Forkhead Transcription Factor FoxO1 in Adipose Tissue Regulates Energy Storage and Expenditure

We purchased anti-FLAG (M2), anti-tubulin, and anti–UCP-1 (M-17) from Sigma; anti-FOXO1 (N18 and H128) and anti–PGC-1 (H-300) from Santa Cruz Biotechnology; anti-HA monoclonal antibody (12CA5) from Roche; and anti–phospho-FOXO1 (pS256 and pT24) and anti–4E-BP1 from Cell Signaling Technology. We used anti-FOXO1 antibody (H128) only when we tried to discriminate endogenous FoxO1 from FLAGΔ256, clearly because the epitope of H128 was the carboxyl terminus of FOXO1, which FLAG-Δ256 did not have.

We cloned a mutant FoxO1 cDNA, bearing a truncated FoxO1 (a.a 1-255) (Δ256) that has no transactivation domain and a FLAG peptide tag, into the SmaI site of plasmid pCMV5-aP2, in which the ClaI and SmaI-treated 5.4-kb promoter/enhancer fragment of the mouse aP2 gene was subcloned into the ClaI/SmaI-treated pCMV5/cMyc vector (19). We excised the transgene with ClaI and XhoI and gel-purified and injected it into fertilized eggs of FVB × Bl6 hybrid mice. The resulting embryos were implanted into CD-1 foster mothers. We screened offspring for transgene transmission by PCR. Of five independent transgenic lines obtained, two founders transmitted the transgene through the germline. Primers for genotypying were FLAG-S1: 5′-atggactacaaagacgatgac-3′ and SE1AS: 5′-gtcgagttggactggttaaac-3′.

We used only male mice for the following analyses. Animals were fed a standard diet and water ad libitum in sterile cages in a barrier animal facility with 12-h/12-h light/dark cycle. All experimental protocols using mice were approved by the animal ethics committee of Kobe University Graduate School of Medicine. A high-fat diet was begun at weaning (4 weeks of age) and continued for 15 weeks. We used the same high-fat diet as described previously (20). We measured glucose levels with a Glutest Pro (Sanwa Kagaku Kenkyusho) and insulin, leptin, and resistin by radioimmunoassay (Linco). We carried out all assays in duplicate. Each value represents the means of two independent determinations. Analysis was limited to male mice, as they are more susceptible to insulin resistance and diabetes. Glucose tolerance and insulin tolerance test were performed as described (21). Hepatic glycogen measurement was performed as described previously (21). Computed tomography was performed using a LaTheta LCT-100M (Aloka).

Six-month-old mice under the normal diet and mice under the high-fat diet for 15 weeks were monitored individually in a metabolic cage (O₂/CO₂ metabolism measuring system for small animals model MK-5000; Muromachi Kikai) with free access to a normal diet and drinking water for 48 h. Each cage was monitored for oxygen consumption at 5-min intervals throughout the 48-h period. Total oxygen consumption was calculated as accumulated oxygen uptake for each mouse divided by its body weight. We performed oxygen consumption of three to four mice in each genotype in each condition (normal-diet or high-fat–dieted condition). Representative graphs were drawn from the mean and SE calculated from data obtained in each measurement.

Isolation of total RNA from tissues and cells was performed using an SV Total RNA Isolation System (Promega) according to manufacturer's protocol. Real-time PCR was also performed as described previously (17). The primers used in this study were described in the online appendix Table 1 (available at http://dx.doi.org/10.2337/db07-0698). Northern blotting was performed according to the standard techniques. For preparation of the probe for FoxO1, we treated pFLAG CMV-2 WT FoxO1 (22) with EcoRI, and the probe β-actin used here was described previously (18).

We homogenized tissues in buffer containing 50 mmol/l Tris/HCl (pH 8.0), 250 mmol/l NaCl, 1% NP40, 0.5% deoxycholate, 0.1% SDS, and protease inhibitors (Roche Diagnostics). After centrifugation to remove insoluble material, each 50 μg of lysate was electrophoresed in 8% or 14% SDS-PAGE, and Western blotting was performed using indicated antibodies.

For immunohistochemistry of WAT and BAT, we incubated WAT and BAT overnight in 4% paraformaldehyde and embedded them in paraffin. We then mounted consecutive 5-μm sections on slides. After rehydration and permeabilization, we immunostained sections using OctA-Probe (d-8) (sc-807; Santa Cruz Biotechnology). After washing with PBS, we incubated the sections sequentially with biotinylated anti-rabbit IgG reagent (Vector Laboratories) and Vectastain Elite ABC reagent (Vector Laboratories) and visualized with the Liquid DAB Substrate Chromogen System (DakoCytomation). For histological analysis, we removed interscapular tissue from 15-week high-fat–dieted mice, fixed the specimens in 10% paraformaldehyde, and embedded them in paraffin. We mounted consecutive 10-μm sections on slides and stained them with hematoxylin and eosin. We calculated adipose cell size with NIH Image 1.62 software by manual tracing of at least 500 adipocytes for each genotype. Measurement of the size of adipocytes in WAT was performed using FLVFS-LS software (Flovel, Tokyo, Japan) by manual tracing of at least 500 adipocytes for each genotype.

After treatment with 1 mg/ml of collagenase I (Roche), cytoplasmic and nuclear extracts of 150 μg of BAT tissues from wild-type and transgenic mice were fractionated by differential lysis of cells in two consecutive steps using the NE-PER extraction reagents (Pierce). Protein concentrations in the cytoplasmic and nuclear extracts were determined using the Micro BSA protein assay kit (Pierce), and aliquots (30 μg) were resolved on 8 or 14% SDS-PAGE, followed by Western blotting using the indicated antibodies.

T37i cells were cultured in Dulbecco's modified Eagle's medium–Ham's F12 medium (Invitrogen) supplemented with 10% FCS, 2 mmol/l glutamine, 100 IU/ml penicillin, 100 μg/ml streptomycin, and 20 mmol/l HEPES and were grown at 37°C in a humidified atmosphere with 5% CO₂. Differentiation into mature adipocytes was achieved under standard conditions by incubating subconfluent undifferentiated T37i cells with 2 nmol/l triiodothyronine (T3; Sigma Chemical, St. Louis, MO) and 20 nmol/l insulin (Invitrogen) for 6 days. Construction of adenoviruses encoding FoxO1 mutants was described elsewhere (18). For infection with adenoviruses, after differentiation into mature adipocytes, cells were infected with adenoviruses at multiplicity of infections between 10 and 20. At 72 h after infection, cells were harvested.

We used RNAi-Ready pSIREN-Shuttle vector (Clontech) with GCTGCAGTACTCTCCTTATGG as the target sequence and selected target sequence of FoxO1 by Block-iT RNAi Designer (Invitrogen). We generated scrambled control plasmid using mismatch sequence GCGGCAGTACTCCCCTTAGGG. After transfection of cells with the knockdown vector using Lipofectamine (Invitrogen), cells were induced by differentiation medium. At day 5 after induction, cells were harvested and real-time PCR or Western blotting was performed.

We calculated descriptive statistics using ANOVA followed by Fisher's test (Statview; SAS Institute). P values of <0.05 were considered significant.

It has been already reported that FoxO family members (FoxOs) were expressed in both WAT and BAT and that haploinsufficiency of FoxO1 altered histological appearance and gene expression in WAT (17). However, it is not known whether FoxO1 has some physiological roles in mature adipose tissues, especially in BAT. To investigate roles of FoxO1 in adipose tissue, we generated adipose tissue–specific FoxO1 transgenic mice (aP₂-FLAG-Δ256) using an adipose tissue–specific 5.4-kb promoter/enhancer fragment of the mouse aP₂, also called fatty acid binding protein 4 (FABP4), and a FLAG-tagged mutant FoxO1 (FLAGΔ256), which lacks carboxyl terminal transactivation domain (18) (Fig. 1A). It has been reported that aP₂ promoter was active in WAT and/or BAT (23–25) and that the FLAGΔ256 can compete with endogenous FoxOs and inhibit expression of FoxOs' target genes (26). We generated two transgenic lines in which FLAGΔ256 was expressed mainly in BAT under the normal-diet condition (Fig. 1B). Transgene expression level in BAT in line 2 was more pronounced than in line 1 (Fig. 1B, lanes 2 and 4). The FLAGΔ256 mRNA in line 2 was expressed in adipose tissue, especially BAT (Fig. 1C, lane 2). The FLAGΔ256 protein was also expressed to a greater extent in BAT than in WAT in both lines (Fig. 1D). Finally, we examined intracellular localization of FLAGΔ256 in BAT from line 2 and detected a nuclear staining of the FLAGΔ256 in BAT from line 2 (Fig. 1E). In BAT, under a normal diet, endogenous FoxO1 is phosphorylated (Fig. 1F, lanes 1 and 2) and localized both in cytosol and in the nucleus (Fig. 1G, upper panel). However, on the high-fat diet, phosphorylation of endogenous FoxO1 is decreased (Fig. 1F, lanes 3 and 4) and FoxO1 is localized mainly in the nucleus (Fig. 1G, upper panel). In contrast, the FLAGΔ256 is localized constitutively in the nucleus in both the normal and high-fat diets (Fig. 1G, middle panel). Therefore, the FLAGΔ256 may compete with endogenous FoxOs and inhibit their transcriptional activity on promoter regions of their target genes in BAT. For subsequent analyses, we used line 2, which has a higher transgene expression levels in BAT than line 1.

To investigate the functional consequences of FLAGΔ256 overexpression in adipose tissue on glucose metabolism and energy homeostasis, we measured body weight, blood glucose, and serum insulin levels and performed intraperitoneal glucose and insulin tolerance tests. The aP₂-FLAG-Δ256 mice have a tendency to have reduced body weight under the normal diet. Especially, from 12 to 14 weeks of age, they showed statistically significant lower body weight than wild-type mice (Fig. 2A). At the age of 3 months, aP₂-FLAG-Δ256 had no significant differences of fed blood glucose and serum insulin levels compared with wild-type mice (data not shown). The intraperitoneal glucose tolerance test, insulin secretion, and the insulin tolerance test did not show any significant differences between wild-type and aP₂-FLAG-Δ256 mice (data not shown). At the age of 6 months, although body weight and fed blood glucose levels of aP₂-FLAG-Δ256 were not different from wild-type mice, fed serum insulin levels of aP₂-FLAG-Δ256 were significantly lower than wild-type mice (Fig. 2B and C). An intraperitoneal glucose tolerance test demonstrated that aP₂-FLAG-Δ256 tended to be more glucose tolerant compared with wild-type mice, but it did not reach statistical significance (online appendix Fig. 1A). However, insulin secretion during the intraperitoneal glucose tolerance test in aP₂-FLAG-Δ256 was significantly lower than in wild-type mice (online appendix Fig. 1B). Insulin tolerance test did not show any significant difference between aP₂-FLAG-Δ256 and wild-type mice (data not shown).

Next, in order to clarify whether the FLAGΔ256 overexpression in adipose tissue may affect glucose metabolism and energy homeostasis, aP₂-FLAG-Δ256 and wild-type mice were fed a high-fat diet for 15 weeks after weaning. Western blotting of endogenous FoxO1 in WAT demonstrated that FoxO1 is phosphorylated under normal diet, but under the high-fat–dieted condition, FoxO1 is dephosphorylated (Fig. 2D). Interestingly, a 15-week high-fat diet induced transgene expression in WAT as well as in BAT (Fig. 2E and F). Real-time PCR revealed that transgene expression in WAT under a high-fat diet increased by fivefold compared with under a normal diet (Fig. 2F). Consistent with the gene expression level of transgene, Western blotting demonstrated that FLAGΔ256 protein could be detected in WAT as well as in BAT (Fig. 2G). Immunohistochemistry showed that the FLAGΔ256 was mainly expressed in the nucleus of white adipocytes (Fig. 2H). Body weight was not different from wild-type mice (Fig. 2A). However, fed blood glucose levels of transgenic mice were significantly lower than wild-type mice (Fig. 2B), and serum insulin levels of transgenic mice tended to be lower than wild-type mice, but it was not statistically significant (Fig. 2C). Food intake and serum leptin level in aP₂-FLAG-Δ256 were not different from wild-type mice both under the normal and high-fat–dieted condition (online appendix Fig. 2). Fasting blood glucose levels of transgenic mice were significantly lower than wild-type mice and an intraperitoneal glucose tolerance test demonstrated that aP₂-FLAG-Δ256 were significantly more glucose tolerant than wild-type mice (Fig. 2I). An insulin tolerance test also demonstrated that insulin sensitivity of aP₂-FLAG-Δ256 was improved significantly compared with wild-type mice (Fig. 2J). To assess hepatic glucose production indirectly, we examined gene expression levels of G6Pase (G6pc) and Pepck (Pck1), which are involved in gluconeogenesis, and measured hepatic glycogen content. In both the normal and high-fat diet, although hepatic G6pc expression in aP₂-FLAG-Δ256 was significantly decreased compared with wild-type mice, Pck1 expression was not changed (Fig. 2K and L). However, the hepatic glycogen content of aP₂-FLAG-Δ256 was not different from wild-type mice (Fig. 2M). These data suggest that overexpression of FLAGΔ256 in adipose tissue ameliorates insulin resistance induced by a high-fat diet.

To investigate the mechanism of how FLAGΔ256 improves glucose tolerance and insulin sensitivity, we examined adiposity, histological analyses, and gene expression in WAT. Computed tomography scanning revealed that visceral fat mass was increased significantly and subcutaneous fat mass of transgenic mice was decreased slightly compared with wild-type mice under a high-fat diet condition, although muscle and whole adipose tissue mass of transgenic mice were not changed (Fig. 3A–D). Interestingly, histological analyses of epididymal fat demonstrated that the size of adipocytes from transgenic mice was significantly smaller than wild-type mice (Fig. 3E–G). We performed real-time PCR in order to investigate the effects of FLAGΔ256 on gene expression of WAT-specific metabolic and inflammatory genes. Under a normal diet, Slc2a4 and resistin (Retn) gene expression levels were increased and fatty acid synthase (Fasn) and tumor necrosis factor (Tnf) gene expression levels in WAT from transgenic mice were significantly decreased compared with wild-type mice (Fig. 3H). Furthermore, in WAT from transgenic mice under a high-fat diet, Fasn, adiponectin, Scl2a4, Retn, and Pparγ (Pparg) gene expression levels were increased, but Tnf gene expression was significantly suppressed (Fig. 3I). Serum resistin concentration of transgenic mice was not different from wild-type mice (online appendix Fig. 3). In addition, the expression level of chemokine (C-C motif) receptor 2 (Ccr2), a receptor for monocyte chemoattractant protein modulating metabolic and inflammatory effects induced by a high-fat diet (27), was significantly decreased in WAT from transgenic mice under both normal and high-fat diet and was significantly decreased compared with wild-type mice (Fig. 3J). These data suggest that overexpression of a mutant FoxO1 affects adipocyte size and gene expression in WAT.

It has been reported that FoxOs have various classes of target genes, which are involved in apoptosis, regulation of cell cycle, stress resistance, and DNA repair, other than glucose metabolism (28) and a constitutively nuclear FoxO1 (T24A/S253D/5316A)-inhibited adipocyte differentiation. Therefore, we hypothesized that overexpression of FLAGΔ256 in WAT may affect gene expression of FoxO target genes and adipocyte differentiation and finally affect adipocyte size and the examined gene expression of FoxO target genes. It can be speculated that gene expression levels of typical FoxO target genes should be suppressed by the FLAGΔ256. Interestingly, in WAT from aP₂-FLAG-Δ256 under a norma diet, gene expression levels of most FoxO1 target genes except p21 (Cdkn1a), Cyclin G2 (Ccng2), and Fas ligand (Fasl) were not affected (Fig. 3K). However, under a high-fat diet, gene expression levels of most FoxO1 target genes, except Cdkn1a and Catalase (Cat), were suppressed (Fig. 3L). These data suggest that overexpression of FLAGΔ256 under a high-fat diet affects the size of adipocytes and gene expression in WAT.

To investigate effects of overexpression of FLAGΔ256 on the physiological function of BAT, we measured resting rectal temperature. Although rectal temperature in aP₂-FLAG-Δ256 was similar to wild-type mice under a normal diet, rectal temperature under a high-fat diet was significantly higher than wild-type mice (Fig. 4A). Furthermore, oxygen consumption was measured using indirect calorimetry. Oxygen consumption of aP₂-FLAG-Δ256 was significantly higher than wild-type mice under both normal and high-fat diet conditions (Fig. 4B and C). Respiratory quotient, which represents utilization of carbohydrate or fat for a fuel source, of aP₂-FLAG-Δ256 was similar to wild-type mice under a normal diet (Fig. 4D and F). However, under a high-fat diet, the respiratory quotient of aP₂-FLAG-Δ256 tended to be higher than wild-type mice, especially in the dark state (Fig. 4E), and the average of the respiratory quotient of aP₂-FLAG-Δ256 was also significantly higher than wild-type mice (Fig. 4F). These data indicate that overexpression of the FLAGΔ256 in BAT increases oxygen consumption and energy expenditure and may reflect increased carbohydrate utilization, which spared triglyceride oxidation and stimulated the accumulation of adipose tissue.

To investigate the effects of overexpression of FLAGΔ256 on BAT, we measured weights of BAT and investigated histological appearance of BAT from wild-type and transgenic mice under a high-fat diet. Weights of BAT in aP₂-FLAG-Δ256 under a normal diet were not different compared with wild-type mice (data not shown). Furthermore, we could not detect any significant difference in weights of BAT in aP₂-FLAG-Δ256 under a high fat-diet compared with wild-type mice (Fig. 5A). However, numbers of brown adipocytes in BAT from aP₂-FLAG-Δ256 were increased significantly compared with wild-type mice (Fig. 5B and C). These data suggest that overexpression of FLAGΔ256 apparently affects the morphological appearance of brown adipocytes and reverses the histological phenotype of BAT induced by a high-fat diet.

To investigate effects of the FLAGΔ256 on BAT-specific gene expression, we performed real-time PCR using total RNA of BAT from wild-type and aP₂-FLAG-Δ256 mice under a normal or high-fat diet condition. Under a normal diet, overexpression of FLAGΔ256 in BAT increased Ppargc1b, Ucp1, andUcp2 gene expression significantly compared with wild-type mice (Fig. 5D). Furthermore, in BAT from aP₂-FLAG-Δ256 mice under a high-fat diet, Ppargc1b, Ucp1, Ucp2, andAdrb3, which have important roles in BAT function, gene expression levels were significantly increased compared with wild-type mice (Fig. 5E). In contrast, under the high-fat diet, overexpression of FLAGΔ256 in BAT significantly decreased 4ebp-1 (Eif4ebp1) gene expression (Fig. 5E). The Eif4ebp1gene has been reported to be a target gene of dFOXO in Drosophila (29,30) and in mice (31). Therefore, it can be speculated that the FLAGΔ256 suppresses Eif4ebp1 gene expression in BAT. Ppargc1a gene expression under a high-fat diet was decreased but not statistically significantly. These data suggest that overexpression of FLAGΔ256 in BAT affects gene expression and may ameliorate BAT function under a high-fat–dieted condition.

It has been reported that targeted disruption of Eif4ebp1 caused increased translation of PGC-1 protein (32). Therefore, it is worth it to investigate PGC-1α protein levels in BAT from aP₂-FLAG-Δ256 mice. To examine whether overexpression of FLAGΔ256 in BAT may affect PGC-1α protein expression levels, we performed Western blotting using lysates of BAT from wild-type and aP₂-FLAG-Δ256 mice under a normal or high-fat–dieted condition. In the normal-diet condition, PGC-1α, 4E-BP1, and UCP-1 protein levels in BAT from aP₂-FLAG-Δ256 tended to be higher, but these effects were not statistically significant (Fig. 5F and G). However, in the high-fat diet, PGC-1α protein levels were increased and 4E-BP1protein levels were decreased significantly compared with wild-type mice (Fig. 5F and H). UCP-1 protein levels in BAT from aP₂-FLAG-Δ256 was also increased but not statistically significantly (Fig. 5F and H). These data indicate that overexpression of FLAGΔ256 in BAT increases PGC-1α protein expression levels under a high-fat diet. Consistent with the findings above, overexpression of FLAGΔ256 in BAT increased gene expression levels for many genes of mitochondrial oxidative phosphorylation that are known to be enriched in BAT, which include Cox5b, CoxIII, Cox8b, Cox4i1, and cytochrome c (Cyc) (33), in both normal and high-fat–dieted conditions (Fig. 5I and J). These data suggest that overexpression of FLAGΔ256 increases mitochondrial gene expression and finally may upregulate physiological function of BAT.

To confirm whether the FLAGΔ256 in BAT increases PGC-1α expression levels and endows BAT-specific phenotypes on cells, we infected differentiated T37i cells with adenovirus encoding LacZ and constitutively nuclear (CN) or Δ256 FoxO1 and analyzed gene expression. T37i cells are derived from the hibernoma of SV40 large T antigen transgenic mice under the control of human mineralocorticoid receptor promoter. T37i cells maintain the ability to undergo terminal differentiation and remain capable of expressing Ucp1 upon retinoic acid and β-adrenergic stimulation (34,35). After T37i cells were differentiated into mature brown adipocytes, we infected cells with adenovirus encoding FoxO1 (Fig. 6B). We used adenoviruses encoding ADA FoxO1 as CN mutant and Δ256 FoxO1 (18). At day 3 after infection, we harvested cells and performed real-time PCR. Overexpression of the CN FoxO1 increased Eif4ebp1 and Ppargc1a gene expression significantly (Fig. 6A). In contrast, Eif4ebp1 and Ppargc1a gene expression levels in cells infected with the Δ256 FoxO1 were not changed compared with the LacZ-infected cells (Fig. 6A). However, although overexpression of the CN FoxO1 did not affect 4E-BP1 and PGC-1α protein levels, overexpression of the Δ256 FoxO1 in cells decreased 4E-BP1 protein levels and increased PGC-1α protein levels (Fig. 6B). Furthermore, overexpression of the Δ256 FoxO1 increased Ucp1 and Adrb3 gene expression in T37i cells (Fig. 6A). These data indicate that inhibition of transcriptional activity FoxO1 in T37i cells increases PGC-1α protein and Ucp1 and Adrb3 gene expression levels.

Because the truncated FoxO1 used in the present study lacks the COOH-terminal transactivation domain but retains the NH₂-terminal and forkhead DNA binding domain, it is possible that this mutant FoxO1 may affect functions of other Fox proteins, including FoxC2 and FoxA2 (36,37) and have effects on some FoxO-binding proteins, such as Sirt1 and PGC-1α (15,38). The best way to elucidate physiological roles of FoxO1 in adipose tissue may be the generation of adipose tissue–specific FoxO1 knockout mice. However, at the present time, we could not generate them. Therefore, we tried to investigate physiological roles of FoxO1 in T37i cells using short-hairpin RNA (shRNA). Transfection with shRNA was performed at the same day as the induction of adipocyte differentiation, and cells were harvested at day 5 after induction. Expression of shRNA reduced expression of FoxO1 by ∼80% (Fig. 6C, left panel, and D). In contrast, other FoxOs, which include FoxO3a and FoxO4, gene expression levels were increased significantly compared with cells transfected with a scrambled control shRNA (Fig. 6C, right panel). Knockdown of endogenous FoxO1 increased gene expression levels of Ppargc1a, Ppargc1b, Ucp1, and Adrb3 significantly and increased UCP-1 protein expression in the presence of isoproterenol (Fig. 6D and E). Pparg gene expression in shRNA-transfected cells increased significantly, and Eif4ebp1 gene expression levels were not changed compared with scrambled control–transfected cells (Fig. 6E). These data demonstrate that knockdown of FoxO1 confers the molecular phenotype of BAT cells onto T37i cells.

In the present study, we generated two lines of adipose tissue–specific mutant FoxO1 transgenic mice (aP₂-FLAG-Δ256) using an adipose tissue–specific 5.4-kb promoter/enhancer fragment of the mouse aP₂ gene. In both lines, FLAGΔ256 is expressed in BAT much more than WAT under a normal diet. The aP₂-FLAG-Δ256 showed improved glucose tolerance and insulin sensitivity under a high-fat diet. Overexpression of FLAGΔ256 in WAT lead to increased visceral fat mass, increased small adipocytes, and altered expression levels of adipose tissue–related genes, such as Fasn, adiponectin, Slc2a4, Retn, Pparg, Tnf, and Ccr2. Furthermore, overexpression of FLAGΔ256 in BAT leads to increased oxygen consumption accompanied with increased expression of PGC-1α, Ucp1, and Adrb3.

Interestingly, in the present study, a high-fat diet increased the expression level of FLAGΔ256 in WAT as well as BAT. Because it has been reported that a high-fat diet induced aP2 gene expression and content (39), it can be speculated that a high-fat diet activated aP2 promoter used in this study and induced transgene expression in WAT. Under a high-fat diet, phosphorylation of endogenous FoxO1 is decreased in adipose tissue. FLAGΔ256 in WAT and BAT from transgenic mice was also present in nucleus. Therefore, it can compete with endogenous FoxOs and inhibit their transcriptional activity. Indeed, expression levels of several FoxO1 target genes, which include p27 (Cdkn1b) (40), pRb (Rb1) (17), p130 (Rbl2) (41), Ccng2 (42), MnSod (Sod2) (43), and Bim (Bcl2l11) (44), were suppressed in WAT. The effects of overexpression of FLAGΔ256 on glucose tolerance and insulin sensitivity under a normal diet were smaller than under a high-fat diet. These can be explained by the findings that endogenous FoxO1 in adipose tissues under a normal diet is phosphorylated and overexpression of the FLAGΔ256 did not result in significant alterations in adipocytic gene expression in adipose tissue under a normal diet. However, under a high-fat diet, endogenous FoxO1 is dephosphorylated and may activate expression of its target genes. In that circumstance, the FLAGΔ256 can inhibit expression of FoxOs' target genes and normalize glucose tolerance, insulin sensitivity, and oxygen consumption.

The aP₂-FLAG-Δ256 mice showed increased expression of PGC-1α protein, which is a key coactivator for adaptive thermogenesis in BAT (9), and increased expression of the Ucp1 gene, body temperature, and oxygen consumption. It has been known that d4E-BP is one of the target genes of dFOXO in Drosophila (29), and recently it has been reported that Eif4ebp1 is a target gene of FoxO1 in C2C12 cells (31). Furthermore, Eif4ebp1^−/− mice exhibited smaller WAT and an increased metabolic rate due to increased expression of PGC-1α, Ucp1, and Adrb3 in WAT, although the mechanism whereby reduced expression of 4E-BP1 leads to enhanced expression of PGC-1α remains to be elucidated (32). Several genes that are involved in mitochondrial oxidative phosphorylation and are known to be enriched in BAT from transgenic mice were increased. Therefore, it is speculated that FLAGΔ256 in BAT may inhibit Eif4ebp1expression and increases PGC-1α protein expression through an unknown mechanism and drives the molecular phenotype of BAT. However, at cellular levels, we could not detect any significant differences of expression levels of Eif4ebp1 between LacZ- and Δ256-infected cells. This discrepancy has been shown by another study (45) that the same Δ256 FoxO1 increased adipoq expression, although it was shown to be a target gene of FoxO1. One of the possibilities might be that FLAGΔ256 used in the present study still had the NH₂-terminus and forkhead DNA binding domain and these domains could bind to some cofactors for transcription of FoxO1 target genes in a tissue-specific or cell type–specific manner. Using knockdown of endogenous FoxO1 by shRNA, we could demonstrate that ∼80% knockdown of endogenous FoxO1 in T37i cells enhanced the molecular phenotype of brown adipocytes through increased expression of Ppargc1a, Ppargc1b, Ucp1, and Adrb3. These data suggest that FoxO1 may suppress gene expression of BAT-specific genes and also suggest the possibility that FoxO1 could affect differentiation of T37i cells into mature brown adipocytes as well as the previous report that CN FoxO1 inhibits differentiation of white preadipocyte cell line 3T3-F442A cells (17) because Pparg expression in T37i cells transfected with sh-FoxO1 vector was increased significantly. Interestingly, although knockdown of FoxO1 increased FoxO3a and FoxO4 expression, the molecular phenotype of T37i cells as brown adipocyte was enhanced. This finding suggests the hypothesis that FoxO1 may play a specific role in brown adipocyte among FoxOs. To elucidate the mechanism how FoxO1 affects differentiation of brown adipocytes, further investigations will be needed.

Several mechanisms of how FLAGΔ256 in adipose tissue improved glucose tolerance and insulin sensitivity can be speculated. We demonstrated that in WAT from transgenic mice, the ratio of small adipocytes was increased significantly compared with wild-type mice. Increased numbers of small adipocytes might contribute to improved glucose tolerance and insulin sensitivity despite increased visceral fat mass. The size of adipocytes is associated with insulin sensitivity (46). Failure in recruitment of new fat cells due to impaired differentiation may lead to a reduction in the capacity of adipose tissue to accumulate lipids and be paralleled by enlargement of the existing adipocytes and also a “spill-over” and ectopic accumulation of lipids in other tissues, such as muscle and liver (47). Suppressed expression of several FoxO1 target genes, such as Cdkn1b, Rb1, Rbl2, and Ccng2, which are involved in cell cycle arrest, might increase proliferation of newly generated adipocytes. In addition, Bcl2l11, which induces cell death, expression level was also suppressed in WAT from transgenic mice. In addition, adipose tissue is an endocrine organ, which secretes several adipokines, such as adiponectin, leptin, and resistin. Enlargement of adipocytes might alter the amount of adipokines released (1). Expression of FLAGΔ256 in WAT affects gene expression, which includes increased expression of Fasn, adiponectin, Slc2a4, and Pparg and decreased expression of the Tnf and Ccr2 genes. These changes of gene expression may improve insulin sensitivity under a high-fat diet. It has been already reported that FoxO1 repressed Pparg promoter activity and transcriptional regulation of Slc2a4 (48). Therefore, the FLAGΔ256 may increase gene expression of Pparg and Slc2a4.

Transgenic mice in this study showed no significant difference of body weight a under high-fat diet, even though energy expenditure of transgenic mice was increased. This odd finding may be due to effects of FLAGΔ256 in WAT. In the present study, smaller adipocytes in WAT from transgenic mice were increased significantly compared with wild-type mice under a high-fat diet. These data suggest the possibility that overexpression of FLAGΔ256 might affect differentiation and proliferation of newly generated adipocytes. Furthermore, the finding that carbohydrate utilization of transgenic mice under a high-fat diet was increased suggests that triglyceride oxidation was spared and triglyceride accumulation was increased in WAT. Indeed, Fasn expression in WAT from transgenic mice was increased. Therefore, autonomous development of WAT may equalize increased energy expenditure and cause unchanged body weight.

The analyses using computed tomography revealed that visceral fat mass of transgenic mice is increased significantly compared with wild-type mice under a high-fat diet, although subcutaneous fat mass of aP₂-FLAG-Δ256 mice is not changed. In general, increased mass of visceral adipose tissue has been known to produce a much greater risk of diabetes, dyslipidemia, and accelerated atherosclerosis than subcutaneous adipose tissue (49) (6). However, adipocytes of visceral fat from transgenic mice were small and could be considered to be insulin sensitive and overcome insulin resistance induced by a high-fat diet. The mechanism of how overexpression of FLAGΔ256 affects visceral and subcutaneous fat mass differentially is still unknown. However, it has been suggested that different adipocyte precursors are responsible for a specific adipose depot development and have a different pattern of gene expression of developmental genes (50). Therefore, these intrinsic differences between subcutaneous and visceral fats might affect differential phenotypes.

The present study demonstrated that the mutant FoxO1 increased energy store in WAT through increased small adipocytes and spared triglycerides and increased energy expenditure in BAT through enhanced thermogenic capacity of brown adipocytes. Therefore, it can be speculated that FoxO1 in adipose tissue might decrease energy store and energy expenditure. FoxO1 in adipose tissues may have an important role for changing energy homeostasis in response to excessive energy intake and identify FoxO1 as a potential target in treatment of obesity and its associated disorders, such as type 2 diabetes.

This work was supported by a grant for the 21st Century Center of Excellence Program “Center of Excellence for Signal Transduction Disease: Diabetes Mellitus as a Model” from the Ministry of Education, Culture, Sports, Science and Technology of Japan; a grant from The Akiyama Foundation to J.N.; a grant from The Mother and Child Health Foundation to J.N.; a grant from Takeda Science Foundation to J.N.; and a grant from Uehara Memorial Foundation to J.N.

We thank Dr. Bruce M. Spiegelman (Dana-Farber Cancer Institute, Boston, MA) for kindly providing the promoter/enhancer fragment of the mouse aP2 gene. We also thank Dr. Susumu Seino (Division of Cellular and Molecular Medicine, Kobe University Graduate School of Medicine) for making his laboratory available to accomplish this work kindly. We are also grateful to Dr. Mutsuo Taiji (Dainippon Sumitomo Pharmaceuticals) for kindly making a metabolic cage available. Y.O. is Research Fellow of the Japan Society for the Promotion of Science.