hnf4g gene invalidation induced an increase of GLP-1–secreting cell number in jejunum (Fig. 3B ) and colon (Supplementary Fig. 3B). PYY-positive cell number was also higher in colons of 6-month-old Hnf4g^−/− mice compared with Hnf4g^+/+ mice (Supplementary Fig. 3C).

To further identify the molecular mechanisms underlying the increase in number of GLP-1–producing cells, the expression of transcription factors known to affect enteroendocrine cell differentiation was quantified. The expression of Foxa1 and Foxa2 was increased by 1.97- and 2.68-fold, respectively, in 2-month-old Hnf4g^−/− mice (Fig. 3C). At 6 months, only Isl1 expression was increased in these mice (1.63-fold) (Fig. 3D). Of note, the mRNA level of Hnf4a was twofold higher in jejunum (Fig. 3C and D) and colon (Supplementary Fig. 3D) of Hnf4g^−/− mice than in Hnf4g^+/+ mice.

To check whether HNF-4α or HNF-4γ could regulate GLP-1 transcription directly, we analyzed their effect on a luciferase reporter gene under the control of the proglucagon promoter (30). In COS-7 cells or enteroendocrine GLUTag cells, no direct activation of the proglucagon promoter by HNF-4α or HNF-4γ was seen, whereas Pax6 transactivated this promoter, as expected (32,33) (Supplementary Fig. 4). The results demonstrate that no direct connection exists between an increase in Hnf4a gene expression and proglucagon gene expression.

Of note, the mRNA levels of the enterohormones proglucagon (Gcg), which gives rise to GLP-1 after posttranslational maturation; Gip; and Cck were not modified in jejunum of Hnf4g^−/− mice compared with Hnf4g^+/+ mice (Fig. 3E). Thus, the higher plasma GLP-1 level relied on the increased GLP-1–positive cell number.

By its trophic effect on pancreatic β-cells, an increased circulating GLP-1 level could also account for an increased β-cell mass and function, leading to the exaggerated insulin peak during OGTT. Having demonstrated an increased number of GLP-1–positive cells and a direct increased incretin effect of GLP-1 in Hnf4g^−/− mice, we hypothesized an augmented GLP-1 trophic effect on pancreatic β-cells. The β-cell fraction of 6-month-old Hnf4g^−/− and Hnf4g^+/+ mice was measured after immunostaining of insulin (Fig. 4A). The morphometric analysis showed that the islet density did not differ in both groups (Fig. 4B, left panel). However, the β-cell fraction (Fig. 4B, middle panel), the islet mean size (Fig. 4B, right panel), and the density of larger-sized islets (Supplementary Fig. 5A) were significantly higher in Hnf4g^−/− than in Hnf4g^+/+ mice. The ex vivo insulin release in response to 2.8 or 16.7 mmol/L glucose did not differ significantly between isolated islets from Hnf4g^+/+ or Hnf4g^−/− mice (Fig. 4C). However, isolated islets from Hnf4g^−/− mice contained twofold more insulin than those isolated from Hnf4g^+/+mice (Fig. 4D). Pancreatic β-cells from Hnf4g^−/− mice could also be more sensitive to GLP-1 and/or express more GLP-1 receptor. We performed a GLP-1 dose-response assessment of glucose-stimulated insulin secretion in isolated islets (Supplementary Fig. 5C) and did not observe any modification of insulin secretion by islets from Hnf4g^−/− compared with Hnfg^+/+ mice. The Glp1r mRNA level was not modified in islets of Hnf4g^−/− compared with Hnf4g^+/+ mice (Supplementary Fig. 5B). These data suggest that the in vivo exaggerated insulin peak in response to glucose in Hnf4g^−/− mice depends on the pancreatic β-cell abundance rather than on β-cell insulin secretory capacity or GLP-1 sensitivity. We did not observe a modification in the expression of Pdx1, NeuroD, Glut2, Ins1, Ins2, and Gck as well as of Hnf4a mRNA in islets from Hnf4g^−/− mice versus Hnf4g^+/+ mice, suggesting that β-cell function was not affected by the HNF-4γ invalidation (Supplementary Fig. 5B). Furthermore, the number of Ki67-positive proliferative cells was fourfold higher in Hnf4g^−/− than in Hnf4g^+/+ pancreatic islets (Fig. 4E). Finally, the massive hyperglycemia provoked by streptozotocin, a β-cytotoxic drug, was delayed by 24 h in Hnf4g^−/− compared with Hnf4g^+/+ mice (Fig. 4F). Such a result was consistent with the increased number of insulin-producing β-cells in the pancreas of Hnf4g^−/− mice.

The results demonstrate that the nuclear receptor HNF-4γ plays a critical role in glucose homeostasis and that loss of HNF-4γ affects the enteroendocrine cell lineage, inducing increases in GLP-1–positive cell numbers and basal and glucose-stimulated GLP-1 plasma levels. This leads to an exaggerated glucose-induced insulin secretion that improves glucose tolerance of Hnf4g^−/− mice through an augmentation of the GLP-1 incretin effect and a trophic impact on pancreatic β-cell mass (Fig. 5).

Mice invalidated for Hnf4g gene expression were reported to present a higher weight and a lower food intake, associated with lower night energy expenditure, than wild-type mice (25). Control of food intake plays a major role in energy homeostasis and weight in mammals. Among hormones that control food intake, leptin, which is secreted by adipose tissue, and PYY, which is secreted by enteroendocrine cells, are anorexigens (34). The current results confirm that Hnf4g^−/− mice are heavier than Hnf4g^+/+ mice, but we did not observe difference in food intake, despite an increased PYY plasma level in Hnf4g^−/− mice (Table 1). One hypothesis is that the level of leptin compensates for the increased level of PYY in Hnf4g^−/− mice, but the plasma leptin level was similar in the two groups of mice (Table 1). Another hypothesis is that increased PYY and GLP-1 levels enhance intestinal transit time and nutrient absorption efficiency, which would in turn lead to weight gain.

The weight gain in Hnf4g^−/− mice was likely due to their larger size and an increased amount of white adipose tissue. The visceral fat augmentation suggested metabolic disorders and prompted us to investigate the insulin sensitivity of Hnf4g^−/− mice. Of note, disruption of the Hnf4g gene led to an increased glucose tolerance that was due to neither the inhibition of glucose absorption nor the higher insulin sensitivity but, rather, to increased insulinemia. These increases of glucose tolerance and insulin secretion in response to glucose were under the control of intestinal functions because they were not observed when intestinal glucose absorption and signaling were bypassed during an intraperitoneal glucose challenge. The specificity of Hnf4g gene disruption for this phenotype was assessed in mice with specific intestinal invalidation of Hnf4a (22). During OGTT, neither glucose tolerance nor plasma insulin levels differed between 6-month-old Hnf4a^lox/lox and Hnf4a^Δint mice (Supplementary Fig. 6).

In response to glucose, incretin hormones, such as GLP-1 and GIP, which are secreted by enteroendocrine cells, potentiate insulin secretion (35). Only the basal and glucose-stimulated plasma GLP-1 levels were increased in Hnf4g^−/− mice compared with Hnf4g^+/+ mice. The increased mean islet size and the higher density of large pancreatic islets in Hnf4g^−/− compared with control mice were consistent with the role of GLP-1 in pancreatic β-cell mass maintenance (36–38) and proliferation (39,40). HNF-4γ loss affected β-cell abundance but not β-cell insulin secretory capacity. Finally, the larger islets led to a resistance to streptozotocin, a β-cell cytotoxic drug, suggesting that HNF-4γ is involved in susceptibility to diabetes.

HNF-4α is expressed in pancreas and induces insulin gene transcription (41) and insulin secretion (42). A compensatory effect of the loss of HNF-4γ by HNF-4α could explain the hyperinsulinemia in response to glucose in pancreas. However, the expression of Hnf4a was not modified in pancreas of Hnf4g^−/− mice (Supplementary Fig. 5B), supporting the impact of HNF-4γ loss and the role of intestine on glucose homeostasis through a direct effect of GLP-1, as assessed by the use of the GLP-1 antagonist exendin (9-39). In accordance with the increased GLP-1 plasma level, we show that Hnf4g gene invalidation leads to modifications of enteroendocrine cell lineage characterized by an increased number of GLP-1–producing cells in jejunum and colon.

The expression of Foxa1, Foxa2, and Isl1 was enhanced, suggesting that modifications of the transcription factor network favored the GLP-1–secreting cell lineage (43,44). FOXA1 and FOXA2 are known to control the expression of Hnf4a (45,46). Furthermore, they are considered pioneer factors that engage chromatin before other transcription factors (47). One may hypothesize that in the absence of HNF-4γ, Foxa1/2 levels increase during development to young adulthood and allow an overexpression of HNF-4α throughout life. Whether the observed effects were the direct consequence of the loss of HNF-4γ or resulted indirectly from the HNF-4α increment in Hnf4g^−/− mouse intestine is difficult to assert.

HNF-4α and HNF-4γ, which are encoded by two different genes, share high homology in their DNA-binding or ligand-binding domains (16,18). In vitro, HNF-4α and HNF-4γ display the same apolipoprotein A-IV promoter transactivation capacity and binding affinity to hormone-responsive elements (29). A chromatin immunoprecipitation sequencing study failed to discriminate between HNF-4α and HNF-4γ binding sites (48). However, these two transcription factors 1) have a different spatial distribution along the crypt-villus axis, with HNF-4α being expressed along the crypt-villus axis and HNF-4γ being restricted to the differentiated villus compartment, and 2) have poor homology in their transactivation and regulation domains. These observations suggest that these two forms of HNF-4 play specific roles. After an inducible and intestine-specific knockout of the Hnf4a gene, we demonstrated that HNF-4α controls the intestinal epithelium homeostasis and cell architecture (22) and the uptake of fatty acids by enterocytes (23). HNF-4α is also required for absorptive function and ion transport in colon (49,50). Despite an ectopic expression of HNF-4γ in crypts of Hnf4a^Δint mice (22), a potential compensatory effect of HNF-4γ on proliferation was weak because Hnf4a^Δint mice display tumors in colon 4–6 months after injection of azoxymethane, a potent carcinogen in colon (24). Of note, the expression of Hnf4a was increased in jejunum and colon of Hnf4g^−/− mice compared with Hnf4g^+/+ mice. Such Hnf4a overexpression could affect the absorptive function and water status in colon.

We previously showed that the enteroendocrine cell population (chromogranin A–positive cells) as well as the level of proglucagon and Gip-encoding mRNA decrease in villi of Hnf4a^Δint mice compared with control mice (22). Altogether, these results suggest that HNF-4γ is a negative modulator of the enteroendocrine cell lineage, whereas HNF-4α is a positive one. Thus, we conclude that the balance between the expression of HNF-4α and HNF-4γ must be finely regulated to maintain enteroendocrine cell lineage and glucose homeostasis (Fig. 5).

Acknowledgments. The authors thank A. Benkhoui (Centre de Recherche des Cordeliers) for hormone dosage by Luminex, S. Saint-Just (Centre de Recherche des Cordeliers) for mouse line maintenance, C. Ayassamy (Centre d’Explorations Fonctionnelles) for mouse care, and C. Amorin (Centre de Génotypage et de Biochimie) for mouse genotyping.

Funding. This work was supported by INSERM, France, and Université Pierre et Marie Curie (UPMC), Paris, France. F.B. received a doctoral fellowship from Cancéropôle, Île-de-France. S.A. received a doctoral fellowship from UPMC.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.B. and A.R. researched data and contributed to the discussion and writing, review, and editing of the manuscript. S.A. and P.S. researched data and contributed to the discussion. V.C., B.B., G.G., M.L., and P.C. researched data and contributed to the discussion and review and editing of the manuscript. C.O. researched data. K.G. researched data and contributed to the review and editing of the manuscript. M.R. contributed to the discussion and writing, review, and editing of the manuscript. A.R. is the guarantor of this work and, as such, had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Part of these data was presented at the Congress of the Société Francophone du Diabète, Paris, France, 11–14 March 2014.