| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Department of Molecular Physiology and Biophysics, Vanderbilt University Medical School, Nashville, Tennessee 37232
Address all correspondence and requests for reprints to: Richard M. O’Brien, Department of Molecular Physiology and Biophysics, 761 Preston Research Building, Vanderbilt University Medical School, Nashville, Tennessee 37232-0615.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
In vivo studies in liver and in situ studies in liver-derived cell lines or in primary hepatocytes have shown that insulin inhibits basal as well as glucose-, glucocorticoid-, cAMP-, and fatty acid-stimulated G6Pase gene expression (see Ref. 4 for individual citations). In HepG2 cells, the effect of insulin on basal mouse and human G6Pase gene transcription is mediated through a multicomponent insulin response unit that is composed of two regions, designated Regions A and B (5, 6, 7). In the mouse G6Pase promoter, Region A is located between –231 and –199, whereas Region B is located between –198 and –159. Region A acts as an accessory element to enhance the effect of insulin on G6Pase expression mediated through Region B (6). The accessory factor that binds Region A is hepatocyte nuclear factor-1 (HNF-1) (6). Region B contains three insulin response sequences (IRSs), designated IRS 1, 2, and 3 (5, 6). All three IRSs can confer an inhibitory effect of insulin on the expression of a heterologous fusion gene but only IRS 1 and 2 are required for the suppression of basal G6Pase gene expression by insulin (8). The sequence of IRS 1 and 2 resembles that of the IRSs identified in the phosphoenolpyruvate carboxykinase (PEPCK) (9, 10), tyrosine aminotransferase (11), and IGF binding protein-1 (IGFBP-1) (12) promoters.
The identity of the factor that binds these motifs and mediates the action of insulin has been elusive (13), but recently, substantial attention has focused on the potential role of the winged helix/forkhead transcription factor FOXO1a (FKHR) and its orthologs, FOXO3a (FKHRL1) and FOXO4a (AFX). Insulin can inhibit FOXO1a-, FOXO3a-, and FOXO4a-mediated transcriptional activation through the phosphatidylinositol 3-kinase-dependent activation of protein kinase B (PKB), which leads to the phosphorylation and nuclear exclusion of these factors (14, 15, 16, 17, 18, 19, 20, 21). FOXO1a can bind G6Pase IRS 1 and 2 as well as the PEPCK and IGFBP-1 IRSs in vitro (8, 14, 22). In addition, when overexpressed, FOXO1a can stimulate G6Pase, PEPCK, and IGFBP-1 fusion gene expression (8, 22, 23, 24).
Despite these data implicating FOXO1a in the regulation of PEPCK, IGFBP-1, and G6Pase gene transcription, the key question is whether FOXO1a is actually the endogenous factor that mediates the action of insulin on these genes. Thus, other factors have been identified that bind the PEPCK IRS, namely C/EBP (25, 26) and FOXA2 (HNF-3ß) (27, 28), and changing the level of these factors also alters PEPCK gene expression (29, 30). However, detailed mutagenesis studies revealed that the effect of insulin on PEPCK gene transcription does not correlate with the binding of these factors (27). Two groups have performed similar analyses investigating the correlation between FOXO binding and insulin-regulated PEPCK and IGFBP-1 gene expression. Guo et al. (17) generated a series of fusion genes comprising a modified IGFBP-1 promoter containing a single IRS with point mutations in individual IRS bases. They then demonstrated that the ability of overexpressed FOXO1a to stimulate fusion gene expression correlated with the ability of insulin to repress basal fusion gene expression in the absence of overexpressed FOXO1a (17). They also showed that expression of antisense FOXO1a RNA disrupted insulin-regulated gene expression, suggesting that FOXO1a was the endogenous factor mediating insulin action (17). In contrast, Hall et al. (31) generated a series of heterologous thymidine kinase (TK) promoter fusion genes containing either the PEPCK IRS or the two IGFBP-1 IRSs again with point mutations in individual IRS bases. They then performed overexpression and gel retardation competition experiments to assess the ability of FOXO3a to bind to these mutated PEPCK and IGFBP-1 IRS motifs in situ and in vitro, respectively (31). Their studies suggested that FOXO3a binding in vitro only correlated with insulin-regulated fusion gene expression when FOXO3a was overexpressed (31). Hall et al. (31) therefore concluded that, whereas FOXO3a could serve as an insulin response factor when overexpressed, it was not the endogenous factor mediating the action of insulin on these genes. The experiments described in this manuscript were designed to address these conflicting reports by determining whether insulin-regulated G6Pase gene transcription correlates with FOXO1a and/or FOXO3a binding.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
). When a labeled double-stranded oligonucleotide, designated G6P WT, representing the wild-type G6Pase promoter sequence from –196 to –155 (Table 1), was incubated with a crude extract, prepared from bacterial cells expressing a glutathione-S-transferase (GST)-FOXO1a fusion protein, a single isopropylthio-ß-D-galactosidase (IPTG)-induced protein-DNA complex was detected (Fig. 1A, arrow). Competition experiments were performed in which a variable molar excess of various unlabeled oligonucleotides representing WT or mutated G6Pase sequences (Table 1) were included with the labeled probe. The mutated G6Pase sequences contain point mutations in IRS base numbers 5, 7, and 10 (Table 1A). This numbering system originates from the initial definition of the PEPCK IRS as a 10-bp motif (27). The experiments described below focus on the effects of mutating bases 5, 7, and 10 on FOXO1a/FOXO3a binding and insulin action because such analyses formed the basis for the conflicting results of Guo et al. (17) and Hall et al. (31).
|
|
, A and B) as did the oligonucleotide containing mutations in base 10 of G6Pase IRS 1 and 2 (Fig. 1A). In contrast, the oligonucleotide containing mutations in base 7 of G6Pase IRS 1 and 2 showed an impaired ability to compete, whereas the oligonucleotide containing mutations in base 5 of G6Pase IRS 1 and 2 failed to compete (Fig. 1, A and B). This latter result is consistent with our previous studies using crude GST-FOXO1a (8).
When these competition experiments were repeated using partially purified GST-FOXO3a the relative efficiency of competition by the oligonucleotides containing mutations in base 5 and 7 was qualitatively similar, however, competition was achieved using a lower molar excess (Fig. 2, C and D). In addition, the oligonucleotide containing mutations in base 5 of G6Pase IRS 1 and 2 now competed for complex formation but less than achieved with the oligonucleotide containing mutations in base 7 of G6Pase IRS 1 and 2; this latter oligonucleotide now competed almost as well as the WT oligonucleotide (Fig. 2, C and D). The high (500x) molar excess of unlabeled competitor required to achieve efficient competition when using crude GST-FOXO1a (Fig. 1, A and B), even with the labeled probe in excess, is consistent with the kinetics of a low affinity binding protein. We therefore hypothesize that partial purification increases binding affinity by removing factors that would otherwise inhibit GST-FOXO3a binding. A similar phenomenon may explain why we and others have been unable to detect FOXO binding in gel retardation assays using crude nuclear extract. Importantly, when these competition experiments were performed using partially purified His-FOXO1a (Fig. 1, E and F) the results were almost identical with those seen with partially purified GST-FOXO3a (Fig. 1, C and D). Most notably, the oligonucleotide containing mutations in base 7 of G6Pase IRS 1 and 2 competed almost as well as the WT oligonucleotide (Fig. 1, E and F).
|
Mutation of G6Pase IRS 1 and 2 at Base 7 Abolishes the Effect of Insulin on G6Pase Fusion Gene Expression
To investigate the correlation between FOXO binding (Fig. 1) and the regulation of G6Pase gene transcription by insulin, a series of G6Pase-luciferase fusion genes were generated that contain the identical IRS mutations as used in the competition experiments described above (Table 1). These mutations were all generated in the context of the –231 to +66 G6Pase-luciferase fusion gene that contains both Regions A and B and therefore mediates a maximal repression of basal fusion gene expression by insulin (6). These fusion gene constructs were transiently transfected into liver-derived human HepG2 cells and the effect of mutating G6Pase IRS 1 and 2 at bases 5, 7, and 10 on the repression of basal G6Pase-luciferase fusion gene expression by insulin was investigated. As shown in Fig. 2A, insulin repressed basal expression of the WT-G6Pase-luciferase fusion gene and mutation of IRS 1 and 2 at base 10 had no effect on the insulin response. In contrast, mutation of IRS 1 and 2 at base 5 or base 7 abolished the insulin response (Fig. 2A). This result suggests a lack of correlation between FOXO binding to the G6Pase promoter and insulin action because the base 7 mutation had little effect on FOXO binding in competition experiments (Fig. 1).
The effect of mutating G6Pase IRS 1 and 2 at bases 5, 7, and 10 on basal fusion gene expression in the absence of insulin is shown in Fig. 2B. Mutation of IRS 1 and IRS 2 at base 5 or 7 resulted in a decrease in basal G6Pase-luciferase fusion gene expression, whereas mutation of base 10 had no effect (Fig. 2B). The combination of the insulin repression and this basal expression data are internally consistent with a model in which insulin inhibits G6Pase gene transcription by disabling the binding of an activator, as previously proposed (8). However, at odds with our previous model (8), the results with the base 7 mutant suggest a lack of correlation between FOXO binding and insulin action, implying that this endogenous activator is a factor distinct from FOXO1a.
Mutation of G6Pase IRS Base 7 Markedly Reduces but Does Not Abolish the Induction of G6Pase Fusion Gene Expression by FOXO1A and FOXO3A
To further explore the correlation, or lack thereof, between FOXO binding and insulin action HepG2 cells were cotransfected with the same G6Pase-luciferase fusion genes described above along with either a control expression vector or the same vector encoding either FOXO1a or FOXO3a (Fig. 3). Both FOXO1a and FOXO3a stimulated expression of the WT G6Pase-luciferase fusion gene and mutation of IRS 1 and 2 at base 10 had no effect on this induction (Fig. 3). In contrast, mutation of IRS 1 and 2 at base 5 abolished this induction, whereas mutation of IRS 1 and 2 at base 7 markedly reduced, but did not abolish, this stimulation (Fig. 3). This result suggested a lack of correlation between the ability of overexpressed FOXO to induce G6Pase fusion gene expression and insulin action. Thus, mutation of IRS 1 and 2 at base 7 impairs FOXO-stimulated fusion gene expression but abolishes insulin action (Figs. 2A and 3).
|
) but had little effect on FOXO binding in gel retardation competition experiments (Fig. 1) and only reduced, rather than abolished, FOXO-stimulated fusion gene expression (Fig. 3). However, because a number of genetic and molecular experiments support a role for endogenous FOXO1a in mediating the effect of insulin on G6Pase gene transcription (see Discussion), we decided to consider an alternate possibility, namely that the results of these FOXO overexpression and gel retardation competition experiments are misleading. Indeed, the magnitude of the effect of the IRS 1 and 2 base 7 mutation was much greater when assessed in the FOXO overexpression experiments (Fig. 3) than in the gel retardation competition experiments using purified FOXO proteins (Fig. 1, C–F), indicating that these data are not internally consistent.
Mutation of G6Pase IRS Base 7 Markedly Reduces FOXO1A and FOXO3A Binding to Labeled Oligonucleotide Probes
Based on a consideration of the kinetics of receptor-ligand interactions (34), we considered the possibility that, whereas the IRS 1 and 2 base 7 mutated oligonucleotide was able to compete efficiently for FOXO binding, it may not be able to form a stable complex with FOXO. Such a situation could arise if both the association and dissociation rate of FOXO binding were increased because this would not alter the dissociation constant (34) but would result in the rapid dissociation of the FOXO-DNA complex upon separation of bound and free probe during electrophoresis. If correct, this would imply that the results of the gel retardation competition experiments do not fully reflect FOXO binding ability.
To investigate this possibility, the oligonucleotides described above (Fig. 1 and Table 1) were labeled and incubated with either crude extract prepared from bacterial cells expressing the GST-FOXO1a fusion protein (Fig. 4, A and B), with partially purified GST-FOXO3a (Fig. 4, C and D), or with partially purified His-FOXO1a (Fig. 4, E and F). In each case, the FOXO proteins bound the WT G6Pase probe and mutation of IRS 1 and 2 at base 10 had little or no effect on this binding (Fig. 4). In contrast, mutation of IRS 1 and 2 at base 5 abolished binding, whereas mutation of IRS 1 and 2 at base 7 markedly reduced binding (Fig. 4). This result confirms the hypothesis that FOXO proteins do not form a stable complex with the IRS 1 and 2 base 7 mutated sequence and that the results of competition analyses are misleading. Such a dissociation between the results of binding and competition experiments has previously been described for signal transducer and activator of transcription 5 binding to the serine protease inhibitor 2.1 promoter (35).
|
), and the ability of overexpressed FOXO to induce G6Pase fusion gene expression (Fig. 3). But although these data correlate, in both assays the magnitude of the effect of the IRS 1 and 2 base 7 mutation is less than that of the base 5 mutation, whereas both mutations abolish the insulin response (Fig. 2A). However, an explanation for this apparent lack of correlation between the overexpression data with insulin action can again be provided by a consideration of the kinetics of receptor-ligand interactions and specifically the intracellular abundance of FOXO proteins (34). Because FOXO proteins are expressed at low levels in hepatoma cells (36), we hypothesize that the IRS 1 and 2 base 7 mutation results in the loss of endogenous FOXO binding, and hence insulin action, whereas overexpression of FOXO alters the binding equilibrium such that limited binding of FOXO can now occur to the IRS 1 and 2 base 7 mutated sequence resulting in a small induction of fusion gene expression (Fig. 3). Similarly, in gel retardation assays in which the probe is in excess, a low level of binding can be detected to the IRS 1 and 2 base 7 mutated probe (Fig. 4), but we hypothesize that this is insufficient to be of functional significance in situ. As such, the observations reported here would be consistent with the conclusion of Guo et al. (17) that endogenous FOXO factors can mediate effects of insulin on gene transcription.
FOXO1a and FOXO3a Bind Cooperatively to G6Pase IRS 1 and 2
Although the data described above support the conclusions of Guo et al. (17), there is a notable difference in the observed effect of the IRS base 10 mutation. Both we (Figs. 2 and 3) and Guo et al. (17) find that mutation of IRS base 5 abolishes the effect of insulin and the ability of overexpressed FOXO1a to induce fusion gene expression. However, Guo et al. (17) found that mutation of IRS base 10 reduced both the effect of insulin and the ability of overexpressed FOXO1a to induce fusion gene expression, whereas we find that this mutation has no effect on either the insulin response (Fig. 2A) or FOXO-induced expression (Fig. 3). One obvious difference in experimental design that might explain this discrepancy is that the G6Pase fusion genes described above contain two adjacent IRSs that both bind FOXO, whereas Guo et al. (17) used fusion genes comprising a modified IGFBP-1 promoter containing a single IRS. We therefore hypothesized that the difference in results might be explained by cooperative binding of FOXO to the two G6Pase IRS motifs such that mutation of the two IRSs at base 10 was insufficient to destabilize FOXO binding.
To investigate whether such cooperative binding occurs, double-stranded oligonucleotides were synthesized, representing the G6Pase promoter sequence from –196 to –155 encompassing both IRS 1 and 2 (Table 1), that contain either a point mutation in G6Pase IRS 2 at base 5 (G6P IRS 2 M5; Table 1) or a block mutation in G6Pase IRS 2 (G6P IRS 2 BM; Table 1). These oligonucleotides were labeled and incubated with either crude extract prepared from bacterial cells expressing the GST-FOXO1a fusion protein (Fig. 5, A and B), with partially purified GST-FOXO3a (Fig. 5, C and D), or with partially purified His-FOXO1a (Fig. 5, E and F). In each case, the mutation of G6Pase IRS 2 at base 5 markedly reduced the level of FOXO binding relative to that achieved with the WT G6Pase probe despite the presence of the intact IRS 1 motif. The fact that binding is reduced by more than 50% indicates that the complex formed between FOXO and IRS 1 is unstable, which suggests that FOXO binds cooperatively to IRS 1 and 2. In fact, competition experiments show that FOXO1a binds with higher affinity to IRS 1 than IRS 2 (8) so, in the absence of cooperativity, the predicted decrease in binding would have been less than 50%. A further decrease in FOXO binding was detected when the oligonucleotide containing a block mutation in G6Pase IRS 2 was used as the labeled probe, suggesting that the point mutation in IRS 2 does not completely inhibit FOXO binding to this element (Fig. 5).
|
), rather than two bands indicative of the binding of one or two FOXO proteins molecules. Multiple bands are detected using partially purified GST-FOXO3a (Fig. 5C) and partially purified His-FOXO1a (Fig. 5E), but this likely represents the occurrence of proteolysis during purification because the relative abundance of these complexes does not change (Fig. 5, C and E).
In contrast with these binding data obtained using the labeled G6P IRS 2 M5 oligonucleotide, gel retardation competition experiments previously indicated that the affinity of FOXO1a binding to this oligonucleotide, that has an intact IRS 1 motif, was only approximately 2-fold lower than binding to the WT sequence (8). As with the labeled G6P IRS 1 and 2 M7 oligonucleotide binding experiments described above (Fig. 4), this observation again reveals a lack of correlation between the results of binding and competition experiments. With regards to the G6P IRS 2 M5 oligonucleotide, this lack of correlation between binding (Fig. 5) and competition experiments (8) can be explained by a requirement for cooperative binding to IRS 1 and 2 for the formation of a stable FOXO-DNA complex. Similarly, with regards to the G6P IRS 1 and 2 M7 oligonucleotide, the simplest explanation for the lack of correlation between binding (Fig. 4) and competition experiments (Fig. 1) is that the mutation of base 7 prevents the formation of a stable FOXO-DNA complex by affecting cooperative binding.
Promoter Context Alters the Specific IRS Base Requirements for FOXO1a-Stimulated Fusion Gene Expression
Given the evidence for cooperative binding of FOXO to G6Pase IRS 1 and 2 (Fig. 5), the effect of mutating G6Pase IRS 1 at base 10 on FOXO1a-stimulated fusion gene expression in the absence of FOXO binding to IRS 2 was assessed. The rationale here was that this created a situation resembling the single IRS-containing fusion gene used by Guo et al. (17) in their analysis of the correlation between FOXO1a-stimulated fusion gene expression and insulin action. We hypothesized that, in the absence of FOXO1a binding to G6Pase IRS 2 and hence cooperativity, mutation of G6Pase IRS 1 at base 10 might now impair FOXO1a-stimulated fusion gene expression.
G6Pase-luciferase fusion genes were constructed that contained a WT IRS 1 motif or a base 10 mutated IRS 1 motif in conjunction with a point mutation in base 5 of IRS 2 or a block mutation in IRS 2 (Fig. 6A). HepG2 cells were cotransfected with these G6Pase-luciferase fusion genes along with either a control expression vector or the same vector encoding FOXO1a. Figure 6A shows that FOXO1a induced expression of the WT G6Pase-luciferase fusion gene, whereas mutation of IRS 1 and 2 at base 5 abolished this induction. In contrast, mutation of IRS 2 at base 5 only blunted this induction (Fig. 6A), as previously reported (8). When a point mutation was introduced into IRS 1 at base 10 in conjunction with the point mutation in IRS 2 at base 5, there was no further reduction in the magnitude of this induction (Fig. 6A). However, the absence of an effect of mutating IRS 1 at base 10 could potentially be explained by residual FOXO binding to the IRS 2 base 5 mutated motif that could still act to stabilize FOXO1a binding to the base 10 mutated IRS 1. The fact that the same result was obtained after the introduction of a block mutation in IRS 2 suggests this is not the case (Fig. 6A). Thus, the block mutation of IRS 2 reduced FOXO1a-induced fusion gene expression to the same extent as the IRS 2 base 5 point mutation and when a point mutation was introduced into IRS 1 at base 10 in conjunction with the block mutation in IRS 2, there was no further reduction in the magnitude of FOXO-induced expression (Fig. 6A). When the analysis, described above, of the effect of base 10 mutation of G6Pase IRS 1 on FOXO1a-induced fusion gene expression was repeated in rat H4IIE hepatoma cells similar results were obtained (Fig. 6B).
|
The fact that mutation of these IRS motifs at base 10 can influence the insulin response, at least in some promoter contexts, is perhaps surprising given that binding site selection experiments indicate that FOXO1a binding to DNA can tolerate variation at this base (33). Interestingly, Hall et al. (31) also noted variable results with respect to mutation of IRSs at base 10; mutation of the PEPCK IRS at base 10 had no effect on the insulin response, whereas mutation of the two IGFBP-1 IRSs at base 10 abolished the insulin response (31).
Importantly, the results of these binding experiments (Fig. 5) broadly correlate with the analysis of FOXO1a-stimulated G6Pase fusion gene expression (Fig. 6). Thus, mutation of IRS 2 reduces both FOXO1a binding (Fig. 5) and FOXO1a-stimulated G6Pase fusion gene expression (Fig. 6) by much more than 50% indicating that FOXO1a binds cooperatively to IRS 1 and 2 both in vitro and in situ. Interestingly, this correlation is not perfect because the block mutation of IRS 2 reduced FOXO1a-induced fusion gene expression to the same extent as the IRS 2 base 5 point mutation (Fig. 6), whereas partially purified FOXO1a binds better to the IRS 2 base 5 point mutation than the IRS 2 block mutant in vitro (Fig. 5, C and D). This result emphasizes the influence of promoter context because it implies that one or more other factors can stabilize FOXO1a binding to IRS 1 in situ and overcome the otherwise expected further reduction of FOXO1a-stimulated fusion gene expression as predicted from the cooperative binding seen in vitro (Fig. 5).
The Effect of IRS Mutations on Basal Fusion Gene Transcription Is Context Dependent
Given the observation that promoter context can influence the results of IRS mutational analyses, we decided to investigate the effect of mutating G6Pase IRS 1 and 2 in the context of the heterologous TK promoter. Much of the original characterization of the PEPCK IRS was performed in the context of the heterologous TK promoter rather than in the context of the native PEPCK promoter (10, 27). We hypothesized that different results might be obtained in the context of the TK promoter. Double-stranded oligonucleotides representing the WT or mutated G6Pase Region B sequence between –197 and –158 (Table 2) were synthesized and ligated into a TK-luciferase vector and the effect of insulin on the expression of the resulting fusion genes was analyzed by transient transfection of the liver-derived rat H4IIE cell line (Fig. 7A). We have previously shown that insulin has little effect on reporter gene expression directed by the native TK promoter in H4IIE cells (10, 27) (Fig. 7A), whereas it stimulates expression in HepG2 cells (data not shown). Figure 7A shows that the WT Region B sequence that contains IRS 1, 2, and 3 (Table 2) mediates an inhibitory effect of insulin on fusion gene expression, whereas mutation of IRS 1 and 2 at base 5 abolishes this effect. We have previously shown that the isolated, multimerized IRS 3 motif can also mediate an insulin effect in this context (8), but this element is apparently not insulin responsive in the context of the mutated Region B oligonucleotide. Consistent with the model, whereby insulin can inhibit gene expression by stimulating the nuclear exclusion of FOXO proteins that act as transcriptional activators, basal gene expression decreased upon mutation of IRS 1 and 2 (Fig. 7B).
|
|
shows that the WT human IGFBP-1 sequence between –124 and –96, that contains the two IRSs (Table 2), mediates an inhibitory effect of insulin on fusion gene expression, whereas mutation of these IRSs at base 5 abolishes this effect. This base 5 mutation has been shown to abolish FOXO1a binding to the IGFBP-1 IRSs (22). In contrast to the data obtained with the G6Pase IRSs (Fig. 7B), the accompanying change in basal gene expression associated with mutation of these IRSs is not consistent with the model whereby insulin can inhibit gene expression by stimulating the nuclear exclusion of FOXO proteins. Thus, basal gene expression actually increased upon mutation of the IGFBP-1 IRSs (Fig. 7D). This unexpected result with the heterologous IGFBP-1 IRS-TK promoter constructs contrasts with what is seen when the IGFBP-1 IRSs are analyzed in the context of the native IGFBP-1 promoter. Thus, deletion of the IGFBP-1 promoter region between –132 and –103, that encompasses the two IGFBP-1 IRSs, abolished the effect of insulin (Fig. 7E) and resulted in a decrease in basal fusion gene expression (Fig. 7F) consistent with the loss of FOXO activator binding. These results suggest that the behavior of the IGFBP-1 IRSs differs in the context of the heterologous and native promoters. Potential explanations for this result are discussed below but additional evidence for a difference in IRS/FOXO function in different promoter contexts is apparent from overexpression studies comparing FOXO1a and FOXO3a. Figure 8 shows that, in H4IIE cells, overexpression of FOXO3a stimulates heterologous G6Pase-TK fusion gene expression more than FOXO1a, whereas in the same cells the effect of FOXO1a on G6Pase-luciferase fusion gene expression is greater than that of FOXO3a. This suggests that the ability of FOXO1a and FOXO3a to induce fusion gene expression is also context dependent.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
show that these conflicting data (17, 31) are not explained by a difference in FOXO1a and FOXO3a binding specificity. Instead, we find that gel retardation competition and binding experiments give different results; the latter reveal a correlation between FOXO binding and the inhibition of basal G6Pase gene transcription by insulin. Several other observations support a role for FOXO1 in the regulation of G6Pase gene transcription. Circumstantial evidence comes from the analyses of mice in which Foxo1 gene expression has been suppressed or in which a constitutively active form of Foxo1 has been overexpressed (40). Both manipulations lead to changes in endogenous G6Pase gene expression (40). However, in these types of experiments one cannot rule out the possibility that the effect of altering Foxo1 expression is indirect. The same caveat applies to the results of Barthel et al. (36) who showed that retroviral overexpression of FOXO1 in H4IIE cells stimulates endogenous G6Pase gene expression. More direct evidence comes from the results of chromatin immunoprecipitation (ChIP) assays. We have previously shown using the ChIP assay that FOXO1a binds the G6Pase promoter within intact HepG2 cells and that insulin treatment reduces this binding (8). The same result is seen in H4IIE cells (8) and mouse immortalized hepatocytes (41).
The studies reported here and the literature cited above strongly support the involvement of FOXO1a in insulin-regulated G6Pase gene transcription. In contrast, whether insulin also acts through FOXO1a to regulate PEPCK gene transcription is controversial. FOXO1a can bind the PEPCK IRS in vitro (22) and, when overexpressed, stimulate PEPCK fusion gene expression in situ (24). In addition, ChIP assays show that Foxo1 binds the PEPCK promoter in hepatocytes (41, 42) and adenoviral overexpression of a Foxo1 carboxy terminal mutant in mice alters PEPCK gene expression (43). However, in 6-month-old mice in which Foxo1 gene expression has been suppressed, or in which a constitutively active form of Foxo1 has been overexpressed, there are changes in G6Pase but not PEPCK gene expression (40). Similarly, Barthel et al. (36) found that retroviral overexpression of FOXO1 in H4IIE cells stimulated G6Pase but not PEPCK gene expression.
These conflicting data regarding the role of FOXO in insulin-regulated PEPCK gene transcription may relate to the observation that insulin can repress the expression of this gene through mechanisms that are independent of the PEPCK IRS motif (9). Thus, alternate PEPCK IRS-independent mechanisms for repression by insulin involving nuclear factor 
B (44), sterol regulatory element binding protein-1c (45), liver-enriched transcriptional-inhibitory protein (46), and Kruppel like factor-15 (47) have been proposed. The existence of alternate mechanisms for transcriptional repression has led to some confusion with several investigators concluding that the PEPCK IRS is not important for the regulation of PEPCK gene expression by insulin (24, 48, 49, 50). A common problem in these studies is that the investigators have studied the ability of insulin to repress the stimulatory effect of glucocorticoids on PEPCK gene transcription. However, the PEPCK IRS is colocalized with an accessory element, designated AF2, that is required for the stimulatory effect of glucocorticoids (51) and mutations that disable the PEPCK IRS but not AF2 have not been identified. Therefore, these studies only show that insulin, working through alternate mechanisms, is able to fully suppress the markedly reduced remaining stimulatory effect of glucocorticoids.
In addition to demonstrating a strong correlation between FOXO binding and insulin-regulated G6Pase gene expression the studies reported here also show 1) that FOXO1a binds cooperatively to G6Pase IRS 1 and 2 in vitro (Fig. 5) and in situ (Fig. 6), 2) that, based on a comparison of the effect of the IRS base 10 mutation reported here (Fig. 3) and by Guo et al. (17), promoter context alters the specific IRS base requirements for FOXO1a-stimulated fusion gene expression, and 3) that promoter context also affects the relative magnitude of FOXO1a- and FOXO3a-stimulated fusion gene expression (Fig. 8). Interestingly, the demonstration of cooperative binding of FOXO1a to G6Pase IRS 1 and 2 provides an explanation for paradoxical observation that FOXO1a binds with higher affinity to IRS 1 than IRS 2 but mutation of IRS 1 or 2 is equally deleterious to the insulin response (8). Whether there is a molecular reason for the differential affinity of FOXO1a binding to G6Pase IRS 1 and 2 is unclear, but it may relate to our recent observation that FOXO1a bound to G6Pase IRS 1 but not IRS 2 acts as an accessory factor for glucocorticoid-stimulated G6Pase gene expression (52). The molecular basis for the observed effect of promoter context on FOXO1a- and FOXO3a-stimulated fusion gene expression is also unclear but presumably relates to the differential interaction with, or modification by, other factors at the two promoters examined. Recent studies have shown that FOXO1a/3a can interact with multiple proteins (53) and are subject to complex regulation by acetylation/deacetylation (54) in addition to phosphorylation. Finally, it should be noted that liver cells are responsive to multiple hormones other than insulin such that the constitutive regulation of FOXO factors by endogenous hormones is likely to be much more complex than suggested by experiments in HepG2 cells in which only insulin action is analyzed.
Perhaps the most surprising result reported in this study is that the analysis of insulin action mediated through the G6Pase and IGFBP-1 IRSs in the context of a heterologous TK promoter reveals that signaling through the latter does not support the model for insulin-stimulated FOXO nuclear exclusion (Fig. 7). Thus, mutation of the G6Pase IRSs in this context results in a decrease in basal fusion gene expression (Fig. 7B), consistent with the loss of FOXO activator binding, whereas mutation of the IGFBP-1 IRSs actually results in an increase in basal expression (Fig. 7D). Rechler and colleagues (55) have previously shown that the inhibition of FOXO1-stimulated gene transcription by insulin does not require nuclear exclusion of FOXO1. However, FOXO1 dissociation from a promoter would still result in the inhibition of gene transcription without requiring nuclear exclusion. Two possible explanations for our unexpected results are that 1) the IRF binding the IGFBP-1 IRSs in this promoter context is not a FOXO factor, and 2) a FOXO factor still binds the IGFBP-1 IRSs in this promoter context but acts by a mechanism that involves insulin-stimulated repression rather than promoter dissociation. There is evidence to support both possibilities. For example, Patel et al. (56) have shown that a mammalian target of rapamycin-dependent pathway mediates the action of insulin on IGFBP-1 but not G6Pase gene expression. This would be consistent with a model in which a FOXO factor mediated insulin-regulated G6Pase but not IGFBP-1 gene expression. However, the regulation of FOXO phosphorylation is complex, and it is apparent that insulin can target this factor through kinases other than PKB (57, 58). Whether these unidentified kinases are a mammalian target of rapamycin dependent and whether phosphorylation can result in repression by FOXO factors in the absence of nuclear exclusion remains to be determined. A number of recent papers give circumstantial support for the alternate explanation for our observations, namely that a FOXO factor could function through a mechanism that involves insulin-stimulated repression rather than promoter dissociation in the heterologous promoter context examined. For example, Zhao et al. (59) have shown that FOXO1 can act as a corepressor and several genes have been identified whose expression are directly repressed by FOXO1 (60, 61, 62, 63). Whether insulin acts on these genes to relieve or augment repression mediated by FOXO1 has not been determined.
The promoter-specific characteristics that determine whether a gene is activated or repressed by FOXO1 remain to be determined. Possibilities include the precise complement of coactivators/repressors at a given promoter, but another important parameter may be the acetylation status of FOXO1. cAMP-response element-binding protein-binding protein has been shown to acetylate FOXO factors, a modification that is reversed by the nicotinamide adenine dinucleotide-dependent deacetylase silent information regulator 2 (Sir2) (64, 65, 66). Most reports suggest that acetylation inhibits FOXO action (64, 65, 67). This would be consistent with the recent observation that acetylation directly inhibits FOXO1 DNA binding affinity as well as promoting the phosphorylation of FOXO1 by PKB and hence promoter dissociation (68). However, one report suggests that acetylation stimulates FOXO action (66). Thus, it is apparent that acetylation modulates FOXO1 action, but it is also possible that it may influence whether FOXO1 activates or represses the expression of specific genes.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
) to generate the plasmids –231 IRS 1 M5 SDM, –231 IRS 2 M5 SDM and –231 IRS 1 + 2 M5 SDM. These mutations were generated in the context of the –231/+66 G6Pase promoter region (8) and were subcloned from the CAT to the pGL3 MOD vector. A three-step PCR strategy (6, 70) was used to generate G6Pase-luciferase fusion genes designated –231 IRS 1 + 2 M7 SDM and –231 IRS 1 + 2 M10 SDM that contain point mutations introduced into IRSs 1 and 2 at base 7 or 10, respectively (Table 1). The same strategy was used to generate the plasmid –231 IRS 2 BM that contains a block mutation of IRS 2 (Table 1) and the plasmids –231 IRS 1 M10 + IRS 2 M5 SDM and –231 IRS 1 M10 + IRS 2 BM that contain a point mutation in base 10 of IRS 1 in conjunction with a point or block mutation in IRS 2, respectively. These mutations were generated in the context of the –231/+66 G6Pase promoter region ligated into pGEM7 (Promega) and were then subcloned into the pGL3 MOD vector. The mutations introduced into these plasmids were identical with those used in gel retardation assays. All promoter fragments generated by PCR were completely sequenced to verify the absence of polymerase errors.
The plasmid TKC-VI contains the herpes simplex virus TK promoter ligated to the CAT reporter gene (27, 71). The TK promoter sequence extends from –480 to +51 and contains a BamHI site located between positions –40 and –35. Various double-stranded oligonucleotides representing the WT or mutated mouse G6Pase promoter sequence between –197 and –158 or the WT or mutated human IGFBP-1 promoter sequence between –124 and –96 (Table 2) were synthesized with BamHI compatible ends and cloned, as a single copy, in the same orientation as found in the G6Pase and IGFBP-1 genes, into BamHI-cleaved TKC-VI by standard techniques (72). The CAT reporter gene was then replaced with the more sensitive firefly luciferase reporter by reisolating the various TKC-VI promoter constructs, as HindIII-BglII fragments, from the plasmids described above and ligating them into HindIII-BglII digested pGL3 MOD.
The generation of human IGFBP-1-CAT fusion genes, containing promoter sequence located between (–132 and +68) and (–103 and +68), relative to the transcription start site, has been described (73) and were generously provided by Dr. David Powell. The CAT reporter gene was replaced with the more sensitive firefly luciferase reporter by reisolating these IGFBP-1 promoter sequences, as HindIII-XhoI fragments, from the plasmids described above and ligating them into HindIII-XhoI digested pGL3 MOD.
For protein expression in HepG2 and H4IIE cells, a plasmid encoding human FOXO3a (pECE-FOXO3a) and the empty pECE vector were generously provided by Dr. Michael Greenberg (18). A vector encoding human FOXO1a (pcDNA3-FKHR) was generously provided by Dr. Frederic Barr (22). The FOXO1a open reading frame was isolated from this plasmid as a KpnI-XbaI fragment and ligated into KpnI-XbaI digested pECE. A pCDM8 (Invitrogen) expression vector encoding the human insulin receptor was generously provided by Dr. Jonathan Whittaker (74).
For protein expression in bacterial cells, a vector encoding a GST-human FOXO1a fusion protein (pGEX-5X-3-FOXO1a) was generously provided by Dr. David Powell (22). A vector encoding a histidine-tagged variant of full-length human FOXO1a (His-FOXO1a) was constructed by isolating the FOXO1a open reading frame as a KpnI-XbaI fragment from pcDNA3-FOXO1a (22). This fragment was blunt ended using Klenow and ligated into XhoI digested, blunt ended pET-15b (Novagen, San Diego, CA). To generate a plasmid encoding a GST-full-length human FOXO3a fusion protein, the FOXO3a open reading frame was isolated from the pECE-FOXO3a plasmid described above as a HindIII-XbaI fragment and ligated into HindIII-XbaI digested pGEM7 (Promega). The FOXO3a open reading frame was then reisolated as a BamHI fragment and ligated into BamHI digested pGEX 2T (Novagen). All plasmid constructs were purified by centrifugation through cesium chloride gradients (72).
Cell Culture and Transient Transfection
Rat H4IIE hepatoma cells were grown in DMEM containing 2.5% (vol/vol) fetal calf serum and 2.5% (vol/vol) newborn calf serum. Human HepG2 hepatoma cells were grown in the same media supplemented with 5% (vol/vol) Nu serum IV (Becton Dickinson, Inc.). H4IIE and HepG2 cells were transiently transfected in suspension with the plasmids indicated in the figure legends using the calcium phosphate-DNA coprecipitation method as previously described (25, 27, 69). Cells (
3 x 106) were transfected with various firefly luciferase plasmids (15 µg) and simian virus 40-Renilla luciferase (0.15 µg), and where indicated with the pECE (2.5 µg), pECE-FOXO1a (2.5 µg), pECE-FOXO3a (2.5 µg), pcDNA3 (2.5 µg), pcDNA3-FOXO1a (2.5 µg), and insulin receptor (5 µg) expression vectors.
Luciferase Assays
Luciferase assays were performed using the Promega Dual-Luciferase Reporter Assay System according to the manufacturer’s instructions. In these experiments, we saw no effect of insulin or FOXO overexpression on firefly luciferase activity in either HepG2 or H4IIE cells. Therefore, for comparisons of basal, insulin-regulated, and forkhead-stimulated gene expression firefly luciferase activity directed by the various fusion gene constructs was expressed relative to simian virus 40-Renilla luciferase activity in the same sample. Each construct was analyzed in duplicate in multiple transfections, as specified in the figure legends, using three independent plasmid preparations.
Gel Retardation Assay
The plasmid encoding the GST-FOXO1a fusion protein (22) was transformed into Escherichia coli (XL1-Blue). Bacteria were grown to an OD600 of approximately 0.7 in LB supplemented with 200 µg/ml ampicillin and GST-FOXO1A expression was induced by incubation with 1 mM IPTG for 2 h at 37 C. Bacteria were pelleted by centrifugation, resuspended in 50 mM HEPES (pH 7.5), 200 mM NaCl, and 1 mM phenylmethylsulfonyl fluoride and lysed by sonication. His-FOXO1a and GST-FOXO3a fusion proteins were grown and partially purified as previously described (52).
Complementary oligonucleotides representing the WT or mutated G6Pase sequence between –196 and –155 (Table 1) were synthesized with BamHI compatible ends, gel purified, annealed, and then labeled with [
-32P]deoxy-ATP (>3000 Ci mmol–1; NEN Life Sciences, Boston, MA) by using the Klenow fragment of E. coli DNA polymerase I to a specific activity of approximately 2.5 µCi/pmol (72). Labeled oligonucleotide (7 fmol) was incubated with bacterial lysate containing GST-FOXO1a or partially purified His-FOXO1a and GST-FOXO3a in a reaction volume of 20 µl containing, at final concentrations, 25 mM HEPES (pH 7.5), 100 mM NaCl, 2 mM dithiothreitol, 10% glycerol (vol/vol), 1 mg/ml BSA and 20 ng of poly(dG-dC)·poly(dG-dC). For competition experiments, unlabeled competitor DNA was mixed with the labeled oligomer at the indicated molar excess before the addition of protein. After incubation for 10 min on ice, the reactants were loaded onto a 6% polyacrylamide gel and electrophoresed at 4 C for 120 min at 190 V in 0.5x TBE (72). Gels were pre-run for 30 min at 100 V. After electrophoresis, the gels were dried, exposed to Kodak XAR5 film, and binding was analyzed by autoradiography. Data were quantitated through the use of a Packard Instant Imager (Meriden, CT).
Statistical Analyses
The transfection data were analyzed for differences from the control values, as specified in the figure legends. Statistical comparisons were calculated using an unpaired Student’s t test. The level of significance was P < 0.05 (two-sided test).
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Disclosure Statement: The authors have nothing to disclose.
First Published Online July 13, 2006
Abbreviations: BM, Block mutation; CAT, chloramphenicol acetyltransferase; ChIP, chromatin immunoprecipitation; ER, endoplasmic reticulum; G6Pase, glucose-6-phosphatase catalytic subunit; GST, glutathione-S-transferase; HNF-1, hepatocyte nuclear factor-1; IGFBP, IGF binding protein; IPTG, isopropylthio-ß-D-galactosidase; IRF, insulin response factor; IRS, insulin response sequence; PEPCK, phosphoenolpyruvate carboxykinase; PKB, protein kinase B; SDM, site-directed point mutation; TK, thymidine kinase; WT, wild type.
Received for publication February 16, 2006. Accepted for publication June 29, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
