In Vitro Transcriptional Induction of the Human Apolipoprotein A-II Gene by Glucose

Transgenic mice were producing human apoA-II at physiological plasma concentrations (line β) (9) and at three times higher levels (line λ) (8). Male β-transgenic mice were used for fasting/refeeding studies. The fasted mice (n = 7) were deprived of food for 36 h, and the fasted/refed mice (n = 9) were fasted for 36 h and then refed for 18 h with the standard carbohydrate-rich diet (51.2% carbohydrate, 23.5% protein, and 5% lipid by weight). For assessment of the correlation between blood glucose and plasma human apoA-II levels, male and female β- (n = 20) and λ-transgenic (n = 24) mice were fed ad libitum with the diet. Mice were anesthetized by intraperitoneal injection of avertine (0.25 mg/g) and were killed between 9 and 11 a.m. Blood was collected onto EDTA, and plasma was isolated. Blood glucose was determined with the Glucotrend plus test (Roche Diagnostics, Mannheim, Germany). Human apoA-II was assayed in plasma by immunonephelometry, using a specific antibody (Dade Behring), as previously described (13). Livers were removed and immediately frozen in liquid N₂ for subsequent RNA analysis. For the glucose tolerance test, male and female β- and λ-transgenic mice and control C57Bl/6J mice (four animals in each group) were fasted overnight. Animals were anesthetized by intraperitoneal injection of 0.15 mg/g Imalgene (Merial, Lyon, France). A bolus of glucose (2 mg/g) was injected intraperitoneally and blood glucose measured from the tail vein at the indicated times. Animal care and the experimental procedures used in this study conformed to French guidelines for animal studies.

Human apoA-II transgenic mouse liver RNA and streptozotocin-induced diabetic rat liver RNA (n = 3), kindly given by I. Guillet-Deniau (14), were analyzed by RT real-time PCR (LightCycler System, Roche). Reverse transcription was performed with 1 μg RNA and Moloney murine leukemia virus reverse transcriptase (Gibco BRL). Human and rat apoA-II mRNA levels were determined with the SYBR Green I dye. We carried out 30 cycles of PCR in a SYBER Green buffer containing 3 mmol/l MgCl₂, as follows: 10 s of denaturation at 95°C and 10 s of annealing at 55°C for human apoA-II or 65°C for rat apoA-II and 10 s of extension at 72°C. The mouse coding strand primer was 5′CGCAGCAACTGTGCTACTCC3′, and the mouse noncoding strand primer was 5′CTCTCCACACATGGCTCCTT3′. The rat coding strand primer was 5′TTTAACTGACTATGGCAAGGATTT3′, and the rat noncoding strand primer was 5′CTGGTTTCTCCTCGGGGCTCATTA3′. As an internal control, we determined levels of 18S RNA with the Ribosomal RNA control kit (18S rRNA; Applied Biosystems).

Hepatocytes were isolated from fed male Wistar rats as previously described (15). Cells were plated at a density of 8 million cells per 100-mm diameter tissue culture plate for RNA analysis and 1 million cells per 60-mm diameter tissue culture plate for transfection experiments and were allowed to adhere for 4 h. For RNA analysis, the medium was replaced with M199 medium (5.5 mmol/l glucose) or M199 medium supplemented with 20 mmol/l glucose (25.5 mmol/l total glucose). The cells were incubated for a further 24 h. Total RNA was extracted from rat hepatocytes in the appropriate amount of RNAplus solution (Appligene) according to the manufacturer’s instructions. Rat hepatocyte RNA samples (15 μg) were analyzed by northern blotting, as previously described (8), and the membrane was hybridized with a rat apoA-II cDNA probe. An 18S probe was used as an internal standard. Autoradiographs were analyzed by densitometry, with densities expressed in arbitrary units. For transfection experiments, the medium was replaced with M199 medium. Then, 30 μg DOTAP liposomal transfection reagent (Boehringer Mannheim) and 5 μg plasmid DNA in 20 mmol/l Hepes, pH 7.4, were added to the medium for overnight transfection, according to the reagent manufacturer’s instructions. After transfection, the medium was removed and replaced with M199 medium or M199 medium supplemented with 20 mmol/l glucose. The cells were then incubated for a further 48 h. Experiments were performed in triplicate and repeated three times.

Human hepatoma HepG2 cells were plated at a density of 0.4 million cells per well in 6-well tissue culture plate. The medium used was Dulbecco’s modified Eagle’s medium (DMEM)-glutamax, 1 g/l glucose (5.5 mmol/l) (Gibco) supplemented with 10% FCS and 1% penicillin/streptomycin, and the cells were allowed to adhere overnight. HepG2 cells were transfected by incubation for 24 h with 1 μg of plasmid DNA in opti-MEM medium supplemented with 20 μg of lipofectin reagent (Gibco BRL) according to the manufacturer’s instructions. After transfection, the medium was replaced with DMEM without sodium pyruvate and glucose, supplemented with glutamine, 1% FCS, 1% penicillin/streptomycin, and the indicated concentration of glucose, and the cells were incubated for a further 24 h. Experiments were performed in triplicate and repeated three to five times.

Transfected rat hepatocytes or HepG2 cells were recovered and chloramphenicol acetyl transferase (CAT) activity (as counts per minute) was measured in a liquid phase assay, as previously described (16). At days 22 and 37, individual liver samples (n = 8) from transgenic mice expressing the CAT gene under the control of the (−911/+29) human apoA-II gene promoter were homogenized and assayed for CAT activity. The proportion of chloramphenicol converted to acetylated was determined by thin layer chromatography (17).

The plasmid containing the CAT gene under the control of the −911/+29 human apoA-II promoter (−911) and plasmids containing human apoA-II promoter constructs with deletions (−614, −440, −230, −80, and E-AB) have been described elsewhere (16,18,19). The plasmid encoding the dominant-negative form of HNF-4α, CD1b, was kindly provided by Dr. M. Hadzopoulou-Cladaras (Department of Biology, Aristotelian University of Thessaloniki, Thessaloniki, Greece) and has been described elsewhere (20). The mutants were constructed by inverse PCR with the Quick Exchange Mutagenesis kit (Stratagene) and sequenced. The M1 mutant was constructed from the −911/+29 plasmid with the coding strand oligonucleotide 5′-GGTGATCAAATGACCAGTAGCCTTCAACCTTTACCCTGG-3′ and the complementary noncoding strand oligonucleotide. The M2 mutant was constructed from the LM2 plasmid (19) with the coding strand oligonucleotide 5′-GAACTGATAAGGTGATCAAAATACCAGTAGCCTTCAACCTTTACCC-3′ and the complementary noncoding strand oligonucleotide. The M3 mutant has been described elsewhere (19) as Km1Lm2. The M4 mutant was constructed from the Km1Lm2 plasmid (19), with the coding strand oligonucleotide 5′-CCAGGTGCCTTCAACGACAGACCTGGTAGAAGCCTCTTATTCACC-3′ and the complementary noncoding strand oligonucleotide.

Results were expressed as means ± SEM. The statistical significance of differences was determined by unpaired t tests after ANOVA, performed with Excel software. The statistical significance of the correlation between plasma human apoA-II concentration and blood glucose and that of the glucose tolerance test in transgenic mice was calculated by two-way ANOVA, carried out with Graph-Pad Prism software.

We investigated the response of apoA-II gene expression to changes in diet by performing in vivo analysis on transgenic mice expressing human apoA-II at physiological levels (line β) (9). A refeeding with a high-carbohydrate diet induced an increase in hepatic mRNA level and plasma concentration of human apoA-II, 2.5- and 1.5-fold, respectively (Fig. 1A). Hepatic mRNA levels of the endogenous mouse genes Glut-2 and l-PK, which are known to be induced by carbohydrates, were increased by factors of 8 and 3.6, respectively, following refeeding (data not shown). Previous studies have reported that apoA-II gene expression is induced during weaning (21). Weaning is characterized by a switch from maternal milk to a high-carbohydrate diet. Moreover, in transgenic mice expressing the CAT reporter gene under the control of the human apoA-II (−911/+29) promoter, the CAT activity was stimulated three times at weaning (Fig. 1B). However, in streptozotocin-induced diabetic rats with hyperglycemia (14), rat apoA-II mRNA levels were not modified when compared with control rats (Fig. 1C).

Since in vivo studies do not allow the determination of the respective role of hormones and dietary substrates on apoA-II gene expression, we looked for a direct effect of glucose on apoA-II gene expression in primary rat hepatocytes. Rat apoA-II mRNA levels were twice as high in the presence of 25.5 mmol/l glucose than in the presence of 5.5 mmol/l glucose (Fig. 2A). This effect was not affected by insulin, as a similar doubling was observed in the absence of insulin and at all insulin concentrations tested. We transiently transfected rat hepatocytes with the CAT reporter gene under the control of the human (−911/+29) apoA-II promoter, which is sufficient to restrict CAT expression to the liver in vivo (17). The CAT reporter gene activity was three times higher in the presence of 25.5 mmol/l glucose than in the presence of 5.5 mmol/l glucose (Fig. 2B). In HepG2 cells, the CAT activity displayed 2-, 1.3-, and 3.7-fold induction by 5 mmol/l glucose with respect to 10 mmol/l lactate/1 mmol/l pyruvate, 1 mmol/l glucose, and no glucose (0 mmol/l), respectively (Fig. 2C). Increasing the concentration of glucose up to 20 mmol/l did not increase the level of induction. As observed with rat hepatocytes, induction of the human apoA-II promoter by glucose in HepG2 cells did not depend on the presence of insulin in the medium (data not shown). Thus, glucose itself stimulates transcription of the apoA-II gene, in vitro, independently of insulin action.

To determine the region(s) of the apoA-II promoter responsible for the response to glucose, we transiently transfected HepG2 cells with the CAT reporter gene under the control of various fragments of the human apoA-II promoter. Deletion of the (−911/–614) apoA-II enhancer abolished the response to glucose of the CAT reporter gene. When the (−911/–614) apoA-II enhancer was fused to the proximal promoter −80/+29 (E-AB), the CAT activity was induced twofold by glucose (Fig. 3B). Thus, the (−911/–614) apoA-II enhancer is required for activation of the CAT reporter gene by glucose.

We then attempted to identify the glucose response element of the apoA-II enhancer. A glucose response element or carbohydrate response element (ChoRE) has been identified in several glucose-regulated genes; the ChoRE consists of two E-boxes (CANNTG) separated by five nucleotides (11). The human apoA-II enhancer contains multiple E-boxes, located between −797 and −721 (Fig. 3A), which bind the transcription factors upstream stimulatory factors (16). To identify a ChoRE, we generated full-length −911/+29 mutants M1 to M3 (Fig. 3A), by inserting specific mutations into the E-boxes. Upstream stimulatory factor binding was impaired (data not shown), but these mutations did not abolish the induction by glucose of CAT reporter activity in transiently transfected HepG2 cells (Fig. 3C). Therefore, these multiple E-boxes, located between −797 and −735, did not behave as a ChoRE.

Optimization of the response of the l-PK gene to glucose stimulation is based on cooperation between the ChoRE and an HRE (22,23). The human apoA-II enhancer possesses an HRE located between −734 and −721 that binds members of the nuclear receptor family (24) and is contiguous to the multiple E-boxes (Fig. 3A). We therefore inserted an additional mutation into mutant M3 (to generate mutant M4, Fig. 3A) to impair HNF-4 binding to the HRE (data not shown). The CAT reporter activity of mutant M4 in transiently transfected HepG2 cells was not induced by glucose (Fig. 3C). We therefore investigated a possible role of HNF-4 in the effect of glucose on apoA-II gene transcription using a dominant-negative mutant of HNF-4α (20). We transiently transfected HepG2 cells with increasing amounts of the mutated form of HNF-4α, CD1b, and the CAT reporter gene under the control of the (−911/+29) human apoA-II promoter. The stimulation of CAT activity by glucose, arbitrarily fixed at 1, was inhibited dose-dependently by CD1b (Fig. 3D). Thus, either a mutation in the HRE or impairment of the transactivation activity of HNF-4α prevented the induction of human apoA-II promoter activity by glucose.

Transgenic mice overexpressing human apoA-II (line λ) display high plasma triglyceride and low HDL levels (8), a dyslipidemic profile similar to that observed in type 2 diabetes. However, glycemia had not yet been determined. To evaluate the pathophysiological relevance of our results, we therefore investigated the relationship between apoA-II gene expression and glucose in two lines of transgenic mice, line β, which produced human apoA-II at physiological concentration (9), and line λ, which overproduced three times the human apoA-II (8). Blood glucose was found to be strongly correlated (R² = 0.588, P < 0.0001) with the level of human apoA-II in transgenic mice of the β- and λ-lines (Fig. 4A). We investigated whether apoA-II was involved in blood glucose kinetics by carrying out a glucose tolerance test in these transgenic mice (Fig. 4B). Mice overproducing human apoA-II (line λ) displayed a marked delay in blood glucose clearance with respect to control mice (C57Bl/6J) and transgenic mice producing physiological levels of human apoA-II (line β).

In this study, we investigated the relationship between apoA-II and glucose levels. We found that glucose stimulated (in vitro) the hepatic transcription of the apoA-II gene. We also found that overexpression of the human apoA-II gene in transgenic mice was accompanied by a marked delay in glucose tolerance. These results are consistent with plasma concentrations of human apoA-II being positively correlated with glycemia in transgenic mice.

It has been shown that plasma apoA-II levels are controlled mainly by its rate of synthesis in the liver rather than its catabolism (25,26). Low plasma concentrations of apoA-II have recently been associated with a common polymorphism in the proximal promoter of the human apoA-II gene that leads to a decrease in apoA-II gene transcription (10). In the present study, we showed that the consumption of a high-carbohydrate diet activates the apoA-II gene transcription in vivo. By contrast, in streptozotocin-induced diabetic rats, the apoA-II mRNA levels remained unaffected despite hyperglycemia. This might be due to the effect of counter-regulatory hormones such as corticosteroids, which are elevated in streptozotocin-induced diabetic animals and are known to decrease apoA-II mRNA levels (27). However, in primary rat hepatocytes and HepG2 cells, glucose stimulated the apoA-II transcription independently of insulin action. The response to glucose required both the enhancer (−911/–614) and the promoter (−80/+29), a region that restricted the human apoA-II gene expression to liver cells (16–19,28). The glucose effect was prevented only when the HRE within the enhancer was mutated and when the active form of HNF-4α is inhibited. HNF-4α has not previously been directly involved in the glucose response but was considered to serve as an accessory factor (22,23). Specific inactivation of HNF-4α gene in the liver abolishes expression of the genes involved in the control of lipid metabolism, such as apoC-III, apoB, and apoA-II, and also inhibits microsomal triglyceride transfer protein (29). Moreover, in humans, mutations in the HNF-4α gene are responsible for type 1 maturity-onset diabetes of the young (MODY1). Significant alterations in triglyceride metabolism have been observed in MODY1 patients and are thought to result from impaired hepatic synthesis of apoA-II, apoB, and apoC-III (30,31). These data indicate that HNF-4α is a major in vivo regulator of genes involved in the control of lipid homeostasis, via the control of apo gene expression.

The modulation of HNF-4α activity by phosphorylation is well documented (32–36). In the fasting state, an increase in serine/threonine phosphorylation of hepatic HNF-4α decreases its DNA binding activity (32). Furthermore, after the consumption of a carbohydrate-rich diet, HNF-4α phosphorylation is decreased and HNF-4α promotes expression of the target genes. The kinases involved in these mechanisms are currently unknown. The AMP-activated protein kinase has been implicated in the inhibition of glucose-dependent gene expression (37) and has been shown to phosphorylate HNF-4 (38,39). Therefore, when glucose is abundant, the increase in binding of HNF-4α to the target gene may be mediated by inactivation of the phosphorylation of HNF-4α.

The overproduction of human apoA-II decreases plasma HDL concentration in a dose-dependent manner (8,40) and induces postprandial hypertriglyceridemia (8,9). This phenotype resembles the dyslipidemia observed in diabetes, suggesting that stimulation of the hepatic synthesis of apoA-II may contribute to the dyslipidemia syndrome that occurs in patients with type 2 diabetes. Several studies in humans have linked the apoA-II gene to type 2 diabetes (41–44). In our transgenic mice, there is a positive correlation between blood glucose and the plasma human apoA-II. A delay in glucose tolerance was impaired in the λ-line overproducing human apoA-II in accordance to the glucose intolerance observed in transgenic mice overproducing murine apoA-II (45). Conversely, apoA-II knockout mice display low plasma levels of free fatty acids, glucose, and insulin, suggesting hypersensitivity to insulin (46). Together with our results, this suggests that the overproduction of apoA-II may contribute to glucose intolerance and that plasma apoA-II levels may be involved in diabetes pathogenesis.

In conclusion, we show that human apoA-II gene transcription is induced by glucose in vitro, and that this regulation requires an HRE that interacts with HNF-4α. Our results allow us to hypothesize a possible pathogenic pathway. An increasing apoA-II level causes hypertriglyceridemia and hypoHDLemia. The dyslipidemia results in glucose intolerance; this promotes a rise in blood glucose, which in turn might increase apoA-II gene transcription, thus establishing a vicious circle that may contribute to type 2 diabetes.

This work was supported by INSERM and Pierre & Marie Curie University (PARIS VI, Paris, France). D.S. is supported by a fellowship from the Fondation Pour la Recherche Medicale.

We thank Nathalie Fournier (Service de Biochimie, Faculté de Pharmacie, Chatenay-Malabry, France) for plasma human apoA-II assays and Margarita Hadzopoulou-Cladaras (Department of Biology, Section of Genetics, Development and Molecular Biology, Aristotelian University of Thessaloniki, Thessaloniki, Greece) for providing the CD1b expression vector. We also thank Maryse Seau and Carole Lasne for technical assistance.