Inhibition by Insulin of Glucocorticoid-Induced Gene Transcription: Involvement of the Ligand-Binding Domain of the Glucocorticoid Receptor and Independence from the Phosphatidylinositol 3-Kinase and Mitogen-Activated Protein Kinase Pathways

Abstract

Insulin can inhibit the stimulatory effect of glucocorticoid hormones on the transcription of genes coding for enzymes involved in glucose metabolism. We reported earlier that insulin inhibits the glucocorticoid-stimulated transcription of the gene coding for liver 6-phosphofructo-2-kinase (PFK-2). To elucidate the mechanism of these hormonal effects, we have studied the regulatory regions of the PFK-2 gene in transfection experiments. We found that both glucocorticoids and insulin act via the glucocorticoid response unit (GRU) located in the first intron. Footprinting experiments showed that the GRU binds not only the glucocorticoid receptor (GR), but also ubiquitous [nuclear factor I (NF-I)] and liver-enriched [hepatocyte nuclear factor (HNF)-3, HNF-6, CAAT/enhancer binding protein (C/EBP)] transcription factors. Site-directed mutational analysis of the GRU revealed that these factors modulate glucocorticoid action but that none of them seems to be individually involved in the inhibitory effect of insulin. We did not find an insulin response element in the GRU, but we showed that insulin targets the GR. Insulin-induced inhibition of the glucocorticoid stimulation required the ligand-binding domain of the GR. Finally, the insulin-signaling cascade involved was independent of the phosphatidylinositol-3-kinase and mitogen-activated protein kinase pathways. Together, these results suggest that insulin acts on the PFK-2 gene via another pathway and targets either the GR in its ligand-binding domain or a cofactor interacting with this domain.

Introduction

Insulin and glucocorticoids exert a long term control on glucose metabolism in liver by regulating the transcription of genes coding for enzymes involved in glycolysis and gluconeogenesis (reviewed in Refs. 1, 2). Insulin stimulates transcription of several genes coding for glycolytic enzymes, such as glucokinase and phosphofructo-1-kinase, while glucocorticoids stimulate transcription of the genes coding for the gluconeogenic enzymes glucose-6-phosphatase (3), phosphoenolpyruvate carboxykinase (PEPCK) (4), tyrosine aminotransferase (TAT) (5), and aspartate aminotransferase (6). Moreover, insulin can inhibit basal and/or glucocorticoid-induced transcription of these gluconeogenic genes. The bifunctional enzyme 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFK-2) catalyzes the synthesis and the degradation of fructose 2,6-bisphosphate, a stimulator of glycolysis and an inhibitor of gluconeogenesis (reviewed in Refs. 7, 8). Transcription of the gene that codes for liver PFK-2 is controlled by insulin and glucocorticoids in a way similar to the genes coding for gluconeogenic enzymes. Liver PFK-2 mRNA is barely detectable in adrenalectomized rats, and it is restored to normal levels upon treatment of these animals with dexamethasone (9). In rat hepatoma FTO-2B cells, glucocorticoids stimulate transcription of the PFK-2 gene and insulin inhibits this stimulation. These effects are rapid and independent of protein synthesis, which indicates that glucocorticoids and insulin are acting on preexisting regulatory proteins (10). The physiological significance of this regulation must be considered by taking into account the posttranslational modifications of the product of the PFK-2 gene (reviewed in Refs. 7, 8). Its phosphorylation induced by glucagon inactivates the 6-phosphofructo-2-kinase activity and stimulates the fructose-2,6-bisphosphatase activity, thereby leading to stimulation of gluconeogenesis via a decrease in fructose 2,6-bisphosphate. On the other hand, insulin treatment stimulates fructose 2,6-bisphosphate synthesis and therefore also glycolysis via stimulation of protein phosphatase 2A, which dephosphorylates the enzyme.

The mechanism by which glucocorticoid hormones regulate transcription is fairly well understood. After binding its ligand, the glucocorticoid receptor (GR) migrates from the cytoplasm to the nucleus and binds to a glucocorticoid response element (GRE) in regulatory regions of target genes. The GR then modulates transcription by interacting with cofactors (11), like glucocorticoid receptor interacting protein-1 or steroid receptor coactivator-1, which link the GR to the general transcription factors without binding themselves to DNA. In several instances, transcription factors bind to DNA in the vicinity of the GRE and cooperate with the GR, thereby constituting a glucocorticoid response unit (GRU). The PFK-2 gene contains a GRU in the first intron (12, 13). Our previous in vivo analysis by genomic footprinting and our transfection experiments showed that in this GRU, hepatocyte nuclear factor (HNF)-3 binds to DNA in a glucocorticoid-dependent way and cooperates with the GR to stimulate transcription (14).

The way in which insulin regulates transcription is less well understood. Upon insulin binding, the insulin receptor (IR) becomes activated. This leads to autophosphorylation of the receptor and to phosphorylation of substrates, thereby initiating the intracellular signaling cascades that control various metabolic processes (reviewed in Refs. 15, 16). In liver, several genes are regulated by insulin at the transcriptional level via a glucose-dependent pathway. In this case, insulin stimulates glucose phosphorylation, and a glucose metabolite indirectly induces a transcriptional response via glucose response elements that bind transcription factors of the basic helix-loop-helix (bHLH) family (reviewed in Ref. 17). In addition, insulin regulates the transcription of several genes via a glucose-independent pathway that targets an insulin response element (IRE). Two pathways that transduce the insulin signal from the IR to the IREs have been identified. The genes coding for PEPCK (18, 19) and for gene 33 (20) are controlled via a phosphatidylinositol-3-kinase (PI3K)-dependent pathway. The genes coding for c-fos (21, 22) and for CAAT/enhancer binding protein (C/EBP)α (23) are controlled via a Ras-Raf-mitogen-activated protein kinase (MAPK) pathway. The transcriptional stimulation of the c-fos gene by insulin also depends on protein kinase C isoforms and on a cross-talk between the Ras and PI3K cascades (24, 25).

Three mechanisms can explain inhibition of transcription by insulin: 1) displacement of a DNA-bound activator by an insulin-regulated transcription factor, 2) active insulin-induced repression by a transcription factor bound to DNA, or 3) inhibition of transcriptional activity via insulin-induced posttranslational modifications or protein-protein interactions involving DNA-bound transcriptional activators, cofactors, or basal transcription factors. The first two, but not the third, of these mechanisms require the presence of an IRE in a regulatory region of the target gene. The PEPCK, TAT, and insulin-like growth factor binding protein-1 (IGFBP-1) genes contain an IRE that mediates an inhibitory effect of insulin, but the transcription factor involved has not been identified (4, 5, 26–28).

In the work reported here we have investigated how glucocorticoids stimulate transcription of the PFK-2 gene and how insulin inhibits this effect. We have shown earlier that insulin controls the PFK-2 gene via a glucose-independent pathway (10). We demonstrate here that both hormones act via the GRU which, in addition to the GR, binds ubiquitous and liver-enriched transcription factors. We also show that these transcription factors modulate the glucocorticoid response, but do not seem to be individually involved in insulin action. We present evidence that insulin targets the GR in its ligand- binding domain (LBD) and that this requires neither the PI3K nor the MAPK pathway.

Results

The GRU of the PFK-2 Gene Mediates the Action of Glucocorticoids and of Insulin

To determine how insulin inhibits the stimulation of the PFK-2 gene by glucocorticoids, we first looked for an IRE in the regulatory regions of that gene. We stably transfected rat hepatoma FTO-2B cells with a luciferase reporter construct driven by the PFK-2 gene promoter alone or linked to the intronic PFK-2 GRU (12). These cells were treated for 24 h with dexamethasone (1 μm) and/or insulin (10 nm) and assayed for luciferase activity. No hormonal response was observed with a construct (pPLLuc138) containing the PFK-2 promoter alone. In contrast, addition of the PFK-2 GRU conferred to the construct (pPLLuc138GRU) the stimulation by dexamethasone and the inhibition of this effect by insulin, without effect of insulin on basal activity (Fig. 1A). Thus, the GRU mediated glucocorticoid and insulin effects similar to those observed on the endogenous PFK-2 gene. To further explore this phenomenon, we turned to transient transfection. We chose FTO-2B cells and a nonhepatic cell line (CHO-IR) that overexpresses the insulin receptor. Since CHO-IR cells lack endogenous GR, they were cotransfected with a GR expression vector. Figure 1B shows that these transiently transfected cells displayed hormonal responses that were qualitatively similar to those of stably transfected cells. The effect of insulin was stronger in transiently transfected CHO-IR than FTO-2B cells, probably because of a higher IR concentration in CHO-IR cells. The effect of dexamethasone was greater in transiently than in stably transfected cells, probably because our nonclonal population of stable transfectants was heterogenous.

Figure 1.

The GRU of the PFK-2 Gene Mediates the Inhibitory Effect of Insulin on Induction by Dexamethasone Relative luciferase activity of stably (A) or transiently (B–D) transfected FTO-2B or CHO-IR cells treated for 24 h with 0.01% ethanol as a control, 1 μm dexamethasone, 10 nm insulin, or 1 μm dexamethasone in the presence of 10 nm insulin. E, Structure of the transfected plasmids. Results are means (± sem) for at least three independent experiments for each construct. The inhibitory effect of insulin on the dexamethasone stimulation was statistically significant for all the GRU-containing constructs (P < 0.05).

Having reproduced in two types of transfected cells the hormonal effects studied here, we tested the specificity of the inhibitory action of insulin. To do so, we transiently transfected CHO-IR cells with a luciferase construct driven by a well-known GRE-containing and glucocorticoid-responsive promoter, namely the mouse mammary tumor virus (MMTV) promoter (pMMTVLuc). As expected, dexamethasone treatment stimulated transcription. However, insulin treatment did not inhibit the glucocorticoid effect (Fig. 1C). Perhaps the response to insulin depended on the PFK-2 promoter and/or the distance between the GRU and the promoter. The PFK-2 GRU was therefore cloned upstream of a shorter PFK-2 promoter (−36 to +86, pPLLuc36GRU), or the thymidine kinase (−38 to +51, pTKLuc38GRU), or α-fetoprotein (−80 to +38, pαFPLuc80GRU) promoters. These constructs were cotransfected with the GR expression vector in CHO-IR cells, and the cells were treated with hormones as indicated in Fig. 1D. The PFK-2 GRU conferred the glucocorticoid stimulation and the insulin inhibition not only to the shorter homologous promoter, but also to the heterologous promoters. We concluded that the glucocorticoid and insulin effects on the PFK-2 gene are mediated solely by the GRU, irrespective of the promoter and of the distance.

Transcription Factor Binding to the PFK-2 GRU

To identify the transcription factors that bind to the PFK-2 GRU, we analyzed this region by in vitro deoxyribonuclease I (DNase I) footprinting with rat liver nuclear extracts. In addition to the previously reported binding of the GR, NF-I, and HNF-3 in the region between coordinates 1 and 70 (Ref. 14 ; Fig. 2, upper panel) a footprint[ nucleotides (nt) 117–151] containing three hyperreactive sites was detected (Fig. 2, upper panel, lane 3). These three hyperreactive sites were localized on sequences that match the HNF-3 consensus (^A/_GCAAA^T/_CA) and were typical of the binding of this factor (29). We therefore performed the DNase I footprinting experiment on the GRU in the presence of competing amounts of a HNF-3-binding oligonucleotide. As shown in Fig. 2 (upper panel, lane 4), this oligonucleotide reduced the intensity of the hyperreactive bands to that seen with naked DNA (upper panel, lane 2). This oligonucleotide also restored the pattern observed without nuclear extracts in the 5′-half of the footprint, although no HNF-3 consensus was found in this region. The latter region contains, on the antisense strand, a AAATC^A/_CATAA consensus-binding site for HNF-6 (30, 30A ). Moreover, the competing oligonucleotide used was known to bind not only HNF-3, but also HNF-6 (31). We therefore used, instead, a competing oligonucleotide known to bind HNF-6 but not HNF-3 (Fig. 2, upper panel, lane 5). This prevented the appearance of the 5′-half of the footprint without preventing the three HNF-3-induced hyperreactive sites in the 3′-half of the footprint. Electrophoretic mobility shift assays (EMSAs), performed on the 117–151 footprint region in the presence of antibodies against HNF-6 or HNF-3 (data not shown), confirmed that HNF-6 and HNF-3 bind to the GRU as indicated in Fig. 2 (upper panel). A region partially protected from DNase I by liver nuclear extracts was detected between nt 90 and 110 (Fig. 2, upper panel). When this region was used as a probe in EMSA with liver nuclear extracts, several complexes were observed (Fig. 2, lower panel, lane 1). These were specific as their appearance was prevented by an excess of cold probe (Fig. 2, lower panel, lane 2). Addition of a competing oligonucleotide known to bind C/EBP (Fig. 2, lower panel, lane 4) also inhibited complex formation whereas an excess of unrelated oligonucleotide upstream stimulatory factor (USF) (Fig. 2, lower panel, lane 3) did not. We conclude from this experiment that the partial footprint detected between nt 90 and 110 is due to binding of C/EBP-related proteins. A schematic representation of the DNA-protein interactions observed in the GRU is shown in Fig. 2 (upper panel).

Figure 2.

The PFK-2 GRU Binds Several Transcription Factors in VitroUpper panel, DNase I footprinting experiments with liver nuclear extracts on a GRU-containing radioactive probe. The probe consisted of an EcoRI-ClaI fragment from pBS.GRU labeled with [α-³²P]dCTP and the Klenow enzyme at the ClaI site on the sense strand. The competitor oligonucleotides (oligo) used are indicated above the lanes. G+A indicates a chemical sequencing of the probe. Bars correspond to protections and arrows to hyperreactive sites. The dashed bar and the open arrow refer to weak protection and hyperreaction, respectively. The coordinates correspond to an arbitrary numbering in the GRU (14 ). A schematic representation of the GRU-protein interactions is shown on the right. The GR binding site indicated could not be seen in DNase I footprinting with crude liver nuclear extracts, but was detected previously with purified GR (14 ). Lower panel, EMSA with a PFK-2 GRU probe (GRU 90/110) and liver nuclear extracts. Addition of competing unlabeled oligonucleotides is indicated above the lanes.

The GRE of the GRU Does Not Function as an IRE

To understand how insulin inhibits the glucocorticoid stimulation of the PFK-2 gene transcription, we tested the possibility that the GRE of the gene is a composite GRE/IRE, i.e. a binding site for the GR and for an insulin-regulated transcription factor that attenuates the glucocorticoid response. To test this, we replaced the PFK-2 GRE sequence by a GRE derived from an insulin-insensitive promoter. If the hypothesis were correct, this swap should abolish the insulin effect. We chose the MMTV GRE, as we demonstrated above (Fig. 1C) that it is insulin insensitive. The construct (pPLLuc138GRU M) containing a GRU modified in this way (mGRU M, Fig. 3A) was stimulated by dexamethasone, as expected. However, it was inhibited by insulin like the wild-type GRU (Fig. 3B). Since the MMTV GRE became sensitive to insulin when placed in the context of the PFK-2 GRU, we concluded that the PFK-2 GRE is not a composite GRE/IRE.

Figure 3.

The GRE of the PFK-2 GRU Does Not Function as an IRE A, Schematic representation of the wild-type (GRU wt) and mutant (mGRU) GRU, and sequences of the GAL4 site (G) and of the GRE in the PFK-2 GRU (wt) and in the MMTV promoter (M). Arrows indicate the imperfect palindromic half-sites of the GRE. These GRUs were cloned upstream of the PFK-2 promoter (−138 to +86) driving the luciferase reporter gene. B, Relative luciferase activity of transiently transfected mGRU M and mGRU G in FTO-2B and CHO-IR cells. Results are means (±sem) for at least three independent experiments for each construct. The inhibitory effect of insulin on the dexamethasone stimulation was statistically significant (P < 0.05).

Another way to check whether the GRE is the IRE was to replace it by an unrelevant sequence that can still bind the GR. We therefore replaced the PFK-2 GRE by a sequence known to bind the yeast GAL4 activator and cotransfected this construct (pPLLuc138GRU G, Fig. 3A) with an expression vector coding for a GAL4/GR fusion protein in which the DNA-binding domain (DBD) of the GR was replaced by the GAL4 DBD (32). As expected, this fusion protein activated the pPLLuc138GRU G reporter construct only in the presence of dexamethasone (Fig. 3B). The glucocorticoid induction of this mutated GRU was also inhibited by insulin. This experiment not only confirmed that the GRE does not function as an IRE, it also showed that the GR DBD is dispensable for this action of insulin.

The Binding Sites for NF-I, HNF-3, C/EBP, and HNF-6 in the GRU Function as Modulators of Glucocorticoid Action but not as IREs

Another mechanism by which insulin could inhibit glucocorticoid-stimulated gene transcription is by targeting one of the transcription factors shown above to bind in the vicinity of the GRE. FTO-2B cells were transiently transfected with constructs containing mutations (Fig. 4A) that prevented these transcription factors from binding to the GRU. The efficiency of the mutations was controlled by in vitro DNase I footprinting experiments (data not shown). After hormonal treatment, luciferase activity was measured in the transfected cells. The results are shown in Fig. 4B. These experiments showed that HNF-3 family members have different functions depending on their location in the GRU. As shown earlier (14), the absence of HNF-3 binding close to the GRE reduced the glucocorticoid stimulation (Fig. 4B, panel e) when compared with the wild-type GRU (Fig. 4B, panel a). In contrast, mutation of the three HNF-3 binding sites at the 3′-end of the GRU (Fig. 4B, panel b) had no effect on the hormonal response. By binding to these three sites, HNF-3 could play a structural role (33) and maintain the GRU in the constitutively open state described previously (14). When the HNF-6 site was mutated, we observed an increase in glucocorticoid stimulation (Fig. 4B, panel c). This phenomenon will be discussed elsewhere (C. E. Pierreux, J. Stafford, D. Demonte, D. K. Scott, J. Vandenhaute, R. M. O’Brien, D. K. Granner, G. G. Rousseau, and F. P. Lemaigre, in preparation). Mutations of the C/EBP (Fig. 4B, panel d) and NF-I (Fig. 4B, panel f) binding sites resulted in a reduced glucocorticoid response, indicating that these proteins were acting as DNA-bound cofactors of the GR. Even though the insulin effect was small in the transiently transfected FTO-2B cells, as discussed above, it was observed with all the mutated constructs (Fig. 4B, panels b–f). To confirm this, we transfected the same constructs in CHO-IR cells and tested their hormonal response. As expected, the effect of insulin in these cells was more clearly demonstrated than in FTO-2B cells (Fig. 4C). Insulin inhibited to the same extent the glucocorticoid stimulation of all the mutated GRUs (Fig. 4C, panels b–f) and of the wild-type GRU (Fig. 4C, panel a). These results demonstrated that the binding sites for NF-I, HNF-3, C/EBP, or HNF-6 do not act as IREs.

Figure 4.

Effects of the Transcription Factors That Bind to the GRU A, Schematic representation of the GRU mutants (mGRU) compared with wild-type GRU (GRU wt). Binding sites are represented by open rectangles or ovals and mutated sites are indicated by black filling. These GRUs were cloned upstream of the PFK-2 promoter (−138 to +86) driving the luciferase reporter gene. B, Relative luciferase activity in FTO-2B cells transiently transfected with the constructs described in panel A. The inhibitory effect of insulin on the dexamethasone stimulation was statistically significant for all the GRU mutants (P < 0.05, paired t test). C, Percentage of inhibition by insulin of the glucocorticoid stimulation of the reporter constructs (described in panel A) transiently cotransfected in CHO-IR cells with the GR expression vector. The results are means (±sem) for at least three independent experiments for each construct. No statistical difference was observed when comparing the various GRU mutants (b–f) to the wild-type GRU (a).

The Ligand-Binding Domain (LBD) of the GR Is Required for the Insulin Effect

In the absence of a detectable IRE in the GRU, we hypothesized that the GR or a GR-bound cofactor might be targeted by insulin. We therefore transfected CHO-IR cells with the pPLLuc138GRU reporter construct and vectors expressing GR deletion mutants (Fig. 5). The reporter activity was measured after treatment with hormones. Consistent with data from Müller et al. (32), the stimulatory activity of the GR varied among mutants (see below). The effect of insulin was therefore expressed as the percentage of inhibition of the GR-induced stimulation. Again consistent with the data from Müller et al. (32), GR₁₋₅₅₀, GR₁₋₅₁₅, and GR₁₋₄₈₈, which are devoid of LBD, were constitutive activators of transcription, while the activity of GR_Δ77−262 and GR_418−777 was glucocorticoid dependent. In the presence of wild-type GR (27-fold induction of transcription) and of amino-terminally deleted GR (GR_Δ77−262 and GR_418−777, which respectively stimulated transcription 9- and 8-fold in response to dexamethasone), insulin inhibited the glucocorticoid stimulation by 60%. In contrast, inhibition by insulin was impaired or absent with GR devoid of LBD (GR₁₋₅₅₀, GR₁₋₅₁₅, and GR₁₋₄₈₈) despite the fact that these GR stimulated reporter activity 16-, 13- and 5-fold, respectively. We concluded from these experiments that the LBD of the GR is required for insulin to inhibit glucocorticoid stimulation and that the GR, or a GR-bound cofactor, is the target of insulin action.

Figure 5.

The GR LBD Is Required for Insulin Inhibition A, Schematic representation of the wild-type and deletion mutants of the human glucocorticoid receptor (hGR) shown with its transactivation domains (τ1 and τ2), DBD, and LBD. Deletion start and end points are indicated. B, Percentage of inhibition by insulin of the glucocorticoid stimulation of the pPLLuc138GRU reporter construct transiently cotransfected in CHO-IR cells with the various GR expression vectors. The results are means (±sem) for at least three independent experiments with each construct. The effect of insulin was significant (P < 0.01) with constructs a, e, and f, but not with constructs b, c, and d.

Insulin Signaling to the GRU Is Transduced via a PI3K- and MAPK-Independent Pathway

As mentioned in the Introduction, regulation of gene transcription by insulin reportedly involves a PI3K- and/or MAPK-dependent pathway. To verify whether this applied to the action of insulin on the PFK-2 GRU, we measured expression of the endogenous PFK-2 gene in FTO-2B cells incubated with hormones and inhibitors of the PI3K and MAPK pathways. The cells were incubated for 24 h in the presence of dexamethasone (1 μm) before addition of the PI3K inhibitors, wortmannin (100 nm or 1μ m, see Ref. 34) or LY294002 (50 μm, see Ref. 35). Fifteen minutes later, insulin (10 nm) was added to the medium containing dexamethasone and the PI3K inhibitor, and the cells were further incubated for 4 h. Wortmannin becomes less effective after 3 h (36). We therefore used two different wortmannin concentrations (100 nm and 1 μm), as a 4-h incubation is required to monitor changes in PFK-2 mRNA levels. Cytoplasmic RNA was extracted, and variations of PFK-2 mRNA concentration were assessed by RT-PCR. As expected (Fig. 6), dexamethasone increased the PFK-2 mRNA content and insulin inhibited this increase. Neither wortmannin (Fig. 6A) nor LY294002 (Fig. 6B) prevented the insulin effect and neither drug affected the glucocorticoid response (Fig. 6, A and B). To verify that these inhibitors did inhibit the PI3K-dependent cascade in FTO-2B cells, we measured the activity of pp70S6 kinase, whose activation requires PI3K (37), rather than the activity of PI3K itself. Indeed, as LY294002 is not stably bound to PI3K, it may be partially washed off the enzyme during the immunoprecipitation step, and this may lead to unduly high PI3K activity (38). FTO-2B cells were incubated with wortmannin (1 μm) or LY294002 (50 μm) 15 min before addition of insulin (10 nm). Activation of pp70S6 kinase was measured after 12 min, which corresponds to the time required for maximal pp70S6 kinase stimulation by insulin under our conditions. The results (Fig. 6C) showed that wortmannin and LY294002 were effective in FTO-2B cells as both inhibited the insulin-induced stimulation of pp70S6 kinase activity.

Figure 6.

Neither the PI3K nor the MAPK Pathway Is Involved in Insulin Signaling to the PFK-2 Gene A and B, PI3K is not involved in the insulin-induced inhibition of PFK-2 gene transcription. FTO-2B cells were incubated for 24 h in the presence of 1 μm dexamethasone (dex) or with 0.01% ethanol as a control (ctrl), before addition of 100 nm (d) or 1 μm (e and f) wortmannin (wt; panel A) or of 50μ m LY294002 (LY; panel B). Insulin (10 nm) was added 15 min later, and incubation was prolonged for 4 h. Variations of PFK-2 liver mRNA concentration in these FTO-2B cells were assessed by RT-PCR and are shown (vertical columns) as the ratio of coamplified liver mRNA (L) and internal standard RNA (st). This ratio was measured by the radioactivity incorporated in the amplified products that were visualized by ethidium bromide staining, as shown at the top of the graphs. C, The stimulation of pp70S6 kinase by insulin is blocked by wortmannin and by LY294002. Immunoprecipitated pp70S6 kinase activity was assayed from FTO-2B cells treated 10 min with the PI3K inhibitors (1 μm wt and 50 μm LY) before a 12-min treatment with insulin (10 nm). D, MEK and MAPK are not involved in the insulin-induced inhibition of PFK-2 gene transcription. Treatment of FTO-2B cells and assay of PFK-2 liver mRNA concentration were performed as in panels A and B. The MEK inhibitor PD098059 was used at 30 μm. E, MAPK activation by insulin is blocked by PD098059. Immunoprecipitated MAPK activity was assayed from FTO-2B cells treated 15 min with 30 μm PD098059 before an 8-min challenge with 10 nm insulin. The data are representative of at least two experiments.

To determine whether the MAPK cascade is involved in transducing the insulin signal to the PFK-2 gene, we used PD098059, an inhibitor of MEK, the kinase that activates MAPK (39, 40). FTO-2B cells were incubated for 15 min with PD098059 (30 μm) before addition of insulin (10 nm). Cytoplasmic RNA was extracted after 4 h, and the relative concentration of PFK-2 mRNA was determined by RT-PCR. As shown in Fig. 6D, the MEK inhibitor did not influence the insulin-induced decrease in PFK-2 mRNA. As a control for PD098059 efficiency in FTO-2B cells, we measured MAPK activation by insulin in the presence or absence of PD098059. The data shown in Fig. 6E demonstrate that PD098059 decreased basal MAPK activity, as assessed on MBP phosphorylation, and blocked insulin-stimulated MAPK activity.

These results demonstrated that the insulin regulation of the PFK-2 GRU is independent from the PI3K and MAPK pathways. These conclusions were confirmed by testing the effect of wortmannin, LY294002, and PD098059 on luciferase activity in FTO-2B cells transiently transfected with pPLLuc138GRU and treated with dexamethasone and insulin (data not shown).

Discussion

We have reported here a mode of action of insulin on gene expression not described so far. Transcription of the gene coding for liver PFK-2 is stimulated by glucocorticoids, and this stimulation is inhibited by insulin (10). The present work shows that these hormonal effects are both mediated by the GRU located in the first intron of the gene and that this GRU binds not only the GR, but also other transcription factors. All of them influenced the glucocorticoid stimulation of the PFK-2 gene. The work also shows that the inhibitory effect of insulin occurs via a PI3K- and MAPK-independent pathway that targets the LBD of the GR.

The regulation of gene transcription by insulin has been studied for several genes in various model systems (reviewed in Ref. 41). The PFK-2 gene regulation by insulin resembles that of the glucose-6-phosphatase, PEPCK, TAT, aspartate aminotransferase, and IGFBP-1 genes. Indeed, these genes are all expressed in the liver, and their transcription is stimulated by glucocorticoids and inhibited by insulin. The glucocorticoid stimulation of these genes involves a cooperation of the GR with liver-enriched (C/EBP, HNF-3) and ubiquitous (NF-I) transcription factors (27, 42). In this respect, the PFK-2 GRU was regulated by the same transcription factors. The cooperation between HNF-3 and the GR could involve the same mechanism in the PFK-2 GRU and in the TAT GRU. Indeed, in both cases HNF-3 binding in vivo is glucocorticoid-dependent (14, 43).

In contrast to these similarities in transcriptional stimulation by glucocorticoids, the mechanism by which insulin inhibited this effect on the PFK-2 gene differed from that described for the PEPCK, TAT, and IGFBP-1 genes. In the latter, the insulin-induced inhibition of the glucocorticoid stimulation is mediated by an IRE that corresponds to a binding site for liver-enriched transcription factors (5, 27). In the PFK-2 gene, our data show that the target of insulin action is the GR. Moreover, deletion analysis of the GR demonstrated that the GR LBD is required for the insulin-induced inhibition of glucocorticoid stimulation. We therefore postulate that insulin inhibits GR activity by directly or indirectly targeting the GR LBD. Phosphorylation or dephosphorylation of the GR LBD could be a mechanism for insulin action. Several reports indicate that the phosphorylation status of the GR can influence its function (44, 45). However, the phosphorylation sites identified so far in the GR are localized in its amino-terminal part, i.e. not in the LBD. We can also rule out that insulin treatment prevents binding of the steroid to the LBD or retains the GR in the cytoplasm since insulin did not prevent dexamethasone from stimulating the MMTV promoter. Recent data on steroid receptor function show that these receptors enhance transcription by recruiting an array of coactivator proteins to the transcription complex (11). Some of these non-DNA-bound coactivators interact with the GR in a ligand-dependent way via the LBD (46). Insulin could thus inhibit glucocorticoid-induced gene transcription by interfering with GR-coactivator interactions. Our model would be in line with that proposed by Nakajima et al. (47) to explain how insulin represses cAMP-responsive genes. These authors demonstrated that repression by insulin occurs through promotion of an interaction between pp90_RSK and the coactivator CBP. The binding of pp90_RSK to CBP may, in some contexts, interfere with CBP activity.

Another contribution of the present work was to demonstrate that the peculiar mode of insulin action defined here hinges on the integrity of the GRU of the PFK-2 gene. While the glucocorticoid-stimulated MMTV promoter was insulin-insensitive, the PFK-2 GRU could confer insulin sensitivity to this promoter and to other heterologous promoters, irrespective of the distance. We tested the possibility that this specificity depends on the sequence of the GRE. Indeed, the GR can adopt different conformations and modes of DNA binding, depending on the nucleotide sequence of the GRE (48). This suggested that only a particular GRE-induced conformation of the GR could be insulin-sensitive. Our results eliminated this possibility, since the insulin-insensitive MMTV-GRE became insulin sensitive in the context of the PFK-2 GRU. Moreover, replacing the PFK-2 GRE by a GAL4-binding site did not prevent insulin from inhibiting glucocorticoid-induced activity of a GAL4/GR fusion protein. We tested a second model, in which the specificity of insulin action relies on a GRU context created by the GR and other factors that bind to the GRU. However, the effect of insulin was not affected by the individual destruction of the binding sites for these transcription factors, ruling out that a single binding site determines the specificity of the insulin effect. The present results are in line with our in vivo methylation protection and DNase I footprinting experiments. Addition of insulin did not modify the in vivo binding pattern seen on the GRU in dexamethasone-stimulated cells (data not shown). We therefore favor a mechanism whereby a combination of transcription factors creates an insulin-responsive unit (IRU) that, for example, recruits specific GR coactivators on which insulin can act to inhibit the glucocorticoid stimulation. To confirm this model, we studied (data not shown) the hormonal response of transfected constructs in which more than one binding site of the GRU had been destroyed. However, these combinations of mutations decreased glucocorticoid stimulation too much to allow the study of insulin inhibition. When insulin inhibits the transcriptional stimulation of the PEPCK, TAT, and IGFBP-1 genes by glucocorticoids, it could well do so by targeting the LBD of the GR as described here for the PFK-2 gene. If so, the sensitivity of these other genes to insulin could rely on their particular IRE rather than on an IRU, as postulated for the PFK-2 gene.

To investigate to what extent the insulin regulation of the PFK-2 gene diverged from that described for other genes, we also studied the signaling pathway that links the insulin receptor to the PFK-2 GRU. We ruled out a role of the PI3K pathway or the MAPK pathway. It is noteworthy that transcription of the PEPCK gene is inhibited by insulin via PI3K (18, 19). Recent data demonstrated that Stat (signal transducers and activators of transcription) proteins can mediate insulin action. Stat 5 interacts with the insulin receptor (49), and it potentiates glucocorticoid action through binding to the GR (50). Stat 3 mediates insulin signaling to a STAT-response element via a PI3K- and MAPK-independent pathway (51). We found in the PFK-2 GRU a sequence (108-TTCTCTGAA-117) that resembles the STAT-binding site consensus. However, this site was not functional (data not shown). Furthermore, overexpression of wild-type or dominant negative Stat 5 did not affect insulin regulation of the PFK-2 gene (data not shown). This ruled out an involvement of Stat 3 or Stat 5 binding in the insulin control of PFK-2 gene expression.

In conclusion, our data provide a new mechanism for the hormonal regulation of gene transcription whereby insulin inhibits glucocorticoid action by targeting the GR LBD. The identification of the protein that interacts with the GR LBD should clarify how insulin exerts its negative effect on the stimulation of gene transcription by glucocorticoids.

Materials and Methods

Plasmids

pPLLuc138, which contains the luciferase gene under the control of the PFK-2 liver promoter (−138 to +86), has been described (52). pPLLuc138GRU (14) contains a RsaI-RsaI fragment of the PFK-2 GRU (12) cloned upstream of the PFK-2 liver promoter (−138 to +86). The RsaI-RsaI fragment of the PFK-2 GRU was also cloned upstream of a shorter PFK-2 promoter (−36 to+ 86, pPLLuc36GRU) or of the thymidine kinase (−38 to +51, pTKLuc38GRU) or α-fetoprotein (−80 to +38, pαFPLuc80GRU) promoters. All the GRU mutants were obtained using PCR-directed mutagenesis and were identical to the wild-type GRU except that in pPLLuc138GRU H the most 5′-HNF-3 binding site of the GRU has been destroyed (14); in pPLLuc138GRU 3H the three HNF-3 sites (5′-TTGTTTGTTTGTTTGTT-3′) were replaced by a GAL4-binding site (5′-CGGAGTACTGTCCTCCG-3′); in pPLLuc138GRU H6 the HNF-6-binding site (5′-AAATCCATA-3′) was destroyed (5′-AAAGTGCTA-3′); in pPLLuc138GRU C the C/EBP binding site (5′-GTTACAGTTT-3′) was destroyed (5′-TACGTCTAGA-3′); in pPLLuc138GRU N the NF-I binding site (5′-TGGCAGAACTTTCA-3′) was destroyed (5′-TTCTAGAACTTTGC-3′); in pPLLuc138GRU M and in pPLLuc138GRU G, the PFK-2 GRE was replaced by the MMTV GRE and by a GAL4 binding site, respectively. pMMTVLuc corresponds to the pMamNEOLuc reporter vector from CLONTECH (Palo Alto, CA). The renilla luciferase internal control (pRL138) was constructed by cloning the PFK-2 promoter (−138 to +86) in the pRLnull vector from Promega (Madison, WI).

Cell Culture and Transfection

Rat hepatoma FTO-2B cells and chinese hamster ovary cells stably transfected with the insulin receptor (CHO-IR) were grown as monolayers in a humidified atmosphere (5% CO₂/95% air) in a 1:1 mixture of DMEM and Ham’s F-12 medium (GIBCO-BRL, Gaithersburg, MD) supplemented with 10% FCS. Stable FTO-2B transfectants were obtained by the calcium phosphate coprecipitation method as previously described (26). For transient transfection, 6 × 10⁵ FTO-2B cells were plated on a 60-mm dish, and the medium was replaced 24 h later by 3 ml of a solution containing 2.4 ml DMEM devoid of serum and of antibiotic, 600 μl OPTIMEM containing 10 μg luciferase reporter plasmid, and 45 μg lipofectin. CHO-IR cells (2 × 10⁵) were plated on a 30-mm dish and transiently cotransfected 24 h later with a solution containing 0.8 ml DMEM devoid of serum and of antibiotic, 200 μl OPTIMEM containing 2 μg luciferase reporter plasmid, 50 ng wild-type GR expression vector (pRShGRα) or deletion mutants (GR₁₋₅₅₀, GR₁₋₅₁₅, GR₁₋₄₈₈, GR_Δ77−262, and GR_418−777), 50 ng renilla luciferase internal control (pRL138), and 10 μg lipofectin. After 16 h, the cells were washed, incubated in DMEM/Ham’s F-12 medium plus 0.1% BSA, and treated with 0.01% ethanol, 1 μm dexamethasone, 10 nm insulin, or 1 μm dexamethasone in the presence of 10 nm insulin. Twenty four hours after addition of the hormones, luciferase reporter activities were measured with a Lumac luminometer and normalized for protein concentration in the FTO-2B cell extracts or for renilla luciferase (dual-luciferase reporter assay system, Promega) activity in CHO-IR cell extracts. In the latter, luciferase activity could not be normalized for protein concentration because of an increase in protein concentration after treatment with insulin.

In Vitro DNase I Footprinting and EMSA

DNase I digestions and EMSAs were performed as described (52) using nuclear extracts from normal rats prepared as described by Hattori et al. (53). The probe used in DNase I footprinting was isolated from pBS.GRU as described (14). The double-stranded oligonucleotide GRU 90/110 used as a probe in EMSA was: 5′-AGCTTAACTGTTACAGTTTCTCTGAAAGA-3′. Double-stranded competing oligonucleotides were: C/EBP, 5′-TTAAGGACTAACGGGTTAACTTAACTAG-3′; HNF-3, 5′-GTTGACTAAGTCAATAATCAGA-3′; HNF-6, 5′-GATCGCTTTGAAATTGATTTCAAAGC-3′; NF-I, 5′-TCGA-ACCATGGCCTGCGGCCAGAGGGC-3′; and USF, 5′-GTAG-GCCACGTGACCGGG-3′.

Enzymatic Activities

For measurement of pp70S6 kinase activity, FTO-2B cells were grown in serum-free DMEM/Ham’s F12 medium containing 0.1% BSA and, after 24 h, incubated for 10 min without or with 1μ m wortmannin (Sigma Chemical Co., St. Louis, MO) or 50μ m LY294002 (BIOMOL Research Laboratories, Plymouth Meeting, PA) before a 12-min treatment with 10 nm insulin. All the subsequent manipulations were conducted at 4 C unless otherwise stated. Cells were washed twice with PBS and lysed in a solution containing 50 mm Tris-Cl, pH 7.5, 100 mm KCl, 1 mm EGTA, 1 mm EDTA, 1% NP-40, 270 mm sucrose, 1 mm Na₃VO₄, 10 mm NaF, 10 mm sodium pyrophosphate, 15 mm sodiumβ -glycerophosphate, 1 mm phenylmethylsulfonyl fluoride, and 1 μg/μl aprotinin and leupeptin. The cell lysates were centrifuged at 10,000 × g for 10 min, and proteins (1 mg) were immunoprecipitated with anti-pp70S6 kinase antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and protein A-Sepharose for 2 h. Immunoprecipitates were washed in buffer A (50 mm Tris-acetate, pH 8, 50 mm NaF, 5 mm sodium pyrophosphate, 1 mm EGTA, 1 mm EDTA), and the pp70S6 kinase activity was assayed as described by Krause et al. (54). For measurement of MAPK activity, the FTO-2B cells were grown in serum-free DMEM/Ham’s F12 medium containing 0.1% BSA and, after 24 h, incubated for 15 min without or with 30μ m PD098059 (BIOMOL). Insulin (10 nm) was then added for 8 min and cells were washed and lysed as for the pp70S6 kinase assay. Proteins (200 μg) were immunoprecipitated with anti-ERK-2 antibody (Santa Cruz Biotechnology) and protein A-Sepharose for 2 h. Immunoprecipitates were washed in buffer A and MAPK activity was assayed in the same buffer supplemented with 1μ m cAMP-dependent protein kinase inhibitor peptide, 0.5 mg/ml myelin basic protein, and 50 μmγ ³²P-ATP. After 20 min at 30 C, the incubation was loaded on a 15% SDS-polyacrylamide gel. The dried gel was quantified with an Instant Imager (Packard Instruments, Meriden, CT).

Relative Concentration of PFK-2 mRNA

This was measured by RT-PCR as described (10). For each reaction, 500 ng FTO-2B cytoplasmic RNA were added to 0.04 pg of in vitro synthesized internal standard RNA. The RT was performed with random hexameric primers. The internal standard RNA and the PFK-2 liver mRNA were coamplified by PCR using two primers that recognize both RNAs and yield products of 210 and 274 bp, respectively. The ratio of the two products reflects the relative concentration of PFK-2 liver mRNA in FTO-2B cells. [α³²P]CTP was added in the PCR reaction to determine the ratio of the two products by measuring the radioactivity of gel slices in a scintillation counter or with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

The technical help of N. Aidant and S. Neou is gratefully acknowledged. We thank G. Verhoeven for pRShGRα, M. Müller for GR mutants, B. Groner and E. Stöcklin for Stat 5 expression vectors, C. Szpirer for the α-fetoprotein promoter, and G. Schütz and E. Clauser for FTO-2B and CHO-IR cells, respectively.

This work was supported by grants from the Belgian State Program on Interuniversity Poles of Attraction, Prime Minister’s Office, Federal Office for Scientific, Technical and Cultural Affairs; from the Délégation Générale Higher Education and Scientific Research, French Community of Belgium; from the Fund for Scientific Medical Research (Belgium); from the National Fund for Scientific Research (Belgium); and from the Fonds de Développement Scientifique (Louvain University). C.E.P. holds a fellowship from the Fonds pour la Formation à la Recherche dans l’Industrie et l’Agriculture (Belgium), and F.P.L. is Research Associate of the National Fund for Scientific Research (Belgium).