Identification of HNF-4α as a key transcription factor to promote ChREBP expression in response to glucose

Transcription factor carbohydrate responsive element binding protein (ChREBP) promotes glycolysis and lipogenesis in metabolic tissues and cancer cells. ChREBP-α and ChREBP-β, two isoforms of ChREBP transcribed from different promoters, are both transcriptionally induced by glucose. However, the mechanism by which glucose increases ChREBP mRNA levels remains unclear. Here we report that hepatocyte nuclear factor 4 alpha (HNF-4α) is a key transcription factor for glucose-induced ChREBP-α and ChREBP-β expression. Ectopic HNF-4α expression increased ChREBP transcription while knockdown of HNF-4α greatly reduced ChREBP mRNA levels in liver cancer cells and mouse primary hepatocytes. HNF-4α not only directly bound to an E-box-containing region in intron 12 of the ChREBP gene, but also promoted ChREBP-β transcription by directly binding to two DR1 sites and one E-box-containing site of the ChREBP-β promoter. Moreover, HNF-4α interacted with ChREBP-α and synergistically promoted ChREBP-β transcription. Functionally, HNF-4α suppression reduced glucose-dependent ChREBP induction. Increased nuclear abundance of HNF-4α and its binding to cis-elements of ChREBP gene in response to glucose contributed to glucose-responsive ChREBP transcription. Taken together, our results not only revealed the novel mechanism by which HNF-4α promoted ChREBP transcription in response to glucose, but also demonstrated that ChREBP-α and HNF-4α synergistically increased ChREBP-β transcription.

In order to further investigate the effect of suppressing HNF-4α expression on ChREBP transcription, we transfected HepG2 cells with two independent siRNAs for HNF-4α and successfully suppressed HNF-4α expression (Fig. 1f). In comparison to the non-targeting control siRNA, both siRNAs for HNF-4α decreased mRNA levels of ChREBP-α , ChREBP-β and total ChREBP in HepG2 cells (Fig. 1f). We also transfected mouse primary hepatocytes with the control shRNA and shRNA for HNF-4α . When compared to the control, HNF-4α knockdown reduced ChREBP mRNA and protein levels in mouse primary hepatocytes (Fig. 1g,h).

HNF-4α and ChREBP expression in mouse primary hepatocytes is up-regulated by glucose.
We treated mouse primary hepatocytes and human liver cancer HepG2 cells with 0 mM, 5.6 mM and 25 mM glucose for 18 hours, respectively. Real time PCR analysis showed that 25 mM glucose was more potent than 5.6 mM glucose to increase mRNA and protein levels of ChREBP-α , ChREBP-β and total ChREBP in comparison to the 0 mM glucose treatment in mouse primary hepatocytes and HepG2 cells (Fig. 2a-c, Supplementary Fig. 2a,b). Our results agree with reported findings that mRNA levels of ChREBP are regulated by glucose or glucose derived metabolites 2,3 .
To investigate the role of HNF-4α in regulating ChREBP expression in response to glucose in vivo, wild type C57Bl/6 mice were fasted or given access to food for 18 hours before mRNA levels of ChREBP in the liver were assayed. Transcription of ChREBP-α , ChREBP-β and total ChREBP was up-regulated in the fed state compared with the fasted control (Fig. 2d). Transcription of target genes activated by ChREBP such as FAS, L-PK and SCD1 were increased whereas those target genes repressed by ChREBP including G6Pase and PEPCK were (c). Real-time PCR analysis for mRNA levels of ChREBP-β at 48 hours after expression plasmids containing control (GFP) or other cDNAs are transfected in 293T cells. *indicates p < 0.05 when compared with the GFPtransfected sample. (d,e). Real-time PCR analysis for mRNA levels of ChREBP-α , ChREBP-β and total ChREBP (d) and western blot analysis for endogenous ChREBP expression (e) at 48 hours after HA-GFP (vector) or HA-HNF-4α expression plasmids are transfected in HepG2 cells. *indicates p < 0.05 when compared with the vector-transfected sample. Tubulin serves as the loading control. (f).Real-time PCR analysis for mRNA levels of HNF-4α , ChREBP-α , ChREBP-β and total ChREBP at 72 hours after control (NC) or two HNF-4α siRNAs (HNF-4α -1 and HNF-4α -2) are transfected in HepG2 cells. *indicates p < 0.05 when compared with the corresponding NC-transfected sample. (g,h). Real-time PCR analysis for mRNA levels of HNF-4α , ChREBP-α , ChREBP-β and total ChREBP (g) and western blot analysis for endogenous ChREBP expression (h) at 48 hours after control (NC) or HNF-4α shRNAs are transfected in mouse primary hepatocytes. *indicates p < 0.05 when compared with the corresponding NC-transfected sample. Actin serves as the loading control.
down-regulated in the fed mouse liver, suggesting that the activity of ChREBP was induced by feeding (Fig. 2e). Interestingly, we observed up-regulation of both ChREBP and HNF-4α protein expression in response to feeding using western blot analysis and immunohistochemistry (Fig. 2f,g).
HNF-4α binds the E-box-containing cis-element located in intron 12 of the ChREBP-α gene. Next we explored the molecular mechanism by which HNF-4α promotes ChREBP-α and ChREBP-β transcription. The 4 kb promoter of ChREBP-α was predicted to contain potential binding sites for HNF-4α . Therefore, we transiently transfected 293T cells with the 4 kb ChREBP-α promoter in pGL3-Basic plasmid along with HNF-4α or control expression plasmids and analyzed luciferase activity 48 hours later. Since LXR was reported to promote ChREBP transcription 21 , it was included as a positive control. Although LXR increased the ChREBP-α promoter luciferase reporter activity, we were not able to detect any induction in luciferase reporter activity by HNF-4α overexpression in comparison to the negative control (Fig. 3a). This finding suggests that the 4 kb promoter of ChREBP-α might not contain potential binding sites for HNF-4α . We next searched the whole human ChREBP-α gene sequence for potential binding sites for HNF-4α , including classical DR1, H4-SBM sites and the non-perfect DR1 site which contains the E-box sequence. We found four E-box-containing non-perfect DR1 sites located in intron 2, intron 6, intron 7 and intron 12 of the ChREBP-α gene. We cloned the 164 bp, 174 bp, 378 bp and 140 bp sequences containing the non-perfect DR1 sites located in intron 2, intron 6, intron 7 and intron 12 of the ChREBP-α gene from human genomic DNA and subcloned them into the pGL3-Promoter vector containing an SV40 promoter upstream of the luciferase reporter gene, respectively. We named these four pGL3-Promoter plasmids containing the 164 bp, 174 bp, 378 bp and 140 bp intronic sequences as plasmid I, II, III and IV, respectively. Insertion of DNA fragments containing potential binding sites for HNF-4α into the pGL3-Promoter vector allowed for screening for putative enhancer elements regulated by HNF-4α . We transiently transfected 293T cells with plasmids I, II, III or IV along with HNF-4α or control expression plasmids and analyzed luciferase activity. Only the luciferase activity of plasmid IV, the pGL3-Promoter plasmid containing the 140 bp sequence in intron 12, was enhanced by HNF-4α overexpression compared with the control (Fig. 3b). We transfected 293T cells with the plasmid IV along with LXR or control expression plasmids and found that LXR was not able to induce luciferase activity, suggesting that the induction . Luciferase activity assay shows that HNF-4α enhances transcriptional activity of the pGL3-Promoter plasmid containing the 140 bp sequence in intron 12 of ChREBP-α (IV) compared with the control. I, II, III and IV are the pGL3-Promoter plasmids containing the 164 bp, 174 bp, 378 bp and 140 bp sequences located in intron 2, intron 6, intron 7 and intron 12 of ChREBP-α . 24 hours after I, II, III or IV and the HA-GFP or HA-HNF-4α expression plasmid are transfected in 293T cells, luciferase activity is analyzed. *indicates p < 0.05 when compared with the HA-GFP-transfected sample. (c). Luciferase activity assay shows that deletion of the E-box in IV reduced induction of the transcriptional activity of IV by HNF-4α compared with the control. 24 hours after the pGL3-Promoter plasmid, IV or IVΔE-box and the HA-GFP, HA-LXR or HA-HNF-4α expression plasmid are transfected in 293T cells, luciferase activity is analyzed. *indicates p < 0.05 when compared with the HA-GFP-transfected sample. Western blot analysis using the anti-HA antibody shows protein levels of ectopically expressed HA-HNF-4α and HA-LXR. (d). ChIP analysis for HepG2 cells using an anti-HNF-4α antibody, nonspecific IgG or anti-histone antibody shows that HNF-4α binds the 140 bp region in intron 12 of the ChREBP-α gene.
in luciferase activity of plasmid IV by HNF-4α should be specific (Fig. 3c). We also deleted the E-box (CACGTG) in the 140 bp sequence in intron 12 and generated the E-box free plasmid IV. We found that loss of the E-box greatly reduced induction of the luciferase activity by HNF-4α overexpression (Fig. 3c).
In order to determine whether endogenous HNF-4α directly bound the E-box-containing cis-element located in intron 12 of the ChREBP-α gene, we performed ChIP analysis for HepG2 cells using an anti-HNF-4α antibody. Quantitative PCR analysis for DNA fragments encompassing the E-box-containing cis-element showed that HNF-4α directly bound the E-box-containing region in intron 12 of the ChREBP-α gene (Fig. 3d).
HNF-4α binds DR1 sites located in the ChREBP-β promoter. To identify possible HNF-4α -binding sites in the promoter of ChREBP-β , we cloned the 2.9 kb ChREBP-β promoter from human genomic DNA and subcloned it into the luciferase reporter vector pGL4-Basic (Fig. 4a). We transiently transfected the 2.9 kb ChREBP-β promoter in pGL4-Basic plasmid into 293T cells along with HNF-4α or control expression plasmids and analyzed luciferase activity 48 hours later. HNF-4α increased the 2.9 kb ChREBP-β promoter luciferase reporter activity in 293T cells by about 12 folds in comparison to controls (Fig. 4b). We next investigated the effect of suppressing HNF-4α expression on activity of human 2.9 kb ChREBP-β promoter. We transfected 293T cells with the 2.9 kb ChREBP-β promoter in pGL4-Basic plasmid along with control siRNAs or siRNAs for HNF-4α and found that HNF-4α knockdown decreased luciferase activity in comparison to the control (Fig. 4c,d).
In order to determine whether endogenous HNF-4α directly bound DR1-B and DR1-C sites of the ChREBP-β promoter, we performed ChIP analysis for HepG2 cells using an anti-HNF-4α antibody. Quantitative PCR analysis for DNA fragments encompassing the nucleotides − 86~− 37 region showed that HNF-4α directly bound this region of ChREBP-β promoter (Fig. 4i).

HNF-4α contributes to glucose-induced ChREBP-α and ChREBP-β transcription. In order to
further investigate the effect of suppressing HNF-4α expression on glucose-induced ChREBP transcription, we transfected HepG2 cells with two independent siRNAs for HNF-4α which successfully suppressed HNF-4α expression and treated cells with 0 mM and 25 mM glucose for 18 hours. In comparison to the non-targeting control siRNA, both siRNAs for HNF-4α decreased glucose-induced transcription of ChREBP-α , ChREBP-β and total ChREBP in HepG2 cells (Fig. 7a). Using the anti-ChREBP antibody which detected both ChREBP-α and ChREBP-β protein, we found that HNF-4α knockdown reduced the protein level of ChREBP induced by 25 mM glucose treatment in HepG2 cells (Fig. 7b). These findings suggest that HNF-4α contributes to glucose-induced ChREBP-α and ChREBP-β transcription.
Next we investigated the mechanism by which HNF-4α mediated the glucose-responsive effect of ChREBP transcription by studying whether and how glucose regulated subcellular localization and DNA binding capacity of HNF-4α . HepG2 cells were treated with 0 mM, 2.5mM, 5.6mM and 25 mM glucose for 18 hours before nuclear and cytoplasmic fractionation. Endogenous HNF-4α in HepG2 cells displayed increased total protein expression and nuclear abundance in response to glucose in a dose dependent manner (Fig. 7c). In order to investigate whether glucose increased HNF-4α binding to the cis-elements in ChREBP-α and ChREBP-β , we performed ChIP analysis for HepG2 cells treated with 0 mM and 25 mM glucose for 18 hours. Quantitative PCR analysis for DNA fragments encompassing the E-box-containing cis-element in ChREBP-α and the nucleotides − 86~− 37 region in ChREBP-β showed that glucose promoted endogenous HNF-4α binding to ChREBP-α and ChREBP-β , respectively (Fig. 7d,e). These results show that HNF-4α mediates the glucose-responsive effect of ChREBP transcription by increasing its nuclear abundance and binding to cis-elements in ChREBP-α and ChREBP-β (Fig. 7f).

Discussion
Transcription of ChREBP-α and ChREBP-β from two promoters is dynamically regulated by various signals including glucose. However, the molecular mechanism triggering glucose-mediated regulation of ChREBP-α and ChREBP-β transcription remains unclear. Here we have identified the new role and mechanism by which HNF-4α promotes ChREBP-α and ChREBP-β transcription in response to glucose, revealed the molecular mechanism by which HNF-4α regulates ChREBP-α and ChREBP-β transcription, and found that ChREBP-α cooperates with HNF-4α in promoting ChREBP-β transcription.
Although much research has been carried out in determining the post-translational control of ChREBP by glucose, only a few studies have addressed the regulation of ChREBP by glucose at the transcriptional level. ChREBP is increased at the mRNA level in liver in response to high carbohydrate diet but not high fat diet 36 . Two nuclear receptors playing important roles in energy homeostasis, namely LXR and TR, have been shown to regulate ChREBP at the transcriptional level in liver 20,21,37 . Fasting and refeeding experiments have shown that ChREBP expression increases to similar levels in refed livers from wild type and TR-β null mice, suggesting that TR-β might not play a role in the nutritional regulation of ChREBP 21 . LXRα can function as a glucose-sensor and 293T cells. (g). Luciferase activity analysis for the 2.9 kb (WT), Δ(− 57~− 53) and Δ(− 42~− 38) ChREBP-β promoter in the pGL4-Basic plasmids at 24 hours after the ChREBP-β promoter plasmids and empty vector or FLAG-HNF-4α expression plasmid are transfected in 293T cells. (h). Luciferase activity analysis for the pGL4, 2.9 kb (WT), 0.3 kb, ΔDR1-C and DR1-C mutant ChREBP-β promoter in the pGL4-Basic plasmids at 24 hours after the ChREBP-β promoter plasmids and empty vector or HA-HNF-4α expression plasmid are transfected in 293T cells. The DR1-C mutant contains 7 underlined point mutations (GCGGACTCTGAAA). (i). ChIP analysis for HepG2 cells using an anti-HNF-4α antibody, nonspecific IgG or anti-histone antibody shows that HNF-4α binds ChREBP-β promoter. *in C-D and F-H indicates p < 0.05 when compared with the corresponding NC or empty vector -transfected sample. convert glucose into triglyceride by inducing lipogenic genes directly or indirectly via ChREBP and SREBP-1c 23 . However, ChREBP expression, its nuclear translocation and the induction of its target genes were not altered by high carbohydrate diet in liver of LXRα /β null mice, suggesting that LXR is not responsible for the effect of glucose on ChREBP 22 . Therefore, LXR and TR can promote ChREBP transcription in liver but they do not mediate glucose-induced ChREBP transcription. Here we have not only revealed the molecular mechanism by which HNF-4α promotes ChREBP-α and ChREBP-β transcription, but also have shown that HNF-4α knockdown reduced the induction of the mRNA and protein expression of ChREBP by glucose in HepG2 cells and mouse primary hepatocytes. Our results suggest that HNF-4α plays an important role in promoting ChREBP-α and ChREBP-β transcription in response to glucose. Moreover, we have revealed that glucose increases HNF-4α mRNA and protein levels, the nuclear abundance of HNF-4α and its binding to the intron of ChREBP-α or the promoter of ChREBP-β . Therefore, our findings have demonstrated that HNF-4α promotes ChREBP-α and ChREBP-β transcription in response to glucose. Glucose-induced endogenous HNF-4α binding to ChREBP-α and ChREBP-β could be due to higher levels of nuclear HNF-4α protein in response to glucose (Fig. 7c-e). Therefore, it is hard to conclude whether glucose promotes DNA binding capacity of HNF-4α . In addition, we have also noticed that USF2 and USF1 increase mRNA levels of ChREBP-α and ChREBP-β , respectively (Fig. 1b,c). It will be intriguing to find out whether USF2 and USF1 regulate transcription of ChREBP-α and ChREBP-β in response to glucose.
HNF-4α is a key transcription factor regulating hepatocyte differentiation and function 32 . HNF-4α can regulate the expression of many liver-specific target genes [25][26][27][28][29] . ChREBP-α and ChREBP-β are highly expressed in liver and our findings of ChREBP-α and ChREBP-β being HNF-4α target genes provide a possible explanation for their liver-enriched expression. HNF-4α promotes ChREBP-α and ChREBP-β transcription via different mechanisms. HNF-4α directly binds DR1 sites in the ChREBP-β promoter and regulates its transcription. However, the 4 kb of ChREBP promoter is not responsible for HNF-4α -induced ChREBP-α transcription. Instead, we have found that HNF-4α , but not LXR, directly binds the E-box-containing region in intron 12 of the ChREBP-α gene. This intronic sequence probably functions as an enhancer and cooperates with ChREBP-α promoter in regulating its transcription. Moreover, ChREBP-α and ChREBP-β genes share intron 12 and the E-box-containing region in intron 12 might also function as an enhancer in regulating ChREBP-β transcription.
The interaction between ChREBP and HNF-4α has been reported 38 . The glycolytic enzyme L-PK is a target gene for both ChREBP and HNF-4α 1,39 . Transcriptional complex containing ChREBP, HNF-4α and the co-activator CBP is necessary for the glucose-mediated induction of the L-PK gene 38 . Here we have identified a transcriptional complex containing ChREBP-α , HNF-4α and Mlx which bind to the ChoRE of the human ChREBP-β promoter and additively increases ChREBP-β transcription. Moreover, we have found that glucose might increase the interaction between HNF-4α and ChREBP-α .

Materials and Methods
Plasmids. We cloned the 4kb human ChREBP-α (− 4000 bp~− 1 bp) and 2.9kb (− 2612 bp~+ 271 bp) human ChREBP-β promoter from human genomic DNA and subcloned them into pGL3-Basic and pGL4-Basic vector (Promega, USA), respectively. Position + 1 is the transcription start site. We cloned the 164 bp, 174 bp, 378 bp and 140 bp DNA fragments containing the non-perfect DR1 sites located in intron 2, intron 6, intron 7 and intron 12 from human genomic DNA and subcloned them into the pGL3-Promoter vector (Promega, USA). The C/EBPα and c-Jun plasmids were provided by Dr. Zhaoyuan Hou and Dr. Man Mohan of Shanghai Jiao Tong University  Mice.  week-old C57BL/6 wild type mice were divided into two groups. One group of three mice were fasted for 18 hours while the other group of three mice were fed with a normal diet for 18 hours before being sacrificed. Liver tissues were removed for further study. All protocols were approved by the Shanghai Jiao Tong University School of Medicine Animal Care and Use Committee. All experiments were carried out in accordance with the approved guidelines.
Primary hepatocyte isolation, culture and treatment. Primary hepatocytes were prepared from male C57/BL6 mice using the collagenase (Gibco, USA) perfusion method 40 and plated in six-well tissue culture plates or 10 cm dish (Corning, USA) at a density of 1 × 10 6 cells /ml in hepatocyte medium (ScienCell, USA) supplemented with 5% FBS, 1% hepatocyte growth supplement and 1% penicillin/streptomycin antibiotic mix. After primary hepatocytes attached, the medium was replaced by DMEM supplemented with 10% FBS, 1% L-glutamine and 1% penicillin/streptomycin antibiotic mix. Primary hepatocytes were cultured in DMEM containing 5.6 mM glucose for 24 hours before 0 mM, 5.6 mM or 25 mM glucose treatment.
Cell lines and glucose treatment. 293T human embryonic kidney cells and HepG2 human hepatocellular carcinoma cells were cultured in DMEM supplemented with 10% FBS, 1% L-glutamine and 1% penicillin/ streptomycin antibiotic mix. Cells at 80% confluency were transfected with plasmids or siRNAs. 48 hours after transfection, cells were harvested for subsequent experiments. For the glucose treatment experiments, cells were cultured in DMEM containing 5.6 mM glucose for 4-12 hours before being treated with 0 mM, 2.5 mM, 5.6 mM or 25 mM glucose medium for 18 hours.
Real time PCR analysis. Total RNA was isolated with Trizol and reverse-transcribed into cDNA using the PrimeScript RT Master Mix kit (Takara Bio, Japan). The expression of specific genes was quantitated by real time PCR using SYBR Premix Ex Taq kit (Takara Bio, Japan) on an ABI 7500HT machine (Applied Biosystems, USA) with β -actin (human) or 18sRNA (mouse) as internal controls.
Luciferase reporter assay. 293T cells were transfected with 0.5 μg of the reporter plasmid, 0.1 μg of the β -gal plasmid and 1 μg of either pcDNA3-HNF-4α or empty vector. 24 hours later, cells were harvested and assayed using the luciferase reporter assay system (Promega, USA) and Beta-Gal Assay Kit (Beyotime, China) according to the manufacturers' instructions.
Co-immunoprecipitation. Cells were freshly lysed in the lysis buffer (1 mM EDTA, 40 mM Tris-HCl, pH 8, 100 mM NaCl, 0.5% NP-40, 1% TritonX-100), and incubated with primary antibodies at 4 °C overnight, followed by an additional 2-hour incubation with protein A/G-agarose beads (Santa Cruz Biotechnology, USA) at 4 °C. The beads were washed with the lysis buffer and boiled in 2× SDS protein loading buffer. Western blot analysis was performed after immunoprecipitation. For quantification, NIH image software was used.
Nuclear and cytoplasmic protein extraction. All the following steps were performed, and all buffers contained the EDTA-free protease inhibitor cocktail (Roche, Switzerland). After PBS rinses, cells were centrifuged at 2000 × g for 5 min at 4 degree. Cell pellets were fractionated as described previously 19 .