|Home | About | Journals | Submit | Contact Us | Français|
The Forkhead transcription factor Foxo1 regulates expression of genes involved in stress resistance and metabolism. To assess the contribution of Foxo1 to metabolic dysregulation during hepatic insulin resistance, we disrupted Foxo1 expression in the liver of mice lacking hepatic Irs1 and Irs2 (DKO-mice). DKO-mice were small and developed diabetes; analysis of the DKO-liver transcriptome identified perturbed expression of growth and metabolic genes, including increased Ppargc1a and Igfbp1, and decreased glucokinase, Srebp1c, Ghr and Igf1. Liver-specific deletion of Foxo1 in DKO-mice resulted in significant normalization of the DKO-liver transcriptome and partial restoration of the response to fasting and feeding, near normal blood glucose and insulin concentrations, and normalization of body size. These results demonstrate that constitutively active Foxo1 significantly contributes to hyperglycemia during severe hepatic insulin resistance, and that the Irs1/2→PI-3K→Akt→Foxo1 branch of insulin signaling is largely responsible for hepatic insulin-regulated glucose homeostasis and somatic growth.
Hyperglycemia and dyslipidemia owing to hepatic insulin resistance are key pathologic features of Type 2 diabetes (Brown and Goldstein, 2008; Zimmet et al., 2001). In mice, near total hepatic insulin resistance can be introduced via the systemic or liver-specific knockout of key insulin signaling genes (Michael et al., 2000; Cho et al., 2001; Okamoto et al., 2007; Mora et al., 2005; Dong et al., 2006). Among these approaches, the compound suppression or deletion of the insulin receptor substrates, Irs1 and Irs2, is the least complicated by defective insulin clearance or liver failure (Taniguchi et al., 2005; Dong et al., 2006). Irs1 and Irs2 link the insulin receptor tyrosine kinase to activation of the PI 3-kinase → Akt cascade, which phosphorylates and inactivates numerous proteins to facilitate adaptation of hepatocytes to the fed state. Targets of Akt include inhibitors of macromolecular synthesis such as GSK3β (glycogen synthesis) and Tsc2 (protein synthesis); it also phosphorylates mediators of fasting gene expression such as Foxo1 and Crtc2 by SIK2, resulting in their degradation or exclusion from nuclei (Barthel et al., 2005; Dann et al., 2007; Dentin et al., 2007; Jope and Johnson, 2004).
The program of gene expression directed by Foxo1 and its cofactors ordinarily protects cells, as well as whole organisms, from the life-threatening consequences of nutrient, oxidative and genotoxic stresses (van der Horst and Burgering, 2007). For example, Daf16—the C. elegans ortholog of Foxo1—extends lifespan in nutrient-deprived worms in part by upregulating superoxide dismutase and catalase expression (Murphy et al., 2003). Foxo1 and paralogous forkhead box O family members counter DNA damage and growth factor withdrawal by suppressing cell cycle progression via upregulation of p27kip, and increasing expression of GADD45 and DDB1 to facilitate DNA repair (van der Horst and Burgering, 2007).
During prolonged starvation, hepatic Foxo1 ensures the production of sufficient glucose to prevent life-threatening hypoglycemia (Matsumoto et al., 2007). In healthy animals, the decreased insulin concentration during fasting promotes the nuclear localization of Foxo1, where it interacts with Ppargc1a and Creb/Crtc2 to increase the expression of the key gluconeogenic enzymes G6pc and Pck1 (Dentin et al., 2007; Koo et al., 2005; Puigserver et al., 2003; Schilling et al., 2006; Barthel et al., 2005; Mounier and Posner, 2006). Foxo1 also coordinates decreased nutrient availability with reduced somatic growth by increasing the hepatic expression of Igfbp1—a secreted factor that limits the bioavailability of Igf1 (Barthel et al., 2005). Finally, in conjunction with Creb/Crtc2, Foxo1 increases the expression of Irs2 and reduces the expression of the Akt inhibitor Trib3, which together can enhance fasting insulin sensitivity and augment the insulin response upon eventual feeding (Canettieri et al., 2005; Matsumoto et al., 2006).
These salutary effects of Foxo1 may, however, be abrogated by the presence of hepatic insulin resistance, in which case the persistent nuclear activity of Foxo1 and its co-factors might block adaptation of hepatocytes back to the fed state (Matsumoto et al., 2007; Samuel et al., 2006; Zhang et al., 2006). To establish genetically the degree to which hepatic Foxo1 alone contributes to hyperglycemia and diabetes during insulin resistance, we disrupted the Irs1 and Irs2 genes in hepatocytes of mice and determined the extent of their recovery upon the coincident disruption of Foxo1.
We used a Cre-loxP strategy to produce mice lacking hepatic Irs1 (LKO1-mice), Irs2 (LKO2-mice) or both genes (DKO-mice) (See Methods and Figure S1 in the Supplemental Data available with this article online). At 6 weeks of age, Irs1 mRNA and protein were undetectable in liver extracts from the LKO1 and DKO-mice, and Irs2 mRNA and protein were undetectable in extracts from the LKO2 and DKO-mice (Figure 1A–C). Irs1 and Irs2 expression was not affected in other tissues including adipocytes, skeletal muscle and brain in the knockout mice compared to the corresponding control mice (data not shown).
To assess insulin signaling in the liver, we analyzed liver extracts from mice treated with insulin or saline. Insulin stimulated the phosphorylation of Akt at the catalytic (T308) and hydrophobic (S473) sites in control liver extracts, and in liver extracts from LKO1 or LKO2-mice (Figure 1D). Consistent with this, the insulin–stimulated phosphorylation of Akt substrates Gsk3(S21), Gsk3β(S9) and Foxo1(T24 and S256) was also observed in control, LKO1 or LKO2 livers (Figure 1D). By contrast, insulin-stimulated phosphorylation of Akt(T308), Gsk3α/β or Foxo1 was undetected in DKO-liver, and the phosphorylation of Akt(S473) was reduced at least 70% (Figure 1D). Insulin failed to stimulate Erk1 phosphorylation (T202/Y204) in the DKO-liver, whereas Erk1 phosphorylation was stimulated 3-fold in the other groups of mice; the moderate insulin-stimulated phosphorylation of Erk2 was also absent in the DKO-liver (Figure 1D). Thus, either Irs1 or Irs2 was needed for the phosphorylation of hepatic Akt(T308) and Erk1/2 during acute insulin treatment; however, some alternative mechanisms might also contribute to about 30% of insulin-stimulated Akt(S473) phosphorylation in the absence of Irs1 and Irs2, including a possible insulin receptor-mediated mechanism as the tyrosine autophosphorylation of IR was stimulated normally by insulin in the liver of DKO-mice (Figure S1C).
DKO-mice developed hyperglycemia by 5 weeks of age when Irs1 and Irs2 deletion was nearly completed (Figure S1D). Consistent with overlapping roles of Irs1 and Irs2 to mediate hepatic insulin signaling, both LKO1 and LKO2-mice displayed normal fasting glucose and normal postprandial glucose and insulin concentrations at 8 weeks of age (Figure S2A–C). Compared with the control mice, fasting and postprandial glucose and insulin blood concentrations were significantly elevated in 8-week old DKO-mice (Figure S2A–C). A glucose tolerance test confirmed that the DKO-mice were diabetic at 8 weeks of age, as fasting and 2-hour blood glucose concentrations were significantly elevated (Figure 1E). By comparison, the LKO1-mice displayed mild glucose intolerance, whereas glucose tolerance of the LKO2-mice was indistinguishable from the control mice (Figure 1E). Consistent with these results, an intraperitoneal injection of insulin failed to reduce blood glucose in DKO-mice, whereas blood glucose decreased normally in the LKO2-mice (Figure 1F). Compared to the control and LKO2-mice, the LKO1-mice displayed a weaker—albeit not significantly weaker—response to the injected insulin (Figure 1F).
To further confirm that diabetes in DKO-mice resulted from loss of hepatic Irs1 and Irs2, we infected the DKO-mice by tail vein injection with adenovirus containing either control green fluorescent protein (Ad-GFP) or Irs1 or Irs2 (Ad-Irs1, Ad-Irs2) coding sequences. Ad-Irs1 or Ad-Irs2 restored the expression of hepatic Irs1 or Irs2, respectively, as well as the phosphorylation of Akt(S473) and Foxo1(S256) until the experiment was terminated (Figure 2A). Moreover, Ad-Irs1 or Ad-Irs2—but not Ad-GFP—restored the fasting blood glucose of DKO-mice to a normal concentration (Figure 2C). We conclude that the DKO-liver was not permanently damaged by 12 weeks of age, as glucose homeostasis was normalized by acute restoration of the Irs1-or Irs2-branch of the insulin signaling cascade.
Since Foxo1 was never inactivated in the DKO-liver by insulin-stimulated phosphorylation, we suppressed its expression with an adenovirus encoding an siRNA against Foxo1 (Ad-siFoxo1) to determine whether glucose homeostasis could be restored. Two days after the tail-vein injection, Ad-siFoxo1, but not Ad-siGFP, reduced Foxo1 expression by at least 75% in control and DKO-liver (Figure 2B, and data not shown). Moreover, Ad-siFoxo1 injection reduced the elevated fasting blood glucose concentration in the DKO-mice to the normal range (Figure 2C). Thus, constitutive Foxo1 activity contributed to the dysregulated glucose homeostasis in the DKO-mice.
We used the Cre-loxP strategy to produce “triple-knockout” (TKO) mice lacking Irs1, Irs2 and Foxo1 in hepatocytes. Irs1, Irs2 and Foxo1 were barely detected in liver extracts of 8 weeks old TKO-mice confirming that the deletion of all three genes was accomplished (Figure 2D). As expected, insulin signaling assessed by the phosphorylation of Akt(T308), Gsk3β (S9), or Erk1/2(T202, Y204) was not detected in liver extracts from TKO-mice (Figure 2E). Whereas insulin-stimulated Akt(S473) phosphorylation was largely impaired in DKO-liver, it was almost completely abolished in the TKO-liver. These results suggest that Foxo1 might play a positive role in Akt(S473) phosphorylation independently of Irs1 or Irs2 (Figure 1D and and2E2E).
Affymetrix GeneChips were used to establish the effect of Foxo1 upon gene expression in the DKO-liver. Analysis of RNA samples from the liver of LKO1, LKO2, DKO and TKO mice and the corresponding control mice revealed 9824 significantly changed probe sets (false discovery rate, FDR<0.05) that corresponded to 5756 annotated genes of which 420 displayed a maximal change of at least 1.5-fold (Table S1). Three principal components (PC)—each with a positively and negatively correlated gene cluster—accounted for 86% of the total expression variance (Figure 3A–C) (Sharov et al., 2005). Most of the significantly changed genes (86.7%) were associated with PC1, including 3531 displaying increased and 593 displaying decreased expression in the DKO-liver (Figure 3A). The dysregulated expression of these genes in the DKO-liver was largely restored in the TKO-liver to the normal range displayed by the control, LKO1 and LKO2-liver (Figure 3A).
PC2 accounted for the expression variance in 440 genes (9.1% of total variance), which responded positively or negatively to feeding regardless of the presence of Irs1, Irs2 or Foxo1 (Figure 3B). Many PC2 genes in the DKO-liver displayed either elevated expression under fasting conditions or an increased response to feeding. Thus PC2 genes are strongly regulated by fasting and feeding, but also modulated by insulin/FOXO signaling. PC3 accounted for the expression variance of 4.2% of the significantly changed genes, which were largely dysregulated in the liver of TKO-mice (Figure 3C). Thus expression of PC3 genes was largely independent of Irs1 or Irs2 signaling, but sensitive to the expression of Foxo1. Foxo1 itself—assessed by 3’-directed Affymetrix probe sets targeted against the deleted exon 2 in Foxo1 gene in the TKO-liver—was found in PC3 (Table S1).
Gene Set Enrichment Analysis (GSEA) revealed at least 50 transcription factor recognition sites which were significantly represented in the set of 5756 significantly changed genes (Table S2). A FOXO recognition site (TTGTTT, p<10−45) was relatively abundant as it occurred in at least 560 genes—69 of the genes that change by at least 1.5-fold (Table S1). Genes regulated by cAMP and glucocorticoids play an important role in the response to fasting; however, the consensus binding sites by CREB or GR were not among the top 50 recognition sites (Table S2). Regardless, two genes contained both FOXO and CREB sites (Ppargc1a and Maf), and two genes contained both FOXO and GR sites (Txnip and Pik3r1). Regardless, 83% of the genes that changed by at least 1.5-fold did not contain a consensus FOXO recognition site. However, it is possible that these 83% significantly changed genes could still be regulated by Foxo1 through either a non-consensus recognition site or protein-protein interaction without Foxo1 direct binding to the gene promoters. Other transcription factors or co-factors such as PGC1α might indirectly contribute to the effect of insulin and Foxo1 upon gene expression.
Next we used real-time PCR to analyze the expression of specific hepatic genes that regulate metabolism and growth, and to measure their response to feeding. As previously shown, feeding control mice increased the expression of Gck and decreased the expression of Pck1, G6pc, and Cpt1a (Badman et al., 2007; Yoon et al., 2001). Gck mRNA was not detected in fasted or fed DKO-liver and feeding failed to reduce the expression of Pck1 and G6pc in DKO-liver; however, Cpt1a responded normally (Figure 3D). Feeding control mice altered several hepatic regulatory factors including decreased expression of Ppargc1a and Fgf21 and increased expression of Srebp1c and Onecut1 (Figure 3E). However, in the DKO-liver the average expression of Srebp1c and Fgf21 decreased but remained sensitive to feeding; Onecut1 mRNA level was undetected; and Ppargc1a mRNA level was strongly increased (Figure 3E). Regardless, the average expression of these genes and sensitivity to feeding was at least partially restored in the TKO-liver (Figure 3D,E). Similar results were observed with the GeneChips for these and other genes (Table S1, and data not shown). Thus, Foxo1 prevented the adaptation of postprandial liver gene expression in DKO-mice, whereas this inhibition was released upon deletion of Foxo1.
Growth of the DKO-mice was retarded prior to the onset of significant hyperglycemia and this growth defect was persistent throughout the period analyzed (Figure 4C, Figure S1D). While DKO and TKO-mice consumed normal amounts of food and had a normal body composition at 3 months of age (Figure S5), the DKO-mice were shorter, had lower bone mineral density and 20% less body mass than the control mice (Figure 4A–C). By contrast, TKO-mice displayed normal body length, mineral density and body mass (Figure 4A–C). Hepatic genes that promote somatic growth—including Ghr, Igf1, and Igfals—were expressed weakly in DKO-mice, but normally in TKO-mice. Moreover, genes that inhibit growth, especially the Foxo1 target Igfbp1, were strongly increased in the liver of DKO-mice and rendered less sensitive to feeding (Figure 3F). The average expression of other genes associated with organismal growth and survival, were also dysregulated in DKO-mice but normalized in the TKO-mice (Table S1). Thus constitutively active hepatic Foxo1 per se was responsible for the observed growth deficit of DKO-mice (Figure 4A–C).
To contrast the response of the DKO and TKO-liver to nutrients, we investigated the phosphorylation of several signaling proteins in the liver of fasted (16 hours) and fed (4 hours after the fast) mice. The phosphorylation of AMPK, a key energy sensor, was not altered in the liver of fasted or postprandial DKO or TKO-mice compared to the control, suggesting that hepatic energy levels were not dramatically altered (Figure 4E). Feeding control mice stimulated the phosphorylation of Akt(S473); however, Akt(T308) and Erk1/2 phosphorylation were not strongly increased (Figure 4E). Phosphorylated Akt (T308 or S473) was undetected in the DKO and TKO-liver, and feeding had no effect upon Erk phosphorylation (Figure 4E). By comparison, feeding strongly stimulated the phosphorylation of S6K1 and ribosomal protein S6 in the liver of control, DKO and TKO-mice, showing that hepatic mTOR (mammalian target of rapamycin)-mediated nutrient sensing was at least partially independent of the Irs1/2→Akt→Foxo1 pathway (Figure 4E).
The dysregulated circulating concentrations of fasting glucose, insulin and adiponectin, and fed glucose were normalized in the TKO-mice (Figure S2A–D). Remarkably, fasting insulin resistance—estimated by the homeostasis model assessment (HOMA2)—was also reduced to the normal range in TKO-mice (Figure 4F). TKO-mice displayed a better response to injected insulin than DKO-mice, although this did not reach significance (Figure 4G). Compared to DKO-mice, glucose tolerance of TKO-mice was dramatically improved, whereas mice lacking hepatic Foxo1 alone (LKOF) were significantly more glucose tolerant than the controls (Figure 4H). Hepatic glycogen was lower in DKO-liver relative to control, but increased toward the normal range in TKO-liver (Figure 4D).
Serum FFA and triglyceride concentrations were low in the 8-week old DKO and TKO-mice compared to the normal serum concentrations in LKO1 and LKO2-mice (Figure 5A,B; Figure S3A,B). Hepatic triglyceride concentration was not significantly changed in 7-week old DKO or TKO-mice (Figure 5E). However, triglyceride secretion in 3-month old DKO mice was reduced by 50% compared with control mice, whereas triglyceride secretion was partially restored in TKO-mice (Figure 5F).
Total serum cholesterol concentrations were also low in the 8-week old DKO and TKO-mice compared to the normal cholesterol levels in LKO1 and LKO2-mice (Figure 5C; Figure S3C). To further examine cholesterol distribution and changes with time, we analyzed total serum cholesterol, high-density cholesterol (HDL), and low and very low-density cholesterol (LDL and VLDL) in the 4-month old control, DKO, and TKO-mice. LDL/VLDL-cholesterol concentrations were normal in DKO and TKO-mice; however, HDL-cholesterol concentrations were significantly lower in the DKO and TKO-mice compared to the control mice (Figure 5D). It is noteworthy that the serum HDL-cholesterol concentrations were increased by 52% (p<0.05) in the TKO-mice relative to the DKO-mice (Figure 5D). Interestingly, the protein levels of ApoA-I, ApoB48/100, and ApoE were indistinguishable between the DKO and TKO-mice and the control mice (Figure 5H). However, the expression of genes involved in cholesterol homeostasis (mainly cholesterol biosynthesis) and lipid synthesis were generally lower at the mRNA levels analyzed by Affymetrix GeneChips and not responsive to feeding in DKO-liver compared to the control liver (Figure 5G). By contrast, expression of the same set of genes was largely normalized in the TKO-liver, which was consistent with the improvement in serum HDL-cholesterol levels and triglyceride secretion in the TKO-mice (Figure 5D,F,G).
Our results suggest that Foxo1 is a dominant regulator of hepatic gene expression that is ordinarily inactivated through the Irs1- or Irs2-branch of the insulin signaling system. Without Irs1 and Irs2 (DKO-mice), the ordinary transition of liver gene expression from the fasted to fed state is inhibited, resulting in glucose intolerance and diabetes. We show that acute or chronic inactivation of Foxo1 in DKO-mice releases this inhibition and allows hepatic gene expression and metabolism to adapt more normally to the nutrient status. This finding is consistent with the important role of dFoxo in Drosophila, which is involved in nutrient response and metabolic adaptation through regulation of numerous biological processes (Gershman et al., 2007).
Our results do not support the conclusion that Foxo1 is preferentially regulated through the Irs2-branch of the insulin signaling cascade, as previously suggested (Wolfrum et al., 2004). Both Irs1 and Irs2 suffice for phosphorylation of Foxo1, and addition of either as an adenovirus to DKO-liver restores Foxo1 phosphorylation and rescues hyperglycemia. Our results do not confirm a predominant role for Irs2 in lipid metabolism, as previously suggested (Taniguchi et al., 2005). Glucose tolerance in LKO2-mice is largely indistinguishable from that of controls, at least at the age analyzed in this report, whereas it is mildly but significantly impaired in LKO1-mice. Thus Irs1 appears to have a greater contribution toward glucose homeostasis than Irs2. Hepatic Irs1 expression is generally stable, whereas Irs2 expression rises during fasting and falls after feeding, which might explain the mild dysregulation of hepatic metabolism in the absence of Irs2. Nevertheless, a uniquely dynamic interaction between Irs2 and IR suggests that Irs2 might play a more important role under some unknown conditions that have high cellular ATP levels (Wu et al., 2008).
Hepatic gene expression is dramatically altered by fasting and feeding owing to the effects of insulin during the postprandial state and the action of counter regulatory hormones including glucagon and glucocorticoids during fasting. Many hepatic genes are dysregulated in DKO-mice confirming that insulin signaling plays a critical role; however, upon deletion of Foxo1 in the TKO-mice, gene expression is largely normalized in both the fasted and postprandial states. Thus the inactivation of Foxo1 is essential before hepatic gene expression can respond to other signals. Although Foxo1 can be regulated by many mechanisms, its inactivation in postprandial liver appears to be under the dominant control of the Irs1/2→Akt cascade (Figure 6). Insulin fails in DKO-mice to stimulate hepatic Akt(T308) phosphorylation, which is required to activate Akt. Consequently, Foxo1 is not phosphorylated in the DKO-mice during intraperitoneal insulin injections or by feeding.
Insulin is also known to promote the phosphorylation of Akt at its hydrophobic C-terminal motif (S473), which contributes to substrate recognition (Manning and Cantley, 2007). Unlike the Akt(T308) phosphorylation, intraperitoneal insulin injections stimulate Akt(S473) phosphorylation in DKO-mice to about 30% of the normal response; however, Akt(S473) phosphorylation is barely detected in the TKO-mice. Although the insulin receptor is functional in the DKO- and TKO-mice, the mechanism coupling it to Akt(S473) phosphorylation in DKO-mice is unclear. One possibility is that the TSC1–TSC2 complex that is normally inhibited by insulin signaling may have a positive effect on Akt(S473) phosphorylation (Huang et al., 2008). Another possibility is that a constitutive Foxo1 in DKO-liver might promote Akt(S473) phosphorylation by modulating expression of Akt regulators such as Trib3 that binds to the Akt catalytic site (Matsumoto et al., 2006); however, the Affymetrix GeneChip data might suggest factors other than Trib3 in this regulation since the mRNA level of Trib3 was strongly increased in the DKO-liver compared to the control liver (Table S1). Regardless of the mechanism, Akt(S473) phosphorylation alone does not sufficiently inactivate Foxo1 because only Foxo1(T24) phosphorylation was abolished in mTORC2-deficient cells (Guertin et al., 2006; Jacinto et al., 2006), and it has been suggested that Foxo1(S256) phosphorylation plays a ‘gatekeeper’ role in the subsequent phosphorylation of other Thr/Ser residues in Foxo1 by insulin (Barthel et al., 2005).
Hepatic deficiency of Irs1 and Irs2 leads to a significant growth retardation in both prediabetic and diabetic DKO-mice. Consistent with this phenotype, the expression of several hepatic genes that promote organismal growth (Igf1, Igfals, and Ghr) were significantly reduced in DKO-mice, whereas Igfbp1 which inactivates circulating Igf1 was increased (Le et al., 2001). The deletion of hepatic Foxo1 (TKO-mice) restores normal expression of these genes, and the TKO-mice grow to a normal size. Thus postprandial inactivation of Foxo1 by insulin is essential for the usual systemic effect of growth hormone and hormonal Igf1 upon body growth (Figure 6).
Numerous gluconeogenic or regulatory genes, including Ppargc1a, Pck1, and G6pc are not suppressed in postprandial DKO-liver, but regain some sensitivity to feeding in TKO-liver. Such genes are likely to be positively regulated by Foxo1 as previously suggested (Altomonte et al., 2003; Daitoku et al., 2003; Matsumoto et al., 2007; Nakae et al., 2001; Puigserver et al., 2003; Zhang et al., 2006), or inaccessible to negative regulation when Foxo1 is active. Similar to our results, the elevated expression of gluconeogenic genes in Insr+/−-mice was normalized upon deletion one allele of Foxo1 (Nakae et al., 2002).
Onecut1 (also known as Hnf6) expression is reduced in DKO-liver and restored in TKO-liver, suggesting that it might be regulated by Foxo1; however, it does not contain a consensus FOXO-binding site in its proximal promoter. It might be regulated indirectly through Ghr→Jak2 signaling since expression of Ghr also increases in the TKO-liver (Figure 6). Other genes including Gck show a clear loss of feeding-induced transcription in DKO-liver, which is consistent with the finding that hepatic Gck is suppressed by a constitutively active transgenic Foxo1 (Zhang et al., 2006). Although Gck does not contain a FOXO core recognition motif, its expression is partially restored in the TKO-liver. The deletion of FOXO1 might indirectly promote Gck expression through Onecut1 (Lannoy et al., 2002). Onecut1 also induces expression of Foxa2 which may promote glycolysis, fatty acid oxidation and ketogenesis, and inhibit glucocorticoid-induced Pck1 expression in liver (Figure 6) (Pierreux et al., 1999; Rausa et al., 1997; Wolfrum et al., 2004). However, a significant change in Foxa2 expression was not detected by the Affymetrix GeneChip analysis.
Igf1 and Igfals expression might also be regulated by Foxo1 (directly or indirectly) in addition to known regulation by Stat5b (Figure 6) (Le et al., 2001). Together, these results reveal that insulin can play a permissive role in the regulation of endocrine growth through Foxo1-regulated gene expression of Ghr, Igf1, Igfbp1, and Igfals in coordination with growth hormone and nutrients (Figure 6). On the other hand, growth hormone also integrates with insulin to regulate metabolism, possibly through Onecut1-mediated gene expression of Foxa2, Gck, and Pck1 (Figure 6) (Lannoy et al., 2002; Pierreux et al., 1999; Rausa et al., 1997).
Similar to mice lacking hepatic insulin receptors (LIRKO) (Michael et al., 2000), DKO-mice have reduced circulating free fatty acid and triglyceride concentrations compared to controls. This effect is consistent with a decreased triglyceride secretion in DKO-mice, as suggested previously for the LIRKO-mice (Biddinger et al., 2008). However, both DKO and LIRKO-mice have nearly normal hepatic triglyceride contents although Srebp-1c expression is decreased in the liver of DKO-mice and LIRKO-mice (Biddinger et al., 2008). DKO-mice also have decreased serum cholesterol concentrations at 2 and 4 months of age, and the deletion of Foxo1 partially increases HDL-cholesterol concentrations in TKO-mice. This response is consistent with the normalized expression of cholesterol biosynthesis genes in TKO-liver. However, there are no significant changes in ApoB48/100, ApoA-I, or ApoE in DKO-mice compared to control mice, whereas ApoB48/100 protein levels were increased in the LIRKO-mice (Biddinger et al., 2008). Whether hepatic insulin resistance might increase susceptibility to atherosclerosis in DKO-mice needs further investigation. Overall, these results suggest that cholesterol homeostasis is also regulated by other insulin-dependent processes in addition to inactivation of Foxo1 (Gibbons, 2003; Horton et al., 2002).
In summary, we find that diabetes from a nearly complete loss of hepatic insulin signaling through the Irs1- and Irs2-branches of the pathway is alleviated by inactivation of Foxo1. Although other factors including Torc2 and Foxa2 play an important role in hepatic gene expression (Dentin et al., 2007; Wolfrum et al., 2004; Zhang et al., 2005), transcriptomic analysis suggests a majority of insulin-regulated genes are directly or indirectly controlled by the PI-3 kinase→Akt→Foxo1 branch of insulin signaling pathway. Thus drugs that interfere specifically with hepatic Foxo1 function are expected to have substantial effects upon dysregulated metabolism and might be of benefit for the treatment of diabetes, as previously suggested (Matsumoto et al., 2007). Finally, hepatic Foxo1 attenuates somatic growth by suppressing genes in the growth hormone signaling cascade, revealing how daily food intake can coordinate the systemic response to growth hormone.
All animal experiments were performed according to procedures approved by the Children’s Hospital Boston Institutional Animal Care and Use Committee. The floxed Irs2 and floxed Foxo1 mice were generated as previously reported (Lin et al., 2004; Paik et al., 2007). To generate liver-specific knockout mice, floxed Irs1, Irs2 and Foxo1 mice were crossed with Albumin-Cre mice (The Jackson Laboratory). The LKO1, LKO2, DKO, and their corresponding control floxed mice were maintained on a C57/BL6 and 129Sv mixed genetic background; the LKOF, TKO and their corresponding floxed mice were maintained on a C57/BL6, 129Sv and FVB mixed background. The procedure for generation of floxed Irs1 mice is described below.
The Irs1 targeting construct (Figure S1A) was based on the pPNT vector modified to contain two loxP sites flanking a neo selectable marker; the neo cassette was immediately preceded by unique NotI and XhoI sites and immediately followed by a unique BamHI site. The right arm of the construct was inserted at the BamHI site and consisted of the first 3 kb of genomic Irs1 DNA immediately 3' of the stop codon. A unique NheI site ~1.6 kb 5' of the Irs1 start codon served as the site of insertion for a third (distal) loxP site, destroying the NheI and creating a new BamHI site. The targeting construct was linearized for transfection into R1 ES cells. Following transfection, potential homologous recombinant ES cells were double selected in G418 and gancyclovir and screened by sequential Southern blotting of KpnI- and BamHI-digested genomic DNA with probes K and B (Figure S1A).
A single homologous recombinant ES cell clone containing the distal loxP was transfected with a Cre-expressing plasmid. The resulting colonies were screened with primers c2 (5'-CAGCAATGAGGGCAACTCCCCAAGACGCTCCA) and rev1 (5'-AGAGAGAAGCCCTTCTGTGGCTGCTCCAAACACA) to identify partial recombinants lacking the neo cassette but retaining Irs1 flanked by loxP sites (floxed Irs1 allele, Irs1L). Several non-unique floxed Irs1 ES subclones were injected separately into C57BL/6 blastocysts to generate highly chimeric mice containing cells heterozygous for the Irs1L alleles. The germlines of these mice were sampled by repeated matings to C57BL/6 females to generate agouti pups heterozygous for the Irs1L alleles.
Adenoviruses carrying either Irs1, Irs2, and GFP coding sequences, or Foxo1 siRNA or GFP siRNA were delivered to 12 week old male mice by tail vein injection at 6.5 × 107 viral particles per gram body weight. Mice were monitored for blood glucose levels at day 0 (pre-injection), 2, 3, and 4 (post-injection). Foxo1 shRNA was constructed using the sense sequence: 5'-gagcgtgccctacttcaag-3', short linker and the corresponding antisense of the above sequence.
Serum samples were collected from mice that were either overnight fasted or fed ad libitum, and were analyzed for insulin, free fatty acids, cholesterol, triglycerides, and adiponectin using commercial kits. Blood glucose levels were measured using a portable glucometer (Bayer). For glucose tolerance tests, mice were fasted overnight and injected with 2 g/kg body weight D-glucose intraperitoneally as previously described (Dong et al., 2006). For insulin tolerance tests, mice were fasted for 3 hours and injected intraperitoneally with 1 U/kg body weight humulin R (Lilly), and blood glucose levels were measured at indicated time points.
For VLDL triglyceride secretion analysis, mice were fasted for 4 hours during daytime and injected with 500 mg/Kg body weight Triton WR1339 via tail vein as previously described (Shachter et al., 1996). Blood samples were collected at 0, 1, and 2 hours after the injections and then analyzed for triglyceride concentrations using a commercial kit (Wako).
Glycogen content in liver was analyzed in 4-hour refed mice after overnight fast as previously described (Dong et al., 2006). Hepatic triglyceride content was analyzed in overnight fasted mice as previously described (Dong et al., 2006).
For insulin signaling analysis, mice were fasted overnight and anesthetized with Avertin (200 mg/kg body weight). A dose of 5 units of insulin was injected via vena cava. Liver, muscle, fat, and brain tissues were harvested after 4 min of stimulation. Tissue lysates were prepared as previously described (Dong et al., 2006). Equal amounts of tissue lysates were resolved by SDS-PAGE and transferred to nitrocellulose for immunoblotting analysis using specific antibodies. For quantitative analysis, enhanced chemiluminescence (ECL) signals on immunoblots were analyzed by Kodak Molecular Imaging Software.
Sixteen control mice (four Irs1L/L, four Irs2L/L, four Irs1L/L::Irs2L/L and four Irs1L/L::Irs2L/L::FoxOL/L) and sixteen liver-specific knockout mice (four LKO1, four LKO2, four DKO and four TKO) at 6 weeks of age were fasted for 16 hours; and half of the mice from each genotype were sacrificed for tissue collection (Fasted groups) and the other half were fed with regular chow for the next 4 hours and then sacrificed for tissue collection (Fed groups). Total RNA was isolated from the liver tissue using TRIzol (Invitrogen), and 15 µg of each RNA sample was used for microarray analysis on the Affymetrix MOE430 2.0 GeneChips by the Harvard Biopolymer facility. Microarray data were archived to the ArrayExpress database in a MIAME compliant format (accession number: X-XXXX-XXXX). Data analysis was detailed in the Supplemental Method for Affymetrix GeneChip Analysis available with this article online.
mRNA levels of genes analyzed in the experiments were quantified by real-time PCR. The procedure was essentially same as previously described (Dong et al., 2006). The relative levels of gene expression were calculated by normalizing to the internal control β-actin.
Data are presented as mean ± s.e.m. For all tests, α=0.05; downward adjustment of α (α*) for error control of multiple comparisons was made as appropriate. Significant differences, if any, are described in the figure legends.
The authors want to thank Dr. Ji-Hye Paik for the help in transferring Foxo1 floxed mice. We also would like to thank Mr. Ömer Hançer and Dr. Nancy J. Fidyk for the help in image processing and Dr. Florian Muller for critically reading this manuscript. This work was funded by the Howard Hughes Medical Institute and US National Institutes of Health grants DK038712 (MFW), DK055326 (MFW), and K99DK077505 (XCD).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.