Introduction

Recent genetic studies [1–3] have considerably expanded the list of known genes that may cause a predisposition to diabetes. However, it remains largely unclear how these genes determine the development of type 2 diabetes. Phosphoprotein enriched in diabetes (PED, also known as phosphoprotein enriched in astrocytes-15 [PEA15]) is a scaffold cytosolic protein widely produced in human tissues [4, 5]. Early studies indicated that PED has an important role in controlling glucose disposal by impacting on protein kinase C signalling [5, 6]. It was later found that PED is commonly overproduced in individuals with type 2 diabetes as well as in their euglycaemic offspring [5, 7]. In these persons PED overproduction causes insulin resistance in GLUT4-mediated glucose disposal. Studies in tissue-specific transgenics and in null mice have indicated that the upregulation of PED observed in type 2 diabetic patients might also contribute to impaired beta cell function [7]. PED cellular levels are regulated by ubiquitinylation and proteosomal degradation [8], but run-on experiments in cultured cells from type 2 diabetic patients have demonstrated that, at least in part, the overproduction observed in these participants is caused by transcriptional abnormalities [5]. The molecular details responsible for these abnormalities and how the PED gene is regulated remain unclear. Moreover, the role of PED in liver glucose metabolism has been less extensively investigated.

We have recently demonstrated that the hepatocyte nuclear factor (HNF)-4α (NR2A1), a highly conserved member of the nuclear receptor superfamily involved in the control of glucose homeostasis [9, 10], regulates transcription of the PED gene by binding to a cis-regulatory element of the PED promoter and represses its transcription [11].

HNF-4α is essential for hepatocyte differentiation at the developmental and the functional levels [12], as well as for accumulation of hepatic glycogen stores and generation of normal hepatic epithelium [13]. Point mutations in HNF-4α impair liver and pancreatic regulation of glucose homeostasis and cause maturity onset diabetes of the young type 1. More recently, genetic and biochemical evidence has been generated indicating that HNF-4α may also have a role in the development of more common forms of type 2 diabetes [14–16], but understanding of the underlying mechanisms is incomplete.

The ability of nuclear receptors to induce specific transcription events depends on their recruitment of chromatin remodelling cofactors and enzymes, and on the assembly of the basal transcription machinery [17]. In the absence of ligand, nuclear receptors recruit co-repressors such as silencing mediator of retinoic acid and thyroid hormone receptor (SMRT) and nuclear receptor co-repressor. These, in turn, bind repressive enzymes such as the histone deacetylases and histone methyltransferases, specifically controlling the methylation state of lysine 9 at histone H3 and certain chromatin remodelling events.

Increasing evidence now indicates that chromatin remodelling is an important mechanism enabling transcription regulation [18]. Chromatin remodelling occurs through different mechanisms. One is the covalent modification of histone tails, including acetylation, methylation and phosphorylation [19]. These changes also affect the ability of chromatin to interact with transcription factors and the basal transcription machinery [20]. Acetylation and methylation of lysine residues at H3 and H4 amino termini represent the most common modifications. Indeed, increased acetylation induces transcriptional activation [21], while reduced acetylation usually signals transcriptional repression [22, 23]. Methylation of H3 lysine 9 is also associated with transcriptional repression [24].

In this work, we have investigated the molecular mechanisms responsible for HNF-4α-dependent silencing of PED expression in liver. We show that by causing epigenetic changes at the PED gene, HNF-4α controls transcriptional activity of PED and may affect glucose metabolism in liver.

Methods

Materials

Media, sera and antibiotics for cell culture, and the lipofectamine reagent were purchased from Invitrogen (Paisley, UK). Goat polyclonal HNF-4α and acetyl-histone H3 (K9/K14) and rabbit polyclonal SMRTe antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Mouse monoclonal histone H3 Lys9 (H3K9) and rabbit monoclonal histone H3 Lys4 (H3K4) antibodies were from Abcam (Cambridge, UK). The PED antibody, the pCDNA3/HNF-4α expression vector and the HNF-4α-specific short hairpin RNA (shRNA) plasmid have been previously described [11]. All short interfering RNAs (siRNAs) were chemically synthesised by Ambion (Austin, TX, USA) as oligonucleotide duplexes. siRNA target sequences for silencing mediator for retinoid and thyroid hormone receptors (SMRT) were directed at regions common to SMRTα and SMRTβ (panSMRT) [25]. As non-specific siRNA controls, the Ambion Silencer 2 negative control was used.

Western blot and electrochemiluminescence (ECL) reagents were from Thermo Scientific (Rockford, Illinois, USA). All other reagents were from Sigma (St Louis, MO, USA).

Animal studies, cell culture, transfections, RT-PCR and western blot assay

The PED transgenic mice and cellular models generation have been previously described [11, 26]. Total RNA extraction, cDNA synthesis, real-time PCR and western blot analysis were performed as described in [11]. Antibodies against SMRT, HNF-4α, PED and actin were used for detection of proteins. All the experiments involving animals were approved by the Local Ethics Committee and conducted in accordance with the Principles of Laboratory Care.

Formaldehyde-assisted isolation of regulatory elements

Whole cells were fixed in growth medium by addition of 37% (vol./vol.) formaldehyde to a final concentration of 1% (vol./vol.) formaldehyde for 10 min. The fixed cells were resuspended in nuclear lysis buffer containing 50 mmol/l Tris–HCl (pH 8.0), 10 mmol/l EDTA, 0.8% (wt/vol.) sodium dodecyl sulphate, 1 mmol/l phenyl methyl sulfonyl fluoride and inhibitors cocktail (Sigma), and then incubated on ice for 10 min. The extracts were then sonicated (Misonix 3000) and microcentrifuged for 10 min at 16,000 g and 4°C. Protein-free DNA was then analysed by SYBR Green real-time PCR [27]. The primer sets used are shown in Electronic supplementary material (ESM) Table 1. The amount of PCR product (representing nucleosome-free DNA) was plotted as a percentage of the input DNA representing total cellular DNA.

Micrococcal nuclease protection assay

Nuclei were isolated from 1 × 108 of the following cells: HeLa wild-type, HeLa overproducing exogenous HNF-4α (\( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \)) HepG2 and HepG2 subjected to silencing of the endogenous HNF-4α (\( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \)). Isolated nuclei were suspended in 1 ml of wash buffer (10 mmol/l Tris–HCl, pH 7.4, 15 mmol/l NaCl, 50 mmol/l KCl, 0.15 mmol/l spermine, 0.5 mmol/l spermidine and 8.5% [wt/vol.] sucrose) and digested with 120 U of micrococcal nuclease (MNase) for 30 min at 37°C. The purified DNA was quantified and identified on an agarose gel, and subsequently amplified by PCR using the C, D and F primer sets. An aliquot of undigested DNA was obtained as a control of total cellular DNA quality.

Chromatin immunoprecipitation and re-chromatin immunoprecipitation assay procedures

Chromatin immunoprecipitation (ChIP) assay was performed as previously described [11]. Sheared chromatin samples were taken as input control or used for immunoprecipitation with the following antibodies: acetyl-histone H3 (sc-8655), dimethylated histone H3 at lysine 9 (H3K9me2; ab-1220), H3K4me2 (ab-7766), HNF-4α (sc-6556), SMRTe (sc-1612) and normal rabbit IgG as a negative control. DNA fragments were recovered and subjected to real-time PCR amplification using primer sequences described in Table 1.

For re-chromatin immunoprecipitation (Re-ChIP) assay, immunoprecipitates with the first antibody were eluted in 50 µl of dithiothreitol 10 mmol/l, diluted tenfold in ChIP dilution buffer supplemented with protease inhibitors and immunoprecipitated with the second antibody. Following immunoprecipitation, samples were processed as described for ChIP assay [11] and eluted DNA was amplified by real-time PCR with specific oligos.

Trichostatin A treatment of cells

HepG2 cells were seeded in 10 cm dishes at a density of 1 × 106 cells 1 day before drug treatment. The cells were treated with 330 nmol/l trichostatin A (Sigma-) for 1 day, and total RNA and chromatin were prepared as described before.

Hepatic glycogen measurement

Cells and tissues were solubilised in 0.1% (wt/vol.) sodium dodecyl sulphate and then further incubated with 1:1 saturated Na2SO4 and 95% (vol./vol.) ethanol. The pellets were rehydrated, and 5% (vol./vol.) phenol and H2SO4 were added. Finally, absorbance at 490 nm was measured. Results are expressed as micrograms of glycogen per milligram of protein or tissues.

Statistical procedures

Statistical analysis was performed with a software package (StatView 5.0; Abacus Concepts, Berkley, CA, USA) using the Student’s t test. Values of p < 0.05 were considered statistically significant.

Results

HNF-4α induces DNA remodelling and nucleosome positioning at the PED promoter

As previously reported, HeLa cells feature high expression of the PED gene, while very low expression was found in the HepG2 cells [11]. Expression of the PED silencer HNF-4α is inversely related to that of PED in these cells [11]. Thus, these cell lines represent an attractive model to investigate the molecular mechanisms responsible for HNF-4α regulation of PED gene transcription. To investigate changes induced by HNF-4α in chromatin structure across the PED promoter, we isolated nucleosome-free chromatin DNA from HeLa cells (either wild-type or expressing an HNF-4α [also known as HNF4A] cDNA) and from HepG2 cells (wild-type and expressing an HNF-4α shRNA) by performing a formaldehyde-assisted isolation of regulatory elements assay [28]. The DNA fragments were amplified using seven sets of PCR primers, covering consecutive 100 bp regions positioned at −500 to 300 in the PED gene (Fig. 1a, ESM Table 1) and PCR amplification was monitored by SYBR green incorporation [29]. The amount of DNA amplified (nucleosome-free DNA) was plotted as percentage of the input DNA representing total cellular DNA.

Fig. 1
figure 1

Enrichment of regulatory DNA across 0.8 kb of PED gene using formaldehyde-assisted isolation of regulatory elements. a Schematic representation of PED showing the HRE, the transcription start site, the exon 2 position and the DNA fragments amplified by primer sets A to F. b Protein-free DNA was extracted from HeLa (black bars) and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) (grey bars), and (c) from HepG2 (white bars) and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) (striped bars) cells as described above and analysed by real-time PCR using the indicated primers. The mean and standard errors of at least three independent experiments each performed in triplicate are shown. Statistical significance was determined by t test (two-tailed) analysis; ***p < 0.001

Full size image

In \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \), the amount of nucleosome-free DNA across the examined regions ranged from 25 to 65% compared with untransfected HeLa cells showing higher free DNA amounts (Fig. 1b). \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) featured high levels and low reduction of nucleosome-free DNA, compared with untransfected HepG2 cells exhibiting high expression levels of HNF-4α and reduced free DNA across the −500 to 300 region of the PED gene (Fig. 1c). In \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) and HepG2 cells, three regions were found to be associated with reduced levels of free DNA. These corresponded to the HNF-4α response element (HRE; at −350, region B), its immediate proximity (−200, region C) and the proximity of the PED transcription start site (−50, region D).

To further investigate nucleosome positioning across the PED promoter, mononucleosomal DNA averaging 150 bp in size was obtained by nuclear MNase digestion of HeLa cells (either transfected with HNF-4α or untransfected) and HepG2 cells (transfected with HNF-4α-sh or untransfected; Fig. 2a). This DNA was subsequently amplified using the PCR primer sets for the C and D regions shown in Fig. 1b. Figure 2b shows the results obtained with primer set C; similar results were obtained with primer set D (data not shown). In agarose gel electrophoresis, the amplification products were barely detectable when using the mononucleosomal DNA from HeLa cells, but were significantly increased with the DNA from the \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) cells (C region). Consistently, only a weak band was obtained in HNF-4α-silenced (\( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \)) cells compared with the untransfected HepG2 cells (C region). The use of undigested DNA in control PCR assays led to the amplification of a 120 bp band with all of the cell lines (C region). Also, upon MNase treatment, the nucleosome-free exon 2 allowed no PCR amplification with F region primers (Fig. 2c), although clear amplification bands were obtained with the undigested DNA (F region). These data suggest that HNF-4α production and binding to the HRE causes nucleosome positioning at C and D regions in the proximal PED gene promoter.

Fig. 2
figure 2

Accessibility of PED gene regions to MNase digestion. HeLa, \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \), HepG2 and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) nuclei were digested with 120 U of MNase for 30 min, which extensively digests the linker and nucleosome-free regions. a Representative image of digested nucleosomes separated on a 2% agarose gel before purification. First lane contains a DNA marker ladder. The arrow indicates mononucleosomes. b PCR amplification with primer sets from Fig. 1a as indicated using mononucleosomal DNA (bottom rows) and genomic DNA as positive control (top rows) for C region and (c) F region. The PCR products are shown for each primer set after agarose gel electrophoresis. Photographs are representative of three (a) and four (b, c) independent experiments

Full size image

HNF-4α induces histone deacetylation and methylation at the PED promoter

We next focused on the significance of HNF-4α cellular levels for the epigenetic state of the PED promoter region. To this end, we looked for typical transcriptional activation and repression marks in the cells with high (\( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \), HepG2) and low (HeLa, \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \)) HNF-4α levels, in which PED transcription is suppressed and active, respectively (Fig. 3a, b). ChIP assays were performed using antibodies recognising histone modifications, as well as normal rabbit IgG as a negative control, with the purified DNA from the immunoprecipitates being used for real-time PCR analysis with primer sets amplifying the C region nucleosome. In \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) and HepG2 cells we saw an almost twofold reduction in histone H3 acetylation at lysines 9 and 14 (a mark of active transcription) compared with HeLa and the \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells (Fig. 3c, e). A further mark of active transcription [30], lysine 4 dimethylated histone H3, was similarly reduced in these cells (Fig. 3g, i). Also, the dimethylation of histone H3 at lysine 9 (H3K9me2), a mark typical of repressed gene promoters [31], was almost threefold higher at the C region nucleosome from \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) and HepG2 as compared with HeLa or \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells (Fig. 3k, m). Similar results were obtained with the D region nucleosome (data not shown). Thus, HNF-4α production and response element binding is accompanied by enrichment in repressive histone modifications together with reduced rate of PED expression.

Fig. 3
figure 3

HNF-4α production determines repressive histone modifications at the PED promoter. a, b Hela and HepG2 cells were transfected with 1 μg of the pCDNA3/HNF-4α expression vector and of the HNF-4α-specific shRNA clone, respectively. At 48 h after transfection, total RNAs were extracted from transfected and non-transfected cells. PED mRNA levels were then quantified by RT-PCR. Data were normalised to β-actin mRNA and are expressed as per cent decrease or increase vs control (untransfected cells). ChIP experiments were performed using antibodies against active histone marks, i.e. (c–f) acetyl-histone H3 and (g–j) H3K4me2, in the PED gene in HeLa (black bars) and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) cells (grey bars), and in HepG2 (white bars) and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) (striped bars) cells. ChIP was also performed against the repressive histone mark H3K9me2 (k–n) in HeLa and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) cells, and in HepG2 and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells. ChIPs were followed by quantitative PCR amplification with primer set for the C region of the PED promoter. Representative gels (d, f, h, j, l, n) are shown. Levels in bar graphs (c, e, g, i, k, m) are presented as per cent enrichment relative to input DNA and corrected for IgG control levels as analysed by quantitative PCR. Bars represent the mean ± SE of three independent experiments each performed in triplicate. Statistical significance was assessed by t test analysis. ***p < 0.001

Full size image

In addition, the \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells, which feature high levels of PED mRNA, exhibited reduced hepatic glycogen content compared with control cells. Similar results were obtained in stably transfected HepG2 cells overexpressing human PED (Fig. 4a).

Fig. 4
figure 4

Effect of PED overproduction on hepatocyte glycogen content. a HepG2, \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) and HepG2/PED cells were assayed for glycogen content after incubation with 100 nmol/l insulin (black bars) as described above. White bars, without insulin. b Glycogen content was also compared in liver tissue from transgenic mice overexpressing PED (Tg-PED) and from their non-transgenic littermates (BDF) maintained under random-feeding conditions. Bars represent the means ± SE of three (a) and four (b) independent measurements each in triplicate. *p < 0.05, **p < 0.01 and ***p < 0.001

Full size image

We also analysed glycogen content in livers from PED transgenic mice. These animals have been previously characterised and described [26]. As shown in Fig. 4b, the abundance of hepatic glycogen content was lower in transgenic mice than in control non-transgenic littermates (Fig. 4b).

HNF-4α recruits repressive enzymes at the PED promoter and silences expression

Since HNF-4α production results in decreased histone acetylation, we investigated the presence of histone-modifying enzymes at the HRE region in the PED promoter in the presence and absence of HNF-4α. Using ChIP assays, we demonstrated the presence of the SMRT co-repressor complex at the HNF-4α in HepG2 cells but not in Hela cells (Fig. 5a). Importantly, HeLa and HepG2 cells feature very comparable levels of SMRT, indicating that the presence of SMRT at the HRE observed in the HepG2 cells is not dependent on SMRT levels (ESM Fig. 1a). The assembly of HNF-4α/SMRT complexes on the PED promoter region was then investigated by Re-ChIP experiments using HNF-4α antibody in the first ChIP and a SMRT antibody in the second ChIP. As shown in Fig. 5b, c, we detected HNF-4α together with SMRT at the HRE in \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) and HepG2 cells, while the complex was barely detectable in cells producing lower amounts of HNF-4α (HeLa and the \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \)). Nuclear extracts from each cell type were immunoprecipitated with the SMRT antibody and the immunocomplexes blotted with the HNF-4α antibody. Again, we detected an HNF-4α/SMRT interaction only in the HNF-4α-producing cells, \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) and HepG2, compared with HeLa and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells, respectively (Fig. 5d), indicating that recruitment of the co-repressor SMRT was entirely dependent on the presence of HNF-4α.

Fig. 5
figure 5

Recruitment of repressive enzymes at the PED promoter is dependent on HNF-4α. ChIP experiments were performed using antibodies against SMRT (a) and HNF-4α (b, c). The binding of SMRT to the HRE of PED was analysed by quantitative PCR in (a) HeLa and HepG2 wild-type cells. HNF-4α binding was analysed by quantitative PCR in the HRE of PED in HeLa (black bar) and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) (grey bar) (b), and (c) in HepG2 (white bar) and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) (striped bar) cells. Re-ChIP experiments were performed using antibodies against SMRT (b, c) to identify protein complexes at the HRE of PED. The amount of precipitated DNA from the first ChIP was used as input. Results are expressed as enrichment relative to input (%) and corrected for IgG control levels. d HeLa, \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \), HepG2 and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells were grown as described and total lysates of the cells immunoprecipitated with antibodies against SMRT and then analysed by immunoblot with antibodies against HNF-4α and SMRT. Filters were revealed by ECL. The autoradiograph shown is representative of four independent experiments. e HepG2 cells were treated with 330 nmol/l trichostatin A (black bars) for 24 h. White bar, vehicle. PED and HNF-4α expression were determined by real-time PCR. Data were normalised to β-actin levels and expressed as relative to vehicle treatments. Bars (a–c, e) represent the mean ± SE of at least three independent experiments each performed in triplicate. Statistical significance was assessed by t test analysis. ***p < 0.001

Full size image

Re-ChIP experiments using antibodies toward the repressive histone mark H3K9me2 followed by HNF-4α or SMRT antibodies revealed that levels of these complexes were reduced in the cells with low HNF-4α levels (HeLa and the \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \)), in which PED transcription is active. This finding indicates that HNF-4α/SMRT assembly to the PED gene is accompanied by histone deacetylation and methylation at its proximal promoter (ESM Fig. 1b).

Since HNF-4α production results in changes of the epigenetic state of chromatin, we investigated the possibility of activating the PED gene by pharmacological inhibition of histone deacetylation. Incubation of HepG2 cells with the deacetylase inhibitor trichostatin A revealed an increase in PED expression in these cells (Fig. 5e). Since HNF-4α levels also significantly increased upon trichostatin A treatment, we sought to verify whether hyperacetylation induced by trichostatin A treatment impairs HNF-4α binding to its response element, as this effect may have been sufficient to enhance PED transcription. We found no difference in HNF-4α at its B region-amplified response element in HepG2 cells, whether preincubated with trichostatin A or not (ESM Fig. 2). This supports the role of histone deacetylation in restraining PED activity. We conclude that histone deacetylation is required for HNF-4α-mediated PED repression.

SMRT silencing rescues PED expression in HepG2 cells

To further address the role of SMRT in HNF-4α-dependent repression of PED function, HepG2 cells were transiently transfected with siRNAs for SMRT (also known as NCOR2)-α and SMRTβ (panSMRT) or a control siRNA [25]. Real-time PCR assays on total RNA from panSMRT-silenced HepG2 cells showed about twofold increased PED RNA levels compared with control siRNA and wild-type cells (Fig. 6a). Western blot analysis of extracts from the different cell types also revealed significantly increased PED and decreased SMRT protein levels upon the panSMRT transfection (Fig. 6b). In addition, ChIP assays in panSMRT-treated HepG2 cells using acetyl-histone H3 (K9/K14) and H3K9me2 antibodies showed significantly increased acetylation and decreased methylation of the histone H3 lysine 9 compared with controls (Fig. 6c).

Fig. 6
figure 6

Effect of SMRT depletion on PED expression in HepG2 cells. HepG2 cells were transfected with Silencer negative control siRNA (grey bars) or siRNAs for panSMRT (black bars). a SMRT and PED mRNA levels were quantified by real-time PCR. Data were normalised to β-actin mRNA and are expressed as fold over control (untransfected cells [NT], white bars). b Lysates prepared from HepG2 cells transfected with siRNAs were western-blotted with PED and SMRT antibodies and further analysed by ECL and autoradiography. The autoradiograph shown is representative of four independent experiments. c ChIP experiments were performed using antibodies against active (acetyl-histone H3 [AcH3]) or repressive (H3K9me2) histone marks. Their binding was analysed by quantitative PCR in the nucleosome C region of PED promoter in control or (panSMRT) HepG2 cells and expressed as fold over control (untransfected cells). d HepG2 cells were co-transfected with 3 μg of the indicated PED promoter–luciferase constructs (or 3 μg of the promotorless pGL3 basic vector) alone or in combination with SMRTα and SMRTβ siRNAs or the Silencer negative control siRNA and 1 μg of the pRSVβ-gal vector DNA. Luciferase activity was assayed as described and is presented as the increase above the activity measured with the control pGL3 basic vector. Bars represent the mean ± SE of three independent experiments each performed in triplicate. Statistical significance was assessed by t test analysis. ***p < 0.001

Full size image

The role of SMRT in HNF-4α regulation of PED expression was further addressed in HepG2 cells transfected with a PED promoter luciferase construct featuring the HRE (pPED477). Co-transfection of the panSMRT siRNA in these cells increased the pPED477 reporter gene activity by twofold. Importantly, the luciferase activity of a vector containing the mutagenised HNF-4α binding sequence (pPED477mut) abolishing its transcriptional activity was increased at similar levels in control and panSMRT transfected HepG2 cells (Fig. 6d) compared with HepG2 wild-type cells transfected with the pPED477 vector alone (Fig. 6d). This indicates that the recruitment of HNF-4α and SMRT to the HRE at the PED promoter is necessary for PED transcriptional inhibition.

SMRT enables HNF-4α inhibition of GRB14 and USP21 genes

To verify whether SMRT action on HNF-4α silencing is unique to the PED gene, we analysed expression of the two other genes repressed by HNF-4α, GRB14 and USP21. Interestingly, GRB14 and USP21 were expressed at reduced levels in the \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) compared with untransfected cells (Fig. 7a). In addition, expression of these genes was rescued by HNF-4α silencing in \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells, further supporting the repressor function of HNF-4α (Fig. 7b). The regulatory region of GRB14 and USP21 has not been reported yet. By in silico analysis, we therefore identified their HREs at −700 and −450 bp upstream of their transcriptional start sites, respectively (Fig. 7c). ChIP and Re-ChIP assays with specific antibodies and HRE primer sets revealed largely increased HNF-4α and HNF-4α/SMRT complex abundance at the HRE in the \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) compared with untransfected cells (Fig. 7d). Consistently, the presence of HNF-4α and HNF-4α/SMRT complex at the response element was significantly reduced in HepG2 cells transfected with the HNF-4α-sh (Fig. 7e). We then examined the acetylation state of the GRB14 and USP21 proximal regulatory regions and performed ChIP assays using acetyl-histone H3 (K9/K14) antibodies and primer sets amplifying a 300 bp region downstream of the HREs. Histone H3 (K9/K14) acetylation was significantly reduced in the \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \), but rescued in the \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) cells (Fig. 7f, g), indicating common mechanisms in the silencing of the PED, GRB14 and USP21 genes by HNF-4α.

Fig. 7
figure 7

HNF-4α-dependent inhibition of GBR14 and USP21 expression. Total RNA preparations were obtained (a) from HeLa (black bars) and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) (grey bars) cells, and (b) from HepG2 (white bars) and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) (striped bars) cells. Levels of GRB14 and USP21 mRNA were then determined by real-time PCR and normalised to β-actin mRNA. mRNA levels are relative to those in control (HeLa, HepG2) cells. c Schematic representation of the GRB14 and USP21 genes, with arrows indicating ChIP primers against the HNF-4α binding site (HRE) and the region downstream the HRE. ChIP experiments were performed using antibodies against HNF-4α (d, e) and (f, g) active histone marks (acetyl-histone H3). HNF-4α binding was analysed by quantitative PCR in the HRE of GRB14 and USP21 in HeLa and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) (d) and HepG2 and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) (e) cells. Acetyl-histone H3 enrichment at the GRB14 and USP21 regulatory regions was analysed by quantitative PCR in HeLa and \( {\hbox{HeL}}{{\hbox{a}}_{{\rm{HNF}} - {4}}}_\alpha \) (f), and HepG2 and \( {\hbox{HepG}}{{2}_{{\rm{HNF}} - {4}}}_{\alpha - {\rm{sh}}} \) (g) cells. Re-ChIP experiments were performed using antibodies against SMRT (d, e) to identify protein complexes at the HRE of GRB14 and USP21 genes. The amount of precipitated DNA from the first ChIP was used as input. Results are expressed as per cent enrichment relative to input DNA and corrected for IgG control levels. Bars represent the mean ± SE of three independent experiments each performed in triplicate. Statistical significance was assessed by t test analysis. **p < 0.01, ***p < 0.001

Full size image

Discussion

Previous work has established the significance of the nuclear receptor superfamily in regulating a broad range of cellular processes by activating or repressing different sets of genes that harbour nuclear receptor-recognising DNA motifs [32]. Several mechanisms leading to transcriptional activation by the nuclear receptors have been elucidated in considerable detail [33]. Against this, the molecular bases of transcriptional repression remain more unclear [34, 35].

In the present study, we show that HNF-4α, a member of the steroid receptor class of nuclear receptors [36, 37], triggers SMRT recruitment to the PED proximal promoter and causes histone tail hypoacetylation at H3-K9/K14, inducing a transcriptionally non-permissive state of the gene. These changes are followed by the appearance of H3K9 hypermethylation and H3K4 hypomethylation, two marks of heterochromatin (Fig. 8). These findings indicate for the first time that, in intact cells, HNF-4α represses gene transcription by directly recruiting SMRT to the promoter region, leading to histone deacetylation-associated remodelling of chromatin.

Fig. 8
figure 8

Epigenetic changes induced by HNF-4α to PED expression. HNF-4α production and binding to its response element causes chromatin packaging at the PED promoter and enrichment of histone H3 methylation at lysine 9 (H3K9me2), a mark typical of repressed gene promoters. HRE binding also causes a decrease in acetylation at lysine 9 and 14 (AcH3K9/K14) and dimethylation of lysine 4 of histone H3 (H3K4me2), two marks of active transcription

Full size image

Previous studies in HNF-4α-producing cells and in cells with no HNF-4α production have shown that SMRT represses HNF-4α-mediated transcription of reporter constructs containing heterologous, as well as native promoters [38]. In addition, using glutathione S-transferase pull-down assays, the same team of authors showed that HNF-4α directly binds the receptor interacting domain 2 of SMRT in vitro [38]. These observations suggest that SMRT repression might represent a physiologically relevant mechanism responsible for HNF-4α regulation and that functional cooperation requires physical interaction of SMRT and HNF-4α. The finding we now report, namely that the HNF-4α/SMRT complex induces GRB14 and USP21 repression in addition to PED gene silencing, further underlines the significance of these mechanisms to the in vivo transcriptional repressor activity of HNF-4α.

HNF-4α has long been known to serve as a transcriptional activator [39]. Indeed, HNF-4α can: (1) activate transcription in the absence of exogenously added ligands [40]; and (2), in mammalian cells, in yeasts and in vitro, respond to several coactivators including glucocorticoid receptor interacting protein 1 (GRIP1), steroid receptor coactivator 1 (SRC-1) and CREB binding protein (CBP)-p300 [41, 42]. However, Ruse et al. reported that HNF-4α regulation by these coactivators can be competitively abolished by interaction of the histone deacetylase-associated co-repressor SMRT with the F domain of HNF-4α [38]. Thus, the nature of the recruited co-regulator and of its association with direct effectors of histone modifications such as histone deacetylases appears to determine the transcriptional activity of HNF-4α.

Recently, it has been shown that a knock-in mouse bearing a mutation in the receptor interacting domain of SMRT (SMRTmRID) that disrupts its interaction with the nuclear hormone receptor [43] develops multiple metabolic defects including altered insulin sensitivity and 70% increased adiposity. These findings indicate that major districts of metabolic regulation such as insulin sensitivity and fat mass depend on the molecular balance of co-repressors and coactivators, which modulate the status of chromatin and nuclear receptor signalling and ultimately regulate gene expression in response to specific physiological situations.

PED expression is upregulated in individuals with type 2 diabetes and in the euglycaemic offspring of these patients, impairing insulin action and insulin secretion [5, 7]. However, the causes for PED upregulation required additional investigation. In the present study, we demonstrated that HNF-4α silencing alters the epigenetic state of the PED gene and determines its overexpression in liver cells. We further show that induction of PED gene expression in hepatocytes is paralleled by reduced hepatic glycogen content. We have also very recently demonstrated that activation of PED expression in HepG2 cells is paralleled by the establishment of a partially dedifferentiated phenotype, accompanied by reduced mRNA levels of genes expressed during normal liver development [11]. Thus HNF-4α-regulated PED gene may have a role in hepatocyte differentiation. Epigenetic changes upregulating its expression might impair development of normal liver function, thereby contributing to progression toward diabetes. Environmental factors have been reported to induce epigenetic changes, leading to the development of abnormal phenotypes [44, 45]. For instance, diet and ageing have been reported to perturb gene expression by inducing DNA methylation and histone modifications [46]. Also, intrauterine growth retardation, a common complication of pregnancy, has been associated with later development of type 2 diabetes, in part through chromatin remodelling effects and epigenetic modulation of the expression of several genes [47]. Type 2 diabetes is a genetically heterogeneous disease resulting from complex interactions of genetic and environmental determinants. Previous studies using different approaches have uncovered a number of type 2 diabetes susceptibility genes, but it remains less clear how these genes determine type 2 diabetes. In particular, the molecular details of gene–environment interplay in diabetes onset have received less attention. Exposure to certain environmental determinants, including maternal obesity [48], intrauterine environment [49, 50] and nutritional factors [51], affects key developmental sequences programming metabolic responsiveness to environmental stimuli during later life. Such concepts are now supported by studies of monozygotic twins, where diabetes risk is discordant and linked to birthweight [52]. Laboratory studies are in progress to assess whether PED may serve as an environment gene target contributing to type 2 diabetes progression.

In conclusion, in the present work, we demonstrate that HNF-4α inhibits PED expression by inducing chromatin remodelling and histone modifications at nucleosomes located in the regulatory region of the gene. Dysregulation of these mechanisms might promote pathogenetic sequences leading to type 2 diabetes.