Coupling between Nutrient Availability and Thyroid Hormone Activation*

Background: Insulin/IGF-1 stimulates thyroid hormone action via type 2 deiodinase (D2). Results: Insulin/IGF-1-induced activation of the PI3K-mTORC2-Akt pathway transcriptionally up-regulates D2. Conclusion: FOXO1 represses DIO2 during fasting, and derepression occurs via nutritional activation of the PI3K-mTORC2-Akt pathway. Significance: This mechanism explains how T3 production, serum T3 levels, and T3-dependent cellular metabolic rate are kept at a level proportionate to the availability of energy substrates. The activity of the thyroid gland is stimulated by food availability via leptin-induced thyrotropin-releasing hormone/thyroid-stimulating hormone expression. Here we show that food availability also stimulates thyroid hormone activation by accelerating the conversion of thyroxine to triiodothyronine via type 2 deiodinase in mouse skeletal muscle and in a cell model transitioning from 0.1 to 10% FBS. The underlying mechanism is transcriptional derepression of DIO2 through the mTORC2 pathway as defined in rictor knockdown cells. In cells kept in 0.1% FBS, there is DIO2 inhibition via FOXO1 binding to the DIO2 promoter. Repression of DIO2 by FOXO1 was confirmed using its specific inhibitor AS1842856 or adenoviral infection of constitutively active FOXO1. ChIP studies indicate that 4 h after 10% FBS-containing medium, FOXO1 binding markedly decreases, and the DIO2 promoter is activated. Studies in the insulin receptor FOXO1 KO mouse indicate that insulin is a key signaling molecule in this process. We conclude that FOXO1 represses DIO2 during fasting and that derepression occurs via nutritional activation of the PI3K-mTORC2-Akt pathway.

The activity of the thyroid gland is stimulated by food availability via leptin-induced thyrotropin-releasing hormone/thyroid-stimulating hormone expression. Here we show that food availability also stimulates thyroid hormone activation by accelerating the conversion of thyroxine to triiodothyronine via type 2 deiodinase in mouse skeletal muscle and in a cell model transitioning from 0.1 to 10% FBS. The underlying mechanism is transcriptional derepression of DIO2 through the mTORC2 pathway as defined in rictor knockdown cells. In cells kept in 0.1% FBS, there is DIO2 inhibition via FOXO1 binding to the DIO2 promoter. Repression of DIO2 by FOXO1 was confirmed using its specific inhibitor AS1842856 or adenoviral infection of constitutively active FOXO1. ChIP studies indicate that 4 h after 10% FBS-containing medium, FOXO1 binding markedly decreases, and the DIO2 promoter is activated. Studies in the insulin receptor FOXO1 KO mouse indicate that insulin is a key signaling molecule in this process. We conclude that FOXO1 represses DIO2 during fasting and that derepression occurs via nutritional activation of the PI3K-mTORC2-Akt pathway.
In mammals, body energy expenditure is gradually decreased with caloric restriction as well as prolonged starvation. This is an adaptive response driven by the hypothalamus that includes behavioral as well as metabolic modifications, such as increased sleep, diminished motor activity, and core temperature (1). During fasting, the hormonal milieu is dominated by low levels of insulin and thyroid hormones, which has been proposed as an explanation for the slowdown in weight loss after a few days of caloric restriction (2,3). Notably, thyroidal activity is restored promptly upon refeeding, as seen in patients recovering from anorexia nervosa (4). In such patients, weight gain and an increase in serum T 3 2 are closely associated with acceleration in energy expenditure (4). These observations indicate that the set point at which the hypothalamic-pituitary-thyroid activity is maintained is affected by nutrient availability, with the lowest (default) levels of activity seen during caloric restriction. Leptin is the key molecule signaling food intake and the availability of energy substrates, which defines the hypothalamic-pituitary-thyroid set point by stimulating secretion of thyrotropin-releasing hormone and thyroid-stimulating hormone and hence thyroidal activity (5,6).
In humans, the thyroid gland contributes with about 20% of the daily T 3 production, and the residual 80% is contributed by two deiodinases that convert T 4 to T 3 outside the thyroid parenchyma (i.e. D1 and D2, with the latter playing the major role) (7). In vivo kinetic studies in rats indicate that serum T 3 levels fall during fasting also as a result of decreased thyroidal T 3 secretion (8). In addition, there is also accelerated thyroid hormone inactivation via D3 expression in multiple tissues as well as accelerated sulfation and glucuronidation in the liver (9,10). However, it is less clear whether the extrathyroidal T 3 production is reduced as well. Whereas hepatic D1 activity is reduced during caloric restriction in rodents (11), this seems to be a consequence rather than a cause of the low serum T 3 , given that Dio1 is highly responsive to T 3 (8,12). On the other hand, D2 is normally stimulated during hypothyroidism (7), and the fact that it is reduced by fasting in the pituitary gland and brown * This work was supported in part by National Institutes of Health, NIDDK, adipose tissue (BAT) (10) indicates a role in the fall of serum T 3 . D2 is an endoplasmic reticulum-resident protein that increases intracellular T 3 concentration and the expression of thyroid hormone-dependent genes (7,13). For example, approximately half of the T 3 present in the brain and in the BAT is originated locally via D2 activity; D2 inactivation in both tissues has been shown to dampen thyroid hormone signaling (7). Therefore, control of the D2 pathway by caloric intake could potentially modulate thyroid hormone signaling by affecting intracellular T 3 concentration as well as circulating T 3 levels.
D2 expression is controlled by transcriptional (i.e. cAMP, FOXO3, and NF-B) (14,15), translational (i.e. ER stress) (16), and post-translational mechanisms, such as ubiquitination and proteasomal degradation (17). In addition, D2 activity in rat BAT is up-regulated by growth factors such as IGF-1 and the multifunctional protein insulin (18,19), which promotes cellular glucose uptake and growth by acting through various nutrient-sensing pathways, such as the PI3K-mTOR signaling pathway. Mammalian target of rapamycin (mTOR) is a signaling pathway that functions as a regulator of translational initiation for cellular metabolism, growth, survival, proliferation, and protein synthesis (20,21). mTOR, a 289-kDa Ser/Thr kinase, serves as the catalytic core of two distinct multicomponent complexes, mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2) (22). Numerous studies have linked mTORC1 to protein synthesis and growth (23) and to the downstream activation of a myriad of metabolic regulatory genes, including HIF-1␣ and SREBP1/2 (24). In addition, it has been determined that mTORC2 phosphorylates the AGC kinases (25) and mediates cell survival and proliferation through phosphorylation of its downstream effector, Akt (26).
The observation that insulin stimulates D2 activity in rat brown adipocytes (19) and that insulin sensitizers stimulate D2 expression in cultures of skeletal myocytes (27) indicate that the nutritional state of an organism can also affect thyroid hormone signaling by acting at the target cell level. The mechanism behind D2 stimulation by insulin remain unknown, despite evidence that an indirect effect via the adrenergic receptors could play a role (28). In the present study, we have examined how nutrient availability and insulin stimulate D2 activity in mice and cell models. We have found that insulin signals through PI3K-mTORC2-Akt to relieve FOXO1-mediated repression of DIO2, indicating that nutrient availability controls serum levels of thyroid hormones not only by modulating the hypothalamicpituitary-thyroid set point but also by regulating extrathyroidal T 3 production. Given that serum T 3 is a major determinant of energy expenditure (29), this explains how the cellular metabolic rate as defined by T 3 signaling is kept at a level that is commensurate to the availability of energy substrates.
Gene Expression Analysis-RNA was extracted from MSTO-211H cells or animal tissue using the RNAqueous Micro Kit (Life Technologies Inc.) or RNeasy Mini Kit (Qiagen), respectively. RNA was quantified with a NanoDrop and reverse transcribed using High Capacity cDNA (Applied Biosystems) or the First Strand cDNA for RT-PCR (AMV) Kit (Roche Applied Science). Genes of interest were measured by qPCR (Bio-Rad iCycler iQ real-time PCR detection system) using the iQ SYBR Green Supermix (Bio-Rad) or qPCR (Applied Biosystems Step One Plus Real-Time PCR System) using the SYBR Green Fast-Mix ROX (Quanta Biosciences). Relative quantitation was using the standard curve method and the iCycler or Step One Plus software.
Western Blot Analysis-Cells/tissues were lysed in 0.25 M sucrose PE containing 10 mM DTT. The lysates were diluted with 4ϫ sample loading buffer (Invitrogen), and 5-25 g of total protein were run on 4 -12% NuPAGE BisTris gels (Life Technologies, Inc.). Samples were transferred to Immobilon-FL PVDF transfer membrane (Millipore, Billerica, MA) and probed with antibodies as indicated at a 1:1000 dilution overnight. Fluorescent labeled secondary antibodies (LI-COR Biosciences, Lincoln, NE) were used at 1:2500 for 1 h. All blots were imaged using the LI-COR Odyssey instrument per the manufacturer's instructions.
T 4 to T 3 Conversion in Intact Cells-The production of T 3 from outer ring-labeled T 4 in intact cells was analyzed by measuring 125 I in the medium as described elsewhere (31) except that the assay was stopped 12 h after the addition of [ 125 I]T 4 , and the free T 4 concentration was 30 pM.
Lentivirus-mediated shRNA Knockdown of Rictor and Raptor-Rictor and raptor knockdown were established by transduction with GIPZ lentiviral shRNAmir vectors in MSTO-211H cells using GIPZ lentiviral shRNAmir for nonsilencing control (NSC), human rictor (clone ID V2LHS_ 120392 and clone ID V3LHS_367492), or human raptor (clone ID V3LHS_636800) from Thermo Scientific (Lafayette, CO). For generation of stable knockdown cell lines, MSTO-211H cells were plated at 1.5 ϫ 10 5 cells/well and transduced with lentiviral particles at a multiplicity of infection of ϳ23 and ϳ15 for rictor, respectively, and a multiplicity of infection of ϳ13 for raptor and diluted in 1 ml of serum-free RPMI containing 8 g/ml Polybrene (Sigma-Aldrich). After 6 h, 1 ml of complete RPMI was added to each well (6-well plates). 72 h later, the transfection mixture was replaced with complete medium containing puromycin (1 g/ml) to select for shRNA-expressing cells. Stable cell lines were generated after puromycin selection for 7 days.
Adenovirus-mediated Transduction of WT and Constitutively Active (CA) FOXO1-Adenoviral particles expressing WT and CA FOXO1 as described (32) were used to generate transient transfections for WT and CA FOXO1 in MSTO-211H. Cells were plated at 2.5 ϫ 10 5 cells/well in 6-well dishes and transduced with adenoviral particles of concentration 2.0 ϫ 10 10 pfu/ml and 1.8 ϫ 10 10 pfu/ml, respectively, that were diluted in 1 ml of serum-free RPMI. Complete medium containing 10% FBS was added after 6 h, and the medium was changed again after 24 h. At 48 h, cells were refed or stimulated with insulin for the experimental procedure.
Animal Studies-All studies performed were approved by the institutional animal care and use committee of the University of Miami in compliance with National Institutes of Health Standards. Male, 3-month-old C57BL/6J mice purchased from Jackson Laboratory were housed at room temperature (22°C) on a 12-h dark/light cycle. Mice were randomly divided into three groups (control, fasting, refeeding) and acclimatized for 5 days. Animals were fed ad libitum with a standard chow diet (3.5 kcal/g, 28.8% protein, 58.5% carbohydrate, 12.7% fat; 5010 Lab-Diet laboratory autoclavable rodent diet; PMI Nutrition, Richmond, IN) and water. When indicated, food was withdrawn at 21:00 h for the fasting and refeeding groups. After 36 h, fasted mice were sacrificed by asphyxiation in a CO 2 chamber. Mice from the refeeding group were weighed, refed for 8 h, and sacrificed. Soleus muscle and cortex were rapidly dissected and frozen in liquid nitrogen and stored at Ϫ80°C until analysis. As indicated, freely feeding 11-week-old male insulin receptor (IR) floxed, liver-specific IR knock-out (LIRKO), and IR/FOXO1 double knock-out (LIRFKO) mice (33) were used as well. They were sacrificed by decapitation following brief sedation with isofluorane.
Blood Biochemistry-Blood was collected by cardiac puncture, and plasma was stored at Ϫ80°C until analysis. Insulin was measured using mouse insulin ELISA (Mercodia, AB). Serum T 3 and T 4 were measured as described previously (34).
Statistical Analysis-All data were analyzed using PRISM software (GraphPad Software) and expressed as mean Ϯ S.E. Student's t test was used to compare two groups. One-way anal-ysis of variance was used to compare more than two groups. Significance was held at p Ͻ 0.05 (two-tailed).
Fresh Serum Induces D2 Expression in a Cell Model-Given that D2 activity is near background level in skeletal muscle cell lines (27), the role played by fasting and refeeding on DIO2 expression was next modeled using MSTO-211H cells, a cell line that endogenously expresses readily measurable D2 activity (30). In this setting, keeping the cells for 48 h in 0.1% FBS caused a progressive reduction in DIO2 mRNA and activity to approximately half of the baseline values seen in cells cultured in 10% FBS ( Fig. 2outf;F2, A and B). Notably, transition to 10% FBS resulted in a ϳ3.5-fold increase in D2 activity (Fig. 2C) that followed a ϳ2-fold increase in D2 mRNA levels (Fig. 2D); both subsided and returned to baseline levels by 24 h (Fig. 2, C and  D). The increases in D2 activity (Fig. 2E) and typical serum-dependent downstream targets (Fig. 2F) were dependent on the amount of FBS contained in the fresh medium. D2 is known for having a post-translational regulation via ubiquitination and proteasomal degradation, explaining its relatively short activity half-life of less than 1 h (35). Thus, we tested whether the surge in D2 activity that follows refeeding could also involve the ubiquitin-proteasomal mechanism. D2 activity half-life was measured by exposing cells in 10% FBS to cyclohexamide, but at no time was there a significant effect on D2 activity half-life (Fig. 2, PI3K-dependent Pathways Stimulate D2 Activity-Serum/ FBS contains a number of growth factors that could potentially induce D2 expression during refeeding, mostly acting via PI3K. Thus, we used insulin and observed a concentration-dependent increase in D2 activity of ϳ50% after 8 h (Fig. 2K). We also observed a time-dependent increase in D2 activity following the stimulation of serum-starved MSTO-211H cells with 50 ng/ml IGF-1 (Fig. 2L). The acceleration in D2 activity increased thyroid hormone activation, given that T 3 production was accelerated 4 -5-fold after the addition of IGF-1 or insulin (Fig. 2M). Next, different strategies were used to dissect the molecular pathway that mediates the insulin/IGF-1 stimulation of DIO2 expression. First, we inhibited PI3K using LY294002, a well known reversible inhibitor of PI3K phosphorylation. Following the standard 48 h starvation period in 0.1% FBS, MSTO-211H cells were treated with 50 M LY294002 and either stimulated with 10% FBS (Fig. 3A) or exposed to insulin (Fig. 3B). Notably, inhibition of PI3K abrogated the increase in D2 activity at all time points studied, which coincided with the inhibition of phosphorylation of the downstream targets Akt and S6rp (Fig. 3, A  and B). Similar results were observed when 0.1-1 M wortmannin was used to inhibit PI3K signaling (Fig. 3C, a and b). Next, we looked further down the pathway and used GSK690693, an inhibitor of all three Akt isoforms (36,37) to examine the involvement of Akt in this mechanism. MSTO-211H cells that had been starved for 48 h were pretreated for 1 h with 10 nM GSK690693 (37) and subsequently exposed to insulin. Again, insulin induction of D2 activity was prevented at all time points tested throughout a 24 h time period (Fig. 3D). Western blot analysis of Akt phosphorylation shows overstimulation of the Akt pathway, a feature commonly observed when small molecule Akt inhibitors are used (36,37). This was coupled with inhibition of S6rp phosphorylation (Fig. 3D), indicating successful Akt inhibition.

Inhibition of mTOR Pathway Prevents FBS and
Insulin-mediated D2 Stimulation-We next used PP242, a selective ATPcompetitive inhibitor of both mTORC1 and mTORC2 (38,39) to test the involvement of the mTOR signaling pathway in the D2 stimulation by FBS or insulin. Treating cells with 2 M PP242 prevented the surge in D2 activity that follows 10% FBS (Fig. 4A) or insulin stimulation (Fig. 4B). The effectiveness of PP242 in blocking both the mTORC1 and mTORC2 signaling pathways was documented by Western blot analysis of downstream targets of mTORC1 (S6rp and 4EBP1) and of mTORC2 (Akt) (Fig. 4, A and B). Next, we used rapamycin, an inhibitor of mTORC1 (40), to further investigate the involvement of the mTOR pathway. Inhibition of mTORC1 delayed but did not prevent the surge in D2 activity caused by 10% FBS (Fig. 4C). Inhibitory selectivity was documented by Western blot analysis of phosphorylated and total S6rp, a downstream marker of mTORC1. Indeed, there was reduction of mTORC1 signaling after 4 h of rapamycin treatment, whereas phosphorylation of Akt via mTORC2 was not inhibited (Fig. 4D). Similar to D2 activity, we also found a significant increase of DIO2 mRNA FIGURE 1. Fasting inhibits Dio2 expression levels and activity in the soleus muscle of mice. A, change in body weight of control, fasted, and refed mice measured before and after the experimental period; n ϭ 7; ***, p Ͻ 0.001. B, serum insulin levels measured in control, fasted, and refed mice; n ϭ 10, 12, 7, respectively; ***, p Ͻ 0.001. C, Western blot analysis of control, fasting, and refeeding groups using ␣-total S6rp, ␣-phospho-S6rp (Ser-235/236/240/244), ␣-pan-Akt, and ␣-phospho-Akt (Ser-473) antibodies. D, CycloA was used as a housekeeping gene for the measurement of relative mRNA levels of Dio2/CycloA in the soleus muscle; n ϭ 3 (control), n ϭ 6 (fasting); ***, p Ͻ 0.001. E, D2 activity measured in soleus muscle of control and fasted mice; n ϭ 3 and 7, respectively; ***, p Ͻ 0.001. F, CycloA was used as a housekeeping gene for the measurement of relative mRNA levels of Dio2/CycloA in the cortex; n ϭ 7. G, RNA polymerase 2 was used as a housekeeping gene; n ϭ 7. H, the measurement of relative mRNA levels of Dio2/RNAPol2 in soleus muscle; n ϭ 7; *, p Ͻ 0.05. I, D2 activity measured in the soleus muscle of fasted and refed mice; n ϭ 7; *, p Ͻ 0.05. Error bars, S.E.  DECEMBER (Fig. 4E). Studies with insulin led to similar observations (Fig. 4F). As further confirmation, we used GIPZ lentiviral shRNA vectors to knockdown raptor (regulatory-associated protein of mTOR), a component of the mTORC1 complex. Despite successful raptor knockdown in these cells, both 10% FBS (Fig. 5A) and the addition of insulin (Fig. 5B) stimulated D2 activity. Altogether, these data suggest that D2 induction by 10% FBS or exposure to insulin is mediated by an mTORC2-dependent pathway.

Nutrient Availability and Thyroid Hormone Activation
To assess the role played by mTORC2, we next used GIPZ lentiviral shRNA vectors to knock down an essential component of the mTORC2 complex, rictor (rapamycin-insensitive component of mTOR) in MSTO-211H cells. Rictor is a core unit of the mTORC2 complex, and its reduction disrupts mTORC2 complex assembly (41). Stably expressing cells were starved for 48 h prior to the addition of 10% FBS (Fig. 5C) or insulin stimulation (Fig. 5D). Inhibition of mTORC2 significantly inhibits the surge in D2 activity at all time points tested, peaking at a 4-fold reduction by 12 h in both conditions as compared with NSCs. The successful knockdown of rictor was documented by Western blot. These data show the dependence of induction of D2 activity on the mTORC2 pathway.
FOXO1 Transcriptionally Represses DIO2-While looking for the mTORC2/Akt-dependent downstream targets that could activate DIO2, we serendipitously observed that Dio2 mRNA levels were low in a mouse with liver-specific inactiva- tion of the insulin receptor (LIRKO) and that Dio2 mRNA levels were restored when this mouse was crossed with a mouse harboring a liver-specific disruption of FOXO1 (LIRFKO) (Fig.  6A). This observation, along with previous studies indicating that FOXO3 interacts with the Dio2 gene (15), suggested that FOXO1 may be the key molecule mediating the effects of nutrient status and insulin signaling on Dio2. This suspicion was strengthened by the observation that there was synchronized   DECEMBER 18, 2015 • VOLUME 290 • NUMBER 51 stimulation of D2 activity and FOXO1 phosphorylation after serum-starved MSTO-211H cells were exposed to 10% FBS (Fig. 6, B and C). Subsequently, the specific FOXO1 inhibitor AS1842856, which prevents Ser-256 phosphorylation, was added to cells during 0.1% FBS starvation (42). Whereas 0.1% FBS resulted in the expected loss of D2 activity, treatment with AS1842856 not only reversed this but increased D2 activity above baseline levels (Fig. 6D). Similar findings were observed in cells that had been starved in 0.1% FBS for 48 h and then treated with AS1842856 (Fig. 6E). Taken together, these results indicate that loss of D2 activity associated with fasting is mediated by FOXO1.

Nutrient Availability and Thyroid Hormone Activation
To further explore the role of FOXO1 in regulating D2 expression, MSTO-211H cells were infected with WT or CA FOXO1-expressing adenovirus and processed through the 0.1%/10% FBS or insulin stimulation protocols. Remarkably, expression of the CA FOXO1 protein prevents all of the induction of DIO2 mRNA (Fig. 6, F and G) and most of the D2 activity (Fig. 6, H and I) caused by 10% FBS and insulin stimulation. This indicates that DIO2 is a probable direct target of FOXO1 and that 10% FBS/insulin acts by relieving FOXO1 gene repression. To demonstrate the binding of FOXO1 to predicted binding sites in the DIO2 promoter region (Fig. 6J), ChIP analysis was performed in MSTO-211H cells in 0.1% FBS and 10% FBS. There is specific binding of FOXO1 to a site in the DIO2 pro-moter located Ϫ570 bp from the transcription start site that is abrogated in refed cells (Fig. 6J).

Discussion
The present studies employed a combination of animal and cell models to establish that nutritional availability affects thyroid hormone activation via induction of D2. This enzyme contributes to about two-thirds of the circulating T 3 in humans (about one-third in rodents) and at least one-half of the T 3 present in murine brain and BAT (7). D2 stimulation by 10% FBS is transcriptional and involves insulin and IGF-1 signaling through PI3K-mTORC2-Akt to relieve FOXO1-mediated DIO2 repression (Fig. 7). These data provide a mechanistic explanation for the observation in humans that fasting is associated with a reduction in skeletal muscle D2 activity that is partially prevented by insulin administration (43). The present findings constitute a new aspect of the overarching homeostatic role played by the hypothalamic-pituitary-thyroid axis to adjust thyroid hormone activation to the appropriate level of energy substrates; thyroidal activity and serum T 3 levels are kept at a minimum (default) during fasting or severe caloric restriction, and both increase as food becomes available.
T 3 receptors (TRs) are ligand-dependent transcription factors that control tissue-specific sets of T 3 -responsive genes. TRs are bound to the promoter of T 3 -responsive genes, and, as opposed to most other receptors, unoccupied TRs actively inhibit transcription, given their affinity for corepressors. Upon bindingofT 3 ,TRsgainaffinityforcoactivatorsandtriggerT 3 -dependent gene transcription. Thus, the role of T 3 is not only to promote expression of T 3 -dependent genes but also to relieve TR-mediated gene repression (44). The present observation that fasting shuts down DIO2 expression (and T 3 production) via shifting the balance between PI3K-mTORC2-Akt and FOXO1 highlights the importance of unoccupied TRs.
D2 expression in the cerebral cortex was not modified by fasting or refeeding (Fig. 1F), indicating that Dio2 regulation by nutrient availability is not universal, probably occurring in tissues where the metabolic pathways are responsive to T 3 and insulin (e.g. BAT and skeletal muscle) (7). D2-generated T 3 in BAT enhances local thyroid hormone signaling and the expression of Ucp1 and Pgc-1␣, both critical for thermogenesis (45). A local role for D2-generated T 3 in skeletal muscle is less clear. The fact that skeletal muscle physiology is T 3 -sensitive (46) sets the stage for a potential D2 contribution, such as during development and response to injury (47). Nevertheless, selective D2 inactivation in the mouse skeletal muscle failed to identify a metabolic phenotype, and serum T 3 was preserved due to increased thyroidal T 3 production (48,49). The present findings are particularly relevant because the thyroid gland does not adjust T 3 production during fasting (8), and serum T 3 drops also as a consequence of FOXO1-mediated DIO2 inhibition.
In addition, the present findings are clinically relevant because there are ϳ10 million hypothyroid individuals in the United States alone who depend Ͼ80% on the D2 pathway to convert levothyroxine to T 3 . Because these patients are hypothyroid, they lack the ability to compensate for FOXO1-mediated DIO2 suppression during fasting. Thus, serum T 3 levels in these individuals could fluctuate substantially with caloric intake. In many aspects, the balance between PI3K-mTORC2-Akt and FOXO1 signaling should provide nutritional input and fine tuning to the regulation of circulating T 3 levels and T 3 -dependent processes.
The data from the LIRKO and LIRFKO mice indicate that insulin is probably a key molecule mediating Dio2 regulation (Fig. 6A), known to provide nutritional input through the PI3K-mTOR signaling pathway in many other systems (50). In the cell model, both insulin and IGF-1 stimulate D2 (Fig. 2, K and L), indicating that this is an insulin/IGF-1-mediated mechanism. mTOR, which is downstream of insulin/IGF-1 signaling, also plays a central role in defining the balance between anabolic and catabolic processes (51). Downstream targets of mTORC2 include the AGC kinases (i.e. Akt, SGK, and PKC␣, which function to promote cell survival, metabolism, and actin organization) (41). In turn, Akt directly phosphorylates FOXO1 (25, 52), a nuclear transcription factor that modulates multiple sets of genes in a tissue-specific manner (53,54).
A functional FOXO3 binding site is located in the 5Ј-UTR of the human DIO2, positively regulating DIO2 expression in developing skeletal muscle (15). Although proteins in the FOXO family can potentially bind to the same DNA cis-elements, the FOXO1 binding site identified in the present studies is within the DIO2 promoter, close to the transcription start site. Binding of FOXO1 to this site (Fig. 6J) suppresses DIO2 gene expression.
In conclusion, the present study reveals that insulin/IGF-1 transcriptionally up-regulates DIO2 expression and activity via signaling through the PI3K-mTORC2-Akt pathway, relieving FOXO1-mediated DIO2 inhibition. D2 is a major contributor to plasma T 3 levels in rodents and in humans and is the predominant pathway producing T 3 in hypothyroid individuals treated with levothyroxine. Therefore, the PI3K-mTORC2-Akt pathway is likely to have major metabolic consequences because it provides a coupling mechanism between nutrient availability and thyroid hormone activation.
Author Contributions-L. J. L. conducted experiments, analyzed and interpreted the data, and prepared the manuscript. J. P. W. and I. O. conducted experiments and analyzed data. T. G. U. created the LIRKO and LIRFKO mice, directed the experiments with these animals, and reviewed/edited the manuscript. A. C. B. directed the studies, interpreted the data, and prepared the manuscript along with L. J. L.