Overexpression of Dominant-Negative Mutant Hepatocyte Nuclear Factor-1α in Pancreatic β-Cells Causes Abnormal Islet Architecture With Decreased Expression of E-Cadherin, Reduced β-cell Proliferation, and Diabetes

Human P291fsinsC–HNF-1α cDNA (9) was ligated to the rat insulin II promoter containing expression vector (kindly provided by Dr. R. Palmiter, University of Washington, Seattle, WA). A 3.4-kb HindIII expression unit was microinjected into the fertilized eggs of (C57BL/6 × SJL)F2 mice. Transgenic mice were selected by Southern blot analysis and polymerase chain reaction (PCR) and were backcrossed to C57BL/6 mice (>4 generations) for experiments. RIP wild type (WT)–HNF-1α transgene (9) was also constructed, and transgenic mice expressing WT–HNF-1α in pancreatic β-cells were generated. Age- and sex-matched transgene-negative littermates were used as control mice throughout the study. The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Osaka University.

Pancreatic islets were isolated from mice by the collagenase digestion method (10). Cells were lysed in extraction buffer [100 mmol/l NaCl, 50 mmol/l Tris-HCl (pH 8.0), 20 mmol/l EDTA, and 1% SDS], and Western blot was performed as described (9). The membrane was incubated with monoclonal anti–HNF-1α antibody (Transduction Laboratories, Lexington, KY) or anti–E-cadherin antibody (kindly provided by Dr. R. Kemler, Max-Planck-Institute, Freiburg, Germany) (11) overnight at 4°C and further incubated with horseradish peroxidase–conjugated anti-IgG antibody (Promega, Osaka, Japan). The binding antibody was visualized using enhanced chemiluminescence Western Blotting detection reagents (Amersham Life Sciences, Little Chalfont, U.K.).

Total RNA was isolated from pancreas, liver, kidney, and brain using TRIzol reagent (Life Technologies, Rockville, Maryland). Contaminating genomic DNA was removed by treatment with DNase I. cDNA synthesis was performed with 1 μg of total RNA using Moloney murine leukemia virus reverse transcriptase (Life Technologies) with dNTPs and oligo dT primers. The following primers were used for the specific PCR amplification of human HNF-1α: 5′-AGGACCTGAGCCTGCCGAGCAAC-3′ and 5′-AGGGCTCTCCATAGGCCCAGGCT-3′ (annealing temperature 60°C and product size 281 bp) and mouse GAPDH: 5′-TGACAACTCACTCAAGATTG-3′ and 5′-CACGTCAGATCCACGACGGA-3′ (annealing temperature 60°C and product size 321 bp).

Glucose tolerance testing was performed in female mice (8 weeks of age) because male mice exhibited severe hyperglycemia. Mice were fasted for 16 h and then loaded with 1 g/kg glucose through the tail vein. Serum insulin levels were determined by Glazyme insulin enzyme immunoassay kit (Wako, Osaka, Japan). Insulin content was measured after extraction by the acid-ethanol method (12).

Male mice were used in all histological studies unless otherwise mentioned. Immunohistochemistry was performed as described (13). The following primary antibodies were used: guinea pig anti-insulin (Dako, Tokyo, Japan), rabbit anti-glucagon (Linco, St. Charles, MO), rabbit anti-somatostatin (Dako), rabbit anti–pancreatic polypeptide (PP) (Dako), rabbit anti–PDX-1 (13), goat anti-glucokinase (Santa Cruz, Santa Cruz, CA), rabbit anti-GLUT2 (kindly provided by Dr. B. Thorens, Institute of Pharmacology and Toxicology, Lausanne, Switzerland), rat anti–N-cadherin (kindly provided by Dr. M. Takeichi, Kyoto University, Kyoto, Japan), and rat anti–E-cadherin (11). The anti-Pax6 antiserum was prepared after immunizing a rabbit with synthetic peptides (QVPGSEPDMSQYWPRLQ, amino acid residues 396 to 422) (14). Immunofluorescence was viewed using a laser scan confocal microscope (LSM510; Carl Zeiss, Jena, Germany) or a light microscope.

Male Tg-1 and control mice (2 days old and 4 weeks old) were used for morphometric analysis. Morphometric analysis was performed as previously described (13). Three sets of 10 serial sections were obtained from each pancreas. The first section in each set was immunostained for insulin, and the second one was stained for glucagon, somatostatin, and PP. The total area of the pancreas was measured under a television monitor. β-Cell and non–β-cell counts were determined in six sections in each pancreas. The relative numbers of β-cells and non–β-cells were expressed as the number per one square millimeter of the pancreatic area.

β-cell and non–β-cell proliferation rates were determined by 5 bromo-2′-deoxyuridine (BrdU) incorporation (13). Male Tg-1 and control mice were injected intraperitoneally with 100 mg/kg of BrdU (Cell Proliferation Kit; Amersham Pharmacia Biotech) and killed 6 h later. Four sets of five serial sections were cut out from a paraffin block. The first section was immunostained for insulin and BrdU, and the second section was stained for non–β cell hormones (glucagon, somatostatin, and PP) and BrdU. Data are shown as the sum of the four separate pancreas sections from four different mice in each group.

Ten islets isolated from 6- to 7-week-old female Tg-1 and control mice were preincubated at 37°C for 60 min in HEPES-balanced Krebs-Ringer bicarbonate (HKRB) buffer (15). The pancreatic islets were then incubated for 20 min in a buffer containing 3.3 or 20 mmol/l glucose.

Isolated islets from Tg-1 mice were plated on microwell dishes (MatTek, Ashland, MA). Cells were loaded with 5 μmol/l fura-2 acetoxymethyl ester (Dojin, Kumamoto, Japan) for 30 min (15). Changes in [Ca²⁺]_i in response to 15 mmol/l glucose and 20 mmol/l KCl were expressed as the ratio of the emitted light intensity (detected at 510 nm) after excitation at 340 and 380 nm (ratio 340/380) (7). Islet cells of control mice were incubated in the presence of rat anti–E-cadherin IgG1 or rat nonimmune IgG1 (200 ng/ml) for 48 h, and changes in [Ca²⁺]_i were also measured.

Islet cells were plated on 24-well dishes and cultured in the presence of rat anti–E-cadherin IgG1 or rat nonimmune IgG1 (200 ng/ml) for another 48 h. Islet cells were perifused at 1 ml/min with HKRB buffer. After preincubation for 60 min, 26-mmol/l glucose stimulation was applied.

Results are expressed as means ± SD. Differences were analyzed using a two-tailed unpaired Student’s t test. P < 0.05 was considered statistically significant.

We generated transgenic mice expressing human dominant-negative mutant (P291fsinsC–HNF-1α) (Fig. 1A) in β-cells. Two founder animals were obtained and bred with C57BL/6 mice to generate transgenic lines. They had different transgene copy numbers (5 Tg-1 and 20 Tg-2 copies). Western blot analysis was performed using proteins prepared from isolated islets. More than 100 islets were generally found after preparation by the collagenase digestion method from a 6- to 8-week-old control mouse, whereas only 10–20 islets could be observed from a transgenic mouse. The levels of expression of mutant proteins in Tg-1 and Tg-2 mice were estimated to be 6- and 24-fold greater than that of endogenous HNF-1α, respectively (Fig. 1B). We also generated transgenic mice with WT–HNF-1α for comparison. The level of expression of the transgene was fivefold greater compared with that of endogenous HNF-1α. Reverse transcription (RT)-PCR analysis was performed to examine the extra-pancreas expression of P291fsinsC–HNF-1α gene (Fig. 1C). Expression of the transgene was not detected in liver and heart, but leaky expression was found in brain, as previously reported (16). P291fsinsC–HNF-1α mice were grossly indistinguishable from their control littermates at birth. Nonfasting blood glucose concentrations and body weights were similar between HNF-1α mutant mice and control littermates at 2 days of age (Fig. 1D). Nonfasting blood glucose levels at 4 weeks were higher in male transgenic mice than in control mice and increased further to 406 ± 154 mg/dl (Tg-1, n = 18, P < 0.001) and 485 ± 147 mg/dl (Tg-2, n = 16, P < 0.001) at 8 weeks of age (Fig. 1E). Nonfasting blood glucose levels of female Tg-1 mice at 4 weeks were comparable with those of control mice, whereas blood glucose concentration of most female Tg-2 mice (13/16) were higher than the average glucose levels (136 mg/dl) in control mice. Nonfasting blood glucose concentrations in female transgenic mice were lower at 8 weeks of age than in male mutant mice, but the concentration was still significantly higher than in control mice (Fig. 1E). Sexual dimorphism of sensitivity to diabetes has been reported in rodent models of diabetes (17).

Insulin levels at 5 min after intravenous glucose load were significantly low both in Tg-1 and Tg-2 mice (Fig. 2A). Pancreatic islets were isolated from 6- to 7-week-old female Tg-1 mice and their littermates, and insulin secretion was examined. Glucose at a concentration of 20 mmol/l stimulated insulin secretion in the islets of control mice by 9.7-fold. Significant impairment of insulin secretion (2.1-fold, P < 0.01) in response to the same dose of glucose, however, was observed in Tg-1 mice (Fig. 2B). Changes in [Ca²⁺]_i after glucose stimulation in β-cells of 6- to 8-week-old female Tg-1 mice were measured to examine whether the reduction in glucose-stimulated insulin secretion is associated with reduced [Ca²⁺]_i changes (Fig. 2C). Changes in [Ca²⁺]_i in islets of transgenic mice were reduced compared with islets in the control mice. However, the responses in [Ca²⁺]_i to 20 mmol/l KCl were similar between control and mutant mice. A defective response to glucose has been reported in HNF-1α knockout mice (7).

We also measured pancreatic insulin content in male mice (Fig. 2D). Insulin content of transgenic mice was already significantly reduced (Tg-1 45% of control P < 0.001 and Tg-2 31% of control, P < 0.001) at day 2. Insulin content in the pancreas of mutant mice was further reduced to 4% (Tg-1 P < 0.001) and 3.8% (Tg-2 P < 0.001) of that in the control mice at 8 weeks of age.

Staining for insulin in the pancreas of 8-week-old Tg-1 mice revealed that islets were small in size and contained fewer insulin-positive cells (Figs. 3A and B). We also analyzed the islets of Tg-1 mice at 2 days and at 4 weeks of age (Figs. 3C– N). Mature islet structure with a core of β-cells surrounded by a mantle of glucagon cells was observed 2 days postnatally in control mice (Figs. 3C and D). In contrast, the endocrine cells did not form well-organized islets in the pancreas of 2-day-old Tg-1 mice (Figs. 3E and F), indicating that the abnormality of islet structure is evident even before the onset of diabetes. Islets of transgenic mice were smaller than those of control mice and morphologically irregular in shape at 4 weeks of age. β-Cells intermingled with non–β-cells within the islet core (Figs. 3K– N). There were no apparent differences in the histological features between Tg-1 and Tg-2 mice or between male and female mutants.

Morphometric analysis showed no significant difference in non–β-cell number between control and Tg-1 mice. However, β-cell numbers in transgenic mice were significantly decreased by 50% (P < 0.05) and 58% (P < 0.05) compared with the respective control at 2 days and at 4 weeks of age (Figs. 4A and B). Pancreatic weight was similar between control and Tg mice [2 days: Tg-1 7.1 ± 0.6 mg (n = 5), Tg-2 7.9 ± 1.4 (n = 10), and control 7.7 ± 1.4 mg (n = 7); 4 weeks: Tg-1 182 ± 44 mg (n = 5), Tg-2 169 ± 18 mg (n = 5), and control 175 ± 35 mg (n = 10)]. Thus, reduction of β-cell number was still significant, even after adjustment for pancreas weight.

We determined the cell proliferation rate in 2-day- and 4-week-old mice by BrdU incorporation. There was no significant difference in non–β-cell proliferation rate between control mice [2 days: 49 of 3,682 (1.3%), 4 weeks: 57 of 9,854 (0.58%)] and male Tg-1 mutant mice [2 days: 64 of 5,361 (1.2%), 4 weeks: 27 of 4,730 (0.57%)]. However, β-cell proliferation rate in Tg-1 mice was significantly reduced to 14.6% at 2 days, relative to the control mice [control: 92 of 6,543 (1.4%), Tg-1: 13 of 6,116 (0.21%); P < 0.01]. The proliferation rate in Tg-1 mice was also significantly reduced at 4 weeks of age [control: 988 of 18,704 (5.2%), Tg-1: 91 of 10,360 (0.87%); P < 0.01], suggesting that β-cell proliferation is reduced in P291fsinsC–HNF-1α transgenic mice. Terminated deoxynucleotidyl transferase ( (TDT)-mediated dUTP nick and labeling (TUNEL)-positive cells were not detected among β-cells of the 4-week-old male control and Tg-2 mice. These data suggest a lack of accelerated apoptosis in β-cells of mutant mice.

We also analyzed the expression of several markers associated with mature functional β-cells. Expression of PDX-1, Pax6, GK (Fig. 5), and Nkx2.2 (data not shown) in transgenic mice was comparable with that of control mice, suggesting that RIP-driven expression of P291fsinsC–HNF-1α does not result in a nonspecific loss of β-cell proteins. GLUT2 facilitates glucose transport in β-cells, and the transcription of GLUT2 is regulated by HNF-1α in vitro (18). The expression of GLUT2 was reduced on the islets of 4-week-old male mutant mice (Fig. 5O). GLUT2 expression was also reduced in β-cells of 4-week-old female Tg-1 and Tg-2 mice with normal glucose levels (data not shown). In sharp contrast, the expression of GLUT2 was not reduced in β-cells of 4-week-old male transgenic mice overexpressing WT–HNF-1α (Fig. 5Q). These results suggest that GLUT2 is a target of HNF-1α in vivo.

Cell adhesion molecules have functional roles in the aggregation and organization of islets (19,20). Cadherins constitute a superfamily of transmembrane glycoproteins that mediate Ca²⁺-dependent homophilic interactions between cells at the level of adherens junctions (21). E- and N-cadherins are both expressed in pancreatic islets and are thought to mediate cell adhesion in islet cells (19,20). Immunostaining with E-cadherin–specific antibody revealed a uniform expression of E-cadherin on the surface of islet cells as well as exocrine cells in control mice. In contrast, the expression of E-cadherin was reduced in cell-to-cell contacts between most β-cells in Tg-1 transgenic mice at 2 days of age (Figs. 6A– D). β-Cells of the transgenic mice were scattered in the islets, and compacted β-cell–to–β-cell contact was abolished. Reduced expression of E-cadherin in β-cells was also observed in 4-week-old transgenic mice (Figs. 6F and H). Tg-2 mice and female mutant mice also showed the abnormality (data not shown). The reduced expression of E-cadherin was confirmed by Western blot analysis using isolated islets (Fig. 6L). In contrast, E-cadherin expression in WT–HNF-1α Tg mice was not reduced (Figs. 6J and K). Expression of N-cadherin, another cadherin in pancreatic islets, appeared normal in transgenic mice (Figs. 6L– O).

It has been suggested that β-cell–to–β-cell contact is required for normal insulin secretion (22,23,24). Mouse islets were cultured in the presence of rat anti–E-cadherin, blocking IgG1 antibody (DECMA-1) or control IgG1, and changes in glucose-stimulated [Ca²⁺]_i and insulin secretion were measured. This antibody can block E-cadherin–mediated adherens junction formation in various types of cells (11). Incubation with DECMA-1 antibody (200 ng/ml) inhibited the formation of the structure (Fig. 7A). In contrast, control rat IgG1 had no effect on the structure. Culture of islet cells with DECMA-1 antibody for 48 h abolished the response of [Ca²⁺]_i to glucose stimulation, whereas 20 mmol/l KCl stimulation increased [Ca²⁺]_i to levels comparable with islet cells incubated with control IgG1 (Fig. 7B). Glucose-stimulated insulin secretion was impaired in islets treated with DECMA-1 antibody (Fig. 7C). These data suggest that reduced expression of E-cadherin might affect insulin secretion as well as islet organogenesis.

P291fsinsC–HNF-1α transgenic mice developed diabetes with impaired glucose-stimulated insulin secretion, which is a characteristic feature of human MODY3 (4,5,6). However, HNF-1α knockout mouse (25) exhibited severe liver and kidney dysfunction, and another kind of HNF-1α null mice (26) exhibited Laron-type dwarfism, infertility, and liver dysfunction. In contrast, human MODY3 patients do not show such signs. Leaky expression of the transgene was detected in the brain of P291fsinsC–HNF-1α transgenic mice, but there were no apparent abnormalities in general appearance or behavior in the mice. Histological studies revealed that P291fsinsC–HNF-1α transgenic mice displayed no abnormalities in liver and kidney (data not shown). Thus, the P291fsinsC–HNF-1α transgenic mouse is a novel animal model of human MODY3. Diabetes in the P291fsinsC–HNF-1α transgenic mouse is more severe than that in the HNF-1α knockout mouse. The mutant HNF-1α protein may sequester other β-cell proteins, and this could account at least in part for the more severe phenotype. HNF-1α forms a heterodimer with structurally related transcription factor HNF-1β, which is also expressed in pancreatic islets. The heterodimer formation between in vitro translated P291fsinsC–HNF-1α and HNF-1β was confirmed by electrophoretic mobility shift assay (data not shown). HNF-1β might be a victim of P291fsinsC–HNF-1α in β-cells.

Expression levels of PDX-1, Pax6, Nkx2.2, GK, and N-cadherin appeared to be comparable in the Tg mice. However, the expression levels of GLUT2 and E-cadherin were reduced in the mutant mice. Because the downregulation of the molecules was found before the onset of diabetes, it is not a secondary effect of hyperglycemia. Reduced levels of expressions of GLUT2 and E-cadherin were not found in WT–HNF-1α Tg mice. These data suggest that these molecules are targets of HNF-1α in vivo. Abnormal islet structure with loosely connected and scattered β-cells was observed in the P291fsinsC–HNF-1α transgenic mice. The alteration of the islet structure is not a general occurrence with RIP-driven transgenic mice (16,27,28). Because E-cadherin is important for proper cell adhesion, its reduced expression in Tg mice could be related to the abnormal structure. Similar abnormalities have been reported in the islets of transgenic mice expressing mutant E-cadherin with a dominant-negative effect (19). Defect in pancreatic islet morphology has been reported in other transcription factor knockout mice (29,30), but there is no information about the expression levels of cadherins in the mice.

Intercellular communication among β-cells appears to be necessary for normal insulin secretion (22,23,24). Treatment of islet cells with the antibody resulted in inhibition of the structure and marked suppression of glucose-stimulated insulin secretion. Thus, the reduced expression of E-cadherin in Tg mice might account at least in part for the impaired glucose-stimulated insulin secretion in the mice. The molecular mechanism of reduced E-cadherin expression in the P291fsinsC–HNF-1α is not currently clear. Chicken E-cadherin (l-CAM) gene has a binding site for HNF-1α in the intron, and HNF-1α acts as an enhancer through the binding site (31). We found a putative HNF-1α binding site in intron 2 of human E-cadherin gene (accession number AC009082) on the human genome sequence database search. Further studies are necessary to clarify the regulation of E-cadherin by HNF-1α.

Our results showed reduced GLUT2 expression in β-cells in transgenic mice before the onset of diabetes. Homozygous mice deficient in GLUT2 develop diabetes with impaired glucose-stimulated insulin secretion (32). Then, loss of GLUT2 on β-cells might also be associated with impaired insulin secretion in P291fsinsC–HNF-1α transgenic mice.

Reduction of β-cell mass and insulin content have been reported in HNF-1α knockout (−/−) mice (7) and was also observed in P291fsinsC–HNF-1α Tg mice. However, the knockout mice demonstrate hepatic dysfunction, phenylketonuria, and renal Fanconi syndrome and show severe growth retardation (25,26). Thus, β-cell mass adjusted for body weight was not changed in the knockout mice (7). In contrast, P291fsinsC–HNF-1α Tg mice did not show such severe growth retardation, and the differences in β-cell mass were still significant, even after adjustment for body weight, suggesting that loss of HNF-1α affects β-cell mass and insulin content. It might be worth noting that the GLUT2 knockout mouse displays a decrease in β-cell mass (32). Because a 90% partial pancreatectomy leads to impaired glucose-induced insulin secretion in rats (33), the decreased insulin content in Tg mice (3.2–4% of the control in 8-week-old mice) could be related to the onset of diabetes.

Earlier studies have suggested that loss of HNF-1α affects glycolysis and mitochondrial oxidation (8,9,18). Our findings suggest that HNF-1α plays a variety of roles in determining normal β-cell function by regulating GLUT2 and E-cadherin expression and β-cell mass. Further studies are necessary to determine whether the same abnormalities are implicated in human HNF-1α diabetes. Detailed studies of Tg mice may lead to a better understanding of the target genes of HNF-1α in β-cells and the molecular basis of MODY3.

This work was supported by grants from the Japanese Ministry of Education, Culture, Sports, Science and Technology, Japan Diabetes Foundation, Yamanouchi Foundation for Research on Metabolic Disorders, Japan Insulin Study Group, and Research for the Future Program of the Japan Society for the Promotion of Science (97L00801).

We thank R. Palmiter (RIP vector), B. Thorens (anti-GLUT2), Y. Kajimoto (anti–PDX-1), R. Kemler (anti–E-cadherin), and M. Takeichi (anti–N-cadherin) for generously providing the respective reagents. We also thank G.I. Bell (University of Chicago, IL) and C.B. Wollheim (University Medical Center, Switzerland) for their continuous encouragement of our work.