Hepatocyte Nuclear Factor-4α Involved in Type 1 Maturity-Onset Diabetes of the Young Is a Novel Target of AMP-Activated Protein Kinase

The -183PK/Luci, (L3)₃-54PK/Luci, and (H4)₄-105TK/Luci constructs have been described elsewhere (26). pHNF4/Luci was obtained by polymerase chain reaction (PCR) amplification of murine HNF-4α promoter (27,28) from mouse-tail genomic DNA with the following primers: 5′-TAC GGT TCA GCT GGA GCA ACC AAG-3′ and 5′-ATG CCG GCA AGG GTT TTA GAG AGT-3′. It was subcloned into the pT-Adv (Clontech) by TA-cloning and then into the pGL3-Basic vector (Promega) at KpnI-XhoI sites. All plasmids were constructed by standard DNA cloning procedures and verified by nucleotide sequencing.

Hepatocytes from Sprague-Dawley rats (160–180 g) and from B6/CBA mice (3–5 months old) were isolated by the collagenase perfusion method (29), followed by low-speed iso-density Percoll centrifugation (30). Animals were fed ad libitum with standard laboratory diet. Hepatocytes were cultured as reported previously (26), using 199 medium (Gibco-BRL) containing 10% (vol/vol) fetal calf serum (Gibco-BRL), 5 mmol/l glucose, 20 nmol/l insulin, 1 μmol/l triiodothyronine, and 1 μmol/l dexamethasone, supplemented with penicillin and streptomycin.

Rat hepatocytes were incubated with AICAR (Sigma), as indicated in the figure legends. A total of ∼30 million cells were used for each assay. Nuclear extracts were prepared according to Hasegawa et al. (31). Electrophoretic mobility shift assays (EMSAs) were performed with 1–5 μg nuclear proteins incubated 20 min on ice with radiolabeled probe, 1 μg poly(dI-dC) (Pharmacia), 10 mmol/l HEPES pH 7.9, 25 mmol/l KCl, 2.5 mmol/l dithiothreitol (DTT), 0.1 mmol/l EDTA, 0.25 mmol/l phenylmethylsulfonyl fluoride, and 5% (vol/vol) glycerol. Gels (5% polyacrylamide 19:1, Tris Borate EDTA 0.25×) were run at room temperature at 120–150 V, dried and autoradiographed.

Hepatocytes in primary culture were scrapped in 2× Laemmli’s buffer [0.125 mol/l Tris-HCl, pH 6.8,  4% SDS, 10% β-mercaptoethanol,  20% glycerol,  0.004% bromophenol blue] or in reporter lysis buffer (Promega). DNA was disrupted by passing the lysate several times through a 22-gauge needle. Samples were boiled 10–15 min at 100°C and centrifuged (12,000g, 5 min). Protein concentration was determined on supernatant by Bradford assay (32). A total of 25–50 μg total protein were resolved by 10% SDS-PAGE and transferred to Hybond-C nitrocellulose membrane (Amersham). The membrane was blocked by incubation in Tris-buffered saline with Tween (TBST) [10 mmol/l Tris-HCl pH 8.0,  150 mmol/l NaCl,  0.05% (vol/vol) Tween 20] supplemented with 5% skimmed milk powder for at least 1 h at room temperature. Polyclonal primary antibodies (HNF-4α [33], kindly provided by Moshe Yaniv, Institut Pasteur, Paris; upstream stimulatory factor (USF) 2a [14]; and annexin V from Santa Cruz Biotechnology) were then added in the same buffer and incubated for 2 h at room temperature. Membranes were then washed three times for 15 min in TBST and incubated in TBST–5% skimmed milk with horseradish-peroxidase–linked anti-rabbit antibody (DAKO) for 45 min at room temperature. Finally, they were washed as above in TBST and revealed by enhanced chemiluminescence (Amersham).

Total RNA was isolated from hepatocytes by lysis in guanidinium thiocyanate, followed by phenol extraction (34). For reverse transcription (RT)-PCR analysis, 50–100 μg total RNA was first treated with 10 U RNase-free DNase I (ROCHE) in 7 mmol/l Tris, 0.14 mmol/l EDTA, 10 mmol/l MgCl₂, 1 mmol/l DTT, 40 U RNAsin (Promega)] in a final volume of 50 μl for 1 h at 37°C. Then, 90 μl of STOP buffer [10 mmol/l Tris, 1 mmol/l EDTA, 0.1% SDS, 0.3 mol/l NaCl] was added (final volume 140 μl), and RNA was extracted with an equal volume of phenol:chloroform:isoamyl alcohol 25:24:1 (vol:vol) pH 8, ethanol-precipitated, and resuspended in a suitable volume of water. First-strand cDNA synthesis was performed as previously described (35). Oligonucleotide primer sequences used for PCR amplification were: l-PK: 5′-AGT CGG AGG TGG AAA TTG TGA AGG G-3′, 5′-CCA GTC CCT CTG GGC CAA TTT TCC G-3′; GLUT2: 5′-CTG GGG CTT TCT GGA CAG AAG AGC AG-3′, 5′-TGT CAG AAG ACA AGA TCA CCG GAA CC-3′; aldolase B: 5′-TGA TGG AGA CCA TGA CCT AGA GCA CT-3′, 5′-CTT GGG CTG CCT GAC AGT TGG CCA-3′; HNF-1α: 5′-TTC TAA GCT GAG CCA GCT GCA GAC G-3′, 5′-GCT GAG GTT CTC CGG CTC TTT CAG A-3′; apoB: 5′-CTT CAG GGA ACA AAG CAG-3′, 5′-TCA AGG GTG AGC TGA TTG-3′; apoCIII: CTC AAT AGC TGG AGT TGG-3′, 5′-CTG AAG AGG TAG AGG GAT-3′; albumin: 5′-CTC CAG CAA ACT GCA GAC TTG CTG CG-3′, 5′-TTT GGA ATC CAT ATT CTC CAA GCT TC-3′; glucokinase: 5′-CGA CTC TGG GGA CCG AAG GCA GAT C-3′, 5′-CTC GGG TGC AGC TTG TAC ACG GAG C-3′; HNF-4α: 5′-CTT CTT TCT TCA TGC CAG-3′, 5′-ACA CGT CCC CAT CTG AAG-3′. Radioactive PCR was performed with 250 ng cDNA as template in 50 μl final volume at different cycles to ensure PCR amplification in the linear range. Conditions were the same for each set of primers: 20–28 cycles of 30 s at 94°C, 30 s at 57°C, and 60 s at 72°C in PCR buffer 1× (Gibco-BRL), 3 mmol/l MgCl₂, 0.5 mmol/l dNTP, 0.5 μCi [α³²P]dATP, 100–200 ng each primer, and 1 U Taq DNA polymerase (Gibco-BRL). Samples were loaded on 5% polyacrylamide gel and TBE 1×, then they were run, dried, and autoradiographed.

For the next assay, 4 μg of luciferase reporter constructs and 500 ng CMV-β-gal as internal transfection control were transfected with cationic liposomes in mouse hepatocytes using N-(2,3-dioleoyloxy) propyl)-N, N, N-trimethylammonium methylsulfate (DOTAP; Roche) according to the manufacturer’s instructions. At 24 h after transfection, the medium containing the liposome-DNA complex was replaced with fresh medium with or without 200 μmol/l AICAR for another 12–16 h. Cells were harvested 36 to 40 h after transfection in 400 μl of reporter lysis buffer (Promega) according to manufacturer’s instructions. Luciferase activity was performed on 20 μl of cellular extract with a luciferase assay system (Promega). For β-galactosidase activity, 50 μl of cellular extract was diluted in 1 ml of assay buffer (PBS, 1 mmol/l MgCl₂, 50 mmol/l β-mercaptoethanol). Then, 200 μl o-nitrophenyl β-d-Galactopyranoside (onpg; Sigma) in 4 mg/ml PBS was added, and the extract was incubated at 37°C for 1 h. The reaction was stopped with 500 μl of Na₂CO₃ 1 mol/l. Optical density was read at 420 nm. Because we noted a consistent twofold diminution in luciferase/β-Gal units from pGL3–control vector (Promega) activity after the addition of AICAR, all promoter activities reported here were normalized with respect to those of pGL3–control vector.

AMPK assays were performed essentially as described previously (19) on whole cell extracts using SAMS (the synthetic peptide HMRSAMSGLHLVKRR) as substrate.

Data are given as the means ± SE of at least three independent experiments. Comparisons between means were performed with Student’s t test for paired data using Microsoft Excel software.

We have previously shown that AMPK activation by 200 μmol/l AICAR in hepatocytes in primary culture inhibited glucose-dependent gene transcription (18). This result was confirmed by others using either AICAR (20) or a constitutively active form of the kinase (21), thereby indicating that the use of AICAR is relatively safe in terms of specificity for AMPK activation. Figure 1A shows that AMPK is efficiently activated in primary hepatocytes in response to 200 μmol/l AICAR.

To test whether AMPK plays a role in transcriptional repression by modulating HNF-4α function, we first assessed HNF-4α binding activity after AMPK activation in a primary culture of hepatocytes. For that purpose, we performed EMSAs using nuclear extracts prepared at different incubation times with 200 μmol/l AICAR. As shown in Fig. 1B (upper panel), we observed a diminution of the HNF-4α–retarded band after 6 h and a complete loss of HNF-4α DNA-binding activity after 12 h of AICAR treatment. Similar results were obtained by using either the HNF-4α binding site from the rat l-PK promoter (Fig. 1B) or the human α1-antitrypsin gene promoter (not shown) as DNA probes. As a control, we monitored DNA-binding activity for the ubiquitously expressed transcription factor nuclear factor Y (NF-Y) and showed that it was not significantly affected by AICAR treatment over time (Fig. 1B). Competition studies using excess of unlabeled oligonucleotides corresponding to consensus HNF-4α binding site and supershift assays using specific HNF-4α antibodies were performed to confirm the identity of the HNF-4α–retarded band (not shown).

The observed loss of the HNF-4α–retarded band could be caused by changes in protein levels. To test this possibility, we performed Western blot analysis to monitor the level of HNF-4α protein in response to AICAR treatment. As shown in Fig. 1B (lower panel), we observed that the HNF-4α signal decreases after 6 h and practically disappears after 12 h of incubation with AICAR. In contrast, the level of the control protein annexin V was not affected by AICAR treatment over time. Thus, activation of AMPK in hepatocytes results in a complete loss of HNF-4α DNA-binding activity that was strictly parallel to a pronounced time-dependent disappearance in HNF-4α protein levels.

Considering the diminution of HNF-4α protein after AMPK activation, it was tempting to speculate that the AMPK effect could be accounted for by suppression of HNF4α gene transcription. To examine this issue, we evaluated HNF-4α mRNA levels using a semiquantitative RT-PCR assay. As shown in Fig. 2A, HNF-4α mRNA was constitutively expressed in the absence and presence of 200 μmol/l AICAR, suggesting that AMPK activation did not affect transcription of the HNF4α gene. As a control, we monitored the mRNA levels for the l-PK and albumin genes, which have previously been found to be sensitive and insensitive, respectively, to AMPK transcriptional repression (18,20). As expected, the expression of the l-PK gene was highly decreased in the presence of AICAR, whereas the expression of the albumin gene was not significantly affected by AICAR treatment (Fig. 2A).

In addition, to confirm that the AMPK effect was not mediated by direct transcriptional repression of the HNF-4α gene promoter, we transiently transfected hepatocytes with a luciferase construct driven by the murine HNF-4α promoter (27,28). Overnight incubation of these hepatocytes with 200 μmol/l AICAR caused no changes in the luciferase reporter enzyme activity when compared with untreated hepatocytes (Fig. 2B). Therefore, the sensitivity of HNF-4α protein abundance to AMPK activation was not related to a direct transcriptional effect on the HNF-4α promoter.

To determine whether HNF-4α might be a bona fide transcriptional target of AMPK, we next monitored the response to AICAR treatment of several transiently transfected HNF-4α–dependent constructs in primary culture of hepatocytes. For this purpose, we used reporter plasmids containing multimerized HNF-4α binding site from either the rat l-PK promoter [(L3)₃-54PK/Luci] or the human α1-antitrypsin promoter [(H4)₄-105TK/Luci]. In addition, we tested as a control transcriptional activity of the rat l-PK promoter regulatory sequences (183PK/Luci) previously identified to be HNF-4α–dependent and AICAR-responsive (18,36). As shown in Fig. 3, in all cases, inhibition was always >70% after overnight incubation with 200 μmol/l AICAR. Thus, our data indicate that AMPK can modulate HNF-4α–dependent gene transcription.

To determine whether the observed effects on artificial promoter constructs were reflective of transcriptional regulation of HNF-4α target genes in the context of native chromatin, we performed semiquantitative RT-PCR analysis of endogenous gene transcription in hepatocytes incubated with increasing amounts of AICAR, ranging from 0 to 200 μmol/l. As shown in Fig. 4A, a frank diminution of known HNF-4α–regulated genes (7,11,12), including l-PK, aldolase B, GLUT2, HNF-1α, apoCIII, and apoB, was detected at an AICAR concentration of 150 mmol/l. Interestingly, the mRNA level of apoAI was not modified by AICAR treatment, nor was it diminished in HNF-4α liver-specific knockout animals (12). For controls, we monitored albumin and glucokinase mRNA levels that were not significantly affected by AICAR treatment (Fig. 4A). In addition, we checked the abundance of the HNF-4α and USF2a transcription factor proteins in hepatocytes incubated overnight with increasing AICAR concentrations. The amount of HNF-4α protein declined for AICAR concentrations >150 mmol/l, whereas the USF2a protein amount was not significantly modified under the same conditions (Fig. 4B). Hence, transcriptional decrease of the HNF-4α–dependent gene observed after AMPK activation parallels the diminution of HNF-4α protein levels, as revealed by Western blot analysis.

It is now well established that AMPK, in addition to its role in fuel partitioning (14), is also involved in the transcriptional control of glucose-regulated genes (18,19,20,21,22). The present study further elucidates the molecular targets involved in transcriptional repression by AMPK. Here, we demonstrated that the MODY1/HNF-4α transcription factor is a novel target of AMPK. The extent of decrease in transcription rate induced by AICAR treatment essentially correlates with the decrease in HNF-4α protein levels. This effect is most likely not related to a transcriptional repression of the HNF4α gene promoter because no significant modification of HNF-4α mRNA abundance can be detected by semiquantitative RT-PCR analysis (Fig. 2A) and the murine HNF-4α promoter appears to be insensitive to AICAR treatment in transfection studies (Fig. 2B).

Cellular processes that limit the transcriptional activity, subcellular location, and abundance of transcription factors play an important role in regulating gene expression. Proteolysis is a prominent mechanism regulating transcription factor, which makes transcriptional activation and activator degradation closely coupled events (37,38). Several proteolytic processes have been implicated in transcription factor destruction, including ubiquitin-mediated proteolysis (39) and cleavage by calpains (40) and lysosomal proteases (41). In the case of transcription factor turnover by the ubiquitin-proteasome pathway, phosphorylation has often been associated as a positive signal for the ubiquitination and targeting of proteins to the 26S proteasome (42). Therefore, we verified that HNF-4α is indeed a target of AMPK phosphorylation in cell-free extracts (not shown) and anticipated that this pathway would be involved in HNF-4α degradation. To test this hypothesis, we treated hepatocytes concomitantly with AICAR and the proteasome inhibitor MG132. We observed that the diminution of HNF-4α protein was not prevented under these conditions (not shown) suggesting that HNF-4α disappearance was independent of the proteasome pathway. Alternatively, it is possible that phosphorylation of HNF-4α by AMPK may introduce a conformational change that makes the protein more susceptible to degradation resulting from proteolytic cleavage by cellular proteases. Efforts to identify the exact mechanisms implicated in this process are underway.

The rapid and controlled adjustment in the amount of transcription factors permits the attenuation of cellular responses to extracellular cues by rapidly modifying key signaling proteins. This provides a fail-safe mechanism against unbridled changes in the level of transcription and allows the transcriptional apparatus to be quickly reprogrammed in the case of unexpected change in the cellular state. AMPK is activated in response to ATP depletion and leads to the switching-off of ATP-consuming pathways and the switching-on of ATP-regenerating pathways. It plays a major role in lipid metabolism and in inhibiting fatty acid, cholesterol, and triglyceride synthesis; it also plays a major role in promoting fatty acid oxidation. This switch from anabolism to catabolism restores the energy status of the cell (25). By decreasing HNF-4α protein levels (as in this study), AMPK decreases the apo synthesis and stops lipid export by the liver, thus preserving this carbon source for ATP repletion at the cellular level. Recently, HNF-4α mutations have been associated with decreased levels of circulating apoAII, apoB, apoCIII, and triglycerides in MODY1 patients and in mice carrying a HNF-4α–null mutation in the liver (10,11).

Genetics studies have shown that heterozygous mutations in HNF-4α result in MODY1, an autosomal-dominant form of type 2 diabetes. MODY1 mutants included nonsense and frameshift mutations that give rise to truncated HNF-4α proteins and missense mutations, which affect different domains of the HNF-4α protein. These mutations impair protein activity by decreasing its binding to target DNA sequences or by affecting protein dimerization and/or the transactivation potential. Surprisingly, none of the MODY1 mutants described so far have exhibited any signs of a dominant-negative effect in functional studies (43). It is speculated that dominant-negative HNF-4α mutants are not compatible with human development, as is the case with HNF4α inactivation in mice (6). Hence, haploinsufficiency or reduced gene dosage may explain the mechanism leading to the MODY1 phenotype. Expression of HNF-4α at a relatively low amount in endocrine pancreatic cells (5) appears to be a limiting factor for efficient transcription of HNF-4α target genes. A 50% reduction in the amount of functional HNF-4α protein would be deleterious in the context of pancreatic β-cells, whereas in hepatocytes it would have less dramatic effects on gene transcription.

The MODY1 phenotype is characterized by a defect in nutrient-induced insulin secretion by the pancreatic β-cells (3) and by a diminution in serum triglycerides, and it accounts for a diminution in apoAII, apoB, and apoCIII synthesis by the liver (10,11). Target genes of HNF-4α in pancreatic β-cells include genes involved in glucose metabolism, e.g., GLUT2, l-PK, and aldolase B (44). Overexpression of wild-type HNF-4α in INS1 β-cells is associated with an increase in insulin secretion in response to glucose, whereas the overexpression of a dominant-negative form of HNF-4α causes a decrease in glucose-induced insulin secretion (44). Endogenous HNF-4α protein levels in islet β-cells must therefore be tightly regulated to maintain normal insulin secretion and euglycemia. This observation suggests that reduced levels of HNF-4α in the pancreas could contribute to the development of β-cell dysfunction. It is noteworthy that AMPK activity is regulated by glucose in pancreatic β-cells and has been implicated in the control of glucose-stimulated insulin secretion (19,24). Whether HNF-4α is also a target of AMPK in islet β-cells is of major interest because any inappropriate activation of the AMPK signaling pathway would downregulate HNF-4α protein levels. Consequently, this would have deleterious effects on glucose-induced insulin secretion and could contribute to the development of type 2 diabetes.

In conclusion, our results add the transcription factor MODY1/HNF-4α to the list of AMPK targets. In addition to its role in the regulation of hepatic gene expression, HNF-4α also regulates the expression of genes that are involved in glucose transport and glucose metabolism in the pancreatic β-cells. This suggests that the modulation of HNF-4α protein levels by AMPK could have important metabolic effects. The potential importance of the loss of HNF-4α protein after AMPK activation is reinforced by the fact that only a minor perturbation of HNF-4α activity results in MODY1 (43). Therefore, we propose that modulation of HNF-4α function by AMPK activity could be of importance in the physiopathology of type 2 diabetes. This issue is an interesting area for future investigation.

This work was supported by INSERM, the Ministère de l’Enseignement Supérieur et de la Recherche, and Aventis. I.L. is a Clinician-Scientist Fellow of the Canadian Institute of Health Research (CIHR).

We thank Michel Raymondjean and Mireille Vasseur-Cognet for helpful discussions.