Regulation of a Transcription Factor Network Required for Differentiation and Metabolism

See allHide authors and affiliations

Science  31 Jul 1998:
Vol. 281, Issue 5377, pp. 692-695
DOI: 10.1126/science.281.5377.692


Hepatocyte nuclear factors (HNFs) are a heterogeneous class of evolutionarily conserved transcription factors that are required for cellular differentiation and metabolism. Mutations in HNF-1αand HNF-4α genes impair insulin secretion and cause type 2 diabetes. Regulation of HNF-4/HNF-1 expression byHNF-3α and HNF-3β was studied in embryoid bodies in which one or both HNF-3α or HNF-3βalleles were inactivated. HNF-3β positively regulated the expression of HNF-4α/HNF-1α and their downstream targets, implicating a role in diabetes. HNF-3β was also necessary for expression of HNF-3α. In contrast, HNF-3α acts as a negative regulator of HNF-4α/HNF-1α demonstrating thatHNF-3α and HNF-3β have antagonistic transcriptional regulatory functions in vivo. HNF-3α does not appear to act as a classic biochemical repressor but rather exerts its negative effect by competing for HNF-3 binding sites with the more efficient activatorHNF-3β. In addition, the HNF-3α/HNF-3β ratio is modulated by the presence of insulin, providing evidence that the HNF network may have important roles in mediating the action of insulin.

Hepatocyte-specific gene expression is controlled primarily at a transcriptional level and relies on the activity of multiple transcription factors including HNF-1, CCAAT/enhancer binding protein (C/EBP), HNF-3, HNF-4, and HNF-6 (1). Our current understanding of transcriptional regulation by these HNFs has been derived primarily from analysis of promoter/enhancer elements of genes selectively expressed in cultured cells with transient transfections (1). Although these approaches have provided useful information about tissue-specific regulation of gene expression, it is important to note that such procedures do not measure transcriptional regulation within a native chromosomal context. This is significant, given that HNF-3 proteins have been shown to modify the nucleosomal organization of the albumin enhancer, a finding consistent with the structure of the HNF-3 DNA binding domain, which is highly similar to that of histone H5 (2). Analyses of specific promoter elements indicate that HNFs act in various combinations to direct cell-specific transcription during cellular differentiation (3). Targeted disruption of transcription factors in mice often results in only moderate reductions in target gene expression, which supports the hypothesis that transcription factors act cooperatively. In contrast to this, we have recently shown that disruption of HNF-4α drastically reduces the expression of numerous target genes, implying that it acts as a central regulator of tissue-specific gene expression (4). In addition, earlier studies had shown that HNF-4α also positively regulated expression of HNF-1α, defining a transcriptional hierarchy involved in maintaining the hepatic phenotype (5). We predicted that factors that control the expression of HNF-4α could have critical functions in directing cell differentiation. Moreover, the recent finding that mutations in the genes encoding HNF-4α and HNF-1α are responsible for two phenotypically indistinguishable forms of early-onset type 2 diabetes, MODY1 and MODY3, respectively, suggested that such upstream regulators of HNF-4α expression could also be necessary for normal pancreatic β cell function and metabolism (6).

Evolutionarily conserved HNF-3 binding sites have previously been identified in a portion of the HNF-4α promoter that is sufficient to drive tissue-specific expression of a transgene in mice (7).HNF-3β and HNF-4α null mice are embryonic lethal and therefore cannot be used for gene expression studies (8). To test whether HNF-3 regulated the expression ofHNF-4α, we therefore generatedHNF-3α and HNF-3β null embryonic stem (ES) cells. ES cells grown in suspension differentiate to form visceral endoderm (VE) that expresses HNF-3α and -3β, as well asHNF-4α and HNF-1α and their target genes. We have previously shown that the regulation of gene expression in the VE directly parallels that in the liver (4). The VE, therefore, provides a suitable system for studying transcriptional regulation of metabolism in the liver and pancreas because they all share expression of the HNF transcription factors (9). To study HNF-3–dependent gene expression, we generated HNF-3α andHNF-3β null ES cells. The strategy for disruption ofHNF-3β in ES cells has been described (8). TheHNF-3α gene was mutated by targeted deletion of the DNA binding domain and the COOH-terminal transcriptional transactivation domains by homologous recombination in ES cells (10). In addition, the Escherichia coli LacZ gene was inserted into exon 2 of HNF-3α, producing a gene fusion that allowed us to follow gene expression from the HNF-3α promoter in embryoid bodies (EBs) and embryos (Fig. 1). Three independent HNF-3α−/− lines were then produced by culturing the HNF-3α+/– ES cells in high concentrations of G418 medium, as described (11).

Figure 1

(A) Expression of LacZ inHNF-3α –/– ES cell EBs. HNF-3α –/– EBs were cultured for 14 days in suspension and then stained for β-galactosidase activity. LacZ expression is restricted to the VE (arrow). (B) HNF-3α –/– ES cell–derived embryos were produced by tetraploid aggregation as described in (16). Embryos were harvested at E8.5 and stained for expression of lacZ. All embryos were indistinguishable from wild types and exhibited β-galactosidase activity in the floorplate of the neural tube and notochord (F/N) as well as gut (G) and developing liver (L).

Steady-state mRNA concentrations of numerous genes expressed in the VE were measured in 21-day-old HNF-3α +/+, +/–, and –/– EBs by reverse transcription polymerase chain reaction (RT-PCR) (Fig. 2). Hypoxanthine phosphoribosyltransferase (HPRT) primers were used to show that each sample contained equivalent amounts of mRNA. Furthermore, mRNA concentrations of VE markers GATA-4 and HNF-1β (12) were similar, demonstrating that each sample contained equal amounts of VE. As expected, no HNF-3α product was amplified from HNF-3α–/– EBs. Surprisingly, mRNAs for apolipoproteins AI, AII, AIV, B, and CII; albumin; glucose transporter 2; and the glycolytic enzymes aldolase B and l-pyruvate kinase, which are putative targets for HNF-3α, were upregulated three- to eightfold inHNF-3α –/– EBs. In addition, mRNA concentrations of HNF-4α and HNF-1α were increased, whereas HNF-3β and -3γ remained relatively unaffected. LacZ expression was similar in HNF-3α +/– and HNF-3α –/– EBs, ruling out any significant transcriptional autoregulation by HNF-3α (13). These results demonstrate that, in a native chromosomal context, HNF-3α functions as a repressor of gene expression, including the expression of transcription factors HNF-4α and HNF-1α.

Figure 2

HNF-3α–dependent gene expression in EBs of wild-type, heterozygous, and HNF-3α null ES cells.HNF-3α +/+ (R1; lane 1), HNF-3α +/– (A13 and A7; lanes 2 and 6), and HNF-3α –/– (A8, A10, and A11; lanes 3 to 5) were assayed for expression of mRNAs from HPRT; GATA-4; HNF-1α, -3α, -3β, -3γ, and -4α; albumin (Alb); apolipoproteins (APo) AI, AII, AIV, B, and CII; aldolase B (Aldo-B), and l-pyruvate kinase (L-PK). HPRT primers were used to show that each sample started with comparable amounts of cDNA (HPRT +RT) and that no product was amplified in the absence of reverse transcriptase (HPRT –RT). Each EB sample contained similar amounts of VE as shown by the comparable amounts of product amplified by GATA-4 primers. RT-PCR was performed as described in (4).

HNF-3α and -3β proteins bind to common DNA recognition sequences as monomers through a conserved DNA binding domain known as the winged helix motif (14). HNF-3α andHNF-3β are sequentially expressed during development in the definitive endoderm, notochord, and floorplate of the neural tube with HNF-3β being expressed first (15). Targeted disruption of HNF-3β produces embryos that lack a notochord and exhibit defects in foregut and neural tube development (8). Because HNF-3α and HNF-3β are coexpressed in these tissues and display similar DNA binding properties, it appeared possible that HNF-3α and HNF-3β have similar functions during early development. To assess whether loss ofHNF-3α would lead to developmental defects similar to those that have been described for HNF-3β null embryos, we generated embryonic day 8.5 (E8.5) HNF-3α –/– embryos by tetraploid aggregation and followed the development of notochord, floorplate of the neural tube, and gut by staining for LacZexpression (16). As shown in Fig. 1B, in contrast toHNF-3β −/− embryos (17), HNF-3α–/– embryos were developmentally indistinguishable from wild-type embryos.

The distinct phenotypes shown by HNF-3α –/– and-3β –/– embryos suggested that HNF-3α and HNF-3β regulated gene expression by different mechanisms. This led us to investigate target gene expression in HNF-3β –/– EBs.HNF-3β null ES cells were generated by two targeting strategies (18). Expression of HNF-4α was markedly reduced in HNF-3β −/−, as were several genes known to be in vivo targets of HNF-4α (Fig. 3). These genes included apolipoproteins AI, AII, AIV, B, CII, and CIII; transthyretin; aldolase B; andl–pyruvate kinase. Expression of HNF-1α was also reduced in the absence of HNF-3β. Strikingly,HNF-3α mRNA was undetectable in HNF-3β null EBs, implying that HNF-3β is absolutely required forHNF-3α expression in EBs. Steady-state levels ofHNF-3γ mRNA were not affected.

Figure 3

HNF-3β–dependent gene expression in EBs of wild-type, heterozygous, and HNF-3β null ES cells. Steady-state mRNA concentrations of putative target genes were measured by RT-PCR (Fig. 2) (4). TTR, Transthyretin.

Our genetic analyses showed that whereas HNF-3α acted as a negative regulator, HNF-3β was a strong activator of the same target genes. This could suggest that HNF-3α acts as a transcriptional repressor of gene expression in a native chromosomal context. Alternatively, because HNF-3α and HNF-3β interact with identical consensus binding sites as monomers, the possibility was raised that HNF-3α could interfere with the ability of HNF-3β to transactivate by competing for HNF-3 binding sites. To distinguish between these models we expressed HNF-3α in HNF-3β –/– EBs and asked whether HNF-3α could restore expression of target genes; if HNF-3α acted as a direct repressor, it should be incapable of achieving this. EBs were generated from wild-type, HNF-3β –/–, and three independent HNF-3β –/– ES cell lines that expressed a rat HNF-3α DNA from the immediate early cytomegalovirus (CMV) promoter (Fig. 4; α3, α4, and α5). The mRNA levels of aldolase B and apolipoprotein CII were restored in HNF-3β –/– EBs that expressedHNF-3α (Fig. 4). Furthermore, this expression depended on the amount of HNF-3α. From this we conclude that both HNF-3α and HNF-3β are activators of gene expression and that in the context of native chromatin HNF-3α is a poorer activator than HNF-3β. Because both of these factors recognize the same cis-acting promoter elements, we propose that HNF-3α inhibits HNF-3β activity by competing for common HNF-3 binding sites.

Figure 4

Expression of HNF-3α inHNF-3β –/– EBs. A transgene expressing the ratHNF-3α cDNA under the control of CMV promoter and containing the hygromycin resistance gene (hyg R) was transfected intoHNF-3β –/– ES cell line B14. Transfectants were grown in the presence of hygromycin (0.4 mg/ml) for 8 days and three resistant lines were used to generate EBs. Steady-state mRNA concentrations were measured by RT-PCR in 14-day-old EBs. Concentrations ofHNF-3α mRNA were moderate in lines α3 and α5 and highest in line α4. The rescue of HNF-3α expression inHNF-3β –/– EBs led to dose-dependent transcriptional activation of the target genes apoCII andaldo-B.

The strong regulation of metabolically controlled genes such as enzymes of glycolysis and gluconeogenesis by HNF-3 and HNF-4 led us to investigate whether HNF-3α/HNF-3β ratios were affected by insulin (19). We measured steady-state concentrations ofHNF-3α and HNF-3β mRNAs in day14 EBs that were cultured for 24 hours in serum-free medium containing 0, 5, or 50 nM insulin and 20 mM glucose. In the presence of insulin,HNF-3β expression was upregulated, whereasHNF–3α was reduced (Fig. 5). Moreover, HNF-4α expression and downstream targets, such as aldolase B andl–pyruvate kinase, were also significantly increased. We conclude that insulin can act as a positive modulator of the HNF transcription factor network.

Figure 5

Insulin regulates expression of HNF-3α, HNF-3β, and downstream target genes. Day 14 EBs were grown in 2.5% fetal calf serum for 12 hours followed by serum-free Dulbecco's minimum essential medium overnight. EBs were then cultured in serum-free medium containing 20 mM glucose and 0, 5, or 50 mM insulin for 24 hours. Steady-state mRNA concentrations of HPRT, GATA-4, HNF-3α, HNF-3β, HNF-4α, Aldo-B, and L-PK mRNA were determined by RT-PCR.

Placement of HNF-3β at the top of the transcription factor hierarchy regulating HNF-4α and HNF-3α is consistent with it being expressed at the earliest stages of fetal development before expression of both HNF-3α andHNF-4α (15). Moreover, whereasHNF-3β –/– embryos exhibit severe defects early in embryogenesis, HNF-3α −/− embryos are normal at the same developmental stage. Further evidence supporting regulation ofHNF-3α by HNF-3β comes from analyses of the HNF-3α promoter, which was found to contain an HNF-3 binding site (20). However, unlike the VE, the definitive endoderm expresses HNF-3α in HNF-3β –/– embryos, suggesting that other transcription factors can compensate for loss ofHNF-3β in this tissue (8).

Mutations in the genes encoding HNF-4α and HNF-1α have been identified in families with an early-onset form of type 2 diabetes (MODY1 and -3, respectively), which is characterized by autosomal-dominant inheritance and defects in insulin secretion (6). It is likely that the MODY phenotype results from a defect in the expression of genes involved in glucose metabolism (4). Our data indicate that the positive and negative impact of HNF-3β and HNF-3α on the expression ofHNF-4α/HNF-1α and their downstream targets are important components of this complex hierarchical circuit regulating pancreatic β cell function. Genetic variation in the genes encoding HNF-3α and HNF-3β may be responsible for other forms of MODY. Moreover, our findings that this pathway is also regulated by insulin implies that this cascade is also important for insulin action. Interestingly, the insulin receptor–like DAF-2 gene inCaenorhabditis elegans has recently been shown to be an upstream regulator of the HNF-3/forkhead homologue DAF-16 in metabolic control of the dauer larval stage, which suggests that this pathway may be evolutionarily conserved (21). Because resistance to insulin is a common feature of late-onset non-insulin-dependent diabetes mellitus, it is possible that dysregulation of the HNF regulatory pathway, whether primary or secondary, can also contribute to this complex metabolic syndrome.

  • * Present address: Department of Cell Biology and Anatomy, Medical College of Wisconsin, Milwaukee, WI 53226, USA.

  • To whom correspondence should be addressed. E-mail: stoffel{at}


View Abstract

Navigate This Article