Interaction of FoxO1 and TSC2 Induces Insulin Resistance through Activation of the Mammalian Target of Rapamycin/p70 S6K Pathway*

Both TSC2 (tuberin) and forkhead transcription factor FoxO1 are phosphorylated and inhibited by Akt and play important roles in insulin signaling. However, little is known about the relationship between TSC2 and FoxO1. Here we identified TSC2 as a FoxO1-binding protein by using a yeast two-hybrid screening with a murine islet cDNA library. Among FoxOs, only FoxO1 can be associated with TSC2. The physical association between the C terminus of TSC2 (amino acids 1280-1499) and FoxO1 degrades the TSC1-TSC2 complex and inhibits GTPase-activating protein activity of TSC2 toward Rheb. Overexpression of wild type FoxO1 enhances p70 S6K phosphorylation, whereas overexpression of TSC2 can reverse these effects. Knockdown of endogenous FOXO1 in human vascular endothelial cells decreased phosphorylation of p70 S6K. Prolonged overexpression of wild type FoxO1 enhanced phosphorylation of serine 307 of IRS1 and decreased phosphorylation of Akt and FoxO1 itself even in the presence of serum. These data suggest a novel mechanism by which FoxO1 regulates the insulin signaling pathway through negative regulation of TSC2 function.

Both TSC2 (tuberin) and forkhead transcription factor FoxO1 are phosphorylated and inhibited by Akt and play important roles in insulin signaling. However, little is known about the relationship between TSC2 and FoxO1. Here we identified TSC2 as a FoxO1-binding protein by using a yeast two-hybrid screening with a murine islet cDNA library. Among FoxOs, only FoxO1 can be associated with TSC2. The physical association between the C terminus of TSC2 (amino acids 1280 -1499) and FoxO1 degrades the TSC1-TSC2 complex and inhibits GTPaseactivating protein activity of TSC2 toward Rheb. Overexpression of wild type FoxO1 enhances p70 S6K phosphorylation, whereas overexpression of TSC2 can reverse these effects. Knockdown of endogenous FOXO1 in human vascular endothelial cells decreased phosphorylation of p70 S6K. Prolonged overexpression of wild type FoxO1 enhanced phosphorylation of serine 307 of IRS1 and decreased phosphorylation of Akt and FoxO1 itself even in the presence of serum. These data suggest a novel mechanism by which FoxO1 regulates the insulin signaling pathway through negative regulation of TSC2 function.
Similarly, dInR phosphorylates and inactivates dFOXO, the Drosophila homologue of DAF-16/FOXO (5). In mammals, InsR/IGF1R-PI3K-Akt signaling inhibits transcription by FoxO1, FoxO3a, and FoxO4 (6). These proteins possess a forkhead DNA binding domain consisting of around 110 amino acids and a transactivation domain in the C terminus. FoxOs bind to consensus FoxO-binding sites (GTAAA(C/T)A, T(G/ A)TTTAC) in the promoter region of their target genes and activate gene expression (7). It has been reported that FoxOs cause cell cycle arrest through induction of p27, p21, cyclin B, polo-like kinase, the retinoblastoma family-related protein p130 and cyclin G2, apoptosis through induction of Fas ligand and Bim, DNA repair through GADD45, stress resistance through MnSOD, and regulation of glucose and lipid metabolism through G6pase, apoC-III, and Igfbp-1 (8). Several FoxObinding proteins, which include co-activators, transcription factors, signaling molecules, and Sirt1, a NAD-dependent deacetylase, have also been identified (8). These FoxO-binding molecules regulate FoxO-dependent transcription and vice versa. However, there have been few reports that described identification of FoxO-binding proteins by comprehensive strategies.
The PI3K 2 -Akt pathway is also important for growth factor stimulation of mammalian target of rapamycin (mTOR) signaling (9). The primary mechanism by which Akt activates mTOR signaling appears to be through direct phosphorylation and inhibition of TSC2 (also known as tuberin). TSC1 (also known as hamartin) and TSC2 were first identified as genes mutated in patients with tuberous sclerosis complex (TSC), an autosomal dominant disease. Affected patients suffer from hamartomas in a wide spectrum of organs. TSC1 and TSC2 physically associate in vivo and form a heterodimeric complex (10 -12). TSC2 has been directly linked to cell size regulation by the discovery that mutation in dTsc2 leads to the gigas (large cell) phenotype (13).
Recent studies revealed that the TSC1-TSC2 complex functions downstream of Akt and upstream of target of rapamycin to restrict cell growth and cell proliferation (14 -17). Akt-phosphorylation of TSC2 leads to the functional inactivation of the TSC1-TSC2 complex and results in mTOR activation leading to phosphorylation of two main mTOR substrates, ribosomal p70 S6 kinase (p70 S6K) and eukaryotic initiation factor 4E-binding protein (4E-BP1), and elevated mRNA translation (18 -20). The TSC2 C-terminal region has homology with the catalytic domain of GTPase-activating proteins (GAPs). An inhibitory target of TSC1-TSC2 has been identified as Ras homologue enriched in brain (Rheb), a small GTPase. GTPbound Rheb is bound to and activates mTOR (21,22).
There have been several reports about molecules, which regulate TSC2 function. Energy depletion inhibits mTOR signaling through AMP-activated kinase phosphorylation of TSC2, although it is not know how AMP-activated kinase phosphorylation of TSC2 enhances the ability of TSC1-TSC2 to inhibit downstream signaling to Rheb (23)(24)(25). The hypoxia-inducible gene, regulated in development and damage responses (REDD1), is also induced by energy depletion, and this leads to inhibition of mTOR complex 1 signaling to p70 S6K in a TSC2dependent manner (26). It is important to identify molecules that regulate TSC2 function because these molecules may affect the activity of mTOR/p70 S6K signaling and finally determine the activity of PI3K/Akt pathway through a negative feedback loop (27).
In this study, we identified TSC2 as a novel FoxO1-binding protein by a yeast two-hybrid screening using a murine islet cDNA library. Binding of FoxO1 to TSC2 in cytoplasm inhibits TSC2 function and results in activation mTOR/p70 S6K and inhibition of Akt activity through negative feedback on IRS protein, leading eventually to feedback activation of FoxO1. Here we demonstrate a novel mechanism by which FoxO1 regulates activity of mTOR/p70 S6K signaling pathway and of FoxO1 itself through association with TSC2.
Construction of Expression Vectors-For mutagenesis of FoxO1, we performed overlap extension PCR using pCMV5/ cMyc/ADA FoxO1 as a template as described previously (35).
Yeast Two-hybrid Screen-Amino acids 424 -550 of the murine FoxO1 were cloned in-frame into the GAL4 DNAbinding domain plasmid pGBKT7 (Clontech). The GAL4 activation domain cDNA library of murine islets was constructed as described previously (36). AH109 yeast strain was used for the library search. The transformation was performed as described in the Clontech Matchmaker two-hybrid system 3 protocol. The transformants were plated on SD/ϪAde/ϪHis/ ϪLeu/ϪTyr plates in the presence of galactose and then were incubated at 30°C for 3-4 days. Positive interaction was identified by strong ␤-galactosidase activity. Individual positive clones were isolated by YEASTMAKER TM yeast plasmid isolation kit (BD Biosciences) and were sequenced by ABI310 automated DNA sequencer and analyzed for homology with sequences in the GenBank TM data base using the BLAST algorithm.
Cell Culture, Transfection, and Viral Transduction-HEK293 cells were cultured in DMEM containing 10% fetal calf serum. SV40-transformed hepatocytes used in these studies have been described in previous publication (29). Human vascular endothelial cells (HUVEC) were cultured in HuMedia-EB2 (KURABO) supplemented with 2% fetal calf serum, 10 ng/ml human recombinant epidermal growth factor, 1 g/ml hydrocortisone, 5 ng/ml human recombinant fibroblast growth factor, and 10 g/ml heparin. Transient transfection was performed using Lipofectamine (Invitrogen) according to manufacturer's protocol. Adenoviral infection was described in a previous publication (30). We transduced SV40-transformed hepatocytes by incubating them with adenoviral preparations at 10 -50 multiplicities of infection for 2 h.
Cell Isolation and Culture-Brown adipocytes and their precursor cells were isolated from newborn wild type mice by collagenase digestion as described previously (37). Preadipocytes were immortalized by infection with the retroviral vector pGCDNsamIRES-Puro, encoding SV40T antigen (38) and selected with puromycin (1 g/ml). Preadipocytes were grown to confluence in culture medium supplemented with 50 nM insulin and 50 nM triiodothyronine (differentiation medium) (day 0). Adipocyte differentiation was induced by treating confluent cells for 24 h in differentiation medium further supplemented with 0.5 mM isobutylmethylxanthine, 0.5 M dexamethasone, and 0.125 mM indomethacin. After induction, cells were changed back to differentiation medium, which was then changed every day. At day 5, cells were harvested and used for experiments.
In Vitro Translation and Glutathione S-Transferase Fusion Protein Pulldown Assay-The TSC2 deletions were generated by PCR using specific primers, and they were cloned in-frame into the EcoRI and SalI sites of pGEX-4T-1. These fusion proteins were expressed in 20 l of 50% slurry beads containing ϳ2 g of protein (either GST or alone, or fused to deleted TSC2 mutants), resuspended in 350 l of binding buffer (50 mM Tris-HCl (pH 8.0), 120 mM NaCl, 1 mM EDTA, and 0.5% Nonidet P-40). This was mixed with 10 l of in vitro translated wild type FoxO1 (Promega TNT reticulocyte lysate system kit). Binding was performed for 6 h at 4°C. The beads were then washed four times with the binding buffer and resuspended in 2ϫ SDS-PAGE sample buffer. Samples were then subjected to SDS-PAGE and transferred to a nitrocellulose membrane; the blot was incubated with anti-FOXO1 antibody and developed with an ECL detection system (Amersham Biosciences).
Measurement of GTP-and GDP-bound Rheb-SV40-transformed hepatocytes were cultured in 6-well plates and co-transfected with pCAG/FLAG-rTSC2DEE and pCA-EGFP-Rheb using Lipofectamine 2000 reagent (Invitrogen) and subsequently transduced with an adenovirus encoding WT FoxO1. At 48 h after transfection, the cells were washed once with phosphate-free DMEM (DMEM without sodium phosphate and sodium pyruvate; Invitrogen) and incubated with 1 ml of phosphate-free DMEM for 90 min. Cells were then incubated with 25 Ci of [ 32 P]phosphate/ml (GE healthcare) for 4 h. After the labeling, cells were lysed with prechilled lysis buffer (0.5% Triton X-100, 20 mM Tris (pH 7.5), 150 mM NaCl, 20 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 10 g of leupeptin/ml, 10 g of aprotinin/ml; 600 l per well of a 6-well plate). To avoid lysing the nuclei, the cells were incubated with lysis buffer for just 30 s with gentle shaking. The lysates were then centrifuged at 12,000 ϫ g for 15 min at 4°C. The supernatant (500 l) was transferred to a fresh tube. Sixteen microliters of NaCl (500 mM) was added to 160 l of supernatant to inhibit GAP activity in the lysates. To immunoprecipitate pCA-EGFP-Rheb, anti-green fluorescent protein and protein A-agarose (GE Healthcare) were added to the supernatant and incubated for 3 h at 4°C. The beads were washed with lysis buffer two times and with wash buffer (20 mM Tris (pH 7.5), 150 mM NaCl, 20 mM MgCl 2 ) one time at 4°C. The Rheb-bound nucleotides were eluted with 15 l of elution buffer (20 mM Tris (pH 7.5), 20 mM EDTA, 2% SDS) at 68°C for 10 min. Ten microliters of eluted nucleotides was then applied onto polyethyleneiminecellulose plates. Before applying sample, the plate was soaked in methanol and dried with a hair dryer. The bottom portion of the plate was immersed in methanol again, and the plate was placed in a sealed chromatography chamber that was filled with 0.75 M KH 2 PO 4 (pH 3.4) to a depth of 1 cm. The chamber was closed, and the solvent was allowed to ascend to the top of the plate. The plate was then removed and air-dried. GTP and GDP resolved by thin layer chromatography were visualized and quantified by a BAS-5000 (Fuji Film).
Design and Transfection of siRNAs-We used BLOCK iT RNAi Designer (Invitrogen) to identify target siRNAs and used Stealth RNAi (Invitrogen). The FOXO1-specific sequence was 5Ј-AACUGCAGAUGUCUGCUGAGCAUGU-3Ј. HUVECs were transfected with Stealth RNAi using Lipofectamine 2000 according to the manufacturer's instruction (Invitrogen). At 48 h after transfection, cells were harvested and used for Western blotting.
Statistics-We calculated descriptive statistics and ANOVA followed by Fisher's test using the Statview software (SAS Institute Inc.).

RESULTS
Identification of TSC2 as a FoxO1-binding Protein-To identify proteins that bind to FoxO1, we performed a yeast twohybrid screen using GAL4-FoxO1 fragment (amino acids 424 -550) as bait and a murine islet cDNA library as prey. About 2 ϫ 10 6 transformants were tested. The strongest colony was isolated and was found to encode the C-terminal fragment of murine TSC2 (amino acids 1280 -1815). To confirm the interaction between TSC2 and FoxO1, we co-transfected HEK293 cells with pCAG/FLAG-rTSC2DEE and pCMV5/cMyc-WT FoxO1, and we immunoprecipitated cell lysates using normal mouse IgG or anti-c-Myc mouse antibody and blotted with anti-FLAG antibody. Reciprocal immunoprecipitation/immunoblotting using anti-FLAG and anti-c-Myc antibodies showed that the exogenously expressed FoxO1 could interact with FLAG-tagged full-length-TSC2 (Fig. 1a).
Furthermore, to investigate whether endogenous FoxO1 is physically associated with TSC2 or not, cell lysates from brown adipocytes immortalized by SV40-T antigen were immunoprecipitated with anti-FOXO1 (Fig. 1b, lane 2) or anti-TSC2 (Fig.  1b, lane 5) and followed by an immunoblotting with using antibodies against TSC2 (Fig. 1b, lanes 1-3, top panel) or FOXO1 (Fig. 1b, lanes 4 -6, top panel). In this cell line, both FoxO1 and TSC2 are expressed abundantly. Our results showed that endogenous FoxO1 was associated with endogenous TSC2 (Fig.  1b). These results suggest that FoxO1 interacts physically with TSC2 in vivo.
TSC2 Binds to Only FoxO1 among FoxO Family Members-FoxOs consist of FoxO1, FoxO3a, and FoxO4. These molecules have highly conserved motifs, which include forkhead DNA binding domain, three Akt phosphorylation sites, and several acetylation sites. Therefore, it is interesting to investigate whether TSC2 is associated with other FoxOs or not. To examine whether all three FoxOs can interact with TSC2 equally, we transfected SV40-transformed hepatocytes with cMyc-FoxO1, -FoxO3a, or FLAG-FOXO4 and immunoprecipitated lysates with antibody against TSC2 or normal rabbit serum and immunoblotted with antibody against c-Myc or FLAG. This co-immunoprecipitation study demonstrated that only FoxO1 could interact with endogenous TSC2 (Fig. 2, lane 3, top panel). These data suggest that FoxO1 interacts with TSC2 among FoxOs specifically.
Identification of FoxO1-binding Site in TSC2 Protein-To examine whether FoxO1 interacts with TSC2 directly, we constructed several glutathione S-transferase (GST) fusion TSC2 fragments. Because the yeast two-hybrid screening identified the C-terminal fragment of TSC2 (amino acids 1280 -1815), we constructed several deleted Tsc2 mutants in this region (Fig.  3a). Using in vitro translated wild type FoxO1, pulldown assays with GST-deleted TSC2 fusion proteins were performed and showed that FoxO1 interacted with the C-terminal TSC2 fragment (amino acids 1280 -1686) directly. The FoxO1-binding site on TSC2 protein encompasses amino acids 1280 -1499 (Fig. 3b, lane 3), located near the GAP domain of TSC2. These data suggest that FoxO1 binds to TSC2 directly.
FoxO1 Co-localizes with TSC2 in Cytoplasm-To examine subcellular localization of the FoxO1/TSC2 interaction, we transfected SV40-transformed hepatocytes with pCAG/FLAG-rTSC2DEE, transduced them with adenovirus encoding with HA-WT or ADA FoxO1 (30), and performed immunofluorescence using anti-TSC2 polyclonal and anti-HA monoclonal antibodies. In this cell line, even in the absence of serum, 70 -80% of wild type FoxO1 was located in cytoplasm until 48 h after transduction (data not shown), where it co-localized with TSC2. In contrast, FLAG-TSC2 failed to co-localize with constitutively nuclear HA-tagged ADA FoxO1 (Fig. 4a). These data demonstrate that FoxO1 co-localizes with TSC2 in cytoplasm.    DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 Furthermore, to examine whether FoxO1 interacts with TSC2 in cytoplasm, we used a constitutively active mutant FoxO1 (3A FoxO1) in which all three Akt phosphorylation sites were mutated to alanine (T24A/S253A/S316A; 3A FoxO1) and performed co-immunoprecipitation experiments in the same cell line. Although exogenous WT FoxO1 interacted with TSC2 as well as in HEK293 cells (Fig. 1a), the 3A FoxO1 interacted with TSC2 weakly compared with wild type FoxO1 (Fig. 4b,  lanes 2 and 3). These data suggest the possibility that FoxO1 may associate with TSC2 in cytoplasm.

TSC2 as a FoxO1-binding Protein
FoxO1 Inhibits TSC2 and Enhances p70 S6K Phosphorylation-TSC2 regulates cellular function mainly by their inhibitory effects on mTOR and its targets p70 S6K and 4E-BP1. It is important to elucidate whether binding of FoxO1 to TSC2 affects the activity of the mTOR/p70 S6K pathway. To investigate the effects of the FoxO1/TSC2 interaction on the mTOR pathway, we transfected SV40-transformed hepatocytes with FLAG-p70 S6K followed by transduction with adenovirus encoding HA-WT FoxO1 or -ADA FoxO1, which is localized in the nucleus and active constitutively and is immunoprecipitated with anti-FLAG monoclonal antibody and blotted with anti-phospho-p70 S6K (Thr(P)-389) antibody. After serum deprivation for 24 h, p70 S6K is dephosphorylated, and insulin increases phosphorylation of p70 S6K (Fig. 5a, lanes 1 and 2 and  lanes 6 and 7). However, even in the absence of insulin, p70 S6K was phosphorylated in a dose-dependent manner of transduced WT FoxO1 (Fig. 5, a, lanes 3-5, and b, left panel). As described above, in this cell line, even after serum deprivation for 24 h, around 70 -80% of transduced wild type FoxO1 is located in the cytoplasm (data not shown) and is phosphorylated (Fig. 5a,  lanes 3-5). These data indicate that FoxO1 is constantly phosphorylated in this cell line even in the absence of serum and insulin. In contrast, p70 S6K was dephosphorylated in cells transduced with the ADA-FoxO1 (Fig. 5, a, lanes 8 -10, and b,  right panel). These data suggest that FoxO1 in cytosol enhances p70 S6K phosphorylation.
To confirm whether enhanced phosphorylation of p70 S6K by FoxO1 is mediated through binding to TSC2, we overexpressed FLAG-TSC2 in SV40-transformed hepatocytes transduced with adenovirus encoding HA-WT FoxO1, and we investigated the effects on phosphorylation of p70 S6K. Overexpression of TSC2 decreased phosphorylation of p70 S6K (Fig. 5c, lanes 1 and 2). Overexpression of WT FoxO1 enhanced phosphorylation of p70 S6K (Fig. 5c, lanes 1 and 3). Even in the presence of WT FoxO1, overexpression of TSC2 decreased phosphorylation of p70 S6K (Fig. 5c, lanes 3 and 4). These data suggest that cytoplasmic FoxO1 enhances phosphorylation of p70 S6K through association with endogenous TSC2. These data also suggest the possibility that cytoplasmic FoxO1 may affect an inhibitory action of TSC2 onto mTOR and activate mTOR.
FoxO1 Functions Upstream of mTOR for Activation of p70 S6K-To investigate whether enhanced phosphorylation of p70 S6K by overexpression of WT FoxO1 is mediated through mTOR activation, we treated cells with rapamycin (25 nM) and examined effects on p70 S6K phosphorylation. Overexpression of WT FoxO1 enhanced phosphorylation of p70 S6K (Fig. 6,  lanes 2 and 3). In contrast, treatment with rapamycin abolished p70 S6K phosphorylation induced by overexpression of FoxO1 (Fig. 6, lane 4). These data suggest that FoxO1 enhances phosphorylation of p70 S6K through activation of mTOR and mTOR functions downstream of FoxO1 for p70 S6K phosphorylation.
Physical Association of FoxO1 with TSC2 Inhibits GAP Activity toward Rheb-From this study, FoxO1 binds to amino acids 1280 -1499 of TSC2, which is near the GAP domain of TSC2. Therefore, it is possible to speculate that binding of FoxO1 may inhibit the GAP activity of TSC2. To investigate whether FoxO1 inhibits the GAP activity of TSC2 toward Rheb, we transfected cells with pCAG/FLAG-rTSC2DEE and pCA-EGFP-Rheb transiently followed by transduction with adenovirus encoding WT FoxO1 in SV40-transformed hepatocytes, and we examined guanyl nucleotide binding by EGFP-Rheb. Overexpression of TSC2 decreased %GTP by 33% compared with nontransfected cells (Fig. 8, a, lanes 1 and 2, and b). In contrast, overexpression of both TSC2 and FoxO1 increased %GTP by 30% compared with TSC2-transfected cells (Fig. 8a, lanes 2 and  3, and b). These data suggest that FoxO1 inhibits the GAP activity of TSC2 toward Rheb.
Knockdown of FOXO1 Reduces Phosphorylation of p70 S6K in HUVEC-If endogenous FoxO1 binds to TSC2 and inhibits its function, decreased expression of FoxO1 activates TSC2 and inhibits mTOR-p70 S6K pathway. To investigate whether endogenous FoxO1 inhibits TSC2, we transfected HUVEC with siRNA of FOXO1. In this cell line, both endogenous FOXO1 and TSC2 are expressed abundantly ( Fig. 9 and data not shown). Knockdown of FOXO1 in HUVEC decreased FOXO1 protein level by 90% (Fig. 9, a, top panel, and b). Knockdown of FOXO1 inhibited phosphorylation of p70 S6K (Fig. 9, 2nd  panel). These data suggest that endogenous FOXO1 inhibits TSC2 and regulates phosphorylation of p70 S6K in vivo.
Prolonged Overexpression of FoxO1 Enhances Phosphorylation of Ser-307 of IRS-1-It has been reported that p70 S6K might be implicated in a negative feedback loop to suppress insulin signaling (27). From this study, we demonstrated that  FoxO1 (lanes 8 -10). At 24 h after transfection, serum deprivation was performed for 16 h, and cells were stimulated with (lanes 2 and 7) or without insulin (100 nM) for 30 min. Cell lysates were immunoprecipitated (IP) with anti-FLAG monoclonal antibody and blotted with anti-phospho-p70 S6K (pT389) (upper panel) or anti-p70 S6K antibody (middle panel). The lower panel shows Western blotting of transduced FoxO1 using anti-HA monoclonal antibody. b, ratio of phospho-p70 S6K to total p70 S6K was calculated by measuring density of bands blotted with anti-phospho-p70 S6K or total p70 S6K using NIH Image 1.62. Data were shown as folds of basal level of phospho-p70 S6K in nontransduced cells in the absence of insulin and represent mean Ϯ S.E. from three independent experiments. Asterisks indicate statistically significant differences compared with the basal state (*, p Ͻ 0.005; **, p Ͻ 0.02; ***, p Ͻ 0.05 by one-factor ANOVA). c, SV40-transformed hepatocytes were transfected transiently with (lanes 2 and 4) or without pCAG/FLAG-TSC2 (lanes 1 and 3) and followed by transduction with adenovirus encoding WT FoxO1 (lanes 3 and 4). Cells were harvested and lysed at 48 h after transduction. Cell lysates were immunoblotted with the indicated antibodies.  DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 cytosolic FoxO1 bound to and inhibited TSC2 and enhanced phosphorylation of p70 S6K. We speculate that enhanced phosphorylation of p70 S6K may lead to phosphorylation of serine 307 in IRS-1, which is one of the phosphorylation sites by p70 S6K (39), and finally to decreased phosphorylation of Akt and FoxO1 itself. To investigate whether FoxO1-TSC2 binding affects phosphorylation of serine 307 of IRS1 or not, we transduced SV40-transformed hepatocytes with adenoviruses encoding LacZ or WT FoxO1 and cultured cells in the presence of serum, and we examined phosphorylation of serine 307 of IRS-1 in a time course study. Phosphorylation of p70 S6K in WT FoxO1-transduced cells at 48 and 72 h is increased compared with LacZ-transduced cells because of inhibition of TSC2 by overexpression of FoxO1 from the previous experiments (Fig. 10a, 3rd top panel). Amounts of total IRS1 protein level showed no significant differences between LacZ-and WT FoxO1-transduced cells (Fig. 10a, lanes 1-6, 2nd top panel). However, phosphorylation of serine 307 of IRS1 in WT FoxO1transduced cells is increased significantly compared with LacZtransduced cells at 72 h after transduction (Fig. 10, a, lanes 3  and 6, top panel, and b). Furthermore, at 72 h after transduction, phosphorylation of both threonine 308 and serine 473 of  P]orthophosphate was immunoprecipitated with anti-GFP antibody, and the ratio of GTP to GDP bound Rheb was determined by PhosphorImager analysis following one-dimensional thin layer chromatography. Representative experiment was shown. b, the values of %GTP were data from three independent experiments and were shown as mean Ϯ S.E. A single asterisk indicates a statistically significant difference between nontransfected and cells transfected with pCAG/ FLAG-rTSC2DEE (p Ͻ 0.001 by one-factor ANOVA). A double asterisk indicates a statistically significant difference between cells transfected with pCAG/ FLAG-rTSC2DEE and with both pCAG/FLAG-rTSC2DEE and WT FoxO1 (p Ͻ 0.005 by one-factor ANOVA). Akt was decreased even in the presence of serum and also phosphorylation of FoxO1 itself was decreased. These data suggest that FoxO1-TSC2 binding leads to enhanced phosphorylation of serine 307 of IRS-1 protein through increased phosphorylation of p70 S6K and finally inhibits phosphorylation of Akt and FoxO1 itself even in the presence of serum.

DISCUSSION
In this study, we identified TSC2 as a novel FoxO1-binding protein by a yeast two-hybrid screening using a murine islet cDNA library. FoxOs interact with several kinds of protein and regulate their function and vice versa. For example, FoxO1 binds to the transcriptional co-activator PGC-1␣, and PGC-1␣ potentiates FoxO1-dependent transcription of gluconeogenic genes (40). Acetylation by Cbp/P300 and deacetylation by Sirt1 regulates transcriptional activity of FoxOs (6). Therefore, it is important for understanding the mechanism of how FoxO1 is regulated to identify FoxO1-binding proteins. However, there are few reports about the identification of FoxO1-binding proteins using comprehensive strategies.
We used a murine islet cDNA library for identification of FoxO1-binding proteins because FoxO1 is expressed in pancreatic ␤-cells abundantly and has been reported already to play an important role for compensatory hypertrophy of ␤-cells under insulin resistance (41)(42)(43). Therefore, we speculated that it might be easy to identify FoxO1-binding proteins by using an islet cDNA library.
One of interesting findings in this study is that only FoxO1 binds to TSC2 among FoxOs. We used a fragment (amino acids 424 -550) of the C terminus of FoxO1 as bait for a yeast twohybrid screening. Amino acid sequences in this region of FoxO1, except the LXXLL motif (amino acids 459 -463) (44), have low similarity among FoxOs. Therefore, it is reasonable for only FoxO1 to associate with TSC2 physically. These findings suggest the hypothesis that FoxO1 may have specific roles in vivo. Studies using genetically modified mice, such as knockout and transgenic mice, support this hypothesis. FoxO1-null mice die at embryonic day 10.5 from defects in angiogenesis (45). Heterozygous mutant mice of FoxO1 are viable and rescue phenotype in heterozygous knock-out mice of insulin receptor or high fat diet-induced mice (46,47). FoxO3a-null mice are viable, and their main defect is an age-dependent female infertility because of premature activation of ovarian follicles (45,48). FoxO4-null mice are also viable and do not show any overt phenotype (45). Each FoxO family member may have a different function in vivo because of distinct patterns or different regulations of each protein.
Several studies in Drosophila demonstrated that TSC2 forms complex with TSC1, and this complex is important for cell growth regulation (13, 49 -51). This complex is assembled with rapid kinetics post-translationally serving to stabilize TSC1 and TSC2, which are ubiquitylated and degraded in their monomeric forms (18,52,53). TSC2 interacts with TSC1 through the N-terminal region and appears to function as a heterodimer (49,50,51). We demonstrated that FoxO1 bound to the C-terminal domain (amino acids 1280 -1499) of TSC2, and thereafter TSC2 was fragmented when FoxO1 was overexpressed in HUVEC as shown in Fig. 7a. This C-terminal fragment of TSC2 was still bound to FoxO1. However, TSC1 could not bind to this C-terminal fragment of TSC2 any more because TSC1 bound to the N terminus of TSC2. After these events, the association between TSC1 and TSC2 was reduced to negatively regulate its function. The C terminus of TSC2 has a GAP domain (amino acids 1517-1674) (54). It has been demonstrated that TSC2 functions as a GAP toward Rheb, which is a small G protein implicated genetically as a positive regulator of mTOR (55)(56)(57)(58), and that TSC2 represses Rheb function (59 -62). This study demonstrated that FoxO1 binds to an adjacent region near the GAP domain of TSC2 directly and inhibits the GAP activity toward Rheb. Inhibition of TSC2 by FoxO1 leads to enhanced phosphorylation of p70 S6K. In contrast, we demonstrated that knockdown of endogenous FOXO1 by around 90% decreased phosphorylation of p70 S6K in HUVEC. These data indicate that endogenous FOXO1 has an important role for regulation of phosphorylation of p70 S6K and might suggest a novel crosstalk between Akt/FoxO1 and mTOR/p70 S6K pathway through TSC2.
It has been reported that p70 S6K may be implicated in a negative feedback loop to suppress insulin signaling. High fatdieted p70 S6K1 Ϫ/Ϫ mice remained insulin-sensitive, and knockdown of p70 S6K1 also potentiates insulin-induced Akt phosphorylation (39). p70 S6K has an inhibitory effect on Akt activation downstream of insulin receptor. p70 S6K enhances IRS-1 serine phosphorylation, which leads to decreased Akt phosphorylation and causes insulin resistance (39). It has been suggested that p70 S6K mediates IRS-1 serine phosphorylation, disrupting its interaction with IR and leading to its degradation (63). Furthermore, degradation of phosphorylated IRS-1 is mediated by its association with a 14-3-3 family member, which relocates IRS-1 from low density microsomes to the cytosol, where it can be accessed and degraded by the 26 S proteasome (64). In this study, overexpression of WT FoxO1 in SV40-transformed hepatocytes enhanced phosphorylation of serine 307 of IRS-1 and decreased Akt phosphorylation at 72 h after transduction. In concordance with this result, phosphorylation of p70 S6K was enhanced. Finally, phosphorylation of WT FoxO1  2 and 3) or HA wild type FoxO1 (lanes 5 and 6) and cultured with complete medium at the indicated time. Culture medium was exchanged every day. Cell lysates were electrophoresed in SDS-PAGE and blotted with the indicated antibodies. b, intensity of bands blotted with anti-phospho-IRS1 (pS307) at 72 h after transduction were measured using NIH Image 1.62. Data represent mean Ϯ S.E. from three independent experiments. An asterisk indicates a statistically significant difference (p Ͻ 0.005 by one-factor ANOVA). DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 itself was blunted even in the presence of serum. These data suggest a novel mechanism in which FoxO1 in cytosol can regulate Akt/FoxO1 through the TSC2/mTOR/p70 S6K/IRS1 pathway (Fig. 11).

TSC2 as a FoxO1-binding Protein
In conclusion, these studies identified TSC2 as a novel interacting protein with FoxO1 and suggested that FoxO1 could negatively regulate TSC2 function. They suggest a novel crosstalk between Akt/FoxO1 and the mTOR/p70 S6K pathway and propose the possibility that FoxO1 can induce insulin resistance not only through increased gene expression in the nucleus but also through down-regulation insulin signaling in the cytosol. Therefore, regulation of the association between FoxO1 and TSC2 should be a target of therapy of type 2 diabetes.