Protein Kinase B-α Inhibits Human Pyruvate Dehydrogenase Kinase-4 Gene Induction by Dexamethasone Through Inactivation of FOXO Transcription Factors

HepG2 cells were grown in minimal essential media supplemented with 10% (vol/vol) fetal bovine serum, 1 mmol/l sodium pyruvate, and 1× antibiotic-antimycotic (GibcoBRL). When the cells reached 80% confluency, various concentrations of dexamethasone and insulin (100 nmol/l) were added to the medium. After 6 h of incubation, total RNA was extracted from cells using Ultraspec RNA isolation system following the manufacturer’s instruction. Forty micrograms of total RNA were used for Northern blotting as previously described (10). The membrane was hybridized with ³²P-labeled hPDK4 cDNA probe and 28S rRNA antisense oligonucleotide as loading control.

The hPDK4 promoter (−1,759/+222 bp and −848/+222 bp) was generated by PCR using human chromosome clone y61-d4 (y16d7) from UK HGMP Resource Centre as template (5) and constructed by cloning into the pGL3-basic (Promega). The BglII-HindIII HSV TK promoter fragment from pRL-TK vector (Promega) was inserted into pGL3-basic (TK-luc). Oligonucleotides with three copies of GRE2 and mutant GRE2 (mGRE2) were synthesized; 3XGRE2: 5′ CTAG AGGACATTTCGTACA AGGACATTTCGTACA AGGACATTTCGTACA 3′; 3XmGRE2: 5′ CTAG AGGTGATTTCGTTGA AGGTGATTTCGTTGA AGGTGATTTCGTTGA 3′. Substituted sequences are indicated with bold letters. Annealed 3XGRE2 and 3XmGRE2 oligonucleotides with NheI-XhoI sites were cloned into the TK-luc polylinker. All constructs were confirmed by sequencing. Expression vectors for human GR (p6RGR) and human constitutively active PKBα/Akt1 were generously provided by Dr. David Crabb (Indiana University) and Dr. Richard Roth (Stanford University) (44), respectively. Expression vectors for constitutively active TSS-Ala FOXO1a and transcriptionally inactive FOXO1a (Helix3mut FOXO1a) have been described previously (23). Expression vectors for wild-type and constitutively active human FOXO3a were gifts from Dr. Michael Greenberg (Harvard Medical School) (28); E1A wild-type and mutant E1A Δ2-36 were gifts from Dr. David Livingstone (Dana-Farber Cancer Institute) (45).

Using METAFECTENE (Biontex, Germany), HepG2 cells with 50–60% confluency in 12-well plates were transfected with the hPDK4 promoter construct and pRL-TK and in some experiments with expression vectors for various proteins as indicated in the figure legends. Cells were subsequently incubated for 24 h with vehicle, dexamethasone (Biomol), RU486 (Sigma), and insulin (Calbiochem) (concentrations indicated in the figures or figure legends) 24 h after transfection. Cells were lysed in 250 μl of 1× Passive reporter lysis buffer (Promega). Luciferase activities were measured using a Dual-luciferase reporter assay system (Promega). All transfection studies were performed in four independent experiments and repeated with plasmids prepared separately at least twice. Data are presented as means ± SE for four independent experiments. Sigma Plot, Version 3.02, was used for statistical analysis.

The HindIII-XbaI fragments of constitutively active FOXO3a cDNA were isolated from expression vectors generously provided by Dr. Michael Greenberg (28) and then cloned into pcDNA3 to use the T7 promoter (FOXO3a/pcDNA3). FOXO3a protein was in vitro translated from FOXO3a/pcDNA3 using TNT T7 coupled reticulocyte lysate system (Promega).

The following oligonucleotides were synthesized for electrophoretic mobility shift assays (EMSAs) (all sequences are sense strands; core sequences for DNA binding are in bold). Consensus GRE: 5′ ACGTTGTACAGGATGTTCT 3′; GRE1: 5′ ACGTAGAATCGCTTGAACC 3′; GRE2: 5′ ACGTAGGACATTTCGTACA 3′; IRS1: 5′ ACGTTTAATTTTGTTTTTAATTA 3′; IRS2: 5′ ACGTTTTAGGTTGTTTATTCCT 3′; IRS3: 5′ ACGTGCCTCCGAGTTGTAAACAAGGGCG 3′. Oligonucleotides were annealed and end labeled with [α-³²P]dATP (DuPontNEN) using Klenow enzyme. The labeled probe (200,000 cpm) was incubated with human recombinant GR (Panvera), human GST-FOXO1a (Upstate), and in vitro translated constitutively active FOXO3a in 10 mmol/l HEPES, pH 7.9, 60 mmol/l KCl, 1 mmol/l EDTA, 7% (vol/vol) glycerol, 1 mmol/l dithiothreitol, 2 μg poly (dI-dC), and 0.5% FBS for 20 min at room temperature. Unlabeled competitor oligonucleotides and 2 μg of each anti-GR (sc-1003X), anti-FOXO1a (sc-11350X), and anti-FOXO3a (sc-11351) (Santa Cruz Biotechnology) were added to the binding mixtures. DNA protein binding was analyzed on a nondenatured (4% wt/vol) acrylamide gel in 0.5× Tris-borate-EDTA buffer.

Mutant constructs of GRE1 (AT at −1418/−1419 bp to TG; AC at −1409/−1410 bp to TG) and GRE2 (AC at −821/−820 bp to TG; AC at −812/−811 bp to TG) were prepared by site-directed mutagenesis (Stratagene) using 1759PDK4-luc as template. Mutant constructs for IRS1 (GT at −736/−735 bp to CA), IRS2 (GT at −531/−530 bp to CA), and IRS3 (CA at −357/−356 bp to GT) were prepared using 848PDK4-luc as template. All mutant constructs were sequenced to confirm mutations. Three different sets of plasmids for transfection were prepared separately using Wizard plus miniprep kit (Promega).

Treatment of HepG2 cells with dexamethasone increased PDK4 mRNA message in a concentration-dependent manner (Fig. 1A, lanes 1–6). Insulin inhibited the effect of dexamethasone (Fig. 1A, lanes 7 and 8), as observed with 7,800 C1 cells (15). To study hPDK4 gene regulation at the transcriptional level, a luciferase reporter plasmid containing the hPDK4 promoter sequence and its 5′ untranslated region extending from −1759 to +222 bp (1759PDK4-luc) was cotransfected with the human GR expression vector, p6RGR, into HepG2 cells (Fig. 1B). As expected from previous findings that HepG2 cells express very low levels of GR (46), dexamethasone had little effect on PDK4 promoter activity without cotransfected GR (data not shown). However, when GR is coexpressed, dexamethasone increased hPDK4 promoter activity over 40-fold (P < 0.001 compared with cells not treated with dexamethasone). Insulin inhibited hPDK4 expression induced by dexamethasone by >40% (P < 0.05, dexamethasone-treated vs. dexamethasone plus insulin-treated cells). To validate that transcriptional activation by dexamethasone is GR-dependent, transfected cells were treated with RU486 (GR antagonist) in the absence or presence of dexamethasone. Although RU486 induced some activation of 1759PDK4-luc expression, as expected from its effects on GR (46), RU486 completely inhibited dexamethasone-stimulated gene expression (Fig. 1C). These findings suggest that glucocorticoids regulate expression of the hPDK4 gene via activation of the GR.

Computer analysis using MatInspector program (www.genomatix.de) revealed two potential GRE sequences in the hPDK4 promoter: the −1422/−1408 bp region (GRE1) and the −824/−809 bp region (GRE2). To determine whether GR binds to these elements, EMSA was carried out with purified GR and its antibody (Fig. 2A). GR binds to a consensus GRE (cGRE) as well as to the putative GRE2 sequence but not to the putative GRE1 (Fig. 2A, lanes 1, 3, and 5). Anti-GR antibody supershifts the bands formed with cGRE and GRE2, confirming that GR binds to the GRE2 (Fig. 2A, lanes 2 and 6). Furthermore, mutation of the GRE2 element in 1759PDK4-luc virtually abrogated induction of PDK4 promoter activity by dexamethasone, whereas mutation of the GRE1 element was without effect (Fig. 2B). Three copies of GRE2 sequences (−824/−809 bp, 3XGRE) and mutant GRE2 (3XmGRE) were cloned into TK-pGL3-basic (see research design and methods) in order to study whether GRE2 element per se activates transcription in response to dexamethasone. 3XGRE2 was strongly activated by dexamethasone (∼60-fold), but no induction was observed with TK-luc or 3XmGRE2 (Fig. 2C). These findings suggest that glucocorticoids induce PDK4 gene expression through GRE2 in a GR-dependent manner.

Because FOXO factors have been shown to contribute to the ability of glucocorticoids to stimulate the expression of other insulin-regulated genes, we asked whether they also contribute to regulation of hPDK4 expression. 848PDK4-luc reporter construct were cotransfected with constitutively active TSS-Ala FOXO1a, which is mutated to alanine at the three phosphorylation sites for PKB, or with transcriptionally inactive helix3mut-FOXO1a, which is mutated to arginine at histidine 215, resulting in loss of DNA binding activity (23,41). TSS-Ala FOXO1a increased not only hPDK4 basal expression but also induction by dexamethasone by about twofold, while helix3mut-FOXO1a had no effect (Fig. 3A). Greater amounts of FOXO1a did not induce further increase in promoter activity, suggesting HepG2 cells may express enough FOXO1a as shown previously (47). FOXO3a was also examined to determine whether it activates hPDK4 gene transcription since FOXO3a is also expressed in HepG2 cells (36). Wild-type FOXO3a (WT-FOXO3a) and constitutively active triple mutant FOXO3a (TM-FOXO3a), which is mutated to alanine at the three phosphorylation sites for PKB (28), activated hPDK4 basal expression up to 13-fold (Fig. 3B). This activation is comparable to the ∼20-fold induction of hPDK4 by dexamethasone. FOXO3a also increased hPDK4 induction by dexamethasone ∼2.7-fold. These findings suggest that FOXO1a and FOXO3a are important for basal as well as dexamethasone-induced hPDK4 expression.

To determine whether insulin might repress hPDK4 expression through PKB, constitutively active PKBα (CA-PKB) was cotransfected with 848PDK4-luc (Fig. 4A). CA-PKBα completely inhibited dexamethasone-stimulated hPDK4 induction in a concentration-dependent manner. As shown in Fig. 3, FOXO factors appear to cooperate with GR to activate hPDK4 gene transcription. Therefore, PKB may inhibit the dexamethasone-stimulated hPDK4 induction through inactivation of endogenous FOXO factors. Therefore exogenously introduced TSS-Ala FOXO1a and TM-FOXO3a, which cannot be phosphorylated by PKB, may prevent hPDK4 repression by PKB. As predicted, TSS-Ala FOXO1a completely prevented PDK4 repression by PKB in a dose-dependent manner (P < 0.001 at the highest tested concentration), whereas helix3mut FOXO1a had no effect (Fig. 4B). Likewise, TM-FOXO3a completely prevented hPDK4 repression by PKB. WT-FOXO3a was less effective than TM-FOXO3a (Fig. 4C), in agreement with the previous finding that PKB phosphorylates and inactivates FOXO3a (28). These results suggest that FOXO factors cooperate with GR to activate hPDK4 gene in response to glucocorticoids and that the ability of PKB to repress hPDK4 induction by dexamethasone may require phosphorylation of FOXO factors.

Computer analysis shows that the hPDK4 promoter has three potential IRSs for FOXO binding. To determine whether FOXO1a and FOXO3a bind to these elements, the three potential IRSs were employed for EMSA with purified GST-FOXO1a or in vitro translated FOXO3a. Probes containing IRS1, IRS2, and IRS3 formed DNA-protein complexes with FOXO1a and FOXO3a that were either supershifted or disappeared with appropriate antibodies (Fig. 5A and B), suggesting that FOXO1a and FOXO3a bind to these regions. Excess unlabelled wild-type probes (Wt) inhibited the DNA-protein interaction, but mutant oligonucleotides (Mt) did not, suggesting TG sequences are important for FOXO binding (Fig. 5A and B). Figure 5C shows the relationship of the GRE and three IRSs in the hPDK4 promoter.

To further investigate whether IRSs are physiologically functional, reporter constructs with mutated IRSs were prepared and transfected into HepG2 cells. Mutation of the IRSs decreased the hPDK4 induction by dexamethasone, suggesting that FOXO binding to IRSs is required for full activation of the hPDK4 gene by dexamethasone (Fig. 6). Although IRS1 mutation alone did not cause a decrease in hPDK4 induction by dexamethasone, the combinational mutations with either IRS2 or IRS3 caused a greater decrease than the IRS2 or IRS3 mutations alone. While mutation of IRS2 alone resulted in an ∼10% decrease in the dexamethasone response, mutation of IRS3 alone caused an ∼50% decrease, suggesting IRS3 contributes the most to the dexamethasone-stimulated hPDK4 induction. The relative contributions of these elements appeared to increase in the following order: IRS1, IRS2, IRS3. The combinational mutations of IRSs caused further decreases, suggesting that IRSs cooperate to activate hPDK4 gene by dexamethasone. Mutations of all three IRSs (mIRS1/2/3) caused 80% decrease in hPDK4 induction by dexamethasone and 35% decrease in basal expression (data not shown). Similarly, CA-PKBα also caused an ∼40% decrease in the hPDK4 basal expression, which was prevented by constitutively active FOXO factors (data not shown), suggesting FOXO binding to IRSs also contributes to the basal expression of hPDK4 gene. Insulin was no longer effective against basal expression (data not shown) as well as dexamethasone-stimulated hPDK4 induction after mutation of all three IRSs, suggesting insulin represses hPDK4 expression through the IRSs.

The above data suggest that insulin exerts its function through IRSs for FOXO binding. Therefore, we examined whether FOXO factors exert their function through IRSs. As this would predict, TSS-Ala FOXO1a completely prevented PKB inhibition with most of the mutant IRS constructs but not with mIRS1/2/3, whose expression is relatively very low and therefore cannot be affected by PKB (Fig. 7A and B). Mutation of all three IRSs virtually abolished transactivation of hPDK4 promoter activity by WT-FOXO3a and TM-FOXO3a in the basal and dexamethasone-induced states (Fig. 7C). These results suggest that FOXO factors exert their function through the three IRSs. Although mIRS1/2/3 contains an intact GRE, its promoter activity was increased only fourfold by dexamethasone. This result suggests again that FOXO binding to IRSs is important for hPDK4 induction by glucocorticoids.

Interactions with p300/CBP are important for the ability of GR to activate its target genes (48). Recent evidence indicates that FOXO factors also interact with p300/CBP to activate transcription (42). Therefore, we examined the role of p300/CBP in promoting hPDK4 expression. E1A and mutant E1A were cotransfected with the hPDK4 promoter. E1A binds to p300/CBP and represses histone acetyltransferase activity of p300/CBP, leading to gene repression (45). Wild-type E1A completely inhibited hPDK4 induction by dexamethasone. E1A Δ2–36, a modified form in which the p300/CBP interacting domain is deleted (45), was not inhibitory, suggesting that the recruitment of p300/CBP to the hPDK4 gene is essential for transcriptional activation by dexamethasone (Fig. 8A). Next we tested whether interaction between FOXO factors and p300/CBP contributes to activate transcription of the hPDK4 gene. Wild-type E1A completely blocked the effects of constitutively active FOXO1a and FOXO3a on PKB inhibition (Fig. 8B and C), but E1A Δ2–36 was totally ineffective. These data suggest that interactions between p300/CBP and GR as well as FOXO factors are important for glucocorticoid-stimulated hPDK4 expression.

We studied the mechanism responsible for hPDK4 gene regulation by glucocorticoids and insulin. Dexamethasone increased the PDK4 mRNA message level in HepG2 cells, although it was not as effective as in rat 7800 C1 cells (15). Transfection studies suggest that this difference between rat and human cells is due to the low expression level of GR in HepG2 cells. Indeed, dexamethasone caused marked activation of an exogenous hPDK4 promoter in HepG2 cells only after coexpression of GR. RU486, a GR antagonist, abolished the effect of dexamethasone, and EMSA studies showed that GR binds to the −824/−809 bp (GRE) region within the hPDK4 promoter. Mutation of the GRE virtually repressed the induction of hPDK4 expression by dexamethasone. Therefore, we conclude that dexamethasone activates PDK4 gene expression through a GRE in a GR-dependent manner.

Consistent with previous findings regarding the regulation of PEPCK and G6Pase (33), insulin inhibits the ability of dexamethasone to stimulate hPDK4 gene expression. Therefore we focused our studies on the mechanism responsible for the insulin inhibition of glucocorticoid-stimulated hPDK4 expression. Wortmanin, a phosphatidylinositol 3-kinase (PI3K) inhibitor, blocks insulin inhibition of PDK4 induction by dexamethasone in 7800 C1 cells (B.H. and R.A.H., unpublished studies), suggesting that insulin inhibits PDK4 expression through the PI3K/PKB pathway. Consistent with this finding, expression of constitutively active PKBα completely repressed hPDK4 induction by dexamethasone. We next examined the role of FOXO factors (FOXO1a and FOXO3a), downstream targets of PKB, for involvement in the hPDK4 gene regulation. FOXO factors exert their function as transcriptional activators, while insulin and PKB inhibit their function by phosphorylation. Both FOXO1a and FOXO3a activate basal hPDK4 expression regardless of increased amounts of FOXO factors. Although FOXO1a, FOXO3a, and FOXO4 are equally effective in activating basal IGFBP-1 expression (42), FOXO3a caused a greater increase in basal hPDK4 expression than FOXO1a (13-fold vs. 2-fold). FOXO3a might have a distinct effect on basal hPDK4 expression. Different levels of FOXO protein expression might also be a factor, particularly since FOXO1a and FOXO3a expression are driven by the cytomegalovirus (CMV) and SV-40 promoters, respectively. FOXO1a and FOXO3a lead to further increases in hPDK4 expression by dexamethasone. Moreover, TSS-Ala FOXO1a and TM-FOXO3a, which cannot be phosphorylated by PKB, completely prevented PKB inhibition. In accord with this observation, we identified three related IRSs that bind FOXO1a and FOXO3a in the 5′ promoter of the hPDK4 gene. Mutation of IRSs disrupts the effects of FOXO factors and impairs the ability of glucocorticoids to stimulate PDK4 promoter activity. Similarly, Yeagley et al. (17) reported that mutation of the IRSs in the PEPCK and IGFBP-1 genes causes 33–67% decrease in induction of these genes by dexamethasone. Mutations in all three IRSs (mIRS1/2/3) caused an 80% decrease in induction of PDK4 gene by dexamethasone. Furthermore, the mIRS1/2/3 promoter activity was increased only fourfold by dexamethasone, despite the presence of an intact GRE, which is capable of activating a target gene by itself, as shown in experiments with 3XGRE-TK-luc. This suggests that FOXO binding to three IRSs are required for full transcriptional activation of hPDK4 gene by glucocorticoids. Several accessory factors, including an IRS for FOXO binding, are required for full activation of PEPCK gene expression by glucocorticoids (49). Likewise, hepatocyte nuclear factor 1 (HNF-1) is necessary for glucocorticoid-stimulated G6Pase (20) and IGFBP-1 (50) expression. The combination of the GRE and three IRSs in the hPDK4 promoter may be important for the ability of insulin to suppress glucocorticoid-stimulated expression.

Next, we asked the question why FOXO factors play an important role in glucocoticoid-stimulated hPDK4 expression. It appears that FOXO factors cooperate with GR through an interaction with p300/CBP as shown for IGFBP-1 gene regulation (42). The viral oncogene E1A, which binds to p300/CBP and represses histone acetyltransferase activity of p300/CBP, completely abolished the dexamethasone effect as well as the effects of TSS-Ala FOXO1a and TM-FOXO3a on PKB inhibition. Based on our findings, we propose a model for hPDK4 gene regulation by glucocorticoids and insulin (Fig. 9). The GR activated by glucocorticoid binding translocates into the nucleus where it forms a homodimer and binds to the GRE. The p300/CBP coactivator complex interacts cooperatively with GR and FOXO factors (Fig. 9A). The p300/CBP may have multiple surfaces that enable cooperative interactions with FOXO factors. The p300/CBP coactivator complex may promote the rate of formation of a stable transcriptional initiation complex and thereby enhance the rate of PDK4 transcription. GR may be primarily responsible for the recruitment of p300/CBP to the hPDK4 gene because mutation of the GRE virtually abolished the PDK4 induction by dexamethasone. The relative contributions of the three IRSs (IRS1< IRS2< IRS3) to the glucocorticoid response appeared to correlate with the distance from the GRE. The IRS that is located furthest from the GRE (IRS3) may play the most important role in mediating the interaction between p300/CBP complex and the general transcription machinery. Insulin inhibition of FOXO bindings to the three IRSs may interfere with the interaction between p300/CBP complex and the general transcriptional machinery. Therefore the rate of formation of a stable transcriptional initiation complex may be significantly decreased, resulting in the suppression of the glucocorticoid response (Fig. 9B). Insulin signaling via PKB causes phosphorylation of FOXO factors that can cause their translocation out of the nucleus (36,37,41,42,47), promotion of the binding of 14-3-3 proteins (31), and a decrease in DNA binding activity (32). Phosphorylation of FOXO1a by PKB also disrupts the interaction of FOXO1a with PPAR-γ coactivator 1α (PGC-1α), leading to repression of the induction of PEPCK and G6Pase gene expression by PGC-1α (33). Therefore, it may also be possible that the interactions between p300/CBP and FOXO factors are disrupted by phosphorylation.

Taken together, our findings suggest that in the absence of insulin and glucocorticoids, FOXO factors contribute to hPDK4 basal expression. In starved and diabetic states, elevated levels of glucocorticoids activate GR and thereby recruit a p300/CBP complex to the hPDK4 gene. FOXO factors as well as GR may cooperatively interact with the p300/CBP complex to make a stable transcription initiation complex, leading to active hPDK4 gene expression. Insulin promotes inactivation of FOXO factors and thereby reduces glucocorticoid-stimulated PDK4 gene activation.

This work was supported in part by U.S. Public Health Service Grant DK 47844, Diabetes Research and Training Grant AM 20542, the Grace M. Showalter Residuary Trust, a predoctoral fellowship from the Indiana University School of Medicine Center for Diabetes Research (B.H.), National Institutes of Health grant DK41430 (T.U.), and support from the Department of Veterans Affairs Merit Review program (T.U.).

We thank Drs. David Crabb (Indiana University School of Medicine), Richard Roth (Stanford University), Michael Greenberg (Harvard Medical School), and David Livingstone (Dana-Farber Cancer Institute) for their generous gifts of plasmids. We thank Dr. Maria Alexander-Bridges (Harvard Medical School) for precious suggestions.