FoxO transcription factors activate Akt and attenuate insulin signaling in heart by inhibiting protein

Yan G. Ni*†, Na Wang*, Dian J. Cao*, Nita Sachan*, David J. Morris*, Robert D. Gerard*‡, Makoto Kuro-o§, Beverly A. Rothermel*, and Joseph A. Hill*‡¶ʈ

¶Donald W. Reynolds Cardiovascular Clinical Research Center, Departments of *Internal Medicine (Cardiology), ‡Molecular Biology, and §Pathology, University of Texas Southwestern Medical Center, Dallas, TX 75390-8573

Edited by Eric N. Olson, University of Texas Southwestern Medical Center, Dallas, TX, and approved November 7, 2007 (received for review November 21, 2006) Insulin resistance and metabolic syndrome are rapidly expanding The O subfamily of Forkhead/winged helix transcription factors public health problems. Acting through the PI3K/Akt pathway, (FoxO) plays important roles in regulating cardiac and skeletal insulin and insulin-like growth factor-1 (IGF-1) inactivate FoxO muscle remodeling (8–11). FoxO proteins are phylogenetically transcription factors, a class of highly conserved proteins impor- conserved and regulate key physiological functions, including cell tant in numerous physiological functions. However, even as FoxO proliferation, cell differentiation, and survival (12–15). In addition, is a downstream target of insulin, FoxO factors also control FoxO functions in complex ways to regulate insulin signaling and upstream signaling elements governing insulin sensitivity and glucose and lipid metabolism (13, 16). In Drosophila, activation of glucose metabolism. Here, we report that sustained activation of dFoxO in fat body represses insulin-dependent signaling and in- either FoxO1 or FoxO3 in cardiac myocytes increases basal levels of creases life span (17). Similarly in mammals, FoxO1 regulates Akt phosphorylation and kinase activity. FoxO-activated Akt di- multiple metabolic pathways in liver and pancreatic ␤-cells (18), and rectly interacts with and phosphorylates FoxO, providing feedback transgenic expression of FoxO1 in various tissues leads to impaired inhibition. We reported previously that FoxO factors attenuate insulin sensitivity and glucose intolerance (18–20). On the other cardiomyocyte (PP2B) activity. We now show that hand, both dFoxO and the mammalian equivalent FoxO1 activate calcineurin forms a complex with Akt and inhibition of calcineurin transcription of the and insulin receptor substrate-2 enhances Akt phosphorylation. In addition, FoxO activity sup- (IRS-2) in Drosophila S2 cells and mouse C2C12 cells, which presses protein 2A (PP2A) and disrupts Akt-PP2A and sensitizes the cellular response to insulin (21). Also, whereas Akt–calcineurin interactions. Repression of Akt–PP2A/B interac- transgenic overexpression of constitutively active FoxO1 (caFoxO1) tions and phosphatase activities contributes, at least in part, to in liver triggers impaired fasting glucose and hyperinsulinemia (18, FoxO-dependent increases in Akt phosphorylation and kinase 22), acute overexpression of caFoxO1 in liver reduces plasma activity. Resveratrol, an activator of Sirt1, increases the transcrip- insulin, glucose, and triglycerides (23). And paradoxically, FoxO3- tional activity of FoxO1 and triggers Akt phosphorylation in heart. deficient mice exhibit impaired insulin sensitivity (24). Thus, many Importantly, FoxO-mediated increases in Akt activity diminish questions remain regarding the roles of FoxO proteins in insulin insulin signaling, as manifested by reduced Akt phosphorylation, signaling and their potential contribution to myocardial insulin reduced membrane translocation of Glut4, and decreased insulin- resistance. triggered glucose uptake. Also, inactivation of the gene coding for An additional layer of complexity exists in that the transcriptional FoxO3 enhances insulin-dependent Akt phosphorylation. Taken activity of FoxO is regulated by insulin through the phosphoinosi- together, this study demonstrates that changes in FoxO activity tide-3 kinase (PI3K)/Akt signaling pathway. Both insulin and have a dose-responsive repressive effect on insulin signaling in insulin-like growth factor-1 (IGF-1) induce PI3K/Akt-dependent cardiomyocytes through inhibition of protein phosphatases, which phosphorylation of FoxO, which facilitates its interaction with 14-3-3 protein, leading to nuclear exclusion and eventual ubiqui- leads to altered Akt activation, reduced insulin sensitivity, and tylation-dependent proteasomal degradation (25). Thus, it is well impaired glucose metabolism. established that Akt (also known as B, or PKB) plays a key role in repressing FoxO transcriptional activity. cardiomyocyte ͉ calcineurin ͉ insulin resistance ͉ cardiomyopathy Immediately upstream of FoxO, the activity of Akt itself is governed by several protein kinases and phosphatases. Akt is ardiovascular disease is the primary cause of death in patients activated by phosphorylation at Thr-308 within its catalytic domain Cwith obesity and type 2 mellitus. Insulin resistance, a by 3-phosphoinositide-dependent protein kinase-1 (PDK1) and by major mechanism underlying these disorders, also contributes to phosphorylation at Ser-473 within a C-terminal hydrophobic motif several cardiovascular disease-promoting factors, which, together, comprise the metabolic syndrome (1). In addition to its deleterious effects on metabolic parameters, such as lipid dyscrasias and blood Author contributions: Y.G.N., B.A.R., and J.A.H. designed research; Y.G.N., N.W., D.J.C., and pressure, insulin resistance of myocardial muscle itself has been D.J.M. performed research; N.S., R.D.G., and M.K.-o. contributed new reagents/analytic tools; Y.G.N. and J.A.H. analyzed data; and Y.G.N. and J.A.H. wrote the paper. documented in patients with type 2 diabetes and in nondiabetic The authors declare no conflict of interest. subjects with angiographically proven coronary artery disease (2). MEDICAL SCIENCES Indeed, strong evidence suggests that diabetes affects cardiac This article is a PNAS Direct Submission. structure and function, independent of blood pressure and coro- †Present address: Division of Cardiovascular Diseases, Merck Research Laboratories, Rah- way, NJ 07065. nary artery disease (3, 4). Insulin resistance has also been docu- ʈTo whom correspondence should be addressed at: Division of Cardiology, University of mented in heart tissue undergoing hypertrophic or atrophic remod- Texas Southwestern Medical Center, NB11.200, 6000 Harry Hines Boulevard, Dallas, TX eling (5, 6). Although much is known already regarding metabolic 75390-8573. E-mail: [email protected]. and insulin signaling pathways in the heart (3, 6, 7), there is This article contains supporting information online at www.pnas.org/cgi/content/full/ continued interest in defining mechanisms that govern insulin 0610290104/DC1. resistance and myocardial plasticity. © 2007 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0610290104 PNAS ͉ December 18, 2007 ͉ vol. 104 ͉ no. 51 ͉ 20517–20522 Downloaded by guest on October 1, 2021 1 P-Akt S473 that stem from supraphysiological levels of expression. To test this O 1 x P-Akt T308

A evitaleR BtkA/tkA-P O o P x F possibility, we evaluated a dose–response relation between FoxO F o a 4 F c ** G ** ** 3 expression and Akt activation. In these experiments, we observed

P-Akt (T308) slevel ** 2 significant increases in Akt phosphorylation even in cells infected P-Akt (S473) 1 with FoxO1 adenovirus at very low titers [multiplicity of infection Akt ϭ Tubulin 0 (MOI) 1, Fig. 1D]. In vitro kinase assays revealed significant GFP FoxO1 caFoxO1 increases in Akt kinase activity in cardiomyocytes overexpressing FoxO1 or caFoxO1 (Fig. 1E). We observed no effects of FoxO 3 C O D x o GFP FoxO1 overexpression on the phosphorylation levels of p38, p42/44 (ERK) P F F a G c 10 3 1 10 3 1 Adenovirus (MOI) MAP kinases, or Jun-amino-terminal kinase (JNK) (Fig. 1F), P-Akt (T308) FoxO1-GFP suggesting specificity of the pathway. P-Akt (S473) FoxO3/1 Next, we tested whether endogenous FoxO was capable of P-Akt (S473) Akt activating Akt. To accomplish this, we treated cells with resveratrol (RVT), an activator of the type III histone deacetylase Sirt1. E F Deacetylation by Sirt1 stabilizes the FoxO1 protein by preventing esanik 1 3 O * 1 x O o its proteasomal degradation (14). In addition, resveratrol increases P x F F o a G F c ytivitca 2 FoxO1 nuclear translocation and consequent expression of FoxO1 tkA * P-p38 MAPK (T180/Y182) target genes in hepatocytes (31). We observed similar results in evitaleR 1 P-p42/44 MAPK (T202/Y204) cardiomyocytes, where resveratrol increased the nuclear (active) 0 P-SAPK/JNK (Thr183/Tyr185) fraction of FoxO1, an effect opposite to that of IGF-1 [supporting GFP FoxO1 caFoxO1 Tubulin information (SI) Fig. 6 A and B]. Resveratrol also increased nuclear FoxO1 in the presence of IGF-1 (SI Fig. 6 A and B), suggesting that Fig. 1. FoxO proteins increase basal levels of Akt phosphorylation and kinase FoxO1 deacetylation in cardiac myocytes can override phosphory- activity. Cardiomyocytes were infected with adenovirus encoding GFP, FoxO1- lation-dependent mechanisms. Further, quantitative real-time PCR GFP, caFoxO1-GFP, or caFoxO3-GFP, and whole cell lysates were harvested 24 h showed that resveratrol elicited a significant increase in the mRNA after infection. (A–C) Western blots (A and C) and densitometric analyses (B) abundance of atrogin-1, a target gene of FoxO1 and FoxO3 (SI Fig. of phospho-Akt (Ser-473 or Thr-308). Results are graphed as mean Ϯ SEM from ␮ at least three independent experiments with duplicate or triplicate samples. 6C). Sustained treatment with resveratrol (20 M, 24 h) induced a (D) Western blot analyses of whole-cell lysates from cells infected with GFP or modest increase in FoxO3 and FoxO1 protein levels as well as in Akt FoxO1-GFP at MOI as indicated. (E) In vitro kinase assay of relative Akt kinase phosphorylation (SI Fig. 6D). In contrast, short-term resveratrol activities in cells overexpressing GFP, FoxO1-GFP, or caFoxO1-GFP. Results are treatment (Ͻ12 h) did not have such an effect (SI Fig. 6E). graphed as mean Ϯ SEM from two independent experiments with duplicate Together, these results suggest that endogenous FoxO, activated by samples. * and ** denote P Ͻ 0.05 and 0.01, respectively. (F) FoxO does not resveratrol, triggers Akt phosphorylation in heart. alter phosphorylation states of p38 MAP kinase, ERK, or JNK. Activated Akt Selectively Phosphorylates and Interacts with FoxO Proteins. To determine the consequences of FoxO-dependent Akt by mammalian target of rapamycin (mTOR) (26, 27). Recent activation, we evaluated multiple Akt targets. As expected, FoxO- studies show that phospho-Thr-308 and phospho-Ser-273 are de- induced increases in Akt activity led to an increase in FoxO protein phosphorylated by 1 (PP1), protein phospha- phosphorylation (both endogenous FoxO and overexpressed tase 2A (PP2A), and pleckstrin homology (PH) domain leucine- FoxO1-GFP) without affecting protein abundance (SI Fig. 7A). rich repeat protein phosphatase (PHLPP), a member of the protein Furthermore, FoxO-activated Akt had no effect on an Akt- phosphatase 2C family (28, 29). Akt regulates a variety of key insensitive mutant (caFoxO1-GFP, mutated at Akt consensus sites physiological functions, and there is strong evidence suggesting that T24A/S319A), even though the wild-type and mutant FoxO1-GFP defective Akt signaling contributes to development of insulin were expressed at similar levels (SI Fig. 7A). Surprisingly, phos- resistance (30). phorylation of two other Akt target proteins, glycogen synthase Thus, although it is clear that FoxO governs multiple events in the kinase-3 (GSK-3) and mammalian target of rapamycin (mTOR), insulin signaling cascade, mediating both positive and negative were not increased. effects, underlying molecular mechanisms are unknown. Here, we We next tested whether Akt interacts with FoxO using an report that sustained activation of FoxO in cardiomyocytes leads to immobilized antibody that binds preferentially to Akt phosphory- increased Akt phosphorylation and kinase activity. These changes lated at 473. Akt coprecipitated with FoxO1 proteins in are mediated, at least in part, by decreased activities of the protein cardiomyocytes expressing FoxO1-GFP but not in control cells phosphatases calcineurin and PP2A and impaired interactions expressing GFP (SI Fig. 7B). These results, then, suggest that between these proteins and Akt. Furthermore, FoxO-induced activated Akt interacts directly with FoxO proteins, leading to their increases in Akt activity attenuate the cellular response to insulin in phosphorylation. Consistent with this, Western blot analysis showed cardiomyocytes and results in decreased glucose uptake. Together, that FoxO proteins that coprecipitate with Akt are phosphorylated these findings provide a mechanism whereby FoxO activity con- (SI Fig. 7B). We did not detect interactions between FoxO and Akt tributes to reduced insulin signaling in heart. in uninfected control cardiomyocytes, where the majority of Akt and FoxO proteins are unphosphorylated (data not shown). Sub- Results cellular fractionation experiments in cardiomyocytes expressing FoxO Enhances Akt Activity. FoxO confers both positive and negative either GFP or caFoxO1 showed that Akt was mainly present in feedback in the insulin/Akt signaling cascade. This fact led us to cytoplasmic fractions (SI Fig. 7C). As expected, we detected study effects of FoxO on Akt, its proximal upstream regulator. We increased levels of phosphorylated Akt in cells expressing caFoxO1. observed that forced expression in cardiomyocytes of wild-type or Thus, FoxO-dependent activation of Akt did not trigger its trans- a constitutively active (ca) mutant FoxO led to increased phos- location to the nucleus. phorylation of Akt at both Thr-308 and Ser-473 without affecting Akt protein abundance (Fig. 1 A and B). Similar effects were also FoxO Diminishes Calcineurin Activity Without Affecting Upstream observed with overexpression of FoxO3 (Fig. 1C), the other major Regulators of Akt. To explore mechanisms underlying FoxO- FoxO isoform in heart. dependent increases in Akt activity, we evaluated the effects of Protein overexpression studies can be limited by spurious findings FoxO on phosphatases and kinases known to target Akt. We first

20518 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0610290104 Ni et al. Downloaded by guest on October 1, 2021 A PP2A A GFP caCnA MCIP1

ytivitca esatahp ytivitca calcineurin

B l

P-Akt (S473) tkA/tkAp evel 3 1.2

Akt evitaleR **

evitaleR ** * 2 0.8 1 ** ** * Con 1µM OA FK506 LY 0.4 C 0

sohp GFP caCnA MCIP1 0.0 P-Akt (S473) GFP caFoxO1 caFoxO3 Akt D l 8 e

vel * evitale 6 B MCIP1 Exon4 MCIP1 Exon1 t Con OA OA CLA k leve 4

E A

0.2µM 1µM / 8 1.5 t * R k 2 l ANR l ** P-Akt (S473) Ap * 6 0 1.0 Akt Con OA FK506 LY 4 1µM m m

ev 0.5 in vitro dephosphorylation assay

i 2 level tkA/ level t **

a ** F l G WCL IP 9 eR 0 0.0 e + + + + CnA v italeR 6 DMSO RVT MG132 DMSO RVT MG132 + + + + IGF-1 *

t 3 kAp * 1 P-Akt (S473) 1 xO 0 O o C D P x F Akt Con OA OA CLA F o a 1 G F c 0.2µM 1µM O 1 x O o 1 2 3 4 5 6 7 8 P x F F o a P-PDK1(S241) G F c Fig. 3. Inhibition of calcineurin increases Akt phosphorylation. (A and B) PP1 PDK1 Western blots and densitometric analysis of total or phosphorylated Akt PP2A PI3Kp85 protein levels in extracts from cardiomyocytes infected with adenovirus en- Calcineurin coding GFP, constitutively active calcineurin (caCnA), or MCIP1.4. (C–F) West- PI3Kp110α Tubulin ern blots (C and E) and densitometric analyses (D and F) showing effects of PI3Kp101 phosphatase inhibitors and LY294002. Serum-starved cardiomyocytes were PI3Kp110γ treated for 1 h with okadaic acid (OA, 0.2 or 1 ␮M), FK506 (10 ␮M), LY294002 Tubulin (LY, 10 ␮M) or for 15 min with calyculin A (CLA, 100 nM). (G) Western blots of whole-cell lysates (WCL, lane 1 and 2) or immunoprecipitated Akt treated with Fig. 2. FoxO decreases protein phosphatase activity. Cardiomyocytes were human recombinant calcineurin (lane 3–8) in vitro. Cells were treated with IGF harvested 24 h after adenovirus infection. (A) Relative phosphatase activity (10 nM, 15min) (lanes 2 and 6–8), and WCL were harvested for Western was normalized to protein content (experiments repeated three times). Re- blotting or for in vitro dephosphorylation assay. Eight (lanes 4 and 7) or 40 sults are graphed as mean Ϯ SEM of a representative experiment with qua- (lanes 5 and 8) units of human recombinant calcineurin (BIOMOL) were used. Ͻ druplicate samples. * and ** denote P 0.05 and 0.01, respectively. (B) *, P Ͻ 0.05 relative to control; **, P Ͻ 0.01 relative to control. Resveratrol treatment decreases mRNA abundance of MCIP1.4 but not MCIP1.1. Data are mean Ϯ SEM (n ϭ 6–8). **, P Յ 0.01. (C) Western blot of the catalytic subunits of PP1, PP2A, and calcineurin in cells expressing FoxO1-GFP Akt Is a Target of Calcineurin. Phospho-Akt is not known to be a or caFoxO1-GFP. (D) Western blot of the protein abundances of the regulatory substrate of calcineurin phosphatase. Therefore, we examined (PI3Kp85, PI3Kp101) and catalytic subunits (PI3Kp110␣, PI3Kp110␥)ofPI3 kinases and the total and phosphorylated protein abundance of PDK1 in cells directly whether perturbation of calcineurin activity affected Akt expressing FoxO1-GFP or caFoxO1-GFP. phosphorylation in cardiomyocytes. Viral overexpression of con- stitutively active calcineurin (33) triggered a significant decrease in Akt phosphorylation, and overexpression of MCIP1.4, an endog- tested the effect of FoxO on PP2A activity (28). Viral expression of enous inhibitor of calcineurin (34), had the opposite effect (Fig. 3 constitutively active FoxO1 or FoxO3 led to modest but consistent A and B). Consistent with these results, FK506 (Fig. 3C) and CSA decreases in PP2A activity (Fig. 2A). In agreement with our (data not shown), selective inhibitors of calcineurin, caused a previous findings (11), forced expression of active FoxO1 or FoxO3 significant increase in the level of phosphorylated Akt. Okadaic also significantly inhibited calcineurin phosphatase activity (Fig. acid, at a concentration that inhibits both PP2A and PP1 (1 ␮M) 2A). We also tested the effect of resveratrol treatment on PP2A/B had a stronger effect (Fig. 3C). As expected, inhibition of PP2A activity and found no changes (data not shown). However, using alone with low concentrations of okadaic acid (0.2 ␮M) also quantitative real-time PCR, a more sensitive assay, we detected in increased Akt phosphorylation, and inhibition of all serine/ resveratrol-treated cells significant decreases in the mRNA abun- threonine phosphatases with calyculin A had effects greater than dance of MCIP1.4, a downstream target of calcineurin, but not that inhibiting each phosphatase alone (Fig. 3 E and F). LY294002, a of MCIP1.1, an isoform whose expression is not driven by cal- selective inhibitor of PI3 kinase abolished phosphorylation of Akt cineurin (Fig. 2B). MCIP1.4 mRNA abundance was increased after in the basal state. Also, inhibition of calcineurin by MCIP1.4 or CSA treatment with MG132, consistent with previous reports that augmented FoxO1-induced increases in Akt phosphorylation when atrogin-1-related proteasomal activity is involved in promoting cells were infected at low titer (MOI ϭ 3) (SI Fig. 8A). Inhibitors calcineurin degradation (11, 32). of calcineurin or PP2A had no effect on the phosphorylation of

Further studies demonstrated that increases in either FoxO1 ERK or p38 under both basal and insulin-stimulated conditions, MEDICAL SCIENCES (Fig. 2C) or FoxO3 (data not shown) activity triggered minimal arguing against nonspecific effects (SI Fig. 8 B and C). decreases in the abundance of the catalytic subunits of these protein To test whether Akt is a direct target of calcineurin, cardiomy- phosphatases. FoxO overexpression did not alter the protein abun- ocytes were treated with IGF-1, and phosphorylated Akt was dance or phosphorylation of PDK1, nor did it affect the abundance immunoprecipitated with an Akt antibody and then exposed to of the major regulatory and catalytic subunits of PI3 kinases in recombinant human calcineurin. Western blot analysis showed that cardiomyocytes (Fig. 2D). These results suggest that FoxO- calcineurin reduced levels of phosphorylated Akt in a dose- dependent increases in Akt phosphorylation are caused mainly by dependent manner (Fig. 3G). Collectively, these results suggest that suppression of PP2A and calcineurin phosphatase activities. Akt is a direct substrate of calcineurin, and reduced calcineurin

Ni et al. PNAS ͉ December 18, 2007 ͉ vol. 104 ͉ no. 51 ͉ 20519 Downloaded by guest on October 1, 2021 y t lanes 7–8). MG132 treatment did not affect basal or IGF-1-induced d u A o C p ib t t in IP t k 1 Akt phosphorylation (SI Fig. 9C), suggesting that upstream acti- A- u 1 t an i p O O u t in x IP x o n o o vators of Akt are not regulated by proteasomal degradation under np N P F P F IP: anti-Akt I A F a F a IP: anti-Akt G c G c pAkt (S473) these conditions. Consistent with the notion that FoxO-induced Calcineurin increases in Akt phosphorylation are not due to changes upstream Calcineurin PP2A of Akt, the PI3 kinase inhibitor LY294002 did not block the Ponceau PP2A enhanced levels of Akt phosphorylation observed in cells expressing caFoxO1 (SI Fig. 9D), whereas it did reduce basal Akt phosphor- PP1 ylation (Fig. 3 C and D).

in Calcineurin and PP2A Form a Complex with Akt. Next, we sought to r B y u d e determine whether endogenous calcineurin interacts directly with o in c tib l Akt. Coimmunoprecipitation analysis showed that anti-Akt anti- t -ca u ti IP: anti- p oan n In N body precipitated both calcineurin and PP2A (Fig. 4A). However, calcineurin A Calcineurin despite multiple attempts, we were unable to detect interactions P-Akt (S473) between Akt and PP1 (Fig. 4A). In addition, reciprocal coimmu- noprecipitation with anti-calcineurin antibody precipitated endog- Fig. 4. Akt interacts with calcineurin and PP2A in cardiomyocytes. (A and B) enous phosphorylated Akt (Fig. 4B). We next tested whether FoxO Coimmunoprecipitation showing interaction of Akt with calcineurin and PP2A affected the interactions between Akt and calcineurin or PP2A. In catalytic subunits but not with PP1. Experiments were repeated at least two cells expressing caFoxO1, we detected lower levels of calcineurin times with independent cell lysates. (C) Decreased association between cal- protein associated with FoxO-activated Akt. FoxO overexpression cineurin and Akt or PP2A and Akt in cardiomyocytes overexpressing caFoxO1 was similarly capable of antagonizing the interaction between Akt relative to that seen in cells overexpressing GFP. and PP2A (Fig. 4C). Similar findings for both calcineurin and PP2A were observed in cells overexpressing caFoxO3 (data not shown). phosphatase activity contributes, at least in part, to the elevated Akt These data, then, provide evidence that calcineurin and PP2A form phosphorylation levels induced by FoxO. a complex with Akt. Further, in addition to reducing total phos- phatase activity, overexpression of FoxO selectively reduces Akt- Previous studies suggest that FoxO proteins inhibit calcineurin associated phosphatase activity. activity through atrogin-1-mediated proteasomal degradation (10, 11, 32). We reasoned that if proteasomal degradation of calcineurin FoxO-Induced Increases in Akt Activity Render Cardiomyocytes Insen- contributes to FoxO-induced increases in Akt phosphorylation, sitive to Insulin. We next sought to determine the effect of FoxO- then inhibition of the ubiquitin-proteasome pathway would prevent dependent increases in Akt activity on insulin responsiveness. As the observed increase in Akt phosphorylation. To test this, we expected, insulin triggered a robust increase in Akt phosphorylation exposed cells to the proteasome inhibitor MG132. Relative to in cardiomyocytes (Fig. 5A). However, in the presence of overex- control cells, FoxO1 triggered time-dependent increases in Akt pressed caFoxO1, insulin-induced increases in Akt phosphorylation phosphorylation (SI Fig. 9 A and B, lanes 1–6). Remarkably, these were greatly diminished (Fig. 5 A and B). Insulin-dependent increases were completely blocked by MG132 (SI Fig. 9 A and B, phosphorylation of GSK-3 was similarly blunted by FoxO (Fig. 5C).

Fig. 5. FoxO attenuates insulin re- sponses in cardiomyocytes. Cardio- A B decudni-nilusnI GFP caFoxO1 dlof( revo )lortnoc myocytes were infected with ade- 16

Con INS Con INS novirus as indicated, and whole-cell tkA/tkAp 12 lysates were harvested at 24 h after P-Akt (S473) α (S9) α/β 8 β P-GSK3 infection. (A and B) Western blots (S21) 4 ** and densitometric analyses show- Akt 0 ing insulin (Ins, 10 nM) -induced GFP caFoxO1 phosphorylation of Akt and GSK-3 in cells expressing GFP or caFoxO1- C D GFP caFoxO1 GFP. Con, vehicle treated control. GFP caFoxO1 Con INS Con INS **, P Ͻ 0.01. (C) Time course studies 0ʼ 5ʼ 15ʼ 30ʼ 60ʼ 0ʼ 5ʼ 15ʼ 30ʼ 60ʼ IGF-1 membrane showing IGF-1 (10 nM) -induced P-Akt (S473) Glut4 phosphorylation of Akt and GSK3 in P-GSK3α/β (S21/9) cells expressing GFP or caFoxO1- Ponceau GFP. (D) Western blots of insulin- Tubulin induced Glut4 membrane translo- cation in cells expressing GFP or E 26h F G Insulin caFoxO1-GFP. (E) FoxO inhibits insu- 1000 )llew/nim 01/lomf( )llew/nim 3h * IR

lin-stimulated glucose uptake.esoculG (Up- ekatpu 800 FoxO3 KO FoxO3 WT per) Glucose uptake measured 3 Con INS Con INS Akt P-Akt *** §§ and 26 h after cells were infected 600 P-Akt (S473) PPs with GFP or caFoxO1 and subse- 400 Akt quently treated with insulin (10 nM, 200 10min). (Lower) Phosphorylated FoxO P-FoxO Akt levels at corresponding time 0 (active) (inactive) pAkt points (experiments repeated twice GFP + + + + with triplicate samples). ***, P Ͻ caFoxO1 + + + + 0.001 (*, P Ͻ 0.01) vs. GFP without Insulin - + - + - + - + INS; §§, P Ͻ 0.01 vs. caFoxO1 with- out insulin. (F) Western blots showing levels of Akt phosphorylation and protein abundance in adult mouse cardiac myocytes isolated from FoxO3 knockout or wild-type littermates and treated with insulin (INS) (10 nM, 10 min). (G) Proposed working model. IR, insulin receptor.

20520 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0610290104 Ni et al. Downloaded by guest on October 1, 2021 Similar effects were observed with overexpression of wild-type on FoxO3 in heart (10) and FoxO1 in liver (23). However, in FoxO1 (data not shown) and when different levels of caFoxO1 contrast with our findings, forced expression of FoxO1 in hepato- proteins were expressed in cardiomyocytes (MOI ϭ 1–10, data not cytes led to increased phosphorylation of both Akt and GSK-3 (23), shown). Time course analysis of IGF-1-induced responses in cells two prominent Akt substrates, whereas we find in heart that expressing caFoxO1 proteins revealed both a delay in the onset (15 FoxO-activated Akt does not lead to increases in the phosphory- min in caFoxO1 cells versus 5 min in GFP control cells) and a lation state of these proteins. One potential explanation for this decrease in the peak response amplitude to insulin (2.4-fold in discrepancy is that in cardiomyocytes, a sequestered subpopulation caFoxO1 cells versus 12-fold in GFP control cells) (Fig. 5 A–C). of activated Akt is formed—and shielded from certain We next tested whether pharmacological suppression of phos- substrates—as a result of reduced interaction with protein phos- phatases mimics the effects of FoxO on insulin responsiveness. phatases and associated scaffolding complexes. Alternatively, dif- Pretreatment with FK506, okadaic acid, or calyculin A each led to ferential expression of Akt isoforms in heart and liver may exist, increases in Akt phosphorylation under both basal conditions and with FoxO preferentially targeting one Akt isoform while leaving on short-term exposure to insulin (SI Fig. 10). Also, overexpression others unchecked (39). That FoxO-induced activation of Akt re- of constitutively active Akt (Myr-Akt) similarly blunted insulin sulted from reduced phosphatase activity rather than increased responsiveness (data not shown), consistent with previous re- kinase activity may also account for the apparent discrepancy that ports (35). PI3 kinase pathway-activated Akt accumulates in the nucleus, and In both Drosophila and mammals, FoxO activates transcription of FoxO-activated Akt is localized in the cytosol. Our finding that the insulin receptor (IR), an effect that is up-regulated during FoxO is capable of binding and sequestering activated Akt—leaving starvation and leads to insulin sensitization (21, 36). In cardiomy- its activity toward other downstream targets unchecked—uncovers ocytes overexpressing FoxO1 and caFoxO1, we found that FoxO a mechanism for additional fine-tuning of the signaling cascade. had no significant effect on IR protein abundance (SI Fig. 11A), IR It was reported recently that tribble 3 (Trb3), a pseudokinase and transcript levels (SI Fig. 11B), or IRS2 protein levels (SI Fig. 11C). modulator of Akt, participates in FoxO1-induced increases in Akt To investigate whether FoxO-induced increases in Akt activity phosphorylation in the liver (23). However, multiple studies have alter glucose metabolism, we measured insulin-induced glucose failed to detect Trb3 in heart (40, 41). Findings reported here point uptake. Forced expression of caFoxO1 attenuated insulin-induced to regulatory mechanisms governing Akt dephosphorylation as membrane translocation of the glucose transporter Glut4 (Fig. 5D), central to FoxO-dependent Akt activation. Similar to our findings consistent with diminished insulin responsiveness. Further, forced in cardiac myocytes, Matsumoto et al. (23) did not detect changes expression of caFoxO1 led to significant decreases in insulin- in PP2A abundance in FoxO1-transduced hepatocytes. We, how- induced glucose uptake in cardiomyocytes (Fig. 5E). Similar results ever, have extended these findings by testing and inhibiting the were observed with caFoxO3 (data not shown). enzymatic activities of PP2A (and other) phosphatases, detecting Finally, to test for effects of endogenous FoxO on insulin significant FoxO-induced declines. These latter experiments are signaling, we studied adult cardiac myocytes isolated from FoxO3- consistent with our prior report of marked FoxO-dependent de- null mice [FoxO1-null mice are embryonic lethal (24)]. Basal levels creases in calcineurin/NFAT signaling (11). These declines in of phospho-Akt were lower in FoxO3-null myocytes relative to wild phosphatase activity likely stem from proteasomal degradation of type (Fig. 5F), consistent with our findings of Akt-activating effects calcineurin through the FoxO target protein atrogin-1 (32). Addi- of FoxO in heart. Moreover, exposure to insulin led to substantially tional evidence for involvement of the proteasome pathway is greater increases in Akt phosphorylation in FoxO3-null hearts provided here in experiments where FoxO-dependent increases in compared with wild type, pointing to insulin sensitization in the Akt phosphorylation were blocked by the proteasome inhibitor FoxO3-null background. Similar results were observed with IGF-1 MG132. (data not shown). These data, then, lend additional support to our By inducing Akt phosphorylation, FoxO is capable of activating conclusions that FoxO activity reduces insulin sensitivity in heart. its own inhibitor. Thus, the existence of simultaneous, opposing, nonequilibrium reactions (a futile cycle) must be considered. How- Discussion ever, phosphorylation of FoxO is a powerful, yet only partial, Insulin resistance is a hallmark of type 2 diabetes and several other mechanism controlling FoxO transcriptional activity. Furthermore, disorders, including obesity, sepsis, and metabolic syndrome. As a FoxO activity is governed by other posttranslational modifications; result, insulin resistance contributes significantly to cardiovascular for example, we find that deacetylation is capable of overriding the morbidity and mortality (37). Here, we elucidate a mechanism inhibitory effects of FoxO phosphorylation, consistent with previ- whereby FoxO transcription factors participate in feedback regu- ous reports (42, 43). lation of insulin signaling in cardiac muscle. The major findings we describe are (i) FoxO activates Akt, and resveratrol, which increases Akt Is a Calcineurin Target. The PI3K/Akt pathway is a key regulator endogenous FoxO activity in cardiomyocytes, has similar effects; of four intersecting biological processes: cell growth and survival, (ii) FoxO-activated Akt forms a complex with FoxO and signals cell-cycle progression, and metabolism. In heart, Akt contributes to selectively to some, but not all, downstream targets; (iii) Akt is a both pathological and physiological cardiac growth, myocyte sur- binding partner and enzymatic target of calcineurin; (iv) FoxO- vival, and contractile function (44). It is not surprising, then, that dependent declines in calcineurin activity and in Akt–calcineurin Akt activity is exquisitely regulated in cardiac tissue. Findings interaction together lead to Akt activation; and (v) FoxO activation reported here highlight the importance of phosphatase actions on attenuates insulin signaling and glucose uptake in cardiac myocytes. Akt and show that, in addition to PP1, PP2A, and PHLPP (28, 29, Together, these results elucidate a previously uncharacterized 45), calcineurin also dephosphorylates Akt. Interestingly, targeted pathway governing insulin signaling in the myocardium, one where inhibition of calcineurin or PP2A alone in myocytes provoked only FoxO selectively fine-tunes insulin responsiveness through regula- modest increases in Akt phosphorylation; in contrast, combined MEDICAL SCIENCES tion of Akt activity. A proposed working model is presented suppression of PP2A and PP1, or all three serine/threonine phos- (Fig. 5G). phatases together, had robust effects, suggesting possible synergistic actions. Feedback Regulation of Akt by FoxO. The transcriptional activity of We did not detect interaction between Akt and PP1. However, FoxO proteins is regulated by posttranslational modifications, it is possible that PP1 also contributes indirectly to FoxO-induced including phosphorylation, acetylation, and ubiquitylation (13, 38). increases in Akt phosphorylation through its interactions with Here, we report that FoxO1 or FoxO3 also trigger increases in Akt calcineurin and other phosphatases. For example, calcineurin has phosphorylation and kinase activity, consistent with prior reports been shown to dephosphorylate and inactivate DARPP-32, a

Ni et al. PNAS ͉ December 18, 2007 ͉ vol. 104 ͉ no. 51 ͉ 20521 Downloaded by guest on October 1, 2021 potent inhibitor of PP1 (46). Recent reports also showed that both insulin responsiveness (54), a finding consistent with our results. calcineurin and PP2A dephosphorylate and inactivate inhibitor-1, a That said, FoxO may also affect steps in the insulin signaling functional homolog of DARPP-32 in cardiomyocytes (47). Thus, cascade that are Akt-independent, such as Ca2ϩ/calmodulin- synergy among Akt phosphatases may arise from reduced cal- dependent protein kinase kinases (CaMKKs) and AMPK (7, 55). cineurin and PP2A activities leading to increased inhibitor-1 func- A recent study described a pathway that selectively targets for tion, diminished PP1 activity, and enhanced Akt phosphorylation proteasomal degradation substrates that are phosphorylated (56). and function. Because FoxO reduces phosphatase activities of both PP2A and PP2B, it may increase the phosphorylation state, and subsequent FoxO and Insulin Resistance. In different tissues, FoxO transcrip- degradation, of proteins that are unrelated to the Akt pathway but tional activity is capable of either sensitizing or inhibiting insulin that are also involved in insulin responsiveness. responsiveness. For example, in both Drosophila and mammals, FoxO1 triggers insulin sensitization because of its actions to up- Perspective. Cardiometabolic risk encompasses a cluster of risk regulate expression of the insulin receptor (36). However, there is factors that predispose individuals to type 2 diabetes and premature precedent in multiple other contexts that sustained increases in cardiovascular disease. In the great majority of instances, insulin FoxO activity induce insulin resistance (18, 19). FoxO proteins are resistance is a major underlying mechanism. Here, we describe a mechanism whereby FoxO transcription factors participate in feed- known as longevity factors (48, 49), and it has been proposed that back regulation of insulin signaling in cardiac muscle, providing an insulin resistance at certain levels may extend lifespan by maintain- additional echelon of control to calibrate this critical pathway. As ing cellular lipid homeostasis and preventing lipid overload (50). a result, these findings prompt consideration of FoxO as a target of Our evidence in heart provides additional evidence that FoxO- therapeutic intervention in metabolic syndrome. dependent mechanisms regulate insulin responses at multiple points. That dFoxO affects multiple steps in the insulin signaling Materials and Methods cascade in Drosophila points to significant evolutionary conserva- Primary Cardiac Cell Preparation. Neonatal rat ventricular myocytes were iso- tion of this mechanism. Interestingly, in db/db mice, hepatic FoxO lated from the ventricles of postnatal day 1–2 Sprague–Dawley rats (11). is nuclear-localized and Akt is hyperphosphorylated in concert with Ϯ dysregulated expression of genes controlling gluconeogenesis, a Statistics. Unless otherwise specified, data are expressed as mean SEM, triplicate samples were used, and each experiment was repeated at least finding that suggests that FoxO-dependent mechanisms contribute twice. Statistical significance was analyzed with a Student’s unpaired t test or to insulin resistance in this model of type 2 diabetes (51). one-way ANOVA, followed by Bonferroni’s method for post hoc pairwise We detected no changes in insulin responsiveness triggered by multiple comparisons. Additional materials and methods are provided in resveratrol (data not shown), which we speculate stems from the SI Materials and Methods. only modest effects it has on Akt phosphorylation. On the other hand, resveratrol increases the activities of the transcriptional ACKNOWLEDGMENTS. We thank Drs. Stephen J. Gold and James Bibb for ␣ helpful discussions, Drs. Diego Castrillon (University of Texas Southwestern) coactivator PGC1 and AMPK (AMP-activated protein kinase), and Ronald DePinho (Harvard University, Cambridge, MA) for providing both of which improve insulin sensitivity (52, 53). Also, PGC1␣ has FoxO3-null mice, and Preston Wong and Jun Cheng for technical assistance. been shown to suppress FoxO3 action in skeletal muscle (9). This work was supported by grants from the Donald W. Reynolds Foundation Together, these actions may antagonize the FoxO-induced inhibi- (to J.A.H.); National Institutes of Health Grants HL-075173, HL-006296, and HL-080144 (to J.A.H.) and HL-072016 (to B.A.R.); and American Heart Associ- tion of insulin responsiveness in cardiomyocytes. ation Grants 0640084N (to J.A.H.), 0565100Y (to Y.G.N.), and 0655202Y Chronic activation of Akt in heart in vivo results in impaired (to B.A.R.).

1. Eckel RH, Grundy SM, Zimmet PZ (2005) Lancet 365:1415–1428. 29. Gao T, Furnari F, Newton AC (2005) Mol Cell 18:13–24. 2. Paternostro G, Camici PG, Lammerstma AA, Marinho N, Baliga RR, Kooner JS, Radda 30. Zdychova J, Komers R (2005) Physiol Res 54:1–16. GK, Ferrannini E (1996) J Clin Invest 98:2094–2099. 31. Frescas D, Valenti L, Accili D (2005) J Biol Chem 280:20589–20595. 3. Poornima IG, Parikh P, Shannon RP (2006) Circ Res 98:596–605. 32. Li HH, Kedar V, Zhang C, McDonough H, Arya R, Wang DZ, Patterson C (2004) J Clin 4. Boudina S, Abel ED (2007) Circulation 115:3213–3223. Invest 114:1058–1071. 5. Paternostro G, Pagano D, Gnecchi-Ruscone T, Bonser RS, Camici PG (1999) Cardiovasc 33. Oh M, Rybkin II, Copeland V, Czubryt MP, Shelton JM, van Rooij E, Richardson JA, Hill Res 42:246–253. JA, De Windt LJ, Bassel-Duby R, et al. (2005) Mol Cell Biol 25, 6629–6638. 6. Doenst T, Goodwin GW, Cedars AM, Wang M, Stepkowski S, Taegtmeyer H (2001) 34. Vega RB, Yang J, Rothermel BA, Bassel-Duby R, Williams RS (2002) J Biol Chem Metab Clin Exp 50:1083–1090. 277:30401–30407. 7. Arad M, Seidman CE, Seidman JG (2007) Circ Res 100:474–488. 35. Nagoshi T, Matsui T, Aoyama T, Leri A, Anversa P, Li L, Ogawa W, del Monte F, 8. Stitt TN, Drujan D, Clarke BA, Panaro F, Timofeyva Y, Kline WO, Gonzalez M, Yanco- Gwathmey JK, Grazette L, et al. (2005) J Clin Invest 115, 2128–2138. poulos GD, Glass DJ (2004) Mol Cell 14 395–403. 36. Puig O, Tjian R (2006) Cell Cycle 5:503–505. 9. Sandri M, Sandri C, Gilbert A, Skurk C, Calabria E, Picard A, Walsh K, Schiaffino S, Lecker 37. Accili D (2004) Diabetes 53:1633–1642. SH, Goldberg AL (2004) Cell 117:399–412. 38. Vogt PK, Jiang H, Aoki M (2005) Cell Cycle 4:908–913. 10. Skurk C, Izumiya Y, Maatz H, Razeghi P, Shiojima I, Sandri M, Sato K, Zeng L, Schiekofer 39. Muslin AJ, DeBosch B (2006) Novartis Found Symp 274:118–126; discussion 126–131, S, Pimentel D, et al. (2005) J Biol Chem 280:20814–20823. 152–155, 272–276. 11. Ni YG, Berenji K, Wang N, Oh M, Sachan N, Dey A, Cheng J, Lu G, Morris DJ, Castrillon 40. Kiss-Toth E, Bagstaff SM, Sung HY, Jozsa V, Dempsey C, Caunt JC, Oxley KM, Wyllie DH, DH, et al. (2006) Circulation 114:1159–1168. Polgar T, Harte M, et al. (2004) J Biol Chem 279:42703–42708. 12. Tran H, Brunet A, Griffith EC, Greenberg ME (2003) SCI STKE Mar 4, RE5. 41. Ord D, Ord T (2003) Exp Cell Res 286:308–320. 13. Accili D, Arden KC (2004) Cell 117:421–426. 42. Brunet A, Sweeney LB, Sturgill JF, Chua KF, Greer PL, Lin Y, Tran H, Ross SE, Mostoslavsky 14. van der Horst A, Burgering BM (2007) Nat Rev Mol Cell Biol 8:440–450. R, Cohen HY, et al. (2004) Science 303:2011–2015. 15. Huang H, Tindall DJ (2007) J Cell Sci 120:2479–2487. 43. Kitamura YI, Kitamura T, Kruse JP, Raum JC, Stein R, Gu W, Accili D (2005) Cell Metab 16. Barthel A, Schmoll D, Unterman TG (2005) Trends Endocrinol Metab 16:183–189. 2:153–163. 17. Hwangbo DS, Gershman B, Tu MP, Palmer M, Tatar M (2004) Nature 429:562–566. 44. Shiojima I, Walsh K (2006) Genes Dev 20:3347–3365. 18. Zhang W, Patil S, Chauhan B, Guo S, Powell DR, Le J, Klotsas A, Matika R, Xiao X, Franks 45. Jiang X, Kim HE, Shu H, Zhao Y, Zhang H, Kofron J, Donnelly J, Burns D, Ng SC, R, et al. (2006) J Biol Chem 281:10105–10117. Rosenberg S, et al. (2003) 299, 223–226. 19. Kamei Y, Miura S, Suzuki M, Kai Y, Mizukami J, Taniguchi T, Mochida K, Hata T, 46. Hemmings HC, Jr, Nairn AC, Greengard P (1984) J Biol Chem 259:14491–14497. Matsuda J, Aburatani H, et al. (2004) J Biol Chem 279:41114–41123. 47. El-Armouche A, Bednorz A, Pamminger T, Ditz D, Didie M, Dobrev D, Eschenhagen T 20. Watamoto K, Towatari M, Ozawa Y, Miyata Y, Okamoto M, Abe A, Naoe T, Saito H (2006) Biochem Biophys Res Commun 346:700–706. (2003) 22:9176–9184. 48. Greer EL, Brunet A (2005) Oncogene 24:7410–7425. 21. Puig O, Tjian R (2005) Genes Dev 19:2435–2446. 49. Morris BJ (2005) J Hypertens 23:1285–1309. 22. Nakae J, Biggs WH, III, Kitamura T, Cavenee WK, Wright CV, Arden KC, Accili D (2002) 50. Unger RH (2006) Nat Med 12:56–57. Nat Genet 32:245–253. 51. Aoyama H, Daitoku H, Fukamizu A (2006) Int J Mol Med 18:433–439. 23. Matsumoto M, Han S, Kitamura T, Accili D (2006) J Clin Invest 116:2464–2472. 52. Baur JA, Pearson KJ, Price NL, Jamieson HA, Lerin C, Kalra A, Prabhu VV, Allard JS, 24. Castrillon DH, Miao L, Kollipara R, Horner JW, DePinho RA (2003) Science 301:215–218. Lopez-Lluch G, Lewis K, et al. (2006) Nature 444:337–342. 25. Matsuzaki H, Daitoku H, Hatta M, Tanaka K, Fukamizu A (2003) Proc Natl Acad Sci USA 53. Lagouge M, Argmann C, Gerhart-Hines Z, Meziane H, Lerin C, Daussin F, Messadeq N, 100:11285–11290. Milne J, Lambert P, Elliott P, et al. (2006) Cell 127:1109–1122. 26. Stokoe D, Stephens LR, Copeland T, Gaffney PRJ, Reese CB, Painter GF, Holmes AB, 54. Matsui T, Nagoshi T, Hong EG, Luptak I, Hartil K, Li L, Gorovits N, Charron MJ, Kim JK, McCormick F, Hawkins PT (1997) Science 277:567–570. Tian R, et al. (2006) Am J Physiol 290:E789–E797. 27. Sarbassov DD, Guertin DA, Ali SM, Sabatini DM (2005) Science 307:1098–1101. 55. Witczak CA, Fujii N, Hirshman MF, Goodyear LJ (2007) Diabetes 56:1403–1409. 28. Beaulieu JM, Sotnikova TD, Marion S, Lefkowitz RJ, Gainetdinov RR, Caron MG (2005) 56. Dorrello NV, Peschiaroli A, Guardavaccaro D, Colburn NH, Sherman NE, Pagano M Cell 122:261–273. (2006) Science 314:467–471.

20522 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0610290104 Ni et al. Downloaded by guest on October 1, 2021