Hypoxia reduces HNF4α/MODY1 protein expression in pancreatic β-cells by activating AMP-activated protein kinase

Hypoxia plays a role in the deterioration of β-cell function. Hepatocyte nuclear factor 4α (HNF4α) has an important role in pancreatic β-cells, and mutations of the human HNF4A gene cause a type of maturity-onset diabetes of the young (MODY1). However, it remains unclear whether hypoxia affects the expression of HNF4α in β-cells. Here, we report that hypoxia reduces HNF4α protein expression in β-cells. Hypoxia-inducible factor was not involved in the down-regulation of HNF4α under hypoxic conditions. The down-regulation of HNF4α was dependent on the activation of AMP-activated protein kinase (AMPK), and the reduction of HNF4α protein expression by metformin, an AMPK activator, and hypoxia was inhibited by the overexpression of a kinase-dead (KD) form of AMPKα2. In addition, hypoxia decreased the stability of the HNF4α protein, and the down-regulation of HNF4α was sensitive to proteasome inhibitors. Adenovirus-mediated overexpression of KD-AMPKα2 improved insulin secretion in metformin-treated islets, hypoxic islets, and ob/ob mouse islets. These results suggest that down-regulation of HNF4α could be of importance in β-cell dysfunction by hypoxia.

Pancreatic ␤-cells sense glucose and secrete insulin to maintain normal glucose levels. Because pancreatic ␤-cells are highly dependent on oxidative phosphorylation for ATP production, they require large amounts of oxygen, especially at increased glucose levels (1,2). Indeed, we and others demonstrated that pancreatic islets and insulin-producing ␤-cell lines become hypoxic under high glucose conditions and that the pancreatic islets of diabetic mice are hypoxic (3)(4)(5). Furthermore, we recently indicated that moderate hypoxia leads to ␤-cell dysfunction with a selective down-regulation of genes including Mafa, Pdx1, and Ins1, which play important roles in pancreatic ␤-cells (6). These results suggest that hypoxia is involved in the deterioration of ␤-cell function.
Hepatocyte nuclear factor 4␣ (HNF4␣) 3 is a transcription factor belonging to the nuclear receptor superfamily and is expressed in the pancreas (including ␤-cells), liver, kidney, and intestine (7). We found that mutations of the human HNF4A gene cause a particular form of maturity-onset diabetes of the young (MODY), that is, MODY1, which is characterized by autosomal dominant inheritance, early age of onset, and pancreatic ␤-cell dysfunction (8). In addition, targeted disruption of HNF4␣ in pancreatic ␤-cells leads to defective insulin secretion in mice (9,10). These findings demonstrate the important role of HNF4␣ in pancreatic ␤-cells.
In the present study, we investigated the impact of hypoxia on HNF4␣ expression in MIN6 cells and mouse islets. We demonstrated that hypoxia decreases HNF4␣ protein expression via proteasome-mediated degradation. The hypoxia-induced down-regulation of HNF4␣ was regulated by the activation of AMP-activated protein kinase (AMPK). This reduction of HNF4␣ protein expression was recovered by inactivation of AMPK and re-oxygenation. Our results suggest that down-regulation of HNF4␣ is a novel mechanism of ␤-cell dysfunction by hypoxia.

Down-regulation of HNF4␣ protein expression by hypoxia
MIN6 cells were cultured under moderately hypoxic conditions (3-7% oxygen tension) for 24 h, and HNF4␣ expression levels were examined by Western blot analysis. Hypoxia significantly decreased HNF4␣ protein levels, but not ␤-actin, in a dose-dependent manner (Fig. 1, A and B). Hypoxia did not affect HNF1␣/MODY3 (11) and HNF1␤/MODY5 (12) protein levels in MIN6 cells (Fig. 1C). Similar to MIN6 cells, decreased HNF4␣ protein levels by hypoxia were observed in pancreatic islets (Fig. 1D). Hypoxia reduces the expression of Hnf3b/Foxa2 mRNA in MIN6 cells (6). We then examined the expression levels of Hnf4a mRNA. Hypoxia for 12 h had no effect on Hnf4␣ mRNA expression in MIN6 cells (Fig. 1E). However, a significant down-regulation of HNF4␣ protein (61.3% of control; p Ͻ 0.01) was detected in MIN6 cells following 5% oxygen tension for 12 h (Fig. 1, F and G), suggesting the post-transcriptional regulation of HNF4␣ expression. In this study, we focused on the hypoxia-induced down-regulation of HNF4␣ at the protein level.

Effect of down-regulation of HNF4␣ expression on ␤-cells
HNF4␣ plays an important role in glucose-stimulated insulin secretion by ␤-cells (8). Suppression of endogenous HNF4␣ consistently reduced insulin secretion in MIN6 cells (supple- Figure 1. Effect of hypoxia on HNF4␣ expression in ␤-cells. A, MIN6 cells were exposed to the indicated oxygen tension (% O 2 ) for 24 h, and HNF4␣ expression was examined by Western blotting. B, relative HNF4␣ protein levels were calculated (n ϭ 3). C, effect of hypoxia for 24 h on HNF1␣ and HNF1␤ expression in MIN6 cells (n ϭ 3). D, isolated mouse islets were incubated at either 5% O 2 or 20% O 2 for 24 h, and HNF4␣ protein levels were evaluated by Western blot analysis (n ϭ 3). E, MIN6 cells were cultured at either 5% O 2 or 20% O 2 for 12 h and 24 h, and Hnf4a mRNA levels were analyzed by qPCR (n ϭ 3). The Hnf4␣ mRNA level was normalized to that of TBP. F and G, MIN6 cells were cultured at the same conditions as in E, and HNF4␣ protein levels were examined by Western blot analysis (n ϭ 3). All data are presented as mean Ϯ S.E. (S.E.; error bars). N.S., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

Regulation of HNF4␣ expression by hypoxia
mental Fig. S1, A and B). In addition, HNF4␣ knockdown significantly increased the number of propidium iodide-positive dead cells (supplemental Fig. S1C). We previously reported that hypoxia induces an insulin secretion defect and cell death in MIN6 cells (6). Down-regulation of HNF4␣ may be involved in this dysfunction by hypoxia. In contrast, cell proliferation rate was unaffected by the down-regulation of HNF4␣ (supplemental Fig. S1D).

Effect of HIF on HNF4␣ expression by hypoxia
Hypoxia-inducible factor (HIF) is a transcription factor that regulates the expression of genes mediating adaptive responses to hypoxia. HIF-1␣ protein is degraded under normoxic conditions, whereas it is stabilized under hypoxic conditions and binds to HIF-1␤ to act as a transcription factor (13). Hypoxia (5% O 2 ) increased the expression of HIF-1␣ protein in a timedependent manner in MIN6 cells ( Fig. 2A). We then investigated the role of HIF-1 in the hypoxia-induced down-regulation of HNF4␣. Overexpression of a constitutively active form of HIF-1␣ by retrovirus infection did not affect HNF4␣ expression (Fig. 2, B and C). Next, we introduced HIF-1␣ short hairpin RNA (shRNA) into MIN6 cells, and successful HIF-1␣ knock-down was confirmed under hypoxic conditions (Fig. 2D). Suppression of HIF-1␣ did not affect the down-regulation of HNF4␣ at the 5% oxygen condition (Fig. 2, D and E). In addition to HIF-1␣, suppression of HIF-2␣, HIF-3␣, or HIF-1␤ failed to reverse the down-regulation of HNF4␣ by hypoxia (supplemental Fig. S2). These findings indicate that HIF is not involved in the down-regulation of HNF4␣ under hypoxic conditions.

Effect of AMPK on HNF4␣ expression by hypoxia
In addition to the stabilization of HIF-1␣, hypoxia is known to activate AMPK (14). Indeed, the phosphorylation of both AMPK and acetyl-CoA carboxylase, a substrate of AMPK, was induced after exposure to hypoxia for 1 h in MIN6 cells (supplemental Fig. S3). Down-regulation of HNF4␣ was subsequently detected after exposure to hypoxia for 12 h. We then examined whether activation of AMPK could affect HNF4␣ expression. Both 5-aminoimidazole-4-carboxamide-1-␤-Dribofuranoside (AICAR), a cell-permeable activator of AMPK, and metformin decreased the expression of HNF4␣ in MIN6 cells (Fig. 3A). Similar to MIN6 cells, metformin also decreased HNF4␣ protein levels in pancreatic islets (Fig. 3, B and C).
The kinase-dead (KD) form of AMPK␣2 (Lys-45 was changed to Arg) reportedly functions as a dominant inhibitory protein that eliminates AMPK activity (15). We examined the effect of KD-AMPK␣2 overexpression by retrovirus on the metformin-induced down-regulation of HNF4␣. Phosphorylation of acetyl-CoA carboxylase was inhibited by KD-AMPK␣2 overexpression (Fig. 3D). In addition, HNF4␣ down-regulation was significantly suppressed by KD-AMPK␣2 overexpression (Fig. 3, D and E). We next examined the effect of AMPK activity on insulin secretion in MIN6 cells. Consistent with the decreased expression of HNF4␣, metformin reduced insulin secretion at 22 mM glucose (Fig. 3, F and G). In accordance with the increased expression of HNF4␣, KD-AMPK␣2 improved insulin release by 22 mM glucose in metformin-treated MIN6 cells. Similarly, adenovirus-mediated overexpression of KD-AMPK␣2 increased HNF4␣ protein expression and improved glucose-stimulated insulin secretion in metformin-treated islets (Fig. 3, H-J).
We also investigated the effect of KD-AMPK␣2 overexpression on the hypoxia-induced down-regulation of HNF4␣. Phosphorylation of AMPK by hypoxia was inhibited by KD-AMPK␣2 overexpression (supplemental Fig. S4). Overexpression of KD-AMPK␣2 in MIN6 cells suppressed the reduction of HNF4␣ protein expression by hypoxia, as with metformin (Fig.  4, A and B). Collectively, these findings indicate that hypoxia decreases HNF4␣ protein levels via the activation of AMPK. Insulin secretion was stimulated 3.4-fold by 22 mM (versus 2.2 mM) glucose in MIN6 cells under 20% O 2 tension, whereas hypoxic MIN6 cells exhibited dysregulated insulin secretion (increased insulin secretion at low glucose and blunted insulin secretion at high glucose) (6) (Fig. 4, C and D). However, the diminished glucose-stimulated changes in insulin secretion under 5% O 2 tension were significantly increased by overexpression of KD-AMPK␣2 (pMX control, 1.9-fold; pMX-KD-AMPK␣2, 2.4-fold, p Ͻ 0.01) (Fig. 4D). Overexpression of KD-AMPK␣2 also increased HNF4␣ expression and enhanced insulin secretion in response to high glucose in islets under hypoxic conditions (Fig. 4, E-G). These results suggest the functional significance of AMPK-and hypoxia-dependent regulation of HNF4␣ expression on insulin secretion.

HNF4␣ protein stability under hypoxic conditions
Hypoxia reportedly leads to the degradation of proteins, such as estrogen receptor ␣ and Na,K-ATPase (16,17). We next evaluated whether HNF4␣ protein stability is influenced by hypoxia. MIN6 cells were cultured under 20 or 5% oxygen ten-sion in the presence of cycloheximide (CHX), a translational inhibitor. During incubation with CHX, HNF4␣ protein decayed rapidly in the 5% oxygen condition (Fig. 5A), indicating that hypoxia decreases the stability of HNF4␣ protein. We also investigated HNF4␣ stability by metformin treatment. Metformin promoted HNF4␣ degradation (Fig. 5B). Then, HNF4␣ protein stability was tested in the presence or absence of the proteasome inhibitor MG132 (Fig. 5C). As shown in Fig.  5, C-F, proteasome inhibition resulted in the stabilization of A, effect of AMPK activators on HNF4␣ protein levels. MIN6 cells were cultured at the indicated concentration of AICAR or metformin for 24 h, and Western blotting was performed. B and C, isolated mouse islets were treated with 2 mM metformin for 24 h, and HNF4␣ protein levels were examined. D and E, MIN6 cells expressing the pMX empty vector or pMX-KD-AMPK␣2 vector were treated with 2 mM metformin for 20 h, and HNF4␣ protein levels were examined (n ϭ 3). F and G, an insulin secretion assay was performed (n ϭ 4 -5), and insulin concentration was determined by an insulin ELISA. Fold-change in glucose-stimulated insulin secretion (insulin level at 22 mM glucose divided by that at 2.2 mM glucose) is shown (n ϭ 4 -5). H, isolated mouse islets expressing either LacZ or KD-AMPK␣2 were treated with 2 mM metformin for 20 h, and HNF4␣ protein levels were examined by Western blot analysis. I and J, islet insulin secretion was examined. Insulin levels are expressed as absolute values or as fold-change (n ϭ 11-12). All data are presented as mean Ϯ S.E. (error bars). N.S., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

Regulation of HNF4␣ expression by hypoxia
HNF4␣ in both hypoxic and metformin-treated conditions (compare lanes 3 and 4 of Fig. 5D). Another proteasome-specific inhibitor, epoxomicin, also stabilized HNF4␣ protein expression under hypoxia (supplemental Fig. S5). These findings indicate that proteasome-mediated degradation is involved in the hypoxia-and AMPK activation-induced down-regulation of HNF4␣.
AMPK activation leads to decreased levels of proteins by regulating the binding of RNA-binding proteins within the 3Ј-untranslated region (3Ј-UTR) of target mRNAs (18). To address the role of the 3Ј-UTR in HNF4␣ mRNA, MIN6 cells were infected with a retroviral HNF4␣-FLAG expression vector (lacking both the 3Ј-UTR and 5Ј-UTR of the HNF4␣ gene). Both hypoxia and metformin decreased HNF4␣ protein levels (Fig. 6, A and B), indicating that the 3Ј-UTR is not involved in down-regulation of HNF4␣ by hypoxia/AMPK activation in MIN6 cells.

Phosphorylation of HNF4␣ by hypoxia
The phosphorylation of proteins triggers protein degradation (19,20). AMPK activation reportedly leads to the phos-phorylation of the liver type of HNF4␣ at serine 304 (serine 291 of HNF4␣7) in vitro (21). Thus, we examined whether this phosphorylation is essential for the hypoxia-induced down-regulation of HNF4␣ in HNF4␣7-FLAG-expressing MIN6 cells. As shown in Fig. 7A, HNF4␣7 with a mutation of serine 291 to alanine (S291A) was still down-regulated by hypoxia. Database screening (ppsp.biocuckoo.org/index.php) identified another potential AMPK phosphorylation site ( 117 TRRSSYEDS 125 , potential phosphorylation serine 121 is underlined) in HNF4␣7. However, the replacement of serine 121 with alanine (S121A) also did not restore HNF4␣7 instability (Fig. 7A). Because both S121A and S291A mutants were as sensitive to hypoxia-induced down-regulation as wild-type HNF4␣7, we investigated whether HNF4␣7 was phosphorylated by hypoxia and AMPK activation in MIN6 cells. Phosphorylated proteins show slower migration in Phos-tag SDS-PAGE due to the selective binding of phosphorylated amino acids to the Phos-tag reagent (22). After treatment with metformin or hypoxia, whole cell lysates of MIN6 cells were separated by SDS-PAGE with Phos-tag, and 20 h, and HNF4␣ protein levels were examined (n ϭ 3). C and D, an insulin secretion assay was performed (n ϭ 6 -9), and insulin concentration was determined by an insulin ELISA. Fold-change in glucose-stimulated insulin secretion (insulin level at 22 mM glucose divided by that at 2.2 mM glucose) is shown (n ϭ 6 -9). E, mouse isolated islets expressing either LacZ or KD-AMPK␣2 were treated with 5% O 2 for 20 h, and HNF4␣ protein levels were examined by Western blot analysis. F and G, islet insulin secretion was examined, and insulin levels are expressed as absolute values or as fold-change (n ϭ 5-10). Data are presented as mean Ϯ S.E. (error bars). N.S., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. blotted with anti-AMPK and anti-HNF4␣ antibodies. Treatment with hypoxia or metformin led to AMPK phosphorylation (Fig. 3). Phosphorylated AMPK following hypoxia or metformin treatment was detected as a band with slower migration using an anti-AMPK antibody (Fig. 7B). It has been reported that protein kinase A phosphorylates HNF4␣ (23). Consistently, a band shift of HNF4␣ was observed by treatment with forskolin, which activates the protein kinase A pathway via cAMP. In contrast, Western blot analysis with an anti-HNF4␣ antibody did not show a detectable band shift by treatment with hypoxia or metformin, suggesting that the HNF4␣ protein is not phosphorylated by metformin and 5% hypoxia in MIN6 cells.

The effect of ERK5 and E3 ubiquitin ligases on HNF4␣ expression by hypoxia
A recent study showed that extracellular signal-regulated kinase 5 (ERK5)/mitogen-activated protein kinase 7 (MAPK7) up-regulates proteasome levels upon mTOR (mechanistic target of rapamycin) signaling inhibition (24). Because hyp- oxia increased the expression of ERK5 protein (Fig. 8A), we examined the effect of ERK5 on HNF4␣ expression. The hypoxia-induced down-regulation of HNF4␣ was similar between control and ERK5 knockdown MIN6 cells (Fig. 8, B and C, supplemental Fig. S6, A and B), indicating that ERK5 does not contribute to the down-regulation of HNF4␣ by hypoxia.

Restoration of HNF4␣ protein expression by re-oxygenation or inactivation of AMPK
We next examined the effect of re-oxygenation on HNF4␣ expression. Expression of HNF4␣ protein was increased in MIN6 cells that were re-oxygenized for more than 3 h (R3-R12) (Fig. 9A). Re-oxygenation for 12 h (R12) also increased HNF4␣ expression in pancreatic islets (Fig. 9, B and C). Consistent with our previous finding that pancreatic islets of diabetic ob/ob mice are hypoxic (3), HNF4␣ protein expression was significantly   decreased in ob/ob mouse islets (Fig. 9, D and E). However, expression levels of the HNF4␣ protein became similar between control and ob/ob islets after incubation at 20% oxygen tension.
Finally, we investigated the effect of KD-AMPK␣2 on insulin secretion in ob/ob islets. Insulin secretion from ob/ob islets was unresponsive to glucose stimulation (28). Consistently, insulin Figure 9. Impact of re-oxygenation on HNF4␣ expression. A, MIN6 cells were cultured at 5% O 2 for 24 h and then cultured at 20% O 2 (re-oxygenation) for the indicated time (0.5-12 h). HNF4␣ expression was examined by Western blot analysis. B and C, isolated mouse islets were cultured at 5% O 2 for 20 h and then cultured at 20% O 2 for 12 h. HNF4␣ protein levels were examined (n ϭ 3). D and E, islets were isolated from C57BL/6J or ob/ob mice. HNF4␣ protein levels were examined immediately after islet isolation and after incubation at 20% O 2 for 12 h (n ϭ 4 -6). F, islets from ob/ob mice were infected with adenovirus expressing LacZ or KD-AMPK␣2 for 2 h, and cultured for 36 h. Expression of AMPK␣ and HNF4␣ was examined by Western blotting. G, insulin secretion assay was performed using ob/ob mouse islets expressing either LacZ (n ϭ 13) or KD-AMPK␣2 (n ϭ 24). Insulin concentration was determined by an insulin ELISA. H, fold-change in glucose-stimulated insulin secretion (insulin level at 22 mM glucose divided by that at 2.2 mM glucose) is shown. All data are presented as mean Ϯ S.E. (error bars). N.S., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

Regulation of HNF4␣ expression by hypoxia
release by high glucose did not differ from that by low glucose in control ob/ob islets. However, overexpression of KD-AMPK␣2 increased the expression of HNF4␣ in ob/ob islets, and insulin secretion by high glucose was significantly improved (Fig. 9, F-H).

Discussion
HNF4␣7 and HNF4␣8 are major isoforms of HNF4␣ in pancreatic ␤-cells, and HNF4␣ plays important roles in the function of these cells (8,29,30). In the present study, we demonstrated that hypoxia leads to the down-regulation of HNF4␣ protein in ␤-cells. HNF4␣ down-regulation was dependent on AMPK activation and was sensitive to proteasome inhibitors, indicating that HNF4␣ proteasomal degradation is regulated, at least in part, by AMPK activation (Fig. 10). A reduction of hepatic HNF4␣ (HNF4␣1 and HNF4␣2 are major hepatic forms) by AICAR treatment has also been reported (31). HNF4␣ plays an essential role in glucose-stimulated insulin secretion by pancreatic ␤-cells (8,9) (supplemental Fig. S1). Metformin treatment attenuated high glucose-stimulated insulin secretion with the decreased expression of HNF4␣. Conversely, overexpression of KD-AMPK␣2 improved insulin secretion with the increased expression of HNF4␣ (Fig. 3). Collectively, these results suggest that the regulation of HNF4␣ expression by AMPK has a functional consequence on insulin secretion. Hypoxia and AMPK activation inhibit multiple steps in the process of insulin secretion (6,(32)(33)(34)(35)(36). Our findings uncover a new role for AMPK in impaired insulin secretion by ␤-cells. Decreased HNF4␣ expression in ␤-cells could reduce oxygen consumption by reducing insulin secretion under hypoxic conditions. Thus, the down-regulation of HNF4␣ may be a fail-safe mechanism against hypoxia. However, prolonged exposure of ␤-cells to hypoxia, in turn, would have deleterious effects on insulin secretion by reducing HNF4␣/MODY1 transcription factor levels and could contribute to the development of type 2 diabetes. AMPK activation reportedly induces the phosphorylation of hepatic HNF4␣ in vitro (21), but the replacement of potential AMPK phosphorylation sites (Ser 121 and Ser 291 ) did not affect HNF4␣ instability in MIN6 cells (Fig. 7). A mobility shift of HNF4␣ after hypoxia or metformin treatment was not detected using Phos-tag SDS-PAGE. In addition, we could not detect a phosphorylation signal in HNF4␣ immunoprecipitates that were prepared from hypoxic MIN6 cells using the commercially available anti-phospho-AMPK substrate motif MultiMab antibody (data not shown). Although the possibility that ␤-celltype HNF4␣ is a direct target of AMPK cannot be excluded, these results suggest that HNF4␣ is not phosphorylated directly by metformin and/or hypoxia in MIN6 cells.
We have demonstrated that hypoxia increases the expression of ERK5, and ERK5 has been shown to up-regulate proteasome levels (24). Moreover, AMPK activation leads to proteasomal degradation of target proteins via the ubiquitin ligases atrogin-1 and Nedd4l (25,26). However, these molecules do not appear to account for the mechanism of HNF4␣ degradation by AMPK (Fig. 8). Further studies are necessary to clarify the role of AMPK on HNF4␣ expression in ␤-cells.
We found that re-oxygenation restored HNF4␣ expression in both control and ob/ob islets (Fig. 9). In addition, restoration of HNF4␣ in ob/ob islets significantly improved insulin secretion. Restoration of HNF4␣ alone may not be sufficient to normalize completely the impaired secretion of insulin because ob/ob islets exhibit additional abnormalities (37). However, our findings suggest that methods to enhance oxygenation in hypoxic ␤-cells could be an effective therapeutic approach to remedy insulin secretion defects in type 2 diabetes by restoring HNF4␣.
There are limitations in this study. First, the mechanisms by which AMPK activation leads to the down-regulation of HNF4␣ are unclear. Second, we do not have direct evidence that hypoxia leads to pancreatic ␤-cell HNF4␣ protein degradation in vivo. Investigation of HNF4␣ protein levels of mouse ␤-cells in a low oxygen environment would be a major undertaking.
In conclusion, we demonstrated that hypoxia decreased HNF4␣ protein expression via proteasome-mediated protein degradation in ␤-cells. Furthermore, the hypoxia-induced down-regulation of HNF4␣ was regulated by the activation of AMPK. Because a small perturbation of HNF4␣ activity in ␤-cells results in MODY1 (38), our findings suggest that downregulation of HNF4␣ could be of importance in ␤-cell dysfunction by hypoxia.

Animals
C57BL/6J and B6.Cg-Lep ob /J (ob/ob) mice were purchased from KBT Oriental Co., Ltd. (Saga, Japan). The mice were kept under specific pathogen-free conditions in a 12-h light (7:00 -19:00)/12-h dark (19:00 -7:00) cycle with free access to water and normal mouse chow (CE-2; CLEA, Tokyo, Japan). Room temperature was maintained at 22 Ϯ 1-2°C. Handling and killing of the mice by cervical dislocation were in compliance with the animal care guidelines of Kumamoto University. This study was approved by the animal research committee of Kumamoto University and all animal experimental protocols were approved by the Kumamoto University Ethics Review Committee for Animal Experimentation.

Western blotting
Western blotting was performed as described previously (41). For specific detection of proteins, the following primary antibodies were used: anti-HNF4␣ (

Insulin secretion assay
Control MIN6 cells or HNF4␣ knockdown MIN6 cells were seeded in a 24-well plate at 5.0 ϫ 10 5 cells/well and maintained for 2-3 days. Then, mouse pancreatic islets from C57BL/6J mice or ob/ob mice were cultured in RPMI-1640 medium for 2 days, and islets of similar size were randomly picked up, divided into 2 groups (20 -30 islets), and separately infected with recombinant adenovirus (LacZ or KD-AMPK␣2, 2.0 ϫ 10 5 infectious units (IFU)) for 2 h. After 24-h incubation in RPMI-1640 medium, these islets were further cultured in 2 mM metformin or 5% O 2 for 20 h, and then the insulin secretion assay was performed until 48 h after adenoviral infection. The islets were preincubated for 30 min in KRBH buffer containing 2.2 mM glucose and 0.5% (v/v) BSA. They were incubated in 2.2 or 22 mM glucose containing KRBH buffer for 30 min, and the culture supernatant was recovered to evaluate insulin secretion. Insulin concentration was determined by a mouse insulin ELISA (TMB) kit (AKRIN-011T; Shibayagi Co., Ltd., Gunma, Japan). For MIN6 cells, insulin levels were standardized by whole cell protein content.

Apoptosis assay
Control MIN6 cells or HNF4␣ knockdown MIN6 cells were incubated for 48 h, and then trypsinized and collected by centrifugation at 6,000 rpm for 10 min. Cells were stained with propidium iodide for 5 min at room temperature in the dark, and stained cells were immediately analyzed using a FACSCalibur flow cytometer (BD Biosciences) and FlowJo software (Tomy Digital Biology, Tokyo, Japan).

Cell proliferation assay
Control or HNF4␣ knockdown MIN6 cells were seeded in a 96-well plate at 1.5 ϫ 10 4 cells/well. After 2 days, a cell proliferation assay was performed for 3 consecutive days using cell proliferation reagent WST-1 (Roche Diagnostic, Mannheim, Germany) and the absorbance (450/655) was measured by iMark microplate reader (Bio-Rad).

Protein degradation assay
MIN6 cells were seeded at 2.0 ϫ 10 5 cells in a 35-mm dish. After MIN6 cells were cultured at 20% O 2 or 5% O 2 condition for 18 h, they were then treated with 10 g/ml of CHX for the indicated time (0 -6 h). To examine the effects of MG-132 or epoxomicin on HNF4␣ protein expression, MIN6 cells were cultured at either 20% O 2 or 5% O 2 for 20 h and then treated with CHX in the presence or absence of MG132 (or epoxomicin) for 4 h. MIN6 cells were also cultured with or without 2 mM metformin for 20 h and then treated with CHX (ϩ)/MG132 (ϩ) or CHX (ϩ)/MG132 (Ϫ) for 4 h. HNF4␣ protein levels were analyzed by Western blotting.

Phos-tag SDS-PAGE
MIN6 cells were treated with 2 mM metformin or cultured under the hypoxic 5% O 2 condition for 12 and 24 h. Whole cells were lysed in EDTA (Ϫ) RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 20 mg/ml of Na 3 VO 4 , 10 mM NaF, 1 mM PMSF, and 1% (v/v) protease inhibitor mixture). Phos-tag-containing polyacrylamide gels were made with Phos-tag acrylamide (40 M; Wako Pure Chemicals, Osaka, Japan) and MnCl 2 (40 M) according to the manufacturer's recommendation. Twenty micrograms of proteins were subjected to 10% SDS-PAGE. Dr. Western (Oriental Yeast Co., Ltd, Tokyo, Japan) was used for a protein ladder marker. After electrophoresis, the Phos-tag gel was soaked in transfer buffer (25 mM Tris and 192 mM glycine) containing 1 mM EDTA for 10 min to remove the Mn 2ϩ before transferring the proteins to a membrane, and then Western blotting was performed.

Statistical analysis
The significance of differences was assessed with an unpaired t test and a value of p Ͻ 0.05 was considered to be statistically significant.
Author contributions-Y. S. and K. Y. designed the study and wrote the paper. Y. S., T. T., C. S., M. F. K., T. Y., and M. I. contributed to the acquisition of data, the statistical data analyses, and drafting of the manuscript. All authors have approved the final version of the manuscript.