Hepatocyte Nuclear Factor-4α Is Essential for Glucose-stimulated Insulin Secretion by Pancreatic β-Cells*

Mutations in the hepatocyte nuclear factor (HNF)-4α gene cause a form of maturity-onset diabetes of the young (MODY1) that is characterized by impairment of glucose-stimulated insulin secretion by pancreatic β-cells. HNF-4α, a transcription factor belonging to the nuclear receptor superfamily, is expressed in pancreatic islets as well as in the liver, kidney, and intestine. However, the role of HNF-4α in pancreatic β-cell is unclear. To clarify the role of HNF-4α in β-cells, we generated β-cell-specific HNF-4α knock-out (βHNF-4αKO) mice using the Cre-LoxP system. The βHNF-4αKO mice exhibited impairment of glucose-stimulated insulin secretion, which is a characteristic of MODY1. Pancreatic islet morphology, β-cell mass, and insulin content were normal in the HNF-4α mutant mice. Insulin secretion by βHNF-4αKO islets and the intracellular calcium response were impaired after stimulation by glucose or sulfonylurea but were normal after stimulation with KCl or arginine. Both NAD(P)H generation and ATP content at high glucose concentrations were normal in the βHNF-4αKO mice. Expression levels of Kir6.2 and SUR1 proteins in the βHNF-4αKO mice were unchanged as compared with control mice. Patch clamp experiments revealed that the current density was significantly increased in βHNF-4αKO mice compared with control mice. These results are suggestive of the dysfunction of KATP channel activity in the pancreatic β-cells of HNF-4α-deficient mice. Because the KATP channel is important for proper insulin secretion in β-cells, altered KATP channel activity could be related to the impaired insulin secretion in the βHNF-4αKO mice.

Mutations in the hepatocyte nuclear factor (HNF)-4␣ gene cause a form of maturity-onset diabetes of the young (MODY1) that is characterized by impairment of glucose-stimulated insulin secretion by pancreatic ␤-cells. HNF-4␣, a transcription factor belonging to the nuclear receptor superfamily, is expressed in pancreatic islets as well as in the liver, kidney, and intestine. However, the role of HNF-4␣ in pancreatic ␤-cell is unclear. To clarify the role of HNF-4␣ in ␤-cells, we generated ␤-cell-specific HNF-4␣ knock-out (␤HNF-4␣KO) mice using the Cre-LoxP system. The ␤HNF-4␣KO mice exhibited impairment of glucose-stimulated insulin secretion, which is a characteristic of MODY1. Pancreatic islet morphology, ␤-cell mass, and insulin content were normal in the HNF-4␣ mutant mice. Insulin secretion by ␤HNF-4␣KO islets and the intracellular calcium response were impaired after stimulation by glucose or sulfonylurea but were normal after stimulation with KCl or arginine. Both NAD(P)H generation and ATP content at high glucose concentrations were normal in the ␤HNF-4␣KO mice. Expression levels of Kir6.2 and SUR1 proteins in the ␤HNF-4␣KO mice were unchanged as compared with control mice. Patch clamp experiments revealed that the current density was significantly increased in ␤HNF-4␣KO mice compared with control mice. These results are suggestive of the dysfunction of K ATP channel activity in the pancreatic ␤-cells of HNF-4␣-deficient mice. Because the K ATP channel is important for proper insulin secretion in ␤-cells, altered K ATP channel activity could be related to the impaired insulin secretion in the ␤HNF-4␣KO mice.
Hepatocyte nuclear factor (HNF) 2 -4␣, a transcription factor belonging to the nuclear hormone receptor superfamily (NR2A1), is expressed in the liver, kidney, intestine, and pancreas (1). Similar to other nuclear receptors, HNF-4␣ has several functional domains, including the N-terminal transactivation domain (AF-1), a DNAbinding domain, a functionally complex C-terminal region that forms a ligand-binding domain, and a dimerization interface and a transactivation domain (AF-2). In the liver, HNF-4␣ plays an important role in regulating various genes involved in glucose, fatty acid, amino acid, and cholesterol metabolism, as well as blood coagulation and hepatic development and differentiation (2,3).
Maturity-onset diabetes of the young (MODY) is a genetically heterogeneous monogenic disorder that accounts for 2-5% of type 2 diabetes. It is characterized by autosomal dominant inheritance and an early age of onset (usually at Ͻ25 years old) (4). We have shown previously that heterozygous mutations of the HNF-4␣ gene can cause a particular form of MODY (MODY1) (5). Functional studies of the mutations in MODY1 patients have shown that the diabetes is caused by loss-offunction mutations (6,7). Clinical studies have also shown that the primary cause of MODY1 is an impairment of glucose-stimulated insulin secretion by pancreatic ␤-cells rather than liver dysfunction (8,9), indicating that loss of HNF-4␣ leads to abnormal insulin secretion from the ␤-cells. In addition, recent genetic studies have shown that single nucleotide polymorphisms in the P2 (pancreatic ␤-cell type) promoter of the HNF-4␣ gene are associated with type 2 diabetes in some populations (10 -12). Thus, not only do HNF-4␣ mutations cause MODY, but variations of the HNF-4␣ gene are associated with a genetic predisposition to common type 2 diabetes. These findings suggest that HNF-4␣ in pancreatic ␤-cells has an important role in maintaining normal glucose metabolism. Previous in vitro studies have revealed that HNF-4␣ regulates the expression of pancreatic ␤-cell genes involved in glucose metabolism (e.g. insulin and glucose transporter-2 (GLUT2)) as well as HNF-1␣ (13). HNF-1␣ is a target gene of HNF-4␣ in the liver (14), and mutations of the HNF-1␣ gene cause type 3 of MODY (MODY3) (15). However, the role of HNF-4␣ in the regulation of pancreatic ␤-cell gene expression and its correlation with metabolism-secretion coupling are still unclear. Targeted disruption of the hnf4␣ gene in mice results in early embryonic death because of dysfunction of the visceral endoderm (2), and this embryonic lethality prevents further analysis of the role of HNF-4␣ in pancreatic ␤-cells. Furthermore, in contrast to humans, heterozygous hnf4␣ (ϩ/Ϫ) mice exhibit normal glucose tolerance (6). To examine the role of HNF-4␣ in pancreatic ␤-cells, we specifically disrupted HNF-4␣ in ␤-cells by employing the Cre-loxP system. Pancreatic ␤-cell-specific HNF-4␣ knock-out mice (␤HNF-4␣KO mice) were viable and exhibited impairment of glucosestimulated insulin secretion, which is a characteristic of human MODY1. The secretory response of insulin and the intracellular Ca 2ϩ response to glucose or tolbutamide were reduced in isolated HNF-4␣ mutant islets. Insulin secretion and the intracellular Ca 2ϩ response to glucose or tolbutamide were decreased in isolated mutant HNF-4␣ islets. In addition, the responsiveness of the K ATP channel current density to high glucose was decreased in ␤HNF-4␣KO mice.

EXPERIMENTAL PROCEDURES
Animals and Genotyping-HNF-4␣ flox/flox mice with loxP sites flanking exons 4 and 5 of the HNF-4␣ gene were generated as described previously (16). Mice expressing Cre recombinase under control of the rat insulin 2 gene promoter (RIP-Cre) have been described elsewhere (17). RIP-Cre mice were backcrossed onto the C57BL/6 genetic background more than seven generations and maintained. Heterozygous (flox/wt) mice carrying one copy of the RIP-Cre transgene were interbred with flox/wt littermates lacking Cre (more than three generations) to generate pancreatic ␤-cell-specific HNF-4␣ knock-out mice (␤HNF-4␣KO) and littermate control mice (flox/flox (FLOX), RIP-Cre (CRE), and wt/wt (WT)). Age-and sex-matched littermates were used as the controls. The mice were genetic hybrids of C57BL/6 and 129/SvJ mice. Genotyping of HNF-4␣ was performed by PCR, as described previously (16). PCR to detect the Cre transgene was achieved with 35 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s using the following specific primers: RIP-CreF (5Ј-TTACCGGTCGATGCAACGAGT-GATG-3Ј) and RIP-CreR (5Ј-TTCCATGAGTGAACGAACCTG-GTCG-3Ј). This study was performed according to the guidelines of the Animal Ethics Committee of Osaka University.
Measurement of Glucose and Insulin Levels-Blood glucose levels were determined with a glucometer (Glutest Ace, Sanwa Kagaku, Japan). Plasma insulin concentrations were determined with a Glazyme insulin enzyme immunoassay kit (Wako, Japan). The pancreatic insulin content or islet insulin content was measured after extraction by the acid-ethanol method (19). A glucose tolerance test was performed at 10 and 24 weeks of age. Mice were fasted for 14 h, and blood was collected before and at 15, 30, 60, and 120 min after the intraperitoneal injection of glucose (1 g/kg or 2 g/kg). For the insulin tolerance test, mice were injected with 1.0 units/kg of regular human insulin after a 4-h fast.
Batch Incubation Experiments-Batch incubation was performed as described previously (19). Briefly, pancreatic islets (10 per tube) were preincubated at 37°C for 60 min in Hepes-Krebs buffer supplemented with 0.5% (w/v) BSA. Then the islets were incubated for 30 min in 350 l of the same buffer containing 500 M tolbutamide or the buffer containing 16.7 mM glucose in the presence or absence of 1 nM GLP-1 (Peptide Institution, Osaka, Japan).
Intracellular Ca 2ϩ Concentration of Pancreatic Islets-The cytosolic Ca 2ϩ concentration was measured with fura-2 acetoxymethyl ester using two-photon excitation imaging as described previously (20,21). In brief, isolated islets were incubated at 37°C for 30 min in serum-free Dulbecco's modified Eagle's medium culture medium supplemented with bovine serum albumin (1 mg/ml), 0.1% Pluronic F-127, and 20 M fura-2 acetoxymethyl ester (Molecular Probes), followed by washing with Sol A (140 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM Hepes-NaOH (pH 7.3), and 2.8 mM glucose). The laser power at the specimen was set at 10 milliwatts, and the excitation wavelength was 830 nm. Images were acquired every 2.5 s. The cytosolic Ca 2ϩ concentrations in response to 20 mM glucose, 1 M glibenclamide, and 30 mM KCl were calculated as (F 0 Ϫ F)/F 0 , where F 0 and F represent the resting and post-stimulation fluorescence, respectively.
Measurement of NAD(P)H Fluorescence-Autofluorescence of NAD(P)H (400 -510 nm) in the islets was imaged using two-photon excitation microscopy at an excitation wavelength of 720 nm (22).

Measurement of the Islet ATP Content-
The ATP content of pancreatic islets was measured by a quantitative bioluminescence assay using a modification of the protocol described previously (23). Experiments were performed on islets from ␤HNF-4␣KO mice and control mice after incubation for 48 h in RPMI 1640 medium containing 11 mM glucose. Groups of 10 islets (diameter 0.15-0.2 mm) were selected manually and preincubated in KRB containing 3 mM glucose for 60 min. After preincubation, the islets were incubated for 30 min in buffer containing 3 or 20 mM glucose. Then ATP was extracted from the islets and assayed using an ATP bioluminescence assay kit (CLS II, Roche Applied Science) according to the manufacturer's instructions.
Patch Clamp Recording-A patch clamp technique (25,26) was used to record whole-cell currents from perforated cells with the pipette solution containing nystatin (150 g/ml) dissolved in 0.1% Me 2 SO. Membrane currents were recorded by using an amplifier (200B Axopatch, Molecular Devices, Foster City, CA), and data were stored on-line in a computer with pCLAMP8 software. Voltage clamp in the perforated mode was defined as adequate when series resistance was less than 20 megohms. The patch pipettes were pulled from glass tubes (Narishige Co. Ltd., Tokyo, Japan). The resistance of each pipette was between 4 and 7 megohms when filled with the following pipette solution (content in mM): K 2 SO 4 40, KCl 50, MgCl 2 5, EGTA 0.5, and Hepes 10 (pH 7.2) with KOH. The perforated whole-cell clamp mode was established within 5-10 min after formation of a gigaohm seal between the patch pipette and the membrane of an isolated ␤-cell. These cells were superfused with BSAfree Hepes-Krebs buffer containing 2.8 mM glucose for at least 30 min before the experiments. The capacitance of cells from control mice and ␤HNF-4␣KO mice was 7.4 Ϯ 2.1 picofarads (n ϭ 20) and 7.6 Ϯ 2.7 picofarads (n ϭ 16), respectively. The ␤-cells were voltageclamped at the holding potential of Ϫ70 mV, stepped to Ϫ100 mV for voltage ramping from Ϫ100 to Ϫ50 mV at a rate of 5 mV/100 ms,  Pancreatic ␤-Cell-specific HNF-4␣ Knock-out Mice FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 and then stepped back to Ϫ70 mV every 10 s. These experiments were performed at room temperature (22-25°C).
Statistical Analysis-Results were expressed as the mean Ϯ S.E. Statistical analysis was performed with the unpaired t test, and differences were considered to be significant at p Ͻ 0.05.
Impaired Glucose Tolerance in ␤HNF-4␣ Knock-out Mice-␤HNF-4␣KO mice were born in a Mendelian fashion (data not shown). On inspection, ␤HNF-4␣KO mice were indistinguishable from their control littermates (wt/wt (WT), flox/flox (FLOX), and RIP-Cre (CRE)). Body weight ( Fig. 2A) and the nonfasting blood glucose concentration (Fig. 2B) were similar in ␤HNF4␣KO mice and control mice at both 10 and 24 weeks. Because random blood glucose levels were similar in the ␤HNF-4␣KO mice and control littermates, the response to intraperitoneal administration of glucose was measured because this is a more sensitive determinant of glucose tolerance. Glucose tolerance was found to be normal in the male ␤HNF-4␣KO mice at 10 weeks (Fig. 2C). However, glucose levels at 15, 30, 60, and 120 min after administration of a glucose load were significantly higher in female KO mice at 10 weeks (Fig. 2D). Female mutant mice also showed a poor initial response of insulin secretion after intraperitoneal injection of glucose. At 24 weeks, both male and female ␤HNF-4␣KO mice exhibited glucose intolerance and an impaired insulin response to glucose loading (Fig. 2E). Most interestingly, the HNF-4␣-deficient mice generated in this study showed no evidence of a decrease in blood glucose or elevation of plasma insulin levels in the fasting state (Fig. 2, C--F), as was reported in another strain of ␤-cell-specific HNF-4␣ knock-out mice (28). Their knock-out mice also exhibited hyperinsulinemic hypoglycemia in the fed state (28), but our mutant animals also did not show hyperinsuline-mic hypoglycemia in the fed state (Fig. 2G). The insulin tolerance test did not reveal any significant difference in insulin sensitivity between ␤HNF-4␣KO mice and control mice (Fig. 2H), excluding the possibility that the mutant mice were insulin-resistant.
Pancreatic Histology and Insulin Content in ␤HNF-4␣KO Mice-MODY3 model mice have an abnormal islet structure (19,23). To determine whether any abnormalities of islet morphology were present, pancreatic sections from 24-week-old mice with different genotypes were examined by immunohistochemistry (Fig. 3A). It was found that the islet architecture was normal in ␤HNF-4␣KO mice, featuring a core of ␤-cells surrounded by a layer of non-␤-cells. Also, the islet levels of insulin, glucagon, somatostatin, and pancreatic polypeptide expression were similar in KO and control animals. Quantitative analysis revealed no significant difference in the islet mass, islet number, islet size distribution, and ␤-cell size between FLOX and KO mice (Fig. 3, B-D and supplemental Fig. 1). Furthermore, the pancreatic insulin content and insulin content of islets of ␤HNF-4␣KO mice were similar to that of control mice (Fig. 3, E and F). These results suggested that loss of HNF-4␣ from pancreatic ␤-cells led to the impairment of ␤-cell function rather than causing abnormalities of islet differentiation or insulin biosynthesis.
Pancreatic Islet Gene Expression-The effect of HNF-4␣ deficiency on gene expression was evaluated in the islets of 24-week-old ␤HNF-4␣ mice and FLOX mice by real time RT-PCR (Fig. 4). Previous in vitrostudies have suggested that insulin, HNF-1␣, GLUT2, and aldolase B are target genes for HNF-4␣ in ␤-cells (13). However, expression of the insulin 1, insulin 2, and HNF-1␣ genes was unchanged in the present study, in agreement with the results mentioned above (Figs. 1B and 3E). Expression of the genes encoding GLUT2, aldolase B, glucokinase, glucose-6-phosphatase, VDCC, syntaxin-1, and SNAP-25, which are involved in insulin secretion by ␤-cells, was also similar in ␤HNF-4␣ islets and control islets. Furthermore, we found that the expression of several transcription factors that are abundant in the islets, including PDX-1, NeuroD, Nkx6.1, Pax6, Foxo1, PGC-1␣, and HNF-1␤, did not differ between ␤HNF-4␣KO mouse islets and control mouse islets. It has been reported that HNF-4␣ may directly regulate the expression of the small heterodimer partner (SHP), an orphan nuclear receptor that forms heterodimers with various nuclear receptors, including HNF-4␣ (29). However, SHP expression was unchanged in ␤HNF-4␣KO mice. PPAR␣ is a transcription factor belonging to the nuclear receptor family that regulates the expression of genes involved in the ␤-oxidation of fatty acids (30). Expression of PPAR␣ was reduced in liver-specific HNF-4␣KO mice (16). Consistent with this result, PPAR␣ gene expression was reduced in ␤HNF-4␣KO islets relative to the level seen in control animals, suggesting that PPAR␣ was a target of HNF-4␣ in pancreatic ␤-cells.
Insulin Secretion Profile of Perifused Islets and Changes of Intracellular Ca 2ϩ -To define the defect of insulin secretion in ␤HNF-4␣KO mice, isolated islets were stimulated with several different secreta- gogues. Glucose stimulates insulin secretion by inducing the closure of K ATP channels via generation of ATP. Glycolysis generates ATP both directly and indirectly via the production of NADH. An increase of the ATP/ADP ratio leads to membrane depolarization and to an increase of cytosolic [Ca 2ϩ ] i , via the VDCC, and eventually results in the exocytosis of insulin-containing secretory granules (31). First, we tested insulin secretion from perifused pancreatic islets after stimulation with glucose. There was no significant difference in basal insulin secretion between ␤HNF-4␣KO islets and control islets in the presence of 2.8 mM glucose. However, insulin secretion in response to a high glucose concentration (20 mM) was significantly reduced in ␤HNF-4␣KO islets compared with control islets (peak insulin response, WT; 102.2 pg/min/islet, FLOX; 141.2 pg/min/islet, CRE; 141.1 pg/min/islet, ␤HNF-4␣KO; 51.1 pg/min/islet, p Ͻ 0.05) (Fig. 5A and data not shown). In contrast, insulin secretion in response to depolarization induced by 25 mM KCl (Fig. 5A) or 20 mM arginine (Fig. 5B) was normal in ␤HNF-4␣KO mice, indicating that there was no defect of the later steps of insulin secretion after elevation of [Ca 2ϩ ] i in ␤HNF-4␣KO islets. GLP-1 (glucagon-like peptide-1) potentiates insulin secretion from pancreatic ␤-cells by increasing intracellular cAMP concentration (32). The insulinotropic action of GLP-1 was also preserved in ␤HNF-4␣KO islets (FLOX, 2.2-fold; KO, 2.5-fold) (Fig. 5C). The glycerol-3-phosphate dehydrogenase reaction forms NADH ϩ H ϩ , which can be oxidized by the respiratory chain in mitochondrion to produce ATP. To evaluate further the reasons for defective glucose-stimulated insulin secretion by ␤HNF-4␣KO islets, we measured NAD(P)H generation by two-photon excitation microscopy (Fig. 6A). Changes in the fluorescence intensity in response to a glucose load were similar for ␤HNF-4␣KO and control islets, suggesting that there was no defect in NADH formation. The ATP content was also measured in islets from ␤HNF-4␣KO mice and control mice after incubation with 3 and 20 mM glucose using a quantitative bioluminescence assay (Fig. 6B). It was found that the ATP levels in ␤HNF-4␣KO islets were similar to those in control islets, suggesting that glucose metabolism was normal in the mutant islets.
Changes of Islet Intracellular Ca 2ϩ -Next, we examined whether the impairment of glucose-stimulated insulin secretion was associated with a reduced intracellular Ca 2ϩ level. Changes of [Ca 2ϩ ] i in ␤HNF-4␣KO islets after a glucose load were significantly smaller compared with those in control islets (Fig. 7A). In contrast, both the amplitude and time course of the KCl-induced increase of [Ca 2ϩ ] i did not differ between the two groups (Fig. 7B). Sulfonylurea-like glibenclamide stimulates insulin secretion by blocking K ATP channels. As was the case for glucose, the [Ca 2ϩ ] i response to glibenclamide stimulation was significantly decreased (Fig. 7C). The rise of [Ca 2ϩ ] i in response to tolbutamide stimulation was also decreased (Fig. 7D). In agreement with the above results, stimulation of ␤HNF-4␣KO islets using tolbutamide only produced 37% of the insulin secretion seen with control islets (Fig. 7E). Taken together, these results indicate that K ATP channel function is defective in ␤HNF-4␣KO mice.
Expression of Kir6.2 and SUR1 Proteins by ␤HNF-4␣KO Islets-The K ATP channels in pancreatic ␤-cells are composed of Kir6.2 and SUR1 (33). Because lack of Kir6.2 or SUR1 expression leads to defective insulin secretion in response to glucose or sulfonylurea stimulation (34,35), we evaluated the level of Kir6.2 and SUR1 expression in ␤HNF-4␣KO islets by Western blotting. Expression of Kir6.2 was significantly reduced in the previously reported ␤-cell-specific HNF-4␣(Ϫ/Ϫ) mice (28). In marked contrast, expression levels of Kir6.2 protein were not reduced in the islets of our ␤HNF-4␣KO mice (Fig. 8, A and B, and supplemental  Fig. 2). Expression of Kir6.2 mRNA was also unchanged in ␤HNF-4␣KO mice (Fig. 8C). Adequate surface expression of K ATP channels is neces-sary for normal insulin secretion (24,36). To examine the expression of Kir6.2 on the cell surface, we performed immunofluorescent staining using Kir6.2 antibody (sc-11228), which recognizes the extracellular domain of Kir6.2 (33). As shown in Fig. 8D, the level of Kir6.2 expression was similar in FLOX and KO islets. In addition, SUR1 protein was also expressed at similar levels in both ␤HNF-4␣KO mice and control mice (Fig. 8, D and E). These results indicated that the K ATP channel defect in ␤HNF-4␣KO mice was not caused by reduced expression of Kir6.2 or SUR1.
Changes of K ATP Channel Current Density in ␤HNF-4␣KO Islets- Fig. 9A displays the K ATP channel currents recorded during voltage ramping from Ϫ100 to Ϫ50 mV. Because the K ATP channel has voltageindependent kinetics and is a major determinant of the resting potential of ␤-cells, the currents elicited by a voltage ramp run through these channels. In control cells exposed to 2.8 mM glucose, the current direction was reversed around Ϫ70 mV, and the current was substantially inhibited from 3 to 10 min after increasing the glucose concentration to 8.3 mM (Fig. 9A, left panel). The inhibited current did not reverse its direction during the ramp procedure, indicating that the resting membrane potential was higher than Ϫ50 mV at this glucose concentration. On the other hand, the ␤-cells from ␤HNF-4␣KO mice yielded a current that crossed 0 even at 8.3 mM, with a subtle shift of the reversal potential despite marked reduction of the current amplitude (right panel). Thus, in ␤-cells from ␤HNF-4␣KO mice, the K ATP channel cur- rent showed a decrease of amplitude in response to 8.3 mM glucose, but the reduction was insufficient to allow membrane depolarization to reach the threshold potential of VDCC and produce an action potential. Among the 10 ␤HNF-4␣KO ␤-cells tested, three cells were responsive to glucose stimulation and showed a concurrent reduction of the K ATP channel current without any action potential (Fig. 9A). Three cells responded to 8.3 mM glucose with an action potential, but the other four cells were unresponsive to glucose stimulation regarding the current inhibition. In contrast to these cells from ␤HNF-4␣KO mice, we observed good glucose responsiveness in all four control ␤-cells, and action potentials occurred after an increase of the glucose concentration to 8.3 mM. The current density recorded from cells of ␤HNF-4␣KO mice at 2.8 mM glucose was significantly increased compared with that from control cells at the same glucose concentration (Fig. 9B). There was a nonsignificant decrease of the current density recorded from ␤HNF-4␣KO ␤-cells as the glucose concentration increased to 8.3 mM, whereas the current density obtained from control cells was significantly inhibited by 8.3 mM glucose. Thus, the responsiveness of K ATP channel current density to 8.3 mM glucose was impaired in ␤HNF-4␣KO mice. Enhanced activity of the K ATP channels in ␤HNF-4␣KO mice was not attributable to reduced ATP sensitivity of the channels, because the 50% inhibitory concentration for ATP was 1.5 M when tested on inside-out membrane patches excised from the ␤-cells of ␤HNF-4␣KO mice (data not shown). This value was rather lower than previously reported for mouse ␤-cells (37). Next, the tolbutamide sensitivity of these ␤-cells was tested (Fig. 9C). The 50% inhibitory concentration was 6.2 and 3.6 M for cells from control and ␤HNF-4␣KO mice, respectively. Therefore, the channels had a similar sensitivity to tolbu- tamide in both types of mice, but the inhibitory effect of 10 -100 M tolbutamide may be insufficient to produce insulin secretion in ␤HNF-4␣KO mice (Fig. 9C).

DISCUSSION
Mutations in the gene encoding HNF-4␣ cause MODY associated with impaired insulin secretion (5). In addition, variation of the ␤-cellspecific promoter in the HNF-4␣ gene is associated with common type 2 diabetes (10 -12), so investigation of the role of HNF-4␣ in pancreatic ␤-cells could lead to a better understanding of type 2 diabetes as well as MODY. In the present study, we generated ␤-cell-specific HNF-4␣ knock-out mice using the Cre-LoxP system with a rat insulin promoterdriven Cre transgene. Expression of HNF-4␣ protein was reduced by about 80% in ␤HNF-4␣KO islets compared with control islets, and the ␤HNF-4␣KO mice exhibited impaired glucose tolerance with defective insulin secretion, which is characteristic of human MODY1. However, the mice did not develop diabetes at 6 months of age. Because a domi-nant negative effect cannot be a major cause of human MODY1 (6,7,38), the reason why these mice did not develop diabetes remains to be solved. Because the insulin response to GLP-1 was normal in ␤HNF-4␣KO islets, the response to incretins may be one of the potential mechanisms compensating for the impaired response to glucose.
HNF-4␣ plays an essential role in hepatic development and organogenesis, and it is also expressed by the developing pancreas in mice (39). 3 Although the insulin promoter-driven Cre expression occurs in neonates (40), the possibility exists that excision of the floxed allele may occur too late to prevent HNF-4␣ from acting in developing ␤-cells. This possibility could have been tested by using another promoter system (e.g. PDX-1-Cre). Although there is a possibility that perturbation of HNF-4␣ activity in another organ such as liver may affect the MODY1 phenotype, our results obtained with isolated islets clearly indicated that HNF-4␣ was necessary for normal insulin secretion by ␤-cells. 3 T. Nammo and K. Yamagata, unpublished observations. Gupta et al. (28) recently reported another line of ␤-cell-specific HNF-4␣ knock-out mice, but there were several important differences between our mutant mice and theirs. First, their mice showed a significant decrease of blood glucose levels in both the fed and fasting states. In contrast, there were no significant differences of blood glucose levels between our ␤HNF-4␣KO and control animals. Second, they reported that plasma insulin levels were significantly elevated in their mutant mice, but the insulin levels in our KO mice were indistinguishable from that of control mice. Third, Kir6.2 expression was significantly reduced in the islets of their HNF-4␣ mutant mice but not in the islets of our animals. Although expression of HNF-4␣ was significantly decreased (Ͻ10%) in the islets of HNF-1␣ knock-out mice, expression of Kir6.2 was also unchanged in the islets (29). At this point, we have no adequate explanation for these differences. Although the same RIP-Cre transgene (17) was used to generate both HNF-4␣ ␤-cell-specific null mouse lines, different background strains of mice were used in the two studies (129 Svj ϫ C57BL/6 mice in our work versus CD1 mice). This may contribute in part to the different phenotypes obtained in the two studies. Because hyperinsulinemic hypoglycemia has not been reported clinically in MODY1 patients (8,9,41), we believe that our new ␤-cell-specific HNF-4␣ knock-out mice is more reflective of human MODY1 and thus may provide a useful animal model of this disease.
Our patch clamp experiments revealed that K ATP channel activity was up-regulated in ␤HNF-4␣KO mice. It has been reported that decreased responsiveness of the K ATP channel to ATP leads to impaired insulin secretion and diabetes in both mice and humans (42,43). However, our data did not support this possibility in ␤HNF-4␣KO mice because their ATP sensitivity was not reduced. The molecular mechanism by which HNF-4␣ regulates the K ATP channel is still unclear, but a defect of this channel could be related to the impaired insulin secretion observed in ␤HNF-4␣KO mice. K ATP channel activity is regulated by various factors, e.g. Kir6.2 is phosphorylated by protein kinase A and its phosphorylation increases K ATP channel activity (44). GTP-binding proteins have also been shown to modulate K ATP channel activity (44). Expression of PPAR␣ was reduced in ␤HNF-4␣KO islets (Fig. 4), but its effect on K ATP channel activity is unknown. Further studies will be necessary to define the mechanism of K ATP channel regulation by HNF-4␣.
Recently, Odom et al. (45) reported that HNF-4␣ showed binding to the promoters of 11% of islet genes. In our mice, however, expression of the many genes involved in insulin secretion was unchanged. Detailed studies of our ␤HNF-4␣KO mice may lead to a better understanding of the target genes of HNF-4␣ in pancreatic ␤-cells, as well as the molecular basis of both MODY and common type 2 diabetes.
Acknowledgments-We thank T. Miki for helpful discussions and T. Tanaka, and T. Kodama for providing HNF-4␣ antibody. We also thank A. Ihara, F. Katsube, Y. Tochino, and M. Onishi for technical assistance. FIGURE 9. Response of the K ATP channel current to glucose and tolbutamide in control and ␤HNF-4␣KO mice. A, K ATP channel currents recorded during voltage ramping from Ϫ100 to Ϫ50 mV at glucose concentrations of 2.8 and 8.3 mM. Voltage-clamped ␤-cells in the perforated mode from a control mouse (left panel) and a ␤HNF-4␣KO mouse (right panel) are shown. Dotted lines indicate the zero current level. Note that calibration bars indicating the current magnitude are different in the two panels. B, comparison of current density response to changes of the glucose concentration in control mice and ␤HNF-4␣KO mice. K ATP channel conductance was measured from the slope of the current trace during voltage ramping, and the current density was obtained as the conductance divided by the cell capacitance. *, p Ͻ 0.05 versus control cells with 2.8 mM glucose. C, dose-response relation between the K ATP channel current density and the tolbutamide concentration. The glucose concentration in the external solution was 2.8 mM throughout these experiments. Curves were fitted by a Michaelis-Menten type equation. Data are shown as the mean Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.02 versus control.