Potential conflict of interest: Dr. Lamers received grants from NWO.
Hepatic HNF4α deficiency induces periportal expression of glutamine synthetase and other pericentral enzymes†
Article first published online: 26 JAN 2007
Copyright © 2007 American Association for the Study of Liver Diseases
Volume 45, Issue 2, pages 433–444, February 2007
How to Cite
Stanulović, V. S., Kyrmizi, I., Kruithof-de Julio, M., Hoogenkamp, M., Vermeulen, J. L. M., Ruijter, J. M., Talianidis, I., Hakvoort, T. B. M. and Lamers, W. H. (2007), Hepatic HNF4α deficiency induces periportal expression of glutamine synthetase and other pericentral enzymes. Hepatology, 45: 433–444. doi: 10.1002/hep.21456
- Issue published online: 26 JAN 2007
- Article first published online: 26 JAN 2007
- Manuscript Accepted: 4 OCT 2006
- Manuscript Received: 15 MAY 2006
- NWO. Grant Number: 902-23-196
In liver, most genes are expressed with a porto-central gradient. The transcription factor hepatic nuclear-factor4α (HNF4α) is associated with 12% of the genes in adult liver, but its involvement in zonation of gene expression has not been investigated. A putative HNF4α-response element in the upstream enhancer of glutamine synthetase (GS), an exclusively pericentral enzyme, was protected against DNase-I and interacted with a protein that is recognized by HNF4α-specific antiserum. Chromatin-immunoprecipitation assays of HNF4α-deficient (H4LivKO) and control (H4Flox) livers with HNF4α antiserum precipitated the GS upstream enhancer DNA only from H4Flox liver. Identical results were obtained with a histone-deacetylase1 (HDAC1) antibody, but antibodies against HDAC3, SMRT and SHP did not precipitate the GS upstream enhancer. In H4Flox liver, GS, ornithine aminotransferase (OAT) and thyroid hormone-receptor β1 (TRβ1) were exclusively expressed in pericentral hepatocytes. In H4LivKO liver, this pericentral expression remained unaffected, but the genes were additionally expressed in the periportal hepatocytes, albeit at a lower level. The expression of the periportal enzyme phosphoenolpyruvate carboxykinase had declined in HNF4α-deficient hepatocytes. GS-negative cells, which were present as single, large hepatocytes or as groups of small cells near portal veins, did express HNF4α. Clusters of very small GS- and HNF4α-negative, and PCNA- and OV6-positive cells near portal veins were contiguous with streaks of brightly HNF4α-positive, OV6-, PCNA-, and PEPCK-dim cells. Conclusion: Our findings show that HNF4α suppresses the expression of pericentral proteins in periportal hepatocytes, possibly via a HDAC1-mediated mechanism. Furthermore, we show that HNF4α deficiency induces foci of regenerating hepatocytes. (HEPATOLOGY 2007;45:433–444.)
The development and maintenance of liver architecture and function is regulated by liver-enriched transcription factors.1 One of these, hepatic nuclear factor 4α (HNF4α; NR2A1) is expressed at high levels in liver, kidney, intestine, and pancreas2, 3 and binds to the promoter of 12% of genes that are expressed in adult liver.4 HNF4α is an orphan member of the nuclear-receptor superfamily.2 Depending on chain length and degree of saturation,5 fatty acyl-coenzyme A thioesters may act as agonistic or antagonistic factors, but whether or not these thioesters function as ligands remains unsettled.2, 6–8
Transcriptional regulation by HNF4α is accomplished by interactions with coactivator or corepressor mediators (e.g., GRIP1, SRC-1, CBP/p300, SMRT).6, 7, 9, 10 The resulting coactivator or corepressor complexes have intrinsic histone acetyltransferase (HAT) and histone deacetylase (HDAC) activity, respectively. Histone modifications play an important role in the regulation of the accessibility of the DNA. They can promote an open chromatin structure, in which the DNA template is accessible for transcription factors, or facilitate chromatin condensation, leading to a transcriptionally nonpermissive state.11 During embryonic development HNF4α supports gastrulation by regulating genes that are expressed in the extraembryonic visceral endoderm.3 In liver organogenesis and regeneration, HNF4α directs hepatoblasts differentiation to hepatocytes and controls the formation of the liver parenchyma.12–15
The adult mammalian liver is actively involved in amino-acid, carbohydrate and lipid metabolism, xenobiotic detoxification, and the production of bile acids and serum proteins. To maintain homeostasis, the complementary routes of the metabolic functions of the liver (e.g., gluconeogenesis and glycolysis, lipid oxidation and lipogenesis) are heterogeneously expressed.16 Amino-acid metabolism, gluconeogenesis, lipid oxidation, energy metabolism and glycogen synthesis from lactate take place in the upstream, periportal hepatocytes. On the other hand, glycolysis, lipogenesis, cytochrome P450-dependent detoxification and glycogen synthesis from glucose are located in the downstream, pericentral hepatocytes. Ammonia detoxification by the urea cycle is found in a wide periportal zone, whereas its detoxification via glutamine biosynthesis is found exclusively in a narrow pericentral zone of hepatocytes.16 We have extensively investigated the regulation of expression of GS as a paradigm for pericentral gene expression. Previous in vitro experiments have shown that interactions between the upstream enhancer located at −2.5 kb and several intronic regulatory elements determine the degree of activation of the GS promoter.17–19 However, the in vivo mechanism regulating pericentral GS expression has remained unclear.
HNF4α has been reported to be involved in the regulation of both periportally expressed genes (e.g., the gluconeogenic enzymes PEPCK20 and glucose-6-phosphatase,21 and the urea cycle enzyme OTC22), and pericentrally expressed genes (e.g., some cytochrome P-450 genes, including Cyp7α and UDP-glucuronyltransferase,23, 24 and apolipoprotein E25). In the course of our analysis of the regulation of GS, we noticed the presence of a potential HNF4α-response element in the GS upstream enhancer. This study explores the regulation of GS expression through this upstream enhancer element and shows that HNF4α binds to the GS upstream enhancer. When we extended the study to HNF4α-deficient livers, we observed that HNF4α not only stimulates expression of periportal genes (PEPCK), but also suppresses the expression of GS and several other pericentrally expressed genes in the periportal areas, suggesting that these genes share a common regulatory mechanism. In addition, we observed that the HNF4α-deficient liver contains regenerative foci.
Materials and Methods
The GS upstream enhancer is a 244bp long element located −2.5kb relative to the transcription-start site.17 This sequence was analyzed for transcription factor-binding sites with MatInspector Professional (release 4.3, Genomatix Software, Genomatix, Munchen, Germany), which is based on the MatInspector program,27 using the vertebrate matrix library and optimized thresholds.
Preparation of Protein Extracts.
Nuclear extracts from rat livers were prepared as described.28 Briefly, livers were homogenized in 2 M sucrose and nuclei were purified through a 2 M sucrose cushion. Chromatin was precipitated in 0.4 M (NH4)2SO4. Nuclear proteins were precipitated from the resulting supernatant by increasing the (NH4)2SO4 concentration to 2.2 M, followed by dialysis against 40 mM KCl.
In Vitro Footprinting.
The DNA matrix was produced by PCR, using a [γ32P]dATP-radiolabelled oligonucleotide primer, and purified on a 2% MS8 agarose gel. The F1 and R1 primer set used for footprinting is depicted in Fig. 1. The reaction mixture contained 20 μg of rat liver nuclear extract or BSA diluted in 15 μl buffer containing 25 mM HEPES pH 7.6, 40 mM KCl, 50 mM (NH4)2SO4, 20% glycerol, 0.1 mM EDTA, 1 mM DTT. After addition of 12.5 μl buffer containing 90 mM HEPES pH 7.6, 144 mM KCl, 24 mM MgCl2, 24% glycerol, 0.36 mM EDTA, 3.6 mM DTT and 10 μg ds poly(dIdC), the mixture was incubated on ice for 10 minutes, followed by the addition of 2 μL probe (20,000 cpm). After incubation on ice for 20 minutes, 0, 0.5, 1.0, 5.0 or 10.0 μg of DNase I was added for exactly 2 additional minutes. The reactions were terminated by addition of 50 μl 50 mM EDTA pH 8.0, 0.2% SDS, 100 μg/ml tRNA, 500 μg/ml proteinase K and incubated at 50°C for 30 minutes. Samples were phenol- extracted and ethanol-precipitated. The pellets of the radiolabelled DNA were dissolved in 4 μl loading buffer (95% formamide, 10 mM EDTA, 0.05% bromophenol blue, 0.05% xylene cyanol), of which 2 μl was separated on a denaturing 6% polyacrylamide gel. Radioactivity was collected on a storage phosphor screen and visualized using a Storm 680 phosphorimager (Molecular Dynamics, Sunnyvale, CA). The results were analyzed with ImageQuant (version 5.0 software).
Electrophoretic Mobility Shift Assay.
Double-stranded probes were designed on basis of the rat GS upstream enhancer sequence. Probes for the intact and mutated HNF4α binding sites were 5′-GGGGCTGACCAAGGGGGGCAAAGCTTCTTGTTTA and 5′-GGGGCTGACCGCGCGCGCGCGCGCGCTTGT-TTA, respectively (Fig. 1). Probes were radiolabelled with [α32P]dATP using the Klenow enzyme and purified on a Sephadex G50. Nuclear extract (5 μg) was pre-incubated on ice for 10 minutes in the presence of 20 mM HEPES pH 7.6, 60 mM KCl, 12% glycerol, 1 mM EDTA, 1 mM DTT, 1 mM spermidine, 1 μg ds poly(dIdC). After addition of 1 μl probe (20,000 cpm), specific complexes were allowed to form on ice for 20 minutes. To perform competition experiments, unlabelled oligonucleotides were added to the reaction mixture in 10, 100 and 1,000-fold excess. The supershift analysis employed 1 μl of goat anti-human HNF4α antibody (Santa Cruz Biotechnology, CA). Unlabelled competitors or antibodies were incubated with the basic reaction mixture for an additional 10 minutes. Samples were loaded onto a native 6% polyacrylamide gel and run at 10 V/cm. The radioactivity on the dried gels was visualized and analyzed as described for in vitro footprinting.
In order to crosslink the transcription complexes in their native nuclear environment, the livers of anaesthetized mice were perfused via the portal vein with, successively, PBS, 1% formaldehyde, and 0.125 M glycine for 10 minutes each. Cross-linked nuclei were purified by centrifugation through a sucrose gradient as described.29 Immunoprecipitations and washings were performed as described.30, 31 Anti-HDAC1 was purchased from Upstate Biotechnology (Milton Keynes, UK), while anti-SMRT and anti-HDAC3 were from Santa Cruz Biotechnology. The anti–small heterodimer partner (anti-SHP) and anti-HNF4α antisera used were described previously.32, 33 Antigen specificities were confirmed by immunoprecipitation-Western-blot assays with cross-linked chromatin or native nuclear extracts and by competition experiments. Real-time PCR analysis of GS upstream nhancer was performed using sense 5′-GCAAGCCAGTTAAGGAGGGA and antisense 5′-CTCCCGTAGCCCTCGA ATAG primers (−97 to 77 bp and 194 to 214 bp relative to Fig. 1).
Construct 6-bGH contains modules of the GS genomic DNA that were cloned in the pSPluc+ plasmid (Promega). The following fragments were placed between the HindIII and NcoI sites of the upstream polylinker: the AflIII-AflIII upstream enhancer of GS (−2677bp to −2374bp relative to the GS transcription-start site, Fig. 1); the BglII-NcoI fragment containing the GS minimal promoter, first exon, 5′ and 3′ parts of the first intron, and the second exon till the beginning of the ORF (Fig. 4A).17, 34 The 305bp bovine growth-hormone polyadenylation sequence (the XbaI–PvuII fragment from pcDNA3 (Invitrogen)) was inserted between the XbaI and EcoRV sites in the polylinker downstream of the luciferase ORF.17 Construct 6A-bGH was made by deleting the AflIII-HindIII fragment from the upstream enhancer of the 6-bGH construct.
The pHNF4α expression plasmid that contained human HNF4α cDNA was kindly provided by Tanja Deurholt, AMC Liver Center, University of Amsterdam.
DNA Transfection and Luciferase Assay.
Plasmids used for transfection were isolated using Nucleobond G-500 columns (Macherey-Nagel). COS-1 cells were grown to 90% confluence in 6-well culture plates. Polyethylenimine (PEI) transfection was performed as described.35 Cells were transfected with 2.9 μg of 6-bGH or 6A-bGH, 0.1 μg of the Renilla luciferase expression vector pRL-CMV (Promega), 0.1 μg of pHNF4α, and 0.9 μg of plasmid pBS (Promega). Culture medium was refreshed 24 hours after transfection. The cells were lysed 48 hours after transfection in “Passive Lysis Buffer” (Promega). Both luciferase activities were measured with the Dual-Luciferase Reporter-Assay System (Promega) and the AutoLumat plus LB953 luminometer (Berthold Technologies). The experiment was performed 3 times in duplicate. Data are presented as average ± SEM. A t test was applied to test HNF4α-dependent differences between 6-bGH and 6A-bGH construct. P less than 0.05 was considered significant.
Immunohistochemistry and in Situ Hybridization.
Adult mouse livers were dissected, fixed overnight in 4% formaldehyde, embedded in paraffin and sectioned at 7 μm thickness. The sections were deparaffinized, hydrated in graded alcohols, heated for 10 min at 120°C, 1kPa in 10 mM Na-citrate (pH 6.0) to retrieve antigens, blocked in Teng-T (10 mM Tris (pH 8.0), 5 mM EDTA, 150 mM NaCl, 0.25% (w/v) gelatin, 0.05% (v/v) Tween-20) and 10% FCS, and incubated overnight with the first antibody. The following antibodies were used: monoclonal anti-rat GS - 1:1,500 (Transduction Laboratories, Lexington, KY), goat anti-human HNF4α - 1:250 (Santa Cruz Biotechnology), rabbit anti-PEPCK - 1:500,20 rabbit anti-OAT – 1:2,000,36 rabbit anti-human TRβ - 1:250,37 rabbit anti-Cyp3a4 1:500 (RDI, Concord, MA), monoclonal OV6 1:250,38, 39 monoclonal anti-β-catenin 1:50 (BD Transduction Laboratories, Bradford, MA), monoclonal anti-PCNA (PC10) 1:1,000 (Santa Cruz Biotechnology). Antibody binding was visualized using ALEXA488- or ALEXA568-labeled (Molecular Probes, Breda, The Netherlands) or alkaline phosphatase-labeled antibodies (Sigma-Aldrich, Zwijndrecht, The Netherlands). Alkaline phosphatase was visualized using nitroblue tetrazolium and 5-bromo-4-choloro-3-indolyl-phosphate (NBT/BCIP 1:50) as substrates (Roche Molecular Biochemicals, Woerden, The Netherlands). For comparisons between H4Flox and H4LivKO sections staining reactions were performed simultaneously and for an equal time period. Fluorescence was analyzed with a BioRad MRC1024 (BioRad, Veenendal, The Netherlands) confocal laser-scanning microscope or a Leica DM RA2 microscope (Rijswijk, The Netherlands).
For in situ hybridization, 4 adjacent sections were stained with digoxigenin (DIG)-labelled complementary RNA probes of glutamine synthetase (GS) and PEPCK.40 Probe bound to the section was immunologically detected by antidigoxigenin Fab fragment covalently coupled to alkaline phosphatase. NBT/BCIP was used as chromogenic substrate.41
In Vitro Footprinting of the GS Upstream Enhancer.
The binding of transcription factors that are present in liver nuclear extracts to the upstream enhancer region of GS was visualized by DNase I footprinting. The sequence tested and primers used for generating the probes are presented in Fig. 1. On both strands, liver nuclear extract decreased DNase I accessibility to specific regions of the DNA template (Fig. 2, nt numbering according to Fig. 1). When the upper strand (F1 primer) was labeled, two footprints and five hypersensitive bands were detected (Fig. 2A). When the lower strand was labeled (R1 primer), binding of the nuclear extract protected four regions and increased sensitivity to DNase I at two positions (Fig. 2B). The protection from position 105 to 130 in the upper strand overlapped with the protected area from nt 105 to 125 and a hypersensitivity at position 105 on the lower strand. The footprint between nt 150 and 210 on the upper strand corresponded with the protection between nt 150 and 220, and the hypersensitivity at position 150 on the lower strand. Proteins causing the hypersensitivities at positions 50, 70 and 85 on the upper strand left a footprint in the region between nt 50 and 100 on the lower strand. Finally, the hypersensitivities at positions 25 and 35 on the upper strand were accompanied by a protection between nt 20 and 30 on the lower strand.
HNF4α Binds to the GS Upstream Enhancer.
According to the MatInspector Professional database,27 the region containing hypersensitivities at position 25 and 35 on the upper strand and a protection between nt 20 and 30 on the lower strand coincides with a potential HNF4α-binding site. Electromobility shift assays, employing rat-liver nuclear extract, were performed to test whether HNF4α indeed recognizes the proposed binding site (GGGGGCaAAGCTT). Addition of nuclear extract to the HNF4 probe resulted in three discrete shifts (Fig. 3A, compare lanes 1 and 2). When HNF4α antibody was added to the reaction mixture, the lowest band disappeared and two new bands with a lower mobility appeared (Fig. 3A, lane 3, bands 4 and 5). Addition of unlabelled intact probe in 10-, 100- and 1000-fold excess (Fig. 3A, lanes 4, 5 and 6, respectively) resulted in a gradual disappearance and finally complete loss of all three shifted bands, whereas a 1000-fold excess of the mutated HNF4 oligo did not influence the shifting of the intact HNF4 oligo (Fig. 3A, lane 7). Nuclear extract and antibodies did neither shift nor supershift the mutated HNF4 oligo (Fig. 3B, lanes 1-3). These results confirmed that HNF4α can bind to the upstream-enhancer sequence at position −2658 to −2645.
HNF4α Represses GS Expression in Transient Transfections.
To determine whether HNF4α directly regulates the expression of GS, we transfected COS-1 cells with an HNF4α expression plasmid and either plasmid 6-bGH (containing the putative HNF4α-response element) or 6A-bGH (lacking the putative HNF4α-response element). Luciferase reporter-gene activity of 6-bGH amounted to 45% of that of 6A-bGH (P < 0.05), showing that HNF4α suppresses GS expression via its binding site in the GS upstream enhancer element.
HNF4α Binds the GS Upstream Enhancer in Vivo.
To verify the in vivo relevance of HNF4α for the regulation of GS gene expression in the context of chromatin, we performed chromatin-immunoprecipitation assays using soluble cross-linked liver chromatin from a liver-specific HNF4α knockout mouse (H4LivKO) and its H4Flox control.13 As shown in Fig. 5, an antibody raised against HNF4α could efficiently immunoprecipitate GS upstream-enhancer DNA, when wild type liver (H4Flox) was used, whereas this was not the case for H4LivKO liver. Additionally, in H4Flox liver, we could demonstrate the presence of histone deacetylase 1 (HDAC1) on the GS upstream enhancer, which was lost in H4LivKO liver (Fig. 5). In contrast, antibodies directed against SMRT, HDAC3 and SHP did not precipitate the GS upstream enhancer (not shown). These findings indicate that the regulation of GS expression in hepatocytes involves HNF4α-dependent recruitment of HDAC1-containing co-repressor complexes.
Periportal Expression of GS in H4LivKO Livers.
To further demonstrate how HNF4α regulates GS expression in vivo, in situ hybridization was performed on H4Flox and H4LivKO liver. The histology of these livers has been reported.13 The control H4Flox liver exhibited the expected pericentral expression of GS mRNA, whereas no GS mRNA could be detected in periportal hepatocytes (Fig. 6A). In contrast, GS was present in almost all hepatocytes of a H4LivKO liver (Fig. 6C), albeit that the ectopic periportal expression of GS was less intense than the expression around the central veins. Furthermore, fewer pericentral hepatocytes expressed high-GS levels in H4LivKO liver than in H4Flox liver and at a lower level. Immunostaining confirmed the exclusive presence of GS protein in the pericentral hepatocytes of control H4Flox liver (Fig. 7A,B). As described for GS mRNA, GS protein was present in almost all hepatocytes of a H4LivKO liver, albeit that its concentration in the periportal hepatocytes was lower than that in the pericentral hepatocytes (Fig. 7C). GS-negative cells were either present as single, large hepatocytes (Fig. 7E,F) or in groups near portal veins (Fig. 7D). The presence of GS mRNA and protein in the periportal zone of the H4LivKO liver shows that HNF4α normally suppresses GS expression in periportal hepatocytes, whereas the decreasd expression of GS in the pericentral zone of the H4LivKO liver shows that HNF4α stimulates its expression in pericentral hepatocytes.
Periportal HNF4α-Positive Hepatocytes in H4LivKO Livers Do Not Express GS.
Staining for HNF4α of control and H4LivKO livers explained the nature of the GS-negative cells in Fig. 7C–F. In the control liver, HNF4α showed a nuclear localization in all hepatocytes (Fig. 8A,B). As expected, H4LivKO liver contained only few HNF4α-positive nuclei (Fig. 8C–F). Immunostaining of serial sections revealed that the HNF4α-positive cells in H4LivKO liver did not contain detectable levels of GS, confirming that HNF4α suppresses GS expression in hepatocytes with a periportal phenotype (compare Fig. 7D,E and Fig. 8D,E). Additionally, we found several clusters of very small cells near portal veins that were negative for both GS and HNF4α (Fig. 7F and Fig. 8F, arrows).
HNF4α Deficiency Also Releases Periportal Expression of OAT and TRβ.
Similar to GS, ornithine aminotransferase (OAT) and thyroid hormone receptor β (TRβ) are normally expressed in a narrow pericentral zone of hepatocytes (Fig. 9A,B). Again similar to GS, they become ectopically expressed in the periportal zone of H4LivKO livers (Fig. 9D,E). This finding indicates that HNF4α is a repressor of periportal expression for several genes that are normally pericentrally expressed.
Deletion of the HNF4α Gene Impairs Periportal PEPCK Expression.
Periportal PEPCK mRNA (Fig. 6B,D) and protein expression (Fig. 10A-E) in the H4LivKO liver was weaker than that in the control H4Flox liver. The comparison with the staining intensity in the few remaining HNF4α-positive hepatocytes underscores this argument (Fig. 8D–F and Fig. 10D–F, black arrows). These observations are in agreement with earlier findings that HNF4α stimulates PEPCK expression.42, 43 Like HNF4α-positive hepatocytes, HNF4α-negative periportal hepatocytes stained stronger for PEPCK than the corresponding pericentral hepatocytes (Fig. 8D,E and Fig. 10D,E, white arrows), suggesting that the extracellular signals that conferred zonation on PEPCK expression remained intact in H4LivKO livers. Interestingly, the very small HNF4α- and GS-negative periportal hepatocytes (Fig. 7F; Fig. 8F) were positive for PEPCK (Fig. 10F). Taken together, these results imply that HNF4α does not influence porto-central gradient of PEPCK expression, but strongly affects the level of its expression.
In contrast to PEPCK, carbamoylphosphate synthetase (CPS) is expressed in a very wide periportal area and is only absent in the GS-positive pericentral hepatocytes (Fig. 9C). The pattern of CPS expression was not disturbed by HNF4α deletion and the hepatocytes around the central vein remained CPS-negative, indicating that periportal CPS expression is not regulated by HNF4α (Fig. 9F). Additionally, less intense CPS staining in H4LivKO liver than in control H4Flox liver is in agreement with earlier findings.22
HNF4α Deficiency Induces DNA Synthesis and Regeneration.
Proliferating-cell nuclear antigen (PCNA) is a marker for DNA-synthesizing cells.44 In control liver, some cells adjacent to central veins had weakly positive nuclei, whereas the rest of the liver was PCNA-negative (Fig. 11A,B). In H4LivKO liver, however, the majority of hepatocytes were weakly PCNA-positive (Fig. 11C,D), suggesting that HNF4α deficiency activates the expression of enzymes involved in DNA synthesis in hepatocytes.
The OV-6 antibody, which recognizes cytokeratin-14 and -19, is a marker for oval or hepatocyte-progenitor cells.38 OV6-positive cells were not detected in control liver (Fig. 11A,B), but nests of very small (∼10 μm diameter), OV-6-positive cells were detected in H4LivKO liver (Fig. 11C,D; Fig. 7F; Fig. 8F; Fig. 10F). In agreement with earlier findings,12, 14, 15 these very small cells did not express HNF4α, but were strongly positive for PCNA. The HNF4α-negative, OV6- and PCNA-positive cells were contiguous with streaks of brightly HNF4α-positive, OV6- and PCNA-dim cells (Fig. 11C,D). This latter group of HNF4α-positive cells weakly expresses PEPCK, but not GS (Fig. 7D; Fig. 8D; Fig. 10D). The near absence of OV6 expression, the bright HNF4α expression and the emerging PEPCK expression identify these cells as newly differentiated hepatocytes45, 46 that do not (yet) have their HNF4α alleles excised in the knockout mice. Together, these findings show that, in contrast to control mice, liver regeneration from oval cells is activated in HNF4α-deficient mice.
HNF4α Suppresses the Expression of a Subset of Pericentrally Expressed Enzymes in the Periportal Zone.
Our study reports HNF4α binding to the glutamine synthetase upstream enhancer and suppression of GS expression in the periportal areas. Although we did not demonstrate HNF4α binding to the TRβ and OAT regulatory regions, the similarity in the change of their expression pattern in the H4LivKO liver suggests that HNF4α suppresses the expression of these and possibly other pericentral genes in the periportal region through a common mechanism. Furthermore, the expression of the periportal enzyme PEPCK was reduced in the H4LivKO liver. However, even though expression of GS, OAT and TRβ was induced periportally and that of PEPCK reduced, HNF4α deletion did not eliminate the porto-central gradient of expression of these genes. These findings demonstrate that HNF4α enhances the expression of PEPCK and other periportal genes (e.g., glucose-6-phosphatase21) and inhibits the expression of several pericentrally expressed genes in the periportal areas.
Northern- and Western-blot quantification of gene expression in the H4LivKO mouse has revealed altered expression levels for many liver-specific genes and an impaired lipid, protein and glycogen metabolism.13, 22 These biochemical studies did not report the residual HNF4α expression in isolated hepatocytes and periportal clusters of small cells. In apparent contrast to our findings, it was also reported that GS mRNA levels were not different between HNF4α-null and control mice.22 Interestingly, our in situ hybridization showed that the GS mRNA concentration in the pericentral area was lower in HNF4α knockout than in control liver. We therefore conclude that the absence of an overall change in GS mRNA levels results from an increase in GS mRNA concentration in the periportal hepatocytes and a compensating decrease in GS mRNA in the pericentral hepatocytes. Our data also show that histochemical techniques are a very useful complement to biochemical techniques, as it produces expression data at the single-cell level and allows observation of changes in the distribution of gene expression. In this respect, the few hepatocytes in which the cre enzyme did not (yet) excise both HNF4α alleles,13 provided valuable tissue-intrinsic controls for the effects of HNF4α deficiency on the level of expression of the genes we studied. Albumin-cre mice express cre recombinase in their hepatocytes from neonatal day 10 onward.26 However, it takes until the animals are 6 weeks old before cre is expressed throughout the liver and HNF4α in the H4LivKO liver is decreased to non-detectable levels.13, 26 The presence of HNF4α in the small, intensely PCNA-positive cells can be understood because these cells do not (yet) express hepatocyte-specific genes and, hence, do not express cre.
Although the similarity in the behavior of GS, OAT and TRβ in H4LivKO livers suggests a common mechanism of periportal suppression for pericentrally expressed genes by HNF4α, we also observed that Cyp3A4 expression, which is expressed in a wide pericentral zone, was not influenced by the HNF4α deletion (unpublished observation). Furthermore, another member of the cytochrome P450 family of enzymes, Cyp7a, which, like GS, OAT and TRβ, is characterized by a very narrow pericentral expression pattern,23 is reportedly down-regulated rather than up-regulated in the HNF4α-deficient livers.13, 47 A similar dichotomy appears to exist with respect to periportal enzymes: some enzymes, like PEPCK and ornithine transcarbamoylase20, 24 are up-regulated by HNF4α, whereas others, including CPS, are not affected. These data therefore suggest that HNF4α is not a prime determinant of the porto-central gradient in gene expression in the liver, but affects a subgroup of hepatocyte-specific genes. In agreement, we noticed that the porto-central gradient in PEPCK expression was also not affected by HNF4α deficiency.
Mechanism of HNF4α-Mediated Periportal Suppression of Gene Expression.
HNF4α is present on approximately 12% of the promoters in liver, 80% of which are also occupied by RNA polymerase II,4 that is, are putatively transcribed. The common determinant of the subgroup of pericentral enzymes, of which expression in the periportal zone becomes released by HNF4α deficiency, remains to be determined.
Our results show that HNF4α binds GS upstream enhancer and suppresses GS expression in the periportal region. The most suggestive proof for a direct mechanism can be found in the immunostainings of H4livKO livers, in which the few remaining HNF4α-positive hepatocytes in the periportal region do not express GS. These HNF4α-positive hepatocytes show a cell-autonomous HNF4α effect on GS expression in the periportal region and do not support an indirect effect due to the overall changed physiology of the H4livKO liver. HNF4α is mostly reported to activate transcription,7, 48 but it can also act as a suppressor.6, 49 Our finding that histone deacetylase 1 (HDAC1) recruitment to the upstream region of GS is dependent on the presence of HNF4α (Fig. 5) suggests that HNF4α-promoted chromatin condensation could well be a potential mechanism. The orphan nuclear receptor SHP, which can directly interact with HDAC1 and the histone-methylating enzyme G9a methyltransferase,50 has been reported to mediate HNF4α repression of gene expression.49 The co-repressor SMRT (silencing mediator of retinoid and thyroid receptors), which interacts with histone deacetylase 3 (HDAC3), can also mediate the HNF4α-promoted suppression of gene expression6, 51 and promotes chromatin condensation.52 Since the ChIP assays with SMRT, HDAC3 and SHP antibodies did not precipitate the GS upstream enhancer, we propose the HNF4α/HDAC1-promoted chromatin condensation, acting via a yet unidentified mediator, as a potential mechanism of the periportal suppression of the GS gene.
HNF4α Deficiency Activates DNA Synthesis and Induces Liver Regeneration from Oval Cells.
We observed an enhanced expression of the DNA-synthesis marker PCNA53 in all HNF4α-deficient hepatocytes. We also observed a very strong PCNA expression in nests of very small (∼10 μm) OV6-positive, HNF4α-negative cells near portal tracts that did not express hepatocyte-specific enzymes. This phenotype identifies these cells as oval cells, i.e., hepatocyte-progenitor cells.54 The spatial continuity of these cells with streaks of OV6- and PCNA-dim, but brightly HNF4α-positive and weakly PEPCK-positive cells implies a steady production of hepatocytes from these hepatic stem cells.13, 55, 56 Altogether, these findings reveal the presence of regenerative foci and suggest de-repression of DNA synthesis in HNF4α-deficient hepatocytes. The regenerative stimulus is unknown, but it is possible that liver failure underlies the very limited life span of liver HNF4α-deficient mice.13, 22
We wish to thank Dr. F.J. Gonzalez (Laboratory of Metabolism, National Cancer Institute, National Institutes of Health, Bethesda, Maryland) for providing the HNF4Flox animals; Dr. T. Matsuzawa (Department of Biochemistry, School of Medicine, Fujita-Gakuen Health University, Toyoake, Japan), Dr. S. Sell (Department of Pathology and Laboratory Medicine, Division of Experimental Pathology, Albany, NY), Dr. B. Christ (Martin Luther Universität Halle-Wittenberg, Halle, Germany) and Dr. O. Bakker (Department of Endocrinology and Metabolism, AMC, University of Amsterdam, Amsterdam, The Netherlands) for their gifts of antiserum against ornithine aminotransferase, OV-6, PEPCK and thyroid-hormone receptor β, respectively.
- 28A Laboratory Guide to in vitro Transcription. Biomethods. Basel: Birkhauser Verlag, 1990..