Hepatocyte Nuclear Factor 4α Regulates the Expression of Pancreatic β-Cell Genes Implicated in Glucose Metabolism and Nutrient-induced Insulin Secretion*

Mutations in the HNF4α gene are associated with the subtype 1 of maturity-onset diabetes of the young (MODY1), which is characterized by impaired insulin secretory response to glucose in pancreatic β-cells. Hepatocyte nuclear factor 4α (HNF4α) is a transcription factor critical for liver development and hepatocyte-specific gene expression. However, the role of HNF4α in the regulation of pancreatic β-cell gene expression and its correlation with metabolism secretion coupling have not been previously investigated. The tetracycline-inducible system was employed to achieve tightly controlled expression of both wild type (WT) and dominant-negative mutant (DN) of HNF4α in INS-1 cells. The induction of WT-HNF4α resulted in a left shift in glucose-stimulated insulin secretion, whereas DN-HNF4α selectively impaired nutrient-stimulated insulin release. Induction of DN-HNF4α also caused defective mitochondrial function substantiated by reduced [14C]pyruvate oxidation, attenuated substrate-evoked mitochondrial membrane hyperpolarization, and blunted nutrient-generated cellular ATP production. Quantitative evaluation of HNF4α-regulated pancreatic β-cell gene expression revealed altered mRNA levels of insulin, glucose transporter-2, L-pyruvate kinase, aldolase B, 2-oxoglutarate dehydrogenase E1 subunit, and mitochondrial uncoupling protein-2. The patterns of HNF4α-regulated gene expression are strikingly similar to that of its downstream transcription factor HNF1α. Indeed, HNF4α changed the HNF1α mRNA levels and HNF1α promoter luciferase activity through altered HNF4α binding. These results demonstrate the importance of HNF4α in β-cell metabolism-secretion coupling.

tations in the human HNF4␣ gene lead to maturity onset diabetes of the young subtype 1 (MODY1), which is characterized by autosomal dominant inheritance and impaired glucosestimulated insulin secretion from pancreatic ␤-cells (4 -6). These MODY1 mutations located in various domains of the HNF4␣ protein result in defective function of the transcription factor (6). The clinical phenotype of MODY1 patients is indistinguishable from that of MODY3 patients who carry mutations in the HNF1␣ gene (5,6). HNF4␣ acts upstream of HNF1␣ in a transcriptional cascade that drives liver-specific gene expression and hepatocyte differentiation (7)(8)(9). A naturally occurring mutation in the HNF4␣-binding site of the HNF1␣ promoter identified in a MODY3 family (10) suggests that the transcriptional hierarchy could also be involved in pancreatic ␤-cell gene expression and function.
HNF4␣ defines the expression of liver-specific genes encoding apolipoproteins, serum factors, cytochrome P-450 isoforms, and proteins involved in the metabolism of glucose, fatty acids, and amino acids (reviewed in Ref. 11). However, clinical characterization of MODY1 subjects reveals that the primary defect is impaired glucose-stimulated insulin secretion from pancreatic ␤-cells rather than liver dysfunction (5,(12)(13)(14). Unfortunately, little is known as to how HNF4␣ regulates ␤-cell-restricted gene expression and glucose metabolism and associated insulin secretion. Targeted disruption of the hnf4␣ gene results in defective gastrulation of mouse embryos due to dysfunction of the visceral endoderm (15). This early embryonic lethality prevents further analysis of the HNF4␣ function in pancreatic ␤-cells. The precise role of HNF4␣ in pancreatic ␤-cells would best be examined by conditional ␤-cell-specific deletion of the mouse hnf4␣ gene. Another alternative is to upand down-regulate HNF4␣ function in pancreatic ␤-cell lines through gene manipulation.
In the present study, the wild type HNF4␣ (WT-HNF4␣) and its dominant-negative mutant (DN-HNF4␣) could be induced in INS-1 cells under tight control of the reverse tetracyclinedependent transactivator (16). DN-HNF4␣ represents the epitope Myc-tagged truncated HNF4␣ mutant protein lacking the first 111 amino acids (myc⌬111HNF4␣) (17). The HNF4␣ protein consists of an N-terminal ligand-independent transactivation domain (amino acids 1-24), a DNA binding domain containing two zinc fingers (amino acids 51-117), and a large hydrophobic portion (amino acids 163-368) composed of the dimerization, ligand binding, cofactor binding, and ligand-dependent transactivation domain (18,19). DN-HNF4␣ therefore suppresses the endogenous WT-HNF4␣ transcriptional activity by the formation of heterodimers lacking DNA binding capacity (17). We have investigated in a quantitative manner the consequences of altered HNF4␣ function on ␤-cell-specific expres-sion of genes implicated in glucose metabolism and insulin secretion. This allowed us to elucidate the molecular basis and HNF4␣ target genes responsible for impaired metabolism secretion coupling in ␤-cells deficient in HNF4␣ function.

EXPERIMENTAL PROCEDURES
Generation of Stable Cell Lines-The rat insulinoma INS-1 cell linederived stable clones were cultured in RPMI 1640 in 11 mM glucose, unless indicated otherwise (20). The establishment of the first step stable clone INS-r3, which expresses the reverse tetracycline-dependent transactivator, was reported previously (21). Plasmids used in the secondary stable transfection were constructed by subcloning the cDNAs encoding the rat WT-HNF4␣ (a generous gift from Dr. Darnell Jr., New York) and DN-HNF4␣ into the expression vector PUHD10-3 (a kind gift from Dr. H. Bujard, University Heidelberg, Germany). DN-HNF4␣ was PCR-amplified from WT-HNF4␣ using the following primers, ctaggatccttccgggctggcatgaagaaagaagcc and ccagaattcctgcagatggttgtcctttag. The PCR fragment was subcloned into pcDNA3.1myc (Invitrogen, Netherlands) and sequenced. Transfection, clone selection, and screening procedures were described previously (21).
Insulin Secretion and Cellular Insulin Content-Cells in 24-well dishes were cultured in 2.5 mM glucose medium with or without indicated doses of doxycycline for 14 or 48 h. Insulin secretion was measured over a period of 30 min, in Krebs-Ringer/bicarbonate-HEPES buffer (KRBH, 140 mM NaCl, 3.6 mM KCl, 0.5 mM NaH 2 PO 4 , 0.5 mM MgSO 4 , 1.5 mM CaCl 2 , 2 mM NaHCO 3 , 10 mM HEPES, 0.1% bovine serum albumin) containing indicated stimulators. Insulin content was determined after extraction with acid ethanol following the procedures of Asfari et al. (20). Insulin was detected by radioimmunoassay using rat insulin as standard (22).
Intracellular ATP-Cells in 6-well dishes were cultured in 2.5 mM glucose medium with or without 500 ng/ml doxycycline for 48 h. The production of ATP was measured during 8 min of stimulation in KRBH. ATP assay was performed as reported previously (22).
Mitochondrial Membrane Potential (⌬ m )-After a 48-h culture period in 2.5 mM glucose medium with or without 500 ng/ml doxycycline, cells were trypsinized (0.025% trypsin, 0.27 mM EDTA), and the cell suspension was maintained for 2 h in a spinner culture with 2.5 mM glucose RPMI 1640 plus 1% newborn calf serum at 37°C. Mitochondrial membrane potential (⌬ m ) was measured as described (25). Briefly, after the spinner culture period, cells were loaded with 10 g/ml rhodamine-123 (Rh-123) for 10 min at 37°C. After centrifugation, the cells were resuspended and transferred to the fluorimeter cuvette at 37°C with gentle stirring in an LS-50B fluorimeter (PerkinElmer Life Sciences), and fluorescence, excited at 490 nm, was measured at 530 nm.
Nuclear Extract Preparation and Electrophoretic Mobility Shift Assay (EMSA)-Cells in 10-cm dishes were grown in culture medium with or without 500 ng/ml doxycycline for 48 h. The following doublestranded oligonucleotides were used as probes, 5Ј-GGCTGAAGTC-CAAAGTTCAGTCCCTTCGC-3Ј (8). EMSA procedures including condi- tions for nuclear extract preparation, probe labeling, binding reactions, unlabeled-probe competition, and antibody supershift were performed as reported previously (22).
Transient transfection experiments and luciferase reporter enzyme assays were carried out as previously reported (22).

WT-HNF4␣ or DN-HNF4␣ Protein Was Induced in INS-1 Cells in a Dose-and Time-dependent Manner-
We have obtained 10 and 8 clones positively expressing WT-HNF4␣ and DN-HNF4␣, respectively. The clones designated as WT-HNF4␣-28 and DN-HNF4␣-26 that displayed highest induction levels of transgene proteins were chosen for the present study. The time course and dose response of doxycycline effect on WT-HNF4␣ and DN-HNF4␣ expression are illustrated in Fig. 1, A and B, respectively. WT-HNF4␣ protein could be induced within a range from 2-to 50-fold above the endogenous protein level (Fig. 1A). Thus, graded overexpression of WT-HNF4␣ could be achieved by culturing the WT-HNF4␣-28 cells with varying doses of doxycycline in a defined period. Similar induction of DN-HNF4␣ protein was detected in the nuclear extracts from DN-HNF4␣-26 cells (Fig. 1B). No leakage of this doxycycline-dependent promoter was observed, since the expression of DN-HNF4␣ protein was not detectable in noninduced DN-HNF4␣-26 cells (Fig. 1B). Therefore, the dominant-negative suppression of HNF4␣ function in INS-1 cells could be rapidly achieved by culturing the DN-HNF4␣-26 cells with a maximum dose of doxycycline (500 ng/ml).
Effects of WT-HNF4␣ and DN-HNF4␣ on Insulin Secretion-Impaired glucose-stimulated insulin secretion from pancreatic ␤-cells is the primary defect causing hyperglycemia in MODY1 patients carrying HNF4␣ mutations. We therefore examined the consequences of induction of WT-HNF4␣ and DN-HNF4␣ on insulin secretion in INS-1 cells. The graded overexpression of WT-HNF4␣ led to a left shift of glucose-stimulated insulin secretion ( Fig. 2A). However, the maximal (above 12 mM) glucose-elicited insulin secretion remained unchanged ( Fig. 2A).
Glucose generates ATP and other metabolic coupling factors important for insulin secretion through glycolysis and mitochondrial oxidation (28). The physiological insulin secretagogue, leucine, is transported directly into mitochondria to provide substrates for the tricarboxylic acid cycle (28). K ϩ causes insulin secretion by depolarization of the ␤-cell membrane, resulting in an increase in cytosolic Ca 2ϩ (28). We therefore examined the insulin secretory responses to these three secretagogues that act at different levels of the signal transduction cascade following induction of DN-HNF4␣. As demonstrated in Fig. 2B, DN-HNF4␣ selectively inhibited glucoseand leucine-stimulated insulin secretion. This could be explained by defective glucose and leucine metabolism.
Effects of DN-HNF4␣ on Cellular ATP Production and Mitochondrial Oxidation-To investigate whether impaired nutrient-evoked insulin secretion is correlated to defective cellular ATP production, we analyzed the impact of DN-HNF4␣ expression on the level of ATP generated by glucose and leucine. As shown in Fig. 3A, induction of DN-HNF4␣ indeed abolished the ATP generation by glucose and leucine. Since the mitochondrial substrate leucine failed to generate ATP after induction of DN-HNF4␣, it would seem that HNF4␣ is required for maintaining normal mitochondrial function.
To test this hypothesis, we examined the consequences of DN-HNF4␣ induction on mitochondrial oxidation of pyruvate. Pyruvate-derived carbons enter the tricarboxylic acid cycle as either acetyl-CoA, catalyzed by pyruvate dehydrogenase, or oxaloacetate via pyruvate carboxylase. By using pyruvate radiolabeled at either the first or second carbon, the putative defects at various steps in pyruvate metabolism can be assessed. The radiolabeled carbon of [1-14 C]pyruvate is lost to CO 2 at the pyruvate dehydrogenase step as pyruvate is converted into acetyl-CoA. Alternatively, if pyruvate enters the tricarboxylic acid cycle via oxaloacetate, the label is lost to CO 2 at isocitrate dehydrogenase within one turn of the cycle. Radiolabeled CO 2 is generated from 2-14 C at either OGDH or isocitrate dehydrogenase when pyruvate enters the tricarboxylic acid cycle as acetyl-CoA. Overexpression of DN-HNF4␣ reduced CO 2 formation from [2-14 C]pyruvate by 41% (Fig. 3B), whereas CO 2 formation from [1-14 C]pyruvate was not different between non-and induced conditions (Fig. 3C). These results suggest that the defect in mitochondrial metabolism is not at the point of entry of pyruvate into the tricarboxylic acid cycle, rather that the defect appears in reactions within the tricarboxylic acid cycle. This is in full agreement with the impairment of leucine stimulation of insulin secretion since leucine metabolism bypasses pyruvate and enters the tricarboxylic acid cycle solely as acetyl-CoA. Decreased isocitrate dehydrogenase activity would also be unlikely since impairment at this step would be observed by both [1-14 C]pyruvate and [2-14 C]pyruvate oxidation. These oxidation experiments suggest that steps following this reaction beginning with OGDH may be responsible for impaired  (Fig. 4A). In cells expressing DN-HNF4␣ (ϩDox), the glucose response was inhibited by 65% (p Ͻ 0.02). Impaired hyperpolarization of ⌬ m was also observed when the glycolysis was bypassed by stimulating cells with the end product of glycolysis pyruvate (Fig. 4B), indicating mitochondrial dysfunction. Direct activation with methyl succinate of the electron transport chain at complex II resulted in a diminished response in DN-HNF4␣-induced cells (Fig. 4C). The amplitude of complete ⌬ m depolarization by FCCP was also reduced in cells treated with doxycycline (Ϫ43%; p Ͻ 0.01), suggesting that the mitochondria were partially uncoupled by suppression of HNF4␣ function (Fig. 4D).
Effects of WT-HNF4␣ and DN-HNF4␣ on Pancreatic ␤-Cell Gene Expression-The expression of genes involved in glucose metabolism (Fig. 5, A and B) or in pancreatic ␤-cell development and differentiation (Fig. 5, C and D) was quantitatively evaluated in WT-HNF4␣-28 (Fig. 5, A and C) and DN-HNF4␣-26 cells (Fig. 5, B and D). As shown in Fig. 5A, WT-HNF4␣ mRNA could be induced by 2-, 8-, and 50-fold above the endogenous level. This graded overexpression of WT-HNF4␣ resulted in a stepwise increase in the expression of three glucose-responsive genes encoding, respectively, GLUT-2, L-PK, and aldolase B (Fig. 5A). However, the mRNA level of glyceraldehyde-3-phosphate dehydrogenase, which is also responsive to glucose, remained unaltered (Fig. 5A). Induction of WT-HNF4␣ also caused incremental expression of OGDH E1 subunit transcript (Fig. 5A). Consistently, The mRNA levels of GLUT-2, aldolase B, L-PK, and OGDH E1 subunit were significantly reduced after induction of DN-HNF4␣ (Fig. 5B). On the other hand, induction of DN-HNF4␣ led to increased UCP-2 mRNA expression (Fig. 5B). Therefore, HNF4␣ regulates the expression of genes involved in both glycolysis and mitochondrial metabolism. The profile of HNF4␣-targeted genes is strik- ingly similar to that of HNF1␣ (29). HNF4␣ may regulate the expression of genes implicated in glucose metabolism through HNF1␣ function as in hepatocytes (7)(8)(9).
Since HNF4␣ is required for liver development and hepatocyte differentiation (9), we investigated whether HNF4␣ regulates the expression of genes important for the pancreatic ␤-cell phenotype. Induction of WT-HNF4␣ (Fig. 5C) or DN-HNF4␣ (Fig. 5D) did not alter the expression patterns of PDX1, Pax4, Pax6, NKx2.2, NKx6.1, and Isl-1, which are necessary for normal pancreatic cell development or differentiation (30). Moreover, HNF4␣ did not regulate the mRNA levels of USF, c-Jun, and C/EBP␤ (Fig. 5, C and D). The expression of these transcription factors appeared to be responsive to glucose (Fig. 5, C  and D). The expression of another glucose-responsive transcription factor, DcoH, was slightly affected by induction of WT-HNF4␣ but not by expression of DN-HNF4␣ (Fig. 5, C and  D), suggesting the involvement of an indirect mechanism. Both Cdk4 and IRS2 are involved in pancreatic ␤-cell development (31,32), but their expression was not regulated by HNF4␣ (Fig.  5, C and D). Induction of DN-HNF4␣ for 48 h caused 50% reduction in insulin mRNA levels (Fig. 5D). This may be secondary to decreased HNF1␣ function, since HNF1␣ is required for insulin gene transcription (29).
HNF4␣ Regulates Pancreatic ␤-Cell Gene Expression through HNF1␣ Function-We performed EMSA for studying HNF4␣ binding activity to HNF1␣ promoter, luciferase reporter enzyme assay for HNF1␣ promoter activity, and Northern blotting for the HNF1␣ mRNA expression. Nuclear extracts were prepared from WT-HNF4␣-28 and DN-HNF4␣-26 cells cultured for 48 h in the presence or absence of 500 ng/ml doxycycline. The murine HNF1␣ promoter segment, which contains the HNF4␣-binding site, was used as probe (8). Induction of WT-HNF4␣ resulted in a dramatic increase in the signal density of HNF4␣ binding (Fig. 6A). On the other hand, induction of DN-HNF4␣ almost completely abolished the binding activity of endogenous HNF4␣ to the HNF1␣ promoter (Fig.  6A). DN-HNF4␣ exerts its dominant-negative function by forming DN-HNF4␣/WT-HNF4␣ heterodimers that lack DNA binding capacity (11). The retarded DNA binding complexes corresponding to endogenous WT-HNF4␣ and/or induced transgene WT-HNF4␣ homodimers were supershifted by a specific antibody against HNF4␣ (Fig. 6A).
Consistently, induction of WT-HNF4␣ resulted in a 2-fold increase in endogenous HNF1␣ mRNA level, whereas DN-HNF4␣ completely eliminated the HNF1␣ expression (Fig. 6B). To confirm that HNF4␣ directly regulates HNF1␣ transcription, we transiently transfected WT-HNF4␣-28 and DN-HNF4␣-26 cells with a luciferase reporter construct containing either the wild type HNF1␣ gene promoter (HNF1␣Luc) or a promoter that lacks a functional HNF4␣-binding site (⌬AHNF1␣Luc). As demonstrated in Fig. 6C, overexpression of WT-HNF4␣ caused a 2.5-fold increase in the luciferase reporter enzyme activity in WT-HNF4␣-28 cells transfected with HNF1␣Luc. Deletion of the HNF4␣-binding site in the HNF1␣ promoter (⌬AHNF1␣) abolished the activation induced by WT-HNF4␣ (Fig. 6C). In contrast, induction of DN-HNF4␣ caused a 71% reduction in wild type HNF1␣ promoter activity (Fig.  6C). The inhibitory effect of DN-HNF4␣ was no longer present in DN-HNF4␣-26 cells transfected with ⌬AHNF1␣Luc (Fig.  6C). Therefore, HNF4␣ directly controls HNF1␣ gene expression in pancreatic ␤-cells as it does in hepatocytes. DISCUSSION It has been demonstrated that HNF4␣ controls the expression of a large array of liver-specific genes encoding several apolipoproteins, metabolic proteins, and serum factors that are essential for hepatocyte differentiation and liver development (9). HNF4␣ is also required for HNF1␣ expression in hepatocytes (7)(8)(9). Another study in embryonic stem cell-differentiated embryoid bodies (33) shows that the absence of HNF4␣ affects the expression of genes encoding GLUT-2, aldolase B, and L-PK, which are involved in glucose transport and glycolysis. However, little is known as to how HNF4␣ regulates pancreatic ␤-cell gene expression. The primary cause of the MODY1 phenotype is impaired glucose-stimulated insulin secretion from pancreatic ␤-cells (5). The present study was therefore designed to investigate the role of HNF4␣ in the regulation of the expression of ␤-cell genes implicated in glucose metabolism and associated insulin secretion.
We found that overexpression of WT-HNF4␣ caused a left shift of glucose-stimulated insulin secretion, whereas dominant-negative suppression of HNF4␣ selectively blunted the insulin release induced by glucose and leucine but not by K ϩ depolarization. The diminished nutrient-evoked insulin secretion is associated with reduced ATP production in DN-HNF4␣expressing cells. The physiological insulin secretagogue leucine raises the cytosolic and mitochondrial Ca 2ϩ concentrations through mitochondrial metabolism downstream of glycolysis (28,34). Therefore, we suggest that loss of HNF4␣ function leads to defective mitochondrial metabolism and, as a consequence, impaired insulin secretion. The reduced mitochondrial oxidation of [2-14 C]pyruvate and the abrogation of mitochondrial membrane hyperpolarization elicited by glucose, pyruvate, and methyl succinate indicate impaired mitochondrial tricarboxylic acid cycle enzyme activity and partial uncoupling of the mitochondrial respiratory chain.
Quantitative Northern blot analysis allows us to identify HNF4␣ target genes responsible for defective metabolism-secretion coupling. HNF4␣ indeed regulates the expression of genes encoding GLUT-2, aldolase B, and L-PK in pancreatic ␤-cells (Fig. 5), as inferred from previous studies in hepatocytes and embryonic stem cell-differentiated embryoid bodies (8,9). Most importantly, we demonstrate that HNF4␣ alters the mRNA expression of mitochondrial OGDH E1 subunit and UCP-2 (Fig. 5), which may indeed contribute more significantly to the impaired metabolism-secretion coupling. The phenotype and gene expression patterns in DN-HNF4␣-expressing cells are strikingly similar to those of DN-HNF1␣-expressing cells (22,29). This prompted us to investigate whether HNF4␣ regulates ␤-cell expression through HNF1␣ function, as in hepatocytes (9). We provide unprecedented evidence that HNF4␣ is required for HNF1␣ expression in pancreatic ␤-cells.
This conclusion is based on the use of an artificial dominantnegative hnf4␣ mutation. The naturally occurring human mutations of HNF4␣ do not function in a dominant-negative manner (6,35). It is to be expected that a mutation with such repressive action on the endogenous HNF4␣ function would cause embryonic lethality, as is the case in the hnf4␣ knock-out mouse (15). Haploinsufficiency or reduced gene dosage of HNF4␣ may thus explain the mechanism leading to the MODY1 phenotype (33). The INS-1 cell line expressing DN-HNF4␣ provides a convenient model to explore the impact of impaired HNF4␣ function on ␤-cell gene expression and metabolism-secretion coupling. This goal cannot be achieved by the introduction of one of the human HNF4␣ mutations into ␤-cell lines. In fact, the induction of a nonsense mutation HNF4␣Q268X to a level similar to DN-HNF4␣ had no detectable consequences on ␤-cell gene expression and metabolismsecretion coupling. 2 MODY1 patients display secretory defects not only in ␤-cells but also in the glucagon-secreting ␣-cells and the pancreatic polypeptide-secreting cells (36,37). However, this general effect on islet hormone release does not seem due to an effect on the development and differentiation of the endocrine pancreas, since altered HNF4␣ function did not affect the expression of PDX1 and other transcription factors determining pancreatic phenotype. On the other hand, loss of HNF4␣ function may cause reduced ␤-cell insulin content secondary to defective HNF1␣ function (22,29).