Reciprocal Regulation of Hepatic TGF-β1 and Foxo1 Controls Gluconeogenesis and Energy Expenditure Free

In response to the change of metabolic states, the liver secretes a group of proteins termed “hepatokines” to exert endocrine, autocrine, or paracrine roles in control of nutrient and systemic homeostasis (1). Several hepatokines, such as fetuin-A, fibroblast growth factor 21, and selenoprotein P, are known to regulate insulin sensitivity and brown adipose tissue (BAT) thermogenesis (1,2). Abnormal secretion of hepatokines has been linked to the development of metabolic diseases, such as obesity, insulin resistance, type 2 diabetes mellitus (T2D), and cardiovascular diseases (3). Thus, understanding hepatokines and their impacts on insulin signaling is critical to bring new insights into the therapeutic interventions for treatment of metabolic diseases.

The O-class of the forkhead transcription factor Foxo1 is a key mediator of insulin signaling. By binding to the insulin response element on the promoter region of target genes, Foxo1 stimulates expression of G6pc (associated with gluconeogenesis) and hemeoxygenase-1 and Tlr4 (associated with inflammation), governing diverse cellular processes (4,5). Upon insulin stimulation, Akt promotes Foxo1 phosphorylation at ser 256, which stimulates Foxo1 localization from the nucleus to the cytoplasm and reduces DNA–protein interactions (6,7). Disruption of hepatic insulin signaling by deletion of insulin receptor substrate 1 (Irs1) and Irs2 (double knockout [DKO]) causes Foxo1 hyperactivation, hyperglycemia, and systemic insulin resistance. These metabolic disorders are largely rescued by further hepatic Foxo1 deletion in DKO mice, suggesting that Foxo1 signaling pathways play crucial roles in insulin resistance and metabolic dysfunction.

TGF-β1 belongs to the TGF-β superfamily and exerts multiple cellular functions, including cell proliferation, differentiation, apoptosis, adhesion, and immune responses (8). Several TGF-β superfamily members, such as activin E, follistatin, and growth differentiation factor-15, function as hepatokines that affect various organs in terms of modulating glucose metabolism, lipid homeostasis, and energy expenditure (9–13). Although several studies have linked TGF-β1 signaling with development of metabolic diseases, including obesity, T2D, and cardiovascular diseases (14–19), the role of hepatocytic TGF-β1 in the pathogenesis of T2D and obesity is not completely understood. In this study, we explored the role of hepatocyte-expressed TGF-β1 in control of blood glucose and energy expenditure in mice, and further investigated underlying mechanisms. We identified a novel reciprocal regulation of hepatic TGF-β1 and Foxo1 in control gluconeogenesis and energy metabolism.

Liver-specific Irs1 and Irs2 DKO and the Irs1, Irs2, and Foxo1 triple-gene knockout (TKO), floxed Foxo1 (Foxo1L/L), and S273A/A mice were generated as previously described (20–22). The floxed Tgfb1 mice (TGF-β1L/L), β1^glo mice, TβRII L/L, albumin-Cre, and db/db mice were purchased from The Jackson Laboratory. To generate liver-specific-gene knockout mice, floxed mice were crossed with albumin-Cre mice. All mice used in this study were male and maintained at 22°C in a 12-h/12-h light/dark cycle and given free access to food and water. For high-fat diet (HFD) treatment, male mice at 8–10 weeks of age were fed an HFD (42% kcal from fat) for 3 months. A low-fat diet (LFD; 13% kcal from fat) was used as the control diet. All animal experiments were performed according to procedures approved by Texas A&M University Institutional Animal Care and Use Committee.

Primary mouse hepatocytes were isolated and cultured in DMEM with 10% FBS, as previously described (21). For gene overexpression, cells were infected with adenovirus (multiplicity of infection, 20) for 16 h and then subjected to further analysis. For gene knockdown, cells were subjected to Lipofectamine 3000 (Life Technologies) with siRNA, according to the manufacturer’s instruction, for 16 h and then subjected to further analysis.

Freshly isolated hepatocytes were cultured in hepatic glucose production (HGP) buffer for 3 h (21). HGP buffer was collected, and glucose levels were measured according to the manufacturer’s protocol, using the Amplex Red Glucose Assay (Invitrogen) (23). For glycogenolysis assay, sodium l-lactate and pyruvate were removed from HGP buffer; the difference between HGP and glycogenolysis was the gluconeogenesis level.

Total RNA was extracted from tissue or cells with TRIzol reagent (Invitrogen) and reversely transcribed to cDNA with the iScript cDNA synthesis system (Bio-Rad Laboratories) according to the manufacturer’s instructions. Quantitative gene expression was measured using gene-specific primers (Supplementary Methods) using the SYBER Green Supermix system (Bio-Rad Laboratories), as previously described (24). Cyclophilin served as the internal control.

Proteins were prepared from cells, liver, or adipose tissue, then resolved by SDS-PAGE and transferred to nitrocellulose membrane for immunoblotting analysis using specific antibodies, as previously described (24). The signal intensity was measured and analyzed by Image J software (U.S. National Institutes of Health).

Mice were recorded hourly for three consecutive days to obtain the following measurements: gas exchange, food intake, and physical activity. Energy expenditure was calculated using regression-based plots and ANCOVA analysis (25). The respiratory exchange ratio was estimated by calculating the ratio of Vco₂ to Vo₂.

Body composition of nonanesthetized mice was measured using the EchoMRI body composition analyzer (Echo Medical Systems). Body temperature was detected during 4°C cold-exposure studies using a rectal thermometer (Physitemp Instruments, model TH-5).

Active TGF-β1 levels in the serum and tissue were measured using the Legend Max Free Active TGF-β1 ELISA Kit (BioLegend), following the manufacturer’s protocol. A free fatty acid (FFA) ELISA kit (Cayman Chemical) was used to measure serum FFA levels; the change in serum FFA level was presented in terms of percentage of the overnight-fasted baseline. Liver triglyceride (TG) levels were measured using Abcam’s Triglyceride Assay Kit – Quantification, following the manufacturer’s instructions.

Blood glucose level was measured using Contour Next One blood glucose meter (Ascensia). For glucose tolerance tests, mice that had undergone a 16-h overnight fast were injected (intraperitoneally) with glucose (2 g/kg body weight). For insulin tolerance tests, mice were fasted for 4 h and then injected (intraperitoneally) with insulin (1 unit/kg body weight). For pyruvate tolerance tests, mice that had undergone a 16-h overnight fast were injected with pyruvate (2 g/kg body weight).

Gene transfection and the luciferase reporter gene assay were performed as previously described (21). Briefly, HepG2 cells were cultured in Opi-MEM medium for 6 h. Cells were then subjected to Lipofectamine 3000 (Life Technologies) with plasmids, according to the manufacturer’s instruction. Then, 36 h after the transfection, cell lysates were used to determine luciferase activity, which was estimated using the Promega Dual-Glo Luciferase Assay System (following manufacturer’s instructions) and measured with an Optocomp I Luminometer (MGM Instruments).

Mouse liver and BAT were fixed in 10% neutral buffered formalin; they were then processed into paraffin blocks, sectioned at 5 microns, and stained with hematoxylin and eosin. Stained sections were scanned using ScanScope (Aperio Technologies Inc.). BAT sections were also stained with Mito-Tracker dye (Invitrogen) to determine mitochondrial density.

Data are generally presented as mean ± SEM. Student two-tailed t tests were used when comparing the differences between two groups to determine the statistical significance. One-way or two-way ANOVA tests were used when comparing two groups among multiple groups. P < 0.05 was considered statistically significant.

The data and resources generated in this study will be shared by the corresponding author upon reasonable request.

We first examined hepatic TGF-β1 mRNA levels in 46 human participants from two public data sets (Gene Expression Omnibus series no. GSE15653 and GSE15653) (26,27), and we found a significant positive correlation between hepatic Tgfb1 mRNA levels and BMI, as well as between hepatic Tgfb1 mRNA levels and insulin resistance (Fig. 1A–C). Moreover, we examined a third data set (Gene Expression Omnibus series no. GSE83452) that included data from 58 obese patients with interventions to manage obesity (28) and noted that hepatic Tgfb1 mRNA levels were significantly lower 12 months after bariatric surgery or after 12 months of dietary restriction (Fig. 1D and E). Consistently, serum and hepatic TGF-β1 levels in mice fed an HFD were significantly higher than levels in LFD-fed mice (Fig. 1F and G). Significant increases of serum and hepatic TGF-β1 levels were also observed in db/db mice versus their lean control mice (db/+) (Fig. 1H and I). These results demonstrate a positive correlation between hepatic TGF-β1 expression and obesity in both humans and mice.

Overnight fasting increased hepatic TGF-β1 levels (Supplementary Fig. 1A). To explore the physiological role of hepatic TGF-β1 in glucose metabolism, we generated liver-specific Tgfb1 knockout mice (L-TGF-β1KO) (Supplementary Fig. 1B). Hepatic TGF-β1 deletion did not significantly alter serum TGF-β1 levels but did decrease postprandial and fasting blood glucose levels (Supplementary Fig. 1C and D). L-TGF-β1KO mice had similar glucose tolerance and insulin sensitivity as control (CNTR) mice but significantly decreased gluconeogenesis and hepatic G6pc levels (Supplementary Fig. 1E–H). No significant differences in body weight, food intake, energy expenditure, BAT morphology, and mitochondrial density were measured between L-TGF-β1KO and TGF-β1 L/L mice (Supplementary Fig. 1I–L). These data suggest that hepatic TGF-β1 may play an important role in maintaining glucose homeostasis under the physiological conditions.

To investigate the role of hepatic TGF-β1 in metabolic regulations in obese mice, we maintained L-TGF-β1KO on an HFD for 3 months. Hepatic Tgfb1 deletion decreased liver and serum TGF-β1 levels (Fig. 2A and B). L-TGF-β1KO mice had lower blood glucose levels and improved glucose tolerance and insulin sensitivity, but they had decreased gluconeogenesis and hepatic G6pc levels (Fig. 2C–G and Supplementary Fig. 2A and B). Moreover, hepatic TGF-β1 deficiency significantly decreased the expression of lipogenic genes Fasn and Srebp1c; fibrogenic genes Acta2, Col1a, and Col3a; and inflammatory genes Mcp1 and Tnfα, whereas expression of the fatty acid oxidation–related gene Cpt1 was increased (Fig. 2H and I). Lower liver weights and hepatic TG levels and reduced liver fibrosis were also measured in L-TGF-β1KO mice (Fig. 2J and K and Supplementary Fig. 2C).

L-TGF-β1KO mice had decreased body weight gain, lower fat mass, and higher energy expenditure (Fig. 2L and M and Supplementary Fig. 2D–I). Along with lower weights of epididymal WAT (eWAT), inguinal WAT (iWAT), and BAT, L-TGF-β1KO mice had smaller brown adipocytes with less lipid accumulation and higher mitochondrial density in BAT and increased expression of thermogenic genes in BAT and iWAT (Fig. 2N–Q and Supplementary Fig. 2J). In addition, we observed improved insulin sensitivity, as indicated by increased pAKT-S473 levels, in eWAT of L-TGF-β1KO mice (Fig. 2R). Together, our results indicate that hepatic TGF-β1 deficiency protects against obesity and improves glucose and energy metabolism in HFD-fed mice.

Obesity is highly associated with insulin resistance. We next tested whether hepatic TGF-β1 had a similar role in an insulin-resistant mouse model. Genetic disruption of hepatic insulin signaling by deletion of Irs1 and Irs2 (DKO mice) causes diabetic hyperglycemia and severe insulin resistance (10,20,29). Here, we measured increased hepatic and serum TGF-β1 levels in DKO mice (Fig. 3A and B). Compared with DKO mice, hepatic and serum TGF-β1 levels in liver-specific Irs1, Irs2, and Tgfb1 triple knockout mice (TKObeta1) were decreased by nearly 60% (Fig. 3A and B). Loss of hepatic TGF-β1 in DKO mice resulted in decreased blood glucose levels, improved glucose tolerance and insulin sensitivity, and reduced gluconeogenesis and hepatic G6pc levels (Fig. 3C–G and Supplementary Fig. 3A). RNA sequencing analyses showed that 279 hepatic genes were differentially expressed in DKO mice compared with CNTR mice but were normalized in TKObeta1 mice (Fig. 3H and Supplementary Fig. 3A). These genes are involved in several metabolic pathways, including the Foxo1 signaling pathway (Fig. 3I).

Hepatic TGF-β1 deficiency improved energy expenditure in DKO mice (Fig. 3J and K and Supplementary Fig. 3B–F). Moreover, hepatic TGF-β1 deficiency ameliorated the alterations in adipose tissue weights in DKO mice (Supplementary Fig. 3H). RNA sequencing analyses demonstrated that hundreds of genes involved in metabolic pathways were differentially expressed in BAT of DKO mice compared with CNTR mice (Supplementary Fig. 3G), including genes related to thermogenesis (Ucp1 and Didea), lipid oxidation (Cpt1b and Hadh), mitochondria (Atp5h, Sdhb, and Cox4i1), and adipogenesis (Lpl) (Fig. 3L). Moreover, we observed hypotrophic brown adipocytes and measured lower mitochondrial density and decreased thermogenic gene expression in BAT of DKO mice (Fig. 3M and N). Importantly, all these changes in BAT of DKO mice were ameliorated or abolished in TKObeta1 mice (Fig. 3M and N). Hepatic TGF-β1 deficiency in DKO mice further increased thermogenic gene expression in iWAT (Supplementary Fig. 3I). Whereas DKO mice developed cold intolerance, TKObeta1 mice had cold tolerance similar to that of CNTR mice (Fig. 3O), indicating that hepatic TGF-β1 deficiency improves thermogenesis in DKO mice. Moreover, hepatic TGF-β1 deficiency attenuated TGF-β1 signaling, improved insulin signaling in eWAT, and restored insulin suppression of serum FFA levels in DKO mice (Fig. 3P and Q).

These data suggest that hepatic TGF-β1 exerts an endocrine effect on eWAT, promoting insulin resistance and impairing insulin-suppressed lipolysis. Collectively, our results suggest that hepatic TGF-β1 links hepatic insulin resistance to systemic metabolic dysfunctions.

We next generated liver-specific Tgfb1–overexpressing mice (L-TGF-β1OE) (Supplementary Fig. 4A), which had higher hepatic and serum TGF-β1 levels (Fig. 4A and B). Upon eating an HFD, hepatic TGF-β1 overexpression increased blood glucose levels, impaired glucose tolerance, decreased insulin sensitivity, and increased gluconeogenesis and hepatic G6pc levels (Fig. 4C–G). Furthermore, L-TGF-β1OE mice had increased mRNA levels of Fasn and Srebp1c and decreased mRNA levels of Acox1, along with increased mRNA levels of Acta2, Elastin, Timp1, Mcp1, and Tnfα in the liver (Fig. 4H and I). Moreover, we also measured significantly higher hepatic TG levels and increased liver fibrosis in L-TGF-β1OE mice (Fig. 4J and K and Supplementary Fig. 4B).

L-TGF-β1OE mice had increased body weight gain, higher fat mass, and lower energy expenditure (Fig. 4L and M and Supplementary Fig. 4C–G). Moreover, L-TGF-β1OE mice had hypertrophy, lower mitochondrial density, and decreased thermogenic genes expression in BAT and iWAT, as well as cold intolerance (Fig. 4N and Q). These results suggest that hepatic TGF-β1 overexpression impairs glucose and energy metabolism and promotes nonalcoholic fatty liver disease development upon an HFD treatment.

Our results showed that hepatic TGF-β1 levels were increased by overnight fasting or by hepatic insulin resistance, suggesting the regulation of TGF-β1 expression by insulin signaling. Indeed, insulin decreased Tgfb1 mRNA expression in wild-type (WT) mice (Fig. 5A). However, deletion of Foxo1 in hepatocytes (Foxo1 knockout [FKO]) significantly reduced Tgfb1 mRNA levels and completely blocked insulin action on hepatic Tgfb1 suppression (Fig. 5A–C). Consistently, loss of Foxo1 dramatically decreased hepatic TGF-β1 levels and hepatocyte TGF-β1 secretion in DKO liver or hepatocytes (Fig. 5D and E). On the contrary, overexpression of Foxo1 in hepatocytes increased Tgfb1 mRNA expression (Fig. 5F and G). We found four putative Foxo1 binding sites (insulin response element: TGTTTTG), located at the upstream −2508-bp to approximately −2487-bp region of the mouse Tgfb1 gene (Fig. 5H). Foxo1 overexpression increased Tgfb1 promoter activity, and this effect was abolished by deletion of the putative insulin response elements at the Tgfb1 promoter region (Fig. 5H). Thus, our results suggest that hepatic TGF-β1 expression is regulated by hepatic insulin→Foxo1 signaling.

TGF-β1–deficient hepatocytes had a 14% reduction in HGP, whereas TGF-β1 promoted HGP by 18% in CNTR cells and increased HGP in L-TGF-β1KO hepatocytes, reaching a similar level as TGF-β1–treated CNTR cells (Supplementary Fig. 5A). Moreover, Tgfb1 deletion or TGF-β1 antibody neutralization attenuated HGP in DKO cells (Supplementary Fig. 5B and C). These results indicate an autonomous regulation of HGP by TGF-β1 in hepatocytes. Furthermore, TGF-β1 significantly increased total HGP and gluconeogenesis levels but barely stimulated glycogenolysis (Fig. 5I and J), suggesting that TGF-β1 stimulates HGP largely via promoting gluconeogenesis.

Interestingly, Foxo1 deletion abolished the TGF-β1–stimulated HGP (Fig. 5K). Moreover, TGF-β1 markedly increased Foxo1 protein levels while barely affecting Foxo1 mRNA levels (Fig. 5L and M). Our previous studies have shown that phosphorylation of Foxo1 (pFoxo1) at Ser253 (pFoxo1-S253) by Akt signaling and at Ser273 (pFoxo1-S273) by cAMP-dependent protein kinase (PKA) signaling controls Foxo1 protein stability in hepatocytes (21,30). Here, we found that TGF-β1 increased Foxo1 and pFoxo1-S273 levels without significantly affecting pFoxo1-S253 levels (Fig. 5M). Foxo1-S273A point mutation, in which Foxo1-S273 genetic loci were replaced by alanine that blocks Foxo1-S273 phosphorylation, completely inhibited TGF-β1–stimulated HGP and G6pc expression (Fig. 5N and Supplementary Fig. 5D). Moreover, TGF-β1 promoted PKA activation, and PKA inhibitor (H89) or siPKAc abolished TGF-β1–stimulated HGP, Foxo1 stability, and nuclear localization (Fig. 5O–R and Supplementary Fig. 5E).

These results indicate that TGF-β1 promotes HGP via PKA→pFoxo1-S273 signaling pathway, which was further evidenced by in vivo results (Supplementary Fig. 5F–L). In hepatocytes, cAMP is the major activator of PKA signaling (31); however, we did not observe significant difference in cAMP levels in hepatocytes with TGF-β1 treatment (Fig. 5S). We found that TGF-β1 promoted binding of PKA-C to Foxo1, which potentially promotes PKA-mediated Foxo1 phosphorylation at Ser 273 (Fig. 5T). As a result, glucagon and TGF-β1 synergistically stimulated HGP (Fig. 5U).

The reciprocal regulation of TGF-β1 and Foxo1 suggests an autoinduction of TGF-β1 via Foxo1 in hepatocytes. Indeed, TGF-β1 significantly induced TGF-β1 levels in CNTR hepatocytes, and this effect was lost in Foxo1 knockout cells (Fig. 5V and W). Interestingly, we also observed induction of TGF-β1 by glucagon in hepatocytes (Supplementary Fig. 5M and N). Thus, we conclude that hepatic TGF-β1→Foxo1→TGF-β1 looping system integrates insulin and glucagon signaling in control of HGP (Fig. 5X).

We next investigated the role of hepatic TGF-β1 action in metabolic regulation and cross talk with Foxo1 by deletion of the TGF-β1 receptor II (TβRII) gene (Tgfbr2) in the liver of mice (L-TβRIIKO) (Supplementary Fig. 6A). The HFD-fed L-TβRIIKO mice had decreased blood glucose levels, improved glucose tolerance, enhanced insulin sensitivity, and decreased gluconeogenesis and hepatic Foxo1 and G6pc levels (Fig. 6A–F). Moreover, L-TβRIIKO mice livers had reduced expression of Fasn, Acc1, Scd1, Pparg, Tgfb1, Acta2, Col1a, Col3a, Tnfa, and Il1b (Fig. 6F and G). L-TβRIIKO mice also had lower liver weights and less hepatic lipid accumulation and fibrosis (Fig. 6H and I).

L-TβRIIKO mice had significantly lower body weights and fat massand higher energy expenditure (Fig. 6J–L and Supplementary Fig. 6B–E). With lower weights of adipose tissues, L-TβRIIKO mice had increased expression of thermogenic genes in BAT and iWAT (Supplementary Fig. 6F and Fig. 6M and N). Moreover, hepatic TβRII deficiency improved cold tolerance (Fig. 6O). Together, our results demonstrated that blockade of hepatic TGF-β1 signaling inhibits the induction of hepatic Foxo1-regulated TGF-β1 expression and improves glucose and energy metabolism upon an HFD treatment.

To establish the role of TGF-β1→PKA→pFoxo1-S273 signaling in glucose metabolism in vivo, we generated liver-specific TGF-β1 overexpression in Foxo1-S273A/A mice (L-TGF-β1OE::S273A/A). Upon an HFD feeding, L-TGF-β1OE::S273A/A mice had similar overnight fasting blood glucose levels, glucose tolerance, and insulin sensitivity compared with S273A/A mice (Fig. 7A–C), suggesting Foxo1-S273A mutation blocks hepatic TGF-β1 overexpression-induced hyperglycemia. Moreover, L-TGF-β1OE::S273A/A mice had similar body weight, fat mass, lean mass, food intake, and energy expenditure compared with S273A/A mice (Fig. 7D–J). These results indicate the key role of Foxo1-S273 phosphorylation in TGF-β1–controlled glucose metabolism and bodily energy expenditure. Of note, L-TGF-β1OE::S273A/A mice had similar serum TGF-β1 levels as those of S273A/A mice (Fig. 7K); this suggests that pFoxo1-S273 is required for endogenous TGF-β1 production. Indeed, we found that S273A/A mice had significantly lower serum TGF-β1 levels compared with WT mice (Fig. 7L). Thus, we expect that elevated hepatic TGF-β1 exerts autocrine effects on hepatocytes to activate PKA→pFoxo1-S273 signaling promoting hyperglycemia, and TGFβ1 enhances hepatic TGF-β1 expression and blood TGF-β1 levels, impairing bodily energy expenditure (Fig. 7M).

In this study, we provide genetic and biochemical evidence establishing that hepatic TGF-β1 acts as an important Foxo1-regulated hepatokine in control of blood glucose and bodily energy expenditure in mice. Our results include these three important findings: 1) hepatic TGF-β1 deficiency ameliorates hyperglycemia and improves energy expenditure in insulin-resistant mice, including the DKO mice and HFD-fed mice; 2) hepatic TGF-β1 maintains glucose homeostasis via upregulation of HGP in lean mice, whereas hepatic TGF-β1 overexpression exacerbates dysregulation of glycemia, fat thermogenesis, and energy expenditure in HFD-fed mice; and 3) hepatic TGF-β1 enhances the pathogenesis of hyperglycemic diabetes, obesity, fatty liver, and fibrosis. These results also suggest that hepatic TGF-β1 and its regulation by Foxo1 serve as potential therapeutic targets for prevention and treatment of T2D, obesity, and associated metabolic diseases.

We uncovered an apparently novel reciprocal regulation between Foxo1 and TGF-β1 in hepatocytes. Foxo1 has been reported as a key mediator of TGF-β1 signaling in control of cell differentiation, cartilage homeostasis, and apoptosis (32–34); here, we identified an autocrine effect of hepatocyte TGF-β1 in control of HGP via Foxo1. We demonstrated that TGF-β1 increased Foxo1 levels by stimulating PKA-mediated Foxo1 phosphorylation at Ser 273. Though Yadav et al. previously reported that TGF-β1 promotes Foxo1 activity via PP2A mediated Foxo1 dephosphorylation at Ser 235 in HepG2 cells (14), we did not observe significant changes of pFoxo1-S253 level by TGF-β1 treatment. The TGF-β1–stimulated HGP observed in this study is likely independent of pFoxo1-S253. Moreover, we revealed a regulation TGF-β1 expression by Foxo1 in hepatocytes. Similarly, Ponugoti et al. (35) showed that Foxo1 upregulates TGF-β1 expression in keratinocytes, indicating that the regulation of TGF-β1 by Foxo1 is conserved. The reciprocal regulation between Foxo1 and TGF-β1 suggests an autoinduction of TGF-β1 in hepatocytes. Of note, hepatic TGF-β1 expression also affects body weight and adiposity in diet-induced obese (DIO) mice. Given that obesity is associated with hyperglycemia, the nonautonomous effect of hepatic TGF-β1 on obesity could partially contribute to the dramatic glucose phenotype observed in our present study. Nevertheless, targeting Foxo1-mediated positive feedback of TGF-β1 expression could be a promising strategy for glycemia control and TGF-β1–related liver diseases.

Although TGF-β is generally believed to act locally, recent studies have shown that TGF-β acts as an endocrine cytokine and influences host metabolism (36). For example, TGF-β2 acts as an adipokine to mediate the effects of exercise on glucose and fatty acid metabolism (37). TGF-β1 generated by a tumor microenvironment targets the liver, pancreas, and skeletal muscle to cause hyperglycemia (36). In the present study, we uncovered an endocrine role of TGF-β1 in mediating cross talk between the liver and adipose tissues in obesity and insulin resistance. Previous studies have shown that TGF-β1 inhibits beige adipocyte differentiation and reduces Ucp1 level in adipocyte precursor cells (38,39). Disruption of TGF-β1 signaling promotes brown and beige adipocyte biogenesis in T2D or DIO mice (14). Here, we found that the hepatocyte-expressed TGF-β1 exerts endocrine effects on BAT and iWAT, impairing thermogenesis and browning in these tissues. Blocking TGF-β1 autoinduction by hepatic TβRII deletion alone promoted BAT thermogenesis and iWAT browning, which is consistent with the findings of Zhao et al. (40). Moreover, we observed an endocrine effect of hepatic TGF-β1 on eWAT insulin signaling and insulin suppression of WAT lipolysis in DKO and HFD-fed mice (Supplementary Fig 4H and I). Such an effect caused impaired suppression of FFA levels by insulin, potentially contributing to the elevated HGP in these mice (41). Of note, hepatic TGFβ1 deficiency results in lower eWAT weight in DIO mice, whereas it protects against eWAT loss in DKO mice. The discrepancy may be due to the different availability of circulating FFAs for TG synthesis in eWAT of DKO and DIO mice. DKO mice have limited FFA availability (dramatically lower circulating FFAs levels than CNTR mice (29)) for TG synthesis; thus, TG breakdown is the major controller of lipid accumulation in eWAT. Hepatic TGFβ1 deficiency recovers insulin signaling and insulin suppression of lipolysis in eWAT, thus protecting against eWAT loss in DKO mice. However, DIO mice had elevated circulating FFA levels compared with lean mice. Hepatic TGFβ1 deficiency decreases circulating FFA levels in DIO mice, which will reduce lipid synthesis and eWAT weight in DIO mice. Together, our study highlights the cross talk between liver and adipose tissue in control of glucose and energy metabolism.

In the liver, nonparenchymal cells, such as macrophages and hepatic stellate cells, also express TGF-β1 (42). In the present study, our results suggest a potential role of hepatocyte TGF-β1 in hepatic macrophage and hepatic stellate cell activation in DIO mice, which could contribute to the altered hepatic and circulating TGF-β1 levels and further affect liver and adipose tissue metabolism. Nevertheless, the dramatic glucose and energy phenotypes observed in our study support the idea that hepatocyte TGF-β1 is the key mediator or initiator of metabolic dysregulations in DIO and DKO mice. Besides TGF-β1, many other TGF-β superfamily members are also associated with obesity, insulin resistance, and diabetes. Although sharing similarities in downstream signal transduction, TGF-β superfamily proteins exert divergent functions in metabolic regulations. For instance, activin E promotes BAT and WAT energy expenditure and improves insulin sensitivity in DIO mice. Although BMP4 inhibits insulin signaling and impairs adipose tissue metabolism, BMP7 and BMP9 improve insulin sensitivity and BAT activity (43). Revealing the molecular mechanisms through which TGF-β superfamily members exert metabolic functions will aid in the development of strategies for treating metabolic disorders. The results of this study suggest targeting the hepatic TGF-β1→Foxo1→TGF-β1 looping system could provide a powerful intervention for prevention and treatment of obesity, T2D, and related metabolic diseases in the future.

Acknowledgments. The authors thank Michael R. Honig, who provided English editing of the manuscript. The use of the Texas A&M Rodent Preclinical Phenotyping Core is acknowledged.

Funding. This work was supported by the National Institutes of Health (grants R01DK095118, R01 DK120968, and R01DK124588), an American Diabetes Association Career Development Award (1-15-CD-09), Faculty Start-up funds from Texas A&M University Health Science Center and AgriLife Research, and the U.S. Department of Agriculture National Institute of Food and Agriculture (grant Hatch 1010958) to S.G. (principal investigator). This work was also partially supported by National Institutes of Health grants R01DK118334 and R01AG064869 to Y.S. and S.G.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Authors Contributions. Q.P. designed the study, carried out research, interpreted the results, and wrote the manuscript, W.A., Y.C., and D.M.K. performed research and reviewed the manuscript. Z.S., W.Y., W.J., Y.S., and S.S. reviewed the manuscript. S.G. designed the study, analyzed the data, and reviewed and revised the manuscript. All authors approved the final version of the manuscript. S.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 79th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 7–11 June 2019.