Polyunsaturated fatty acyl coenzyme A suppress the glucose-6-phosphatase promoter activity by modulating the DNA binding of hepatocyte nuclear factor 4 alpha.

Glucose-6-phosphatase confers on gluconeogenic tissues the capacity to release endogenous glucose in blood. The expression of its gene is modulated by nutritional mechanisms dependent on dietary fatty acids, with specific inhibitory effects of polyunsaturated fatty acids (PUFA). The presence of consensus binding sites of hepatocyte nuclear factor 4 (HNF4) in the -1640/+60 bp region of the rat glucose-6-phosphatase gene has led us to consider the hypothesis that HNF4 alpha could be involved in the regulation of glucose-6-phosphatase gene transcription by long chain fatty acid (LCFA). Our results have shown that the glucose-6-phosphatase promoter activity is specifically inhibited in the presence of PUFA in HepG2 hepatoma cells, whereas saturated LCFA have no effect. In HeLa cells, the glucose-6-phosphatase promoter activity is induced by the co-expression of HNF4 alpha or HNF1 alpha. PUFA repress the promoter activity only in HNF4 alpha-cotransfected HeLa cells, whereas they have no effects on the promoter activity in HNF1 alpha-cotransfected HeLa cells. From gel shift mobility assays, deletion, and mutagenesis experiments, two specific binding sequences have been identified that appear able to account for both transactivation by HNF4 alpha and regulation by LCFA in cells. The binding of HNF4 alpha to its cognate sites is specifically inhibited by polyunsaturated fatty acyl coenzyme A in vitro. These data strongly suggest that the mechanism by which PUFA suppress the glucose-6-phosphatase gene transcription involves an inhibition of the binding of HNF4 alpha to its cognate sites in the presence of polyunsaturated fatty acyl-CoA thioesters.

. The expression of its gene is increased during diabetes and fasting and normalized upon insulin treatment and refeeding, respectively, in all three gluconeogenic tissues (3,4). An increase in the Glc6Pase flux (5,6) and maximal velocity (7) has also been strongly suggested to account for increased glucose production and hepatic insulin resistance in type 2 diabetes mellitus.
The Glc6Pase gene expression is also modulated by nutritional mechanisms dependent on dietary fatty acids. In the liver of rats, Glc6Pase mRNA and protein contents are increased upon high fat feeding (8) and upon elevation in plasma fatty acid levels (9). Under these nutritional conditions, the suppression of hepatic glucose production by insulin is impaired (10). This suggests that a high plasma fatty acid level may contribute increased production of glucose via increased expression of Glc6Pase, resulting in the development of liver insulin resistance (9,11,12). In vitro, the treatment of fetal hepatocytes with a high concentration (500 M) of long chain fatty acids (LCFA), such as oleic and linoleic acids, increases the Glc6Pase mRNA content (13). We have shown that the likely mechanism involves a stabilizing effect on Glc6Pase mRNA (13). On the other hand, in vivo, the presence in high fat diets of substantial amounts of polyunsaturated fatty acids (PUFA), such as that in soybean oil (rich in :6 fatty acids) or fish oil (rich in :3 fatty acids), does not result in an increase in Glc6Pase activity in rats (8,14). This suggests that among LCFA, PUFA may have specific inhibitory effects on Glc6Pase gene expression. Noteworthy, it is also well known that diets rich in PUFA have only weak deleterious effects on insulin sensitivity, as compared with diets enriched in saturated fat (15,16). It is interesting to note that PUFA may also have opposing effects in regard to hepatic insulin resistance at the level of Glc6Pase activity. Indeed, we have previously shown that PUFA, either added to endoplasmic reticulum membranes or naturally associated to glycogen granules, specifically inhibit the Glc6Pase activity (17,18). To date, no molecular mechanism involved in the specific inhibitory effect of PUFA on the Glc6Pase gene expression has been described.
In the liver, although peroxisome proliferator-activated receptors (PPARs) have emerged as an important factor in the fatty acid regulation at the transcription level, recent evidence indicates that the DNA binding activity and/or the abundance of other factors, such as sterol regulatory element-binding protein 1c (SREBP1c) or hepatocyte nuclear factor 4 (HNF4), may be affected by fatty acids or their metabolites (for reviews, see Refs. 19 -21). Interestingly, Hertz et al. (22) have recently reported that the transcriptional activity of HNF4␣ (previously considered an orphan nuclear receptor) could be differentially modulated by LCFA, depending on their unsaturation level, through the binding of their acyl-CoA thioesters to the ligand domain of HNF4␣. For example, saturated LCFA such as C16:0 * This work was supported by a grant from the "Fondation de France" and from the University Claude Bernard Lyon I (Bonus Qualité Recherche). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
(palmitic acid) might activate the transcriptional activity of HNF4␣ in cellular transfection assays, and its palmitoyl-CoA derivative might enhance the binding of HNF4␣ to its DNAbinding site in vitro in bandshift assays. On the contrary, PUFA, such as C18:3(n-3) or C20:5(n-3), might inhibit the transcriptional activity of HNF4␣, and their respective acyl-CoA derivatives might decrease the binding of HNF4␣ to its site. The latter data were obtained using a synthetic promoter containing three copies of the HNF4-binding sequence of the human apoCIII gene regulatory region (22). Noteworthy, as consensus sequences for the binding of HNF4␣ are present in the 5Ј-flanking region of the rat Glc6Pase gene (23), this has given us, using a natural promoter, the opportunity to examine the original hypotheses: 1) that a differential regulation of Glc6Pase gene expression by saturated and polyunsaturated LCFA could occur at a transcriptional level and 2) that HNF4␣ might be involved in the mechanism. We studied the effects of LCFA on the Glc6Pase promoter activity in HepG2 hepatoma and HeLa cells. Our results strongly suggest that the Glc6Pase promoter activity is negatively regulated by PUFA through the modulating effect of their CoA-derivatives on HNF4␣ activity.

EXPERIMENTAL PROCEDURES
Construction of the Reporter Gene Plasmids-The 5Ј-flanking region of the rat Glc6Pase gene up to nucleotide Ϫ1640, relative to the transcription start site, was cloned into the "pGL2 basic" vector (Promega) upstream of a luciferase reporter gene and into the "pGL2 enhancer" vector (Promega), which also contains the SV40 enhancer. The Ϫ1640/ ϩ109 region of the Glc6Pase gene was amplified by PCR from rat genomic DNA using the primers 5Ј-AAGCTTAAGGTAACTGAGT-GAA-3Ј sense and 5Ј-CCAAAGTCGTGGAGCACGTTC-3Ј antisense (23) and cloned into the pBluescript SKϩ vector to generate the Ϫ1640/ ϩ109SKϩ plasmid. The Ϫ1640/ϩ66LUC plasmid was constructed by insertion into pGL2 vector of the SmaI/KpnI fragment obtained by digestion of the Ϫ1640/ϩ109SKϩ plasmid. A series of truncated Glc6Pase promoter-LUC constructs with progressive 5Ј-end deletions of the Glc6Pase promoter region was generated by either restriction enzyme or PCR using Ϫ1640/ϩ109SKϩ plasmid as a template (as denoted in Fig. 1). Site-directed mutagenesis of the HNF4␣ sequences was generated using the GeneEditor TM in vitro site-directed mutagenesis system (Promega) using the antibiotic resistance selection oligonucleotide provided in the kit and specific mutagenic oligonucleotides (denoted in Fig. 7 insets). All PCR-derived constructs and mutations were confirmed by sequencing using the T7 sequencing kit (Amersham Biosciences) to ensure the absence of polymerase errors. Plasmid constructs were purified by the Plasmid Maxi Kit (Jet Star, Genomed).
Cell Culture and Transient Transfection Assays-HepG2 human hepatoma cells and HeLa human epithelial cervical carcinoma cells were grown in Dulbecco's modified Eagle's medium supplemented with 6% (for HepG2 cells) or 10% (for HeLa cells) fetal bovine serum, 5 mM glutamine, streptomycin (1 g/ml), and penicillin (1 unit/ml) in a 5% CO 2 atmosphere at 37°C. 1 day before the transfection, 200,000 cells were plated out in 35-mm wells in six-well cell culture plates. The complete medium was refreshed 1 h prior to transfection. Cells were transfected by the calcium phosphate transfection method with 1 g of the Glc6Pase-LUC plasmid, 0.2 g of pRSV-CAT to correct for transfection efficiency, and 0.5 g of a HNF4␣ expression vector pCR3-HNF4 (obtained from M. Raymondjean and B. Viollet (24)) or 1 g of a HNF1 expression vector pRSV-HNF1 (obtained from M. Yaniv (25)) as indicated. The total amount of DNA (2.2 g) was kept constant by the addition of pBluescript SKϩ plasmid. For transfection of HepG2 cells, calcium phosphate-DNA coprecipitates were formed in Hepes buffer, and for HeLa cells, precipitates were formed in BES buffer (26). The precipitate was removed after 20 h by the addition of EDTA (3.5 mM final) for 1 min, and the cultures were further incubated for another 24 h in normal grown medium. Cells were treated for 6 h by 200 M fatty acid (Sigma) diluted in ethanol (0.1% final) in the presence of 0.4% bovine serum albumin, in serum-free medium either alone or supplemented with indomethacin (10 M, Sigma) or NDGA (10 M, Sigma) or acetylsalicylic acid (aspirin, 200 M, Sigma). In some experiments, treatment was performed for 24 h in the presence of 5 ϫ 10 Ϫ4 M Trolox (Sigma). Cells were treated with indomethacin or NDGA or acetylsalicylic acid for 1 h before adding fatty acid. The cells were then washed three times with phosphate-buffered saline and lysed with reporter lysis buffer (Promega). After a 15-min incubation, cells were scraped and centrifuged at 10,000 ϫ g for 5 min at 4°C to eliminate cell debris. Luciferase activity was determined with a BCLBook luminometer (Promega) using the Luciferase Assay Reagent (Promega). For CAT activity, the cell extract was treated at 60°C for 10 min to inactivate endogenous deacetylase activity, and CAT activity was determined as described by Newmann et al. (27). The levels of luciferase activities were normalized by means of the CAT activities.
Statistical analyses were performed using Student's t test for unpaired data.

FIG. 1. Functional analysis of rat Glc6Pase promoter constructs in HepG2 and HeLa cells.
Constructs based on the progressive 5Ј-end deletion of the Glc6Pase gene fused to a luciferase reporter gene (in pGL2"basic" vector, denoted by "B" in the construct name) are shown on the left. The black boxes indicate putative HNF4 binding sites (site 1 (ϩ9/ϩ15), site 2 (Ϫ511/Ϫ516), site 3 (Ϫ667/Ϫ672), site 4 (Ϫ713/Ϫ718)), and the TATA box is symbolized by a hatched box. Numbers on the left represent the position of the end of the fragment relative to the transcription start site (ϩ1, denoted by the arrow). All fragments were obtained by PCR or by restriction digestion as indicated in parentheses. HepG2 cells (hatched bars) and HeLa cells (black bars) were transiently transfected with each promoter-LUC plasmid together with a pRSV-CAT plasmid as a control for correction for transfection efficiency. LUC activity was determined 48 h after transfection and was normalized relative to the level of CAT activity. The transcriptional activity of each construct is expressed relative to the LUC activity of pGL2"basic" and is the mean Ϯ S.E. of at least three independent experiments performed in duplicate.
Gel Shift Mobility Assays-Double-stranded oligonucleotides used were described in Fig. 3A. Recombinant HNF4␣ protein was expressed in Escherichia coli from pRSET-HNF4 (obtained from M. Raymondjean and B. Viollet) and purified as described by Viollet et al. (24).
End-labeled oligonucleotide probes (0.1 ng, 30,000 -50,000 cpm) were incubated for 15 min at 4°C in the binding reaction buffer (10 mM Hepes, 50 mM KCl, 50 mM NaCl, 4 mM spermidine, 0.1 mM EDTA, 2 mM dithiothreitol, 100 g/ml bovine serum albumin, pH 7.6) in the presence of 50 ng of bacterial recombinant HNF4␣ protein and 100 ng of poly(dI-dC) (24). Free DNA and DNA-protein complexes were separated on a 5% nondenaturing polyacrylamide gel (acrylamide/bisacrylamide, 29:1) in 0.2ϫ Tris borate-EDTA buffer. In competition experiments, the competitor DNA was incubated in the mixture prior to the addition of the probe. The antiserum anti-HNF4 used for gel supershifts (28) and/or purified acyl-CoA (Sigma) were preincubated with proteins at 4°C for 15 min before the addition of free probe. After electrophoresis, the gels were dried, and the quantification of the DNA binding complexes was monitored using a PhosphorImager and performed using ImageQuant (Molecular Dynamics, Inc., Sunnyvale, CA).
Western Blotting-50 g of whole cell protein extracts from HeLa cells, HeLa cells transfected by pCR3-HNF4 for 48 h, HepG2 cells, or rat liver nuclear extract were separated by SDS-10% polyacrylamide gel electrophoresis and transferred to Immobilon membrane. Immunoblotting was performed as described by B. Viollet et al. (24) using anti-HNF4 antibodies (28) at a 1:500 dilution.

RESULTS
The Glc6Pase Promoter Activity Is Inhibited in the Presence of PUFA-All Glc6Pase promoter-LUC constructs conferred a significant promoter activity in HepG2 cells but were almost totally inactive in HeLa cells (Fig. 1). In HeLa cells, the addi-tion of an SV40 enhancer was not sufficient to confer significant promoter activity. Indeed, only a weak LUC activity was observed with constructs of the Glc6Pase promoter containing the SV40 enhancer (Ϫ1640/ϩ60E to Ϫ80/ϩ60E constructs) (see Fig. 3). This strongly suggested that some hepatic specific factors were required for Glc6Pase promoter activity. A promoter fragment from Ϫ80 to ϩ60 bp (minimal promoter construct) containing the TATA box was sufficient to confer significant transcriptional activity in HepG2 cells (Fig. 1). A strong maximal LUC activity was obtained with a construct containing the Ϫ500/ϩ60 bp region of the Glc6Pase promoter. In contrast, the Ϫ320/ϩ60B and the Ϫ730/ϩ60B constructs presented a lower activity, which was very similar to that of the minimal promoter construct (Fig. 1). This suggested that the Ϫ320/Ϫ500 bp region might contain enhancer elements and that the Ϫ500/Ϫ730 bp might contain suppressive regulatory elements able to counterbalance the inductive action of the upstream region. Other enhancer regulatory elements might exist within the Ϫ1320 to Ϫ1480 bp region, since both the Ϫ1640/ϩ60B and the Ϫ1480/ϩ60B constructs exhibited a sim- HeLa cells were transiently transfected with the 5Ј-end-deleted Glc6Pase promoter fragments fused to a luciferase reporter gene into the pGL2"enhancer" plasmid containing the SV40 enhancer (denoted by "E" in the construct name), pRSV-CAT (used as internal control), without (closed bars) or with pRSV-HNF1␣ (open bars) or pRSV-HNF4␣ (hatched bars). LUC activities were determined 48 h after transfection and were normalized with regard to the level of CAT activity. The transcriptional activity of each construct is expressed relatively to the LUC activity of pGL2"enhancer" and is the means Ϯ S.E. of at least three independent experiments performed in duplicate. The transactivation rate is indicated on the left as -fold induction over the basal condition for each construct. ilar substantial promoter activity higher than that of the Ϫ1320/ϩ60B construct (Fig. 1).
The suppression of the Glc6Pase promoter activity by PUFA was not affected in the presence of indomethacin (10 M), NDGA (10 M), or aspirin (200 M) (Fig. 2B). In a similar manner, the PUFA inhibition effect at a longer time point (24 h) in the presence of the antioxidant agent Trolox (50 M) was in the same order range as that occurring for a 6-h incubation (Fig. 2C). This allowed us to rule out the hypotheses that the PUFA-inhibitory effects could be dependent either on further metabolism through the cyclooxygenase or lipooxygenase pathways or on the deleterious accumulation of peroxidation products (29).
HNF4␣ Is Involved in the PUFA-induced Inhibition of Glc6Pase Promoter Activity-The involvement of HNF1 in the Glc6Pase gene transcription has been previously documented (30 -32). In contrast, the involvement of HNF4␣ has not yet been described. To assess the putative role of the latter, HeLa cells were co-transfected with Glc6Pase promoter-LUC constructs and eucaryotic expression vectors of either HNF4␣ or HNF1␣ as a control (Fig. 3). In HNF4␣-expressing HeLa cells, the Glc6Pase promoter activity was markedly induced by about 10 times from the shortest construct (Ϫ80/ϩ60 bp), suggesting the presence of at least one HNF4-binding site in this proximal region. This was in agreement with the presence of a potential HNF4-binding site in this region of the rat Glc6Pase promoter (23). In HNF1␣-expressing HeLa cells, the Glc6Pase promoter activity was induced by 3-5 times for the constructs containing the Ϫ320/ϩ60 up to the Ϫ1640/ϩ60 bp regions (Fig. 3). In contrast, HNF1␣ did not significantly transactivate the Glc6Pase promoter sequence between Ϫ160 to ϩ60 bp (Fig. 3). These results suggested the presence of a first HNF1-binding site localized just upstream the Ϫ160 bp of the Glc6Pase promoter. This was in agreement with the presence of HNF1binding sequence identified between Ϫ220 and Ϫ210 bp on the mouse, human, and rat Glc6Pase promoters (30,31).
The amount of HNF4␣ expressed in the different cell models used above was quantitatively analyzed by Western blotting. Noteworthy, HNF4␣ was expressed in substantial amount in HepG2 cells (Fig. 5). In contrast, HNF4␣ was not detected in HeLa cells, whereas the amount expressed in HeLa cells transfected with pCR3-HNF4␣ was about 2-2.5 times that in HepG2 cells (Fig. 5). It could be noted that the HNF4␣ level found in HNF4␣-transfected HeLa cells was in the same order range as that present in a rat liver extract analyzed in parallel (Fig. 5). Taken together, the data presented in Figs. 2, 4, and 5 were in Fatty Acyl-CoA Regulation of Glucose-6-phosphatase Promoter keeping with a key role of HNF4␣ in the PUFA-inhibitory effects on the Glc6Pase promoter activity in both HepG2 and pCR3-HNF4␣-transfected HeLa cells.
Functional Identification of Two HNF4␣ Binding Sites on the Glc6Pase Gene-Four potential HNF4␣Ϫbinding sites could be found in the rat Glc6Pase gene regulatory region using the SignalScan program, at positions ϩ9, Ϫ511, Ϫ667, and Ϫ713 (sites 1-4, respectively) (Ref. 23; see also Fig. 1). The four radiolabeled oligonucleotide probes matching these sites (Fig.  6A) were able to bind purified bacterially expressed HNF4␣ in gel shift mobility assays, and a supershift was obtained in the presence of anti-HNF4 antiserum (Fig. 6B). The binding of each probe was competed by the respective unlabeled oligonucleotide and by an oligonucleotide matching the specific HNF4␣-binding site of the PEPCK promoter (Fig. 6B). In contrast, an oligonucleotide matching a specific HNF1␣-binding site was not able to compete for the binding of each probe to HNF4␣ (Fig. 6B). Among the four potential HNF4␣-binding sites of the Glc6Pase promoter, three of them (sites 1, 3, and 4) were able to compete efficiently for the binding of the radiolabeled PEPCK-specific binding site probe to HNF4␣ (Fig. 6C). The affinity of the binding was lower, however, since at least 500 ng of unlabeled Glc6Pase oligonucleotides were required to obtain the same competition effect as that produced by 100 ng of unlabeled PEPCK oligonucleotide (Fig. 6C). These data suggested that HNF4␣ could bind to at least three specific binding sites (sites 1, 3, and 4) on the Glc6Pase gene promoter.
The functional activity of these HNF4 binding sites in the Glc6Pase gene transcription was further characterized via deletion or site-directed mutagenesis as indicated in Fig. 7. The Ϫ694/ϩ60E construct exhibited the same induction by HNF4␣ as that of the Ϫ730/ϩ60E construct, indicating that the presence of the HNF4␣ site 4 was not essential for the HNF4␣induced transactivation of the Glc6Pase promoter. The mutation of the single half-site 1a on the Ϫ694/ϩ60 fragment decreased the induction by HNF4␣ by 45%, and the same mutation on the minimal promoter fragment (Ϫ80/ϩ60) resulted in a near total loss in the transactivation effect. In contrast, the mutation of the half-site 1b had no effect (Fig. 7A). These results strongly suggested that the site 1a is crucial for the HNF4␣ transactivation and for the basal promoter activity. The mutation of the site 3a on the Ϫ694/ϩ60 fragment decreased the HNF4␣ transactivation by about 25% and induced a 25% additional inhibition of the activity of the same fragment mutated on site 1a (i.e. both mutations of site 1a and site 3a on the Ϫ694/ϩ60 fragment resulted in a decrease in the HNF4␣ transactivation by about 75%) (Fig. 7A). In contrast, mutations of site 3b had no substantial effect on the Ϫ694/ϩ60 fragment or additional inhibition effect on the Ϫ694/ϩ60 fragment mutated on site 1a (Fig. 7A). These data strongly suggested that the half-site 3a is crucial for the HNF4␣-transactivation of the Glc6Pase promoter.
HN4␣ Sites 1 and 3 Both Confer Inhibition of Transactivation by PUFA-We next sought to determine whether the Fatty Acyl-CoA Regulation of Glucose-6-phosphatase Promoter HNF4␣ binding sites were responsible for the PUFA suppression of the Glc6Pase promoter activity. A significant 20 -35% of inhibition by C20:4(n-6) of the promoter activity was observed for Ϫ730/ϩ60, Ϫ694/ϩ60, Ϫ694/ϩ60mut3a and -3b, and Ϫ80/ ϩ60 fragments (i.e. all constructs having a functional site 1) (Fig. 7B). The activity of the construct having an inactivated mutated site 1 (mut1a) and a functional site 3 was also significantly suppressed by 10% by C20:4(n-6) treatment. In contrast, the inactivation of both sites 1 and 3 (Ϫ694/ϩ60mut1a ϩ mut3a) resulted in a complete loss of the suppression of promoter activity by C20:4(n-6) (Fig. 7B). These results strongly suggested that individually each of the two sites 1 and 3 alone was able to confer (and that both together were sufficient to fully account for) suppression of transcription of the Glc6Pase promoter by PUFA.
Polyunsaturated Fatty Acyl-CoAs Inhibit the Binding of HNF4␣ to Its Cognate Site-We further studied the effects of fatty acyl-CoA thioesters on the binding of HNF4␣ to site 3 (i.e. the binding site exhibiting the highest affinity in vitro) by gel shift mobility assay. The binding was markedly inhibited in a dose-dependent manner in the presence of oleoyl-CoA (C18:1(n-9)), linolenoyl-CoA (C18:3(n-3)), and arachidonoyl-CoA (C20: 4(n-6)) (Fig. 8). A maximal inhibitory effect of about 60% was observed within a concentration range of 1-10 M. Half-maximal inhibition was obtained at about 2 M for arachidonoyl-CoA, 5 M for linolenoyl-CoA, and 7 M for oleoyl-CoA (Fig. 8B). In contrast, palmitoyl-CoA (C16:0) and stearoyl-CoA (C18:0) had no inhibitory effects on the HNF4␣ binding within the same range of concentration, whereas a weak significant activation effect was obtained at 10 M stearoyl-CoA (Fig. 8B). These data strongly suggested that the mechanism by which PUFA suppress the HNF4␣-induced Glc6Pase promoter activ-ity involves an inhibition of the binding of HNF4␣ to its cognate sites in the presence of polyunsaturated fatty acyl-CoA thioesters.

DISCUSSION
The results presented here constitute the first demonstration that LCFA are able to regulate the expression of the Glc6Pase gene at a transcriptional level. We report that PUFA, and to a lesser extent monounsaturated fatty acid (oleic acid), may exert specific suppressive effects on the Glc6Pase promoter activity with regard to saturated fatty acids. In addition, we elucidated the likely molecular mechanism of such an effect, strongly suggesting that it may be mediated via a control of the transactivation effect induced by a liver-specific transactivation factor (i.e. HNF4␣). More specifically, we have shown 1) that HNF4 ␣ plays a crucial enhancer role in the Glc6Pase promoter activity, through the binding to two specific DNA cognate sites, and 2) that LCFA, via their intracellular metabolites (e.g. fatty acyl-CoA thioesters) are able to modulate the enhancing action of HNF4 ␣, by means of a modulation of its DNA binding activity.
In regard to the original demonstration of the involvement of HNF4␣ in the regulation of the Glc6Pase promoter, it must be noted that 2 out of 4 putative HNF4␣-binding sites predicted from homology with the consensus AGGTCA sequence (33) seem unlikely to have a key role in the HNF4␣ transactivation of the Glc6Pase promoter. Indeed, an oligonucleotide matching site 2 was unable to compete for the binding of HNF4␣ to its specific binding sequence of the PEPCK promoter in gel shift assay, and the deletion of site 4 from the Ϫ730/ϩ60B construct had no significant effect on the transactivation induced by HNF4␣ (results of Fig. 6 and 7A). This is in keeping with the  1-4). Oligonucleotides used for directed mutation are indicated in the insets and compared with the wild type sequence (WT seq.). In column A, LUC activity was determined 48 h after transfection and was normalized relative to the level of CAT activity. The induction by HNF4␣ of each construct is expressed relative to the induction by HNF4␣ of the Ϫ694/ϩ60E construct. In the column B, LUC activity was determined 6 h after treatment by arachidonic acid (200 M) and referred to the activity in the cells incubated in the absence of fatty acids (condition with 0.1% EtOH). The results are expressed as percentage of inhibition induced by C20:4. In both columns, the results are the means Ϯ S.E. of at least three independent experiments performed in duplicate. Column A, * and **, significantly different from Ϫ694/ϩ60E activity; #, different from mut1a and mut1a ϩ 3b. Column B,°and°°, significantly different from EtOH value; p Ͻ 0.05 and p Ͻ 0.01, respectively. nd, not determined. observation that neither site 2 nor site 4 is an integral part of a direct repeat of the DR-1 type (see Fig. 6A). In contrast, both predicted sites 1 and 3 are an integral part of a classical DR-1 repeat, with higher homology with the AGGTCA consensus in the 3Ј half-site (see sequence alignments in Table I). In agreement with a crucial role of these sites in the HNF4␣-transactivation of the Glc6Pase promoter, the invalidation by mutation of each or both results in marked suppressions in the HNF4␣induced transcription of the Ϫ694/ϩ60B and Ϫ80/ϩ60B Glc6Pase promoter constructs. Interestingly, in both cases, the 3Ј-half sites (1a and 3a) of highest homology with the consensus appear to be the most crucial in the HNF4␣ transactivation. It must be mentioned that HNF4␣ still transactivates, albeit weakly, the Ϫ694/ϩ60B construct having sites 1 and 3 invalidated. This might suggest the presence of another uncovered HNF4␣ binding site within this promoter region. Another possibility, because HNF4␣ also transactivates the HNF1␣ gene, might be an indirect effect of HNF4␣ mediated by HNF1␣ on the Glc6Pase transcription. A functional HNF1␣ binding site has indeed been described in the Ϫ220/Ϫ210 bp region of the rat Glc6Pase promoter (30,31).
Regarding to the regulation of Glc6Pase transcription by LCFA, it is of note that each from both sites 1 and 3 is able to confer the suppression of the HNF4␣-induced transactivation by PUFA and that the invalidation of both sites results in a total loss of the PUFA-inhibitory effect. However, an intriguing observation has been that site 1 apparently has a prominent role compared with site 3 in transfection experiments, with regard to both the transactivation efficiency (about 50% of total transactivation effect of the Ϫ694/ϩ60B construct for site 1 versus 25% for site 3) and to PUFA inhibition (about 20 -35% for site 1 versus 10% for site 3; see Fig. 7). In contrast, site 3 exhibits a much better affinity in HNF4␣ binding in gel shift assays (see e.g. Figs. 6C and 8 for inhibition of the binding in the presence of PUFA-CoA). We have no definitive explanation for the latter. It seems likely that the putative involvement of HNF4␣-transcriptional co-activators (e.g. PGC-1) (34) might play a key role in cells, whereas they are absent in in vitro assays.
Taken together, the results presented herein are in keeping with the molecular mechanism previously proposed by Hertz et al. (22) (i.e. that the LCFA regulatory effects are mediated via a modulation of the binding of HNF4␣ to its DNA cognate sites induced by their intracellular CoA-thioesters derivatives). More specifically, our results are consistent with the previous ones of Hertz et al. (22) in regard to two points: 1) we have found that PUFA inhibit the HNF4␣-induced transactivation of the Glc6Pase promoter in cells, and 2) we have found that polyunsaturated fatty acyl-CoA thioesters inhibit the binding of HNF4␣ to its cognate sites in gel shift mobility assays. Our results are, however, somewhat in disagreement in regard to some other important points: 1) we have not found any enhancement of HNF4␣-induced transactivation by palmitate in HeLa cells or of the binding of HNF4␣ to its DNA binding sites in the presence of C:16-CoA in the gel shift mobility assay; 2) we have not found that stearate and stearoyl-CoA had inhibitory effects similar to those of PUFA and PUFA-CoA, respectively; and 3) we have found that oleate and oleoyl-CoA can induce, at least under some conditions, suppressive effects similar to those of PUFA and PUFA-CoA (with a weaker efficiency, however). We have no definitive explanation for these discrepancies. That the experiments herein have been carried out with a natural gene promoter while Hertz et al. studied a synthetic promoter could explain at least some of them. With the exception of the slight enhancing effect of stearoyl-CoA on the HNF4␣ binding in gel shift assay (see Fig. 8), which has been suggested to be possibly due to a detergent-like effect on HNF4␣ oligomerization (35), we have never found enhancer effects induced by any LCFA. Therefore, we agree with the  opinion of Sladeck and co-workers (35) that LCFA-CoA thioesters cannot be considered as a classical ligand for HNF4␣. There is no doubt, however, that HNF4␣ exhibits specific affinity binding sites for LCFA-CoA thioesters (22). Therefore, the consistent suppressive effects induced by PUFA in transfection experiments and PUFA-CoA in binding assays, respectively, led us to propose that PUFA-CoA may be considered as a possible physiological inhibitor of HNF4␣-induced transactivation processes. Because some of the most potent inhibitory PUFA (e.g. arachidonic acid) are possible precursors of eicosanoids, it is of note that the experiments in the presence of various inhibitors of the cyclooxygenase and/or lipooxygenase pathways allowed us to definitively rule out the hypothesis that the further metabolism of PUFA via these pathways could be involved in the effects observed. It should also be mentioned that the expression of another liver gene, the L-pyruvate kinase gene, has been reported to be inhibited by PUFA at the level of transcription (36). The molecular mechanism had not been understood in the latter study, but it is interesting to notice that the promoter region conferring the PUFA response overlaps a HNF4-binding sequence (36).
Hertz et al. (22) argue in their paper that dietary stearate (C18:0) and PUFA have been reported having effects 1) similar between them, on the one hand, and 2) opposed to those of dietary palmitate (C16:0), on the other hand, in some physiological processes such as blood coagulability or the level of blood lipids. Our findings that palmitate and stearate and their respective CoA-derivatives may rather be ranged within the same class with regard to the effects induced on the HNF4␣ transactivation seem, in our opinion, more in keeping with the similarity in their physicochemical properties in opposition to their unsaturated counterparts. It seems indeed obvious that there exists a striking parallelism between the level of unsaturation of LCFA and their respective CoA-thioester derivatives and the effects induced on either the transactivation of the Glc6Pase promoter by HNF4␣ in cells or the binding of HNF4␣ to its DNA cognate sites in vitro. 1) Saturated LCFA (C16:0 and C18:0) did not suppress HNF4␣-induced transactivation, and their respective CoA-counterparts did not inhibit the HNF4␣binding. 2) The monounsaturated LCFA (C18:1(n-9)) exhibited a weak suppressing effect on the HNF4␣ transactivation in HeLa cells, and C18:1-CoA had a significant inhibitory effect on the HNF4␣-DNA binding only at the highest concentrations studied in gel shift mobility assays (see Fig. 8B). 3) PUFA markedly suppressed the HNF4␣-transactivating effect on the Glc6Pase promoter, and their respective CoA-thioesters derivatives inhibited the binding of HNF4␣ to its cognate DNA binding sites from the lowest concentrations studied. This suggests that the inhibitory action of LCFA-CoA could be at least in part dependent on either the number of double bonds in the carbon chain or on the conformational constraints dictated by these double bonds.
We have also taken in consideration the hypotheses that either PPAR␣ or SREBP1c (see Introduction) could be involved in the PUFA modulation of Glc6Pase gene transcription. However, in none of our transfection experiments did clofibrate (0.5-500 M) alter the Glc6Pase promoter activity (data not shown). In addition, there was no difference in the abundance of the Glc6Pase mRNA in the liver in PPAR␣ null (Ϫ/Ϫ) mice (37) as compared with wild-type mice, whether the animals were fed with fenofibrate or not (data not shown). These data strongly suggested that PPAR␣ might not be involved in the PUFA modulation of the Glc6Pase gene transcription. Furthermore, our preliminary results have shown that the overexpression of SREBP1c in HepG2 cells or HeLa cells did not alter the activity of the Glc6Pase promoter (data not shown). This al-lowed us to rule out the hypothesis that PUFA-induced suppression of the Glc6Pase promoter activity might be dependent on a decrease in the abundance of SREBP1c, as has been suggested, accounting for the PUFA-suppressive effects on the transcription of genes involved in fatty acid synthesis (19,20).
In conclusion, we report here that PUFA are able to moderate the Glc6Pase gene transcription by means of a specific PUFA-CoA thioester-mediated inhibition of the binding of HNF4␣ to specific enhancer sites on the Glc6Pase promoter. These results further extend the previous ones of Hertz et al. (22). Because of the crucial importance of the Glc6Pase gene on the one hand and of the nutritional facts on the other hand in the context of hepatic insulin resistance and type 2 diabetes, the data presented herein may likely explain at least in part the beneficial effects of PUFA on insulin resistance at the level of endogenous glucose production.