Glucose regulates MafA transcription factor abundance and insulin gene expression by inhibiting AMP-activated protein kinase in pancreatic β-cells

Insulin mRNA expression in pancreatic islet β-cells is up-regulated by extracellular glucose concentration, but the underlying mechanism remains incompletely understood. MafA is a transcriptional activator specifically enriched in β-cells that binds to the insulin gene promoter. Its expression is transcriptionally and posttranscriptionally regulated by glucose. Moreover, AMP-activated protein kinase (AMPK), a regulator of cellular energy homeostasis, is inhibited by high glucose, and this inhibition is essential for the up-regulation of insulin gene expression and glucose-stimulated insulin secretion (GSIS). Here we mutagenized the insulin promoter and found that the MafA-binding element C1/RIPE3b is required for glucose- or AMPK-induced alterations in insulin gene promoter activity. Under high-glucose conditions, pharmacological activation of AMPK in isolated mouse islets or MIN6 cells by metformin or 5-aminoimidazole-4-carboxamide riboside decreased MafA protein levels and mRNA expression of insulin and GSIS-related genes (i.e. glut2 and sur1). Overexpression of constitutively active AMPK also reduced MafA and insulin expression. Conversely, pharmacological AMPK inhibition by dorsomorphin (compound C) or expression of a dominant-negative form of AMPK increased MafA and insulin expression under low-glucose conditions. However, AMPK activation or inhibition did not change the expression levels of the β-cell-enriched transcription factors Pdx1 and Beta2/NeuroD1. AMPK activation accelerated MafA protein degradation, which is not dependent on the proteasome. We also noted that MafA overexpression prevents metformin-induced decreases in insulin and GSIS-related gene expression. These findings indicate that high glucose concentrations inhibit AMPK, thereby increasing MafA protein levels and activating the insulin promoter.

Pancreatic islet ␤-cells maintain blood glucose homeostasis by secreting the glucose-lowering hormone insulin in response to a rise in blood glucose levels within minutes. This short-term response of ␤-cells to glucose, called glucose-stimulated insulin secretion (GSIS), 2 requires normal function of ␤-cell-enriched gene products such as glut2, sur1, and kir6.2 (1,2). Glucose is transported into ␤-cells through the transmembrane glucose transporter Glut2 and is metabolized by the glycolytic pathway and the Krebs cycle to generate ATP. The rise in ATP concentration triggers closure of the ATP-sensitive K ϩ channel Sur1/ Kir6.2 and subsequent plasma membrane depolarization and Ca 2ϩ influx, which ultimately leads to insulin secretion (1,2).
In addition to GSIS, long-term (for hours) exposure of ␤-cells to high or low glucose concentrations also influences insulin (preproinsulin) gene expression (3). It has been suggested that the ␤-cell-enriched transcription factors Pdx1, Beta2/ NeuroD1, and MafA mediate glucose-regulated expression of insulin mRNA. Pdx1 binds to the cis-regulatory enhancer elements A1, A3, and GG2 within the insulin promoter region (4 -7), and Beta2/NeuroD1 and MafA bind to the E1 and C1/RIPE3b elements (8 -10), respectively (Fig. 1A). These transcription factors cooperate to establish ␤-cell-restricted expression of the insulin gene (11,12). However, the mechanisms underlying glucose regulation of insulin gene expression remain incompletely understood.
MafA, a member of the large Maf family of basic leucine zipper transcriptional activators, was identified as the ␤-cellspecific factor binding to the C1/RIPE3b element (11)(12)(13). The C1/RIPE3b element is one of the cis-regulatory elements mediating ␤-cell-restricted and glucose-regulated expression of insulin (14,15). MafA activates not only the transcription of insulin but also of other genes involved in ␤-cell-specific functions, particularly GSIS, playing an important role in the maturation of ␤-cells (16 -18). During mouse embryonic development, MafA expression is restricted to insulin-positive immature ␤-cells in the pancreas, but it remains low, and the cells lack the ability for GSIS (19 -21). Upon terminal differentiation, ␤-cells acquire this ability in parallel with elevated MafA expression. Ablation of the mafA gene in mice does not affect ␤-cell differentiation but leads to age-related glucose intolerance (16). During the progression of type 2 diabetes in human patients or animal models, such as the db/db mouse, ␤-cells gradually lose their insulin expression and GSIS ability. This dysfunction of ␤-cells, called ␤-cell failure, is preceded by reduced expression of the MafA, Pdx1, and Nkx6.1 transcription factors (22,23), but transgenic expression of MafA in ␤-cells partially rescues ␤-cell function in mice (24).
AMP-activated protein kinase (AMPK) is a heterotrimeric serine/threonine kinase consisting of catalytic ␣ (␣1 or ␣2) and non-catalytic ␤ (␤1 or ␤2) and ␥ (␥1, ␥2, or ␥3) subunits (29). AMPK is activated by intracellular ATP depletion and the consequent increase in AMP and thus functions as an intracellular energy sensor, regulating cellular metabolism (30). AMP allosterically activates AMPK enzyme activity by binding to the ␥ subunit and also promotes activating phosphorylation at Thr-172 of the ␣ subunit (31). Activated AMPK phosphorylates acetyl-CoA carboxylase (ACC) and hydroxymethylglutaryl-CoA reductase, thereby inhibiting fatty acid and cholesterol synthesis, respectively, and prevents ATP consumption (32,33). Previous studies have demonstrated that AMPK activity in ␤-cells is inhibited under high-glucose conditions and affects GSIS and insulin gene expression (34,35). We therefore examined whether AMPK regulates MafA and found that AMPK negatively regulates MafA protein stability. We demonstrate here that, under high glucose conditions, AMPK activity is suppressed, which leads to MafA protein accumulation and subsequent up-regulation of insulin mRNA expression as well as that of GSIS genes, including glut2 and sur1.

Glucose regulates insulin gene promoter activity predominantly through the C1/RIPE3b element
To examine the roles of cis-regulatory elements in the insulin promoter in its regulation by glucose, luciferase assays were performed using a human insulin promoter construct (Fig. 1A). Locations of the binding sites for Pdx1 (A1, A3, and GG2), MafA (C1/RIPE3b), and Beta2/NeuroD1 (E1) are also indicated. B, the firefly luciferase reporter constructs (0.5 g) indicated in A and a Renilla luciferase expression plasmid, pEF-Rluc (0.5 g), were transfected into MIN6 cells, and the cells were incubated in medium containing high (Hi, 20 mM) or low (Lo, 2 mM) glucose for 16 h. Data are expressed relative to the activity in cells that received pGL4/h-ins-p (WT) under high-glucose concentration. Data are mean Ϯ S.E. of five independent experiments. *, p Ͻ 0.05; n.s., not significant (Student's t test). C and D, MIN6 cells (C) or isolated mouse islets (D) were incubated in high-glucose (Hi, 20 mM) or low-glucose (Lo, 2 mM) medium for 16 h, and whole-cell extracts were analyzed by immunoblotting using the indicated antibodies. TBP was analyzed as a loading control. The data in C and D are representative of two independent biological experiments. The intensity of the bands was quantified using ImageJ software, and the relative amounts (averages of two independent experiments) are indicated below the bands.

Glucose regulation of MafA via AMPK
The reporter plasmid was transfected into the glucose-responsive insulinoma-derived cell line MIN6 and cultured in a medium containing high (20 mM) or low (2 mM) glucose concentrations. Consistent with data reported previously (9), luciferase reporter activity was higher in a high-glucose environment than in a low-glucose environment (Fig. 1B). We mutated three major cis elements (GG2, C1/RIPE3b, and E1) that are bound by the Pdx1, MafA, and Beta2/NeuroD1 transcription factors, respectively, and assayed them for regulation by glucose (Fig. 1B). Mutation of the GG2 or E1 element (mutGG2 or mutE1) resulted in lower luciferase activity, but a glucose response remained. The promoter construct containing a mutation in the C1/RIPE3b (mutC1) also exhibited decreased activity, but it became completely unresponsive to glucose concentration, indicating that the C1/RIPE3b element is essential for glucose responsiveness of the insulin promoter.
We next examined glucose-induced changes in the expression levels of transcriptional activators that bind to GG2, C1/RIPE3b, and E1 elements, namely, Pdx1, MafA, and Beta2/NeuroD1, respectively, in MIN6 cells (Fig. 1C) and primary isolated islets (Fig. 1D). MafA protein expression was lower in MIN6 cells and islets incubated in low-glucose medium. By contrast, the expression levels of Pdx1 and Beta2/NeuroD1 proteins were unchanged. These observations suggest that the C1/RIPE3b-binding factor MafA mediates glucose-induced alterations in insulin gene expression.

Glucose regulates MafA and insulin expression through AMPK
Previous studies have demonstrated that AMPK activation reduces insulin gene expression and insulin promoter activity in ␤-cells (35). To activate AMPK, we used two mechanistically distinct activators of AMPK, AICAR and metformin. AICAR is converted to AICA-riboside monophosphate in the cell, and AICA-riboside monophosphate directly binds to and activates AMPK by mimicking AMP. By contrast, metformin indirectly activates AMPK in part through inhibiting complex I in the mitochondrial respiratory chain and subsequent ATP depletion. Both compounds thus induce activating Thr-172 phosphorylation of the catalytic AMPK␣1/␣2 subunit. Similar to the previous results, quantitative RT-PCR analysis showed that mRNA expression of insulin (ins-I and ins-II) decreased after treatment of MIN6 cells with metformin or AICAR ( Fig. 2A). Among selected ␤-cellenriched and GSIS-related genes, glut2 and sur1 were also down-regulated by metformin or AICAR treatment ( Fig. 2A).
A luciferase assay using the insulin promoter reporter construct also showed that treatment of MIN6 cells with metformin or AICAR reduced insulin promoter activity (Fig. 2B), recapitulating the previous results (35). To determine the cisregulatory elements responsible for the regulation of the insulin promoter by AMPK, we evaluated the expression from luciferase reporter plasmids with mutations in the GG2, C1/RIPE3b, or E1 elements (Fig. 2B). Luciferase activities generated by the mutGG2 or mutE1 promoter constructs were lower after metformin or AICAR treatment. By contrast, mutation of the C1/RIPE3b (mutC1) resulted in loss of responsiveness to metformin or AICAR, indicating that negative regulation of the insulin promoter by AMPK requires the C1/RIPE3b element. Thus, the C1/RIPE3b element is essential for regulation of the insulin promoter by both glucose and AMPK. We next examined the expression levels of Pdx1, MafA, and Beta2/NeuroD1 in metformin-or AICAR-treated MIN6 cells (Fig. 2C) and primary isolated islets (Fig. 2D) and found that MafA, but not Pdx1 or Beta2/NeuroD1, was decreased upon AMPK activation using these compounds.
To confirm the specific role of AMPK in mediating the effect of metformin on MafA, we blocked AMPK activity using dorsomorphin (compound C), an ATP-competitive inhibitor of AMPK kinase activity. Simultaneous treatment of MIN6 cells with dorsomorphin inhibited the decrease of MafA by metformin (Fig. 2E). Although dorsomorphin is also known to inhibit the type I bone morphogenetic protein receptor, this result suggests that metformin decreases MafA through AMPK activation.
To directly test the relationship of AMPK activity and MafA abundance, MIN6 cells were transfected with expression plasmids for full-length (FL) or constitutively active (CA) forms of AMPK␣1 and cultured in high-glucose medium. We then analyzed whole-cell lysates by immunoblotting and found that both FL and CA forms of AMPK␣1 increased ACC␣ phosphorylation and decreased MafA protein levels (Fig. 2F).
To further establish the role of AMPK in the regulation of MafA by glucose, MIN6 cells ( Fig. 3A) or primary isolated mouse islets (Fig. 3B) cultured in high-or low-glucose medium were treated with dorsomorphin. Under high-glucose conditions, where AMPK activity is low, dorsomorphin treatment further decreased AMPK Thr-172 phosphorylation and AMPK activity (phospho-ACC␣), and increased MafA protein levels. Under low-glucose conditions, where AMPK activity is high, dorsomorphin reduced AMPK activity and increased MafA protein levels. Therefore, AMPK inhibition counteracted the low glucose-induced decrease of MafA.
We next examined the insulin promoter activity in MIN6 cells cultured in high-or low-glucose medium with or without dorsomorphin by luciferase assay (Fig. 3C) and found that dorsomorphin treatment rescued the promoter activity under lowglucose conditions. Quantitative RT-PCR analysis also showed that mRNA expression of insulin (ins-I and ins-II), glut2, and sur1 was restored by dorsomorphin treatment under low-glucose conditions (Fig. 3D).
Furthermore, when a kinase-dead dominant-negative (DN) form of AMPK␣1 was expressed in MIN6 cells, MafA protein levels increased even under low-glucose conditions (Fig. 3E). However, the expression levels of Pdx1 and Beta2/NeuroD1 were not affected. These observations indicate that a low-glucose environment decreases MafA protein levels and expression of glucose-responsive genes, including

Glucose regulation of MafA via AMPK
ins-I, ins-II, glut2, and sur1, at least in part, by activating AMPK activity.

AMPK activation destabilizes MafA protein
Previous studies have shown that low glucose not only downregulates mafA mRNA levels but also destabilizes MafA protein. Quantitative RT-PCR analysis showed that the mafA mRNA was decreased under low-glucose conditions (Fig. 4A), as reported previously (9). However, mafA mRNA was not affected by AMPK activation by metformin or AICAR (Fig. 4B).
To further elucidate the mechanism of regulation of MafA protein abundance by AMPK, the half-life of MafA was determined in MIN6 cells treated with or without metformin (Fig.  4C). The half-life of MafA protein was about 4 h in untreated MIN6 cells cultured in high-glucose medium, but it was less than 30 min in metformin-treated cells, indicating that AMPK activation accelerates MafA protein degradation. These results indicate that glucose regulates both mafA mRNA and MafA protein levels but that AMPK only influences MafA protein levels.
We also examined the effect of AMPK activation or inhibition on exogenous HA-tagged MafA protein expressed under the control of the EF1␣ promoter. Similar to endogenous MafA, metformin and AICAR both decreased exogenous HA-MafA protein (Fig. 5A). Dorsomorphin treatment increased HA-MafA levels even under low-glucose conditions (Fig. 5B). Coexpression of the FL or CA forms of AMPK␣1 also reduced HA-MafA (Fig. 5C), and AMPK␣1-DN increased HA-MafA under low-glucose conditions (Fig. 5D). These data also support a role for AMPK in the regulation of MafA protein stability.  t test). B, luciferase assay. MIN6 cells were transfected with the insulin promoter reporter constructs used in Fig. 1, A and B (0.5 g) and pEF-Rluc (0.5 g). The cells were then treated with metformin (500 M) or AICAR (200 M) under high-glucose conditions for 16 h. Luciferase activities are presented relative to that of the pGL4/h-ins-p (WT) vector under high-glucose conditions. Data are mean Ϯ S.E. of five independent experiments. *, p Ͻ 0.05; n.s., not significant (Student's t test). C and D, whole-cell extracts obtained from MIN6 cells (C) or primary mouse islets (D) grown in high-glucose medium containing metformin (500 M) or AICAR (200 M) for 16 h were analyzed by immunoblotting using the indicated antibodies. E, MIN6 cells grown in high-glucose medium were treated with metformin (500 M) with or without the AMPK inhibitor dorsomorphin (Dorso, 10 M) for 16 h. Cell extracts were then subjected to immunoblot analysis. F, expression plasmids for the FL or CA forms of AMPK␣1 were transfected into MIN6 cells grown under high-glucose (Hi, 20 mM) conditions, and the cell extracts were analyzed by immunoblotting. The data in C-F are representative of two independent biological experiments. The intensity of the bands was quantified using ImageJ software, and the relative amounts (averages of two independent experiments) are indicated below the bands.

Glucose regulation of MafA via AMPK AMPK facilitates MafA degradation through a proteasomeindependent mechanism
Previous studies have demonstrated that multiple phosphorylation events occur at Ser-49, Thr-53, Thr-57, Ser-61, and Ser-65 and play major roles in the proteasome-dependent degradation of MafA (25). Furthermore, oxidative stress promotes proteasome-mediated human MafA degradation through phosphorylation at Thr-134 (corresponding to Thr-131 in mouse MafA) (26). To determine whether AMPK facilitates MafA degradation through a similar mechanism, MIN6 cells were treated with metformin or the oxidative stress-inducing reagent tertbutylhydroquinone (tBHQ), with or without the proteasome inhibitor MG132 (Fig. 6A). Endogenous MafA protein levels were higher after MG132 treatment, indicating that steadystate degradation of MafA in MIN6 cells is proteasome-depen-dent. Treatment of the cells with tBHQ reduced MafA protein, but this reduction was completely rescued by MG132 treatment, as reported previously (26). By contrast, metformin-mediated MafA reduction was only partially rescued by MG132, indicating that metformin decreases MafA protein through a pathway distinct from oxidative stress and independent of the proteasome.
To determine the role of MafA phosphorylation in its AMPK-mediated degradation, the phosphoacceptor Ser and Thr residues of MafA were mutated to Ala (Fig. 6B). The 5A mutant had five Ala substitutions at Ser-49, Thr-53, Thr-57, Ser-61, and Ser-65, and the T131A mutant had a single Thr-131-to-Ala mutation. The ⌬N164 mutant lacks the amino-terminal domain. These HA-tagged MafA variants were expressed in MIN6 cells, and the cells were treated with or without met-  's t test). E, the DN form of AMPK␣1 was overexpressed in MIN6 cells grown under low-glucose conditions, and the cell extracts were analyzed by immunoblotting. The data in A, B, and E are representative of two independent biological experiments. The intensity of the bands was quantified using ImageJ software, and the relative amounts (averages of two independent experiments) are indicated below the bands.

Glucose regulation of MafA via AMPK
formin and/or MG132 (Fig. 6, C and D). HA-tagged WT MafA protein expression was lower after metformin treatment, and this was only partially rescued by MG132, as was the case for endogenous MafA. Abundance of the T131A mutant was also reduced by metformin and was partially rescued by MG132 treatment.
The 5A mutant was more abundant than WT MafA, and this was not affected by MG132 treatment, confirming the role of these five phosphorylation sites in the proteasome-dependent steady-state degradation of MafA. Importantly, protein expression of this mutant was lower after metformin treatment, and this decrease was not rescued by MG132. Similarly to the 5A mutant, the ⌬N164 mutant was also more abundant and was not increased by MG132 treatment, indicating that this mutant escaped from proteasome-mediated degradation. The expression levels of this mutant were reduced by metformin, and this decrease was not rescued by MG132. Therefore, the aminoterminal region and the phosphorylation events at Ser-49, Thr-53, Thr-57, Ser-61, Ser-65, and Thr-131 are not involved in the regulation of MafA by AMPK. These results indicate that AMPK facilitates MafA degradation through a previously unrecognized mechanism that does not depend on phosphorylation or the proteasome. In ␤-cells, MafA protein is thus degraded through proteasome-dependent and AMPK-facilitated pathways in parallel.
To further examine whether glucose regulates MafA protein levels through AMPK, HA-tagged MafA and the mutants (5A, T131A, and ⌬N164) were expressed in MIN6 cells grown in high-or low-glucose medium (Fig. 6E). Similar to metformin treatment, the expression levels of these mutants were reduced under low-glucose conditions. These results support that low glucose decreases MafA protein levels through AMPK activation.

AMPK decreases insulin expression through MafA down-regulation
We next examined whether ectopic expression of MafA restores the expression levels of ␤-cell-specific mRNAs under AMPK-activated conditions. MIN6 cells were transfected with an empty expression vector or HA-tagged MafA expression plasmid. The cells were then treated with or without metformin under high-glucose condition, followed by analysis of protein expression by immunoblot (Fig. 7A) and mRNA expression by quantitative RT-PCR (Fig. 7B). Immunoblot analysis using anti-HA antibody (detecting exogenous HA-MafA protein) and anti-Maf (detecting both exogenous HA-MafA and endo genous MafA proteins) revealed that transfection of the HA-MafA expression plasmid achieved an excess amount of HA-MafA protein relative to endogenous MafA levels (Fig. 7A). Without metformin, expression of HA-MafA increased the glucose-responsive genes ins-I, ins-II, glut2, and sur1 (Fig. 7B). The expression levels of these mRNAs were reduced by metformin treatment but restored by HA-MafA expression. These results suggest that AMPK activation by a pharmacological activator decreases expression of insulin and other glucose-responsive mRNAs at least in part through downregulating MafA protein.

Discussion
In this study, we have shown that glucose regulates insulin gene expression and MafA protein abundance by modulating AMPK activity in ␤-cells. As shown previously, MafA protein levels increase (25) and AMPK activity decreases (34,35) under high-glucose conditions in isolated mouse primary islets and MIN6 insulinoma-derived cells. Under high-glucose conditions, pharmacological activation of AMPK by metformin or AICAR treatment or overexpression of a CA form of AMPK reduced MafA protein and insulin gene expression. By contrast, pharmacological inhibition of AMPK by dorsomorphin treatment or ectopic expression of a DN form of AMPK increased MafA protein and insulin gene expression under low-glucose conditions.
We have shown here that the MafA-binding element C1/RIPE3b in the insulin promoter region is critical for the effects of both glucose and AMPK, suggesting that they regulate insulin gene expression mainly through MafA. Previous studies have shown that prolonged AMPK activation attenuates GSIS in ␤-cells (36,37). Conversely, MafA is reportedly required for GSIS because it regulates ␤-cell-specific genes such as glut2 and sur1 in addition to insulin (ins-I and ins-II) (16 -18). We further demonstrated that introduction of exogenous MafA rescued the expression levels of ins-I, ins-II, glut2, and sur1 mRNAs in MIN6 cells treated with metformin. These findings together suggest that AMPK activation, induced by metformin or low-glucose conditions, decreases insulin expression and GSIS by down-regulating MafA.
MafA, Pdx1, and Beta2/NeuroD1 have been identified as glucose-responsive transcriptional activators of insulin gene expression. We have shown here that the expression levels of MafA, but not Pdx1 and Beta2/NeuroD1, were lower under Cell extracts were then subjected to immunoblot analysis. C, the HA-MafA expression plasmid was co-transfected into MIN6 cells with an expression plasmid for FL or CA AMPK␣1. Cell extracts were analyzed by immunoblotting. D, MIN6 cells were transfected with expression plasmids for HA-MafA and a DN form of AMPK␣1, and the cells were incubated in medium containing the indicated concentrations of glucose for 16 h. Whole-cell extracts were subjected to immunoblot analysis. The data in A-D are representative of two independent biological experiments. The intensity of the bands was quantified using ImageJ software, and the relative amounts (averages of two independent experiments) are indicated below the bands.

Glucose regulation of MafA via AMPK
low-glucose conditions or after AMPK activation. However, it has been shown that glucose regulates co-factor recruitment and the transcriptional activity of Pdx1 and the subcellular localization of Beta2/NeuroD1 (38 -40). Therefore, such activities or behaviors of Pdx1 and Beta2/NeuroD1 might be regulated by AMPK. Our mutational analysis of the insulin promoter does not exclude a possible involvement of Pdx1 and/or Beta2/NeuroD1 in glucose-or AMPK-mediated regulation of the insulin promoter activity because we did not mutate all of the binding elements of Pdx1 and Beta2/NeuroD1 within the insulin promoter region to eliminate the influence of these transcriptional activators.
Our gene expression analyses revealed that low glucose reduced mafA mRNA, as reported previously (9), whereas AMPK activation did not. Our observation that AMPK activation decreases MafA protein but not mafA mRNA indicates that AMPK regulation is only a part of the glucose response in ␤-cells. Expression of mafA mRNA is regulated by many ␤-cellenriched transcription factors, including FoxA1, Nkx2.2, and Pdx1, through directly binding to its enhancer regions (41). Thus, glucose might regulate mafA expression by modifying the activities of these transcription factors independent of AMPK.
We have shown here that AMPK activation accelerates MafA protein degradation. Previous studies have demonstrated that MafA degradation depends on its phosphorylation status and the proteasome (25,26). Consistent with this, oxidative stressinduced MafA degradation was attenuated by treatment of cells with the proteasome inhibitor MG132. However, MG132 treatment did not inhibit AMPK-induced MafA degradation. A previous study using mass spectrometry has identified 18 phosphorylation sites of MafA in ␤-cells (42). Among these phosphorylation sites, we mutated functionally relevant sites and found that phosphorylation events at Ser-49, Thr-53, Thr- shown in B were expressed in MIN6 cells. The electrophoretic mobility of the 5A mutant is higher than WT MafA because of loss of phosphorylation events at five Ser/Thr residues. Cells were then treated with or without metformin (500 M) and MG132 (5 M) for 16 h and were analyzed by immunoblotting. Short exposure of the boxed area of the anti-HA blot (D) was also indicated. E, WT or mutant forms of HA-tagged MafA (5A, T131A, and ⌬N164) were expressed in MIN6 cells. Cells were then grown in high-glucose (Hi, 20 mM) or low-glucose (Lo, 2 mM) medium for 16 h and analyzed by immunoblot analysis. The data in A and C-E are representative of two independent biological experiments. The intensity of the bands was quantified using ImageJ software, and the relative amounts (averages of two independent experiments) are indicated below the bands.

Glucose regulation of MafA via AMPK
57, Ser-61, Ser-65, and Thr-131 of MafA are not required for its AMPK-induced degradation. Substitution of the other phosphoacceptor residue in MafA, Ser-342, with Ala also did not abrogate the AMPK-induced degradation. 3 Therefore, MafA protein is destabilized by AMPK through a previously unrecognized mechanism. As far as we have examined, AMPK does not bind MafA directly in a co-immunoprecipitation assay. 3 Furthermore, none of the previously identified phosphorylation sites of MafA is canonical AMPK target site. We thus favor a model whereby AMPK destabilizes MafA via an indirect interaction. The mechanism whereby AMPK promotes MafA degradation remains to be elucidated.
The AMPK activator metformin has long been proposed as an anti-type 2 diabetes drug because it increases glucose utilization in peripheral insulin-sensitive tissues (43,44). However, it has also been shown that AMPK activation reduces GSIS from ␤-cells (34,35). The data presented here suggest how AMPK activation leads to the inhibition of GSIS in ␤-cells. AMPK activation decreases MafA abundance and thereby reduces the expression of insulin and genes involved in GSIS, such as glut2 and sur1. Elucidation of the precise molecular mechanism of MafA regulation by AMPK should be particularly useful as part of the process of developing better strategies for the treatment of type 2 diabetes.

Cells and transfection
The mouse insulinoma-derived cell line MIN6 was a generous gift from Dr. Jun-ichi Miyazaki (Osaka University). The cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 25 mM glucose supplemented with 15% fetal bovine serum (FBS) in a humidified atmosphere containing 5% CO 2 at 37°C. For transfection, 1 ϫ 10 6 cells were mixed with transfection mixture (total 3.6 g of plasmid, 7.2 l of P3000 reagent, and 4 l of Lipofectamine 3000 (Invitrogen)) and seeded into 24-well plates. When glucose conditions were 3 R. Iwaoka and K. Kataoka, unpublished observation.

Glucose regulation of MafA via AMPK
altered, the medium was changed to DMEM containing high (20 mM) or low (2 mM) glucose supplemented with 0.5% FBS.

Isolation and culture of primary islets
Mouse primary islets of Langerhans were isolated by a standard collagenase digestion method (47) and incubated in RPMI 1640 medium containing 10% FBS. When glucose conditions were altered, the medium was changed to DMEM containing high (20 mM) or low (2 mM) glucose supplemented with 10% FBS. All experiments were performed according to the guidelines for the care and use of laboratory animals of the Yokohama City University. The experiments were approved by the Animal Research Committee of Yokohama City University (T-A-16-1004).

Luciferase assay
MIN6 cells were transfected with 3.6 g plasmid DNA (0.5 g of luciferase reporter plasmid, 0.5 g of pEF-Rluc, and 2.6 g of expression plasmid) using 4 l of Lipofectamine 3000 and 7.2 l of P3000 reagents (Invitrogen). Cells were harvested 48 h after transfection. Firefly and Renilla luciferase activities were measured using a dual luciferase assay system (Promega), and firefly luciferase activities were normalized on the basis of Renilla luciferase activities. Data represent the mean Ϯ S.E. of five independent experiments. Statistical significance was calculated using Student's t test.
Author contributions-R. I. performed most of the experiments and analyzed the data. K. K. designed the research and wrote the manuscript with R. I. Both authors reviewed the results and approved the final version of the manuscript.