Acute Deletion of the FOXO1-dependent Hepatokine FGF21 Does not Alter Basal Glucose Homeostasis or Lipolysis in Mice

Abstract

The hepatic transcription factor forkhead box O1 (FOXO1) is a critical regulator of hepatic and systemic insulin sensitivity. Previous work by our group and others demonstrated that genetic inhibition of FOXO1 improves insulin sensitivity both in genetic and dietary mouse models of metabolic disease. Mechanistically, this is due in part to cell nonautonomous control of adipose tissue insulin sensitivity. However, the mechanisms mediating this liver-adipose tissue crosstalk remain ill defined. One candidate hepatokine controlled by hepatic FOXO1 is fibroblast growth factor 21 (FGF21). Preclinical and clinical studies have explored the potential of pharmacological FGF21 as an antiobesity and antidiabetic therapy. In this manuscript, we performed acute loss-of-function experiments to determine the role of hepatocyte-derived FGF21 in glucose homeostasis and insulin tolerance both in control and mice lacking hepatic insulin signaling. Surprisingly, acute deletion of FGF21 did not alter glucose tolerance, insulin tolerance, or adipocyte lipolysis in either liver-specific FGF21KO mice or mice lacking hepatic AKT-FOXO1-FGF21, suggesting a permissive role for endogenous FGF21 in the regulation of systemic glucose homeostasis and insulin tolerance in mice. In addition, these data indicate that liver FOXO1 controls glucose homeostasis independently of liver-derived FGF21.

The liver is an important metabolic organ that orchestrates systemic carbohydrate and lipid metabolism. One of the liver’s major roles is to respond to periods of nutrient abundance and scarcity to control systemic metabolism. For example, during periods of starvation, the liver produces glucose to meet the body’s metabolic demands by increasing hepatic glucose production. On the other hand, postprandial nutrient availability leads to insulin secretion from the pancreas and subsequent activation of hepatic insulin signaling. In the liver, insulin signaling suppresses glucose production by inhibition of glycogenolysis and gluconeogenesis. In addition to glucose metabolism, insulin increases de novo lipogenesis, export triglyceride-rich lipoproteins in the liver, and suppresses fatty acid oxidation (1). However, in metabolic diseases such as type 2 diabetes mellitus and obesity, insulin-dependent regulation of these processes is abnormal, leading to sustained hyperglycemia and lipid synthesis in the face of hyperinsulinemia and insulin resistance (2, 3). Thus, understanding the specific mechanisms underlying hepatic insulin action and the control of systemic glucose and lipid metabolism is of paramount importance.

The serine/threonine-protein kinase B, also known as AKT, is a critical node in hepatic insulin cascade signaling. In mammals, there are 3 isoforms of AKT (AKT1, AKT2, and AKT3). AKT1 and AKT2 are the dominant isoforms in insulin-responsive metabolic tissues such as muscle, adipose tissue, and liver (4-9). One of the key targets of AKT is the family of forkhead box O (FOXO) transcription factors. There are multiple isoforms of the FOXO transcription factors family in mammals, including FOXO1, FOXO3a, FOXO4, and FOXO6, with FOXO1 serving as the primary FOXO isoform in the liver. These transcription factors are critical regulators of several genes involved in the adaptation to fasting and reduced insulin levels (10, 11). In response to insulin action on hepatocytes, AKT directly phosphorylates 3 conserved residues (FOXO1: T24, S256, and S319). Phosphorylation of 2 of these sites (FOXO1: T24 and S256) allows the binding of the 14-3-3 family of phospho-binding proteins, which induces an acute translocation of FOXO proteins from the nucleus into the cytoplasm, thus inhibiting the FOXO transcriptional program (12, 13).

The AKT-dependent inhibition of FOXO1 is central to the control of hepatic lipid synthesis and glucose production (14). Loss of hepatic AKT results in impaired glucose tolerance and systemic insulin resistance; however, concomitant hepatic deletion of FOXO1 normalizes these metabolic disturbances. This is due, in part, to the FOXO1-dependent cell-nonautonomous control of hepatic glucose production via inhibition of white adipose tissue (WAT) lipolysis (15-17). Mechanistically, several potential candidates have been proposed to link excess hepatic FOXO1 activity with adipose tissue insulin resistance and systemic glucose intolerance (16-18).

The hepatokine fibroblast growth factor 21 (FGF21) is implicated in the control of carbohydrate and lipid metabolism (19). Circulating FGF21 levels are mainly derived from the liver (20) and signal through the cell-surface FGF receptor 1c (FGFR1c) isoform and its coreceptor β-klotho, a single-pass transmembrane protein (21-24). FGF21 activation of the FGFR1c/β-klotho complex leads to several signaling events, including regulation of peripheral metabolism via central effects in the central nervous system (25). FGF21 levels increase in response to several physiological challenges, including nutrient restriction such as fasting and low-protein diets (26-28), as well by carbohydrate intake, including glucose, sucrose, and fructose consumption (29-33). Recently, work by our group and others demonstrated that Fgf21 gene expression is increased following genetic deletion of hepatic FOXO1 (18, 34-36), consistent with the insulin-dependent upregulation of FGF21 observed in humans (37). These data suggest that hepatic insulin signaling via FOXO1 is an essential regulator of hepatic FGF21 expression.

Over the last several years, FGF21 has emerged as a potential treatment for obesity and type 2 diabetes mellitus because of the insulin-sensitizing effects seen in obese and insulin-resistant mice and nonhuman primates (38-42). Even though the metabolic effects of pharmacological FGF21 administration have been fairly well studied, gaps remain in our understanding of how hepatic insulin signaling contributes to the regulation of endogenous hepatic FGF21 expression and its physiological role as a FOXO1-dependent hepatokine in the regulation of metabolism. Relatively limited studies have addressed the role of liver-derived FGF21 in systemic metabolism; however, these studies employed chronic models of FGF21 deficiency from birth that may lead to secondary effects due to sustained FGF21 deficiency (18, 20). Therefore, we evaluated the effects of acute deletion of hepatocyte-specific FGF21 on glucose homeostasis in control and mice lacking hepatic AKT-FOXO1 signaling. Interestingly, acute deletion of FGF21 does not alter glucose homeostasis, insulin tolerance, or adipocyte insulin sensitivity in either control or mice lacking hepatic AKT and FOXO1 signaling, suggesting a minimal role for endogenous FGF21 in the systemic control of basal glucose homeostasis and adipocyte lipolysis in mice.

Materials and Methods

Experimental Model and Subject Details

Mice

Akt1 ^loxp/loxp , Akt2^loxp/loxp; Akt1^loxp/loxp, Akt2^loxp/loxp, Foxo1^loxp/loxp; Akt1^loxp/loxp, Akt2^loxp/loxp, Foxo1^loxp/loxp, Fgf21^loxp/loxp; Fgf21^loxp/loxp, and Foxo1^loxp/loxp male mice (7, 36, 43) were injected at age 6 to 14 weeks with 10¹¹ genomic copies of adeno-associated virus (AAV) expressing either GFP or Cre recombinase under the liver-specific promoter, thyroxine-binding globulin (TBG), per mouse to generate control, L-AktDKO, L-AktFoxo1TKO, L-QKO, L-Fgf21KO, and L-Foxo1KO mice. The control group consisted of GFP-injected littermates floxed for each group. Experiments in nonobesogenic diets were performed 2 to 3 weeks post AAV injection. All mice were housed at room temperature (RT) (22 °C). To induce diet-induced obesity, control and L-Foxo1KO mice were put on a high-fat diet (60 kcal% fat; D12492i, Research Diets) after AAV injection for 26 to 28 weeks. For fasted/re-fed experiments, mice were fasted for 16 hours for the fasted conditions and then placed under ad libitum fed conditions for 4 hours to complete the 4 hours re-fed condition. Animal use followed all standards and guidelines of the Institutional Animal Care and Use Committee at the University of Pennsylvania.

Western blotting

Liver and adipose tissue samples were homogenized in radioimmunoprecipitation assay (50mM Tris HCl, 1% Triton x100, 0.5% Sodium deoxycholate, 0.1% sodium dodecyl sulfate, 150 mM NaCl, 2 mM EDTA) buffer supplemented with protease inhibitor cocktail tablets (Roche; 04693159001), and phosphatase inhibitor cocktail II and III (Sigma-Aldrich; P5726; P0044) in a TissueLyser (Qiagen). Samples were separated on 4% to 15% Mini-PROTEAN TGX precast gels (BioRad; 4561084). Primary antibodies used were HSP90 (1:1,000; catalog No. 4874; RRID:AB_2121214), AKT2 (1:1,000; catalog No. 2964; RRID:AB_331162), P-AKT2 (1:1000; catalog No. 8599; RRID:AB_26303478), IGFBP1 (1:1000; catalog No. sc-6000; RRID:AB_2123087), glucokinase (GCK; gift from Dr Magnuson). P-AKT1/2 (1:1000; catalog No. 4060; RRID:AB_2315049), AKT (1:1000; catalog No. 4691, RRID:AB_915783), P-S6 (1:1000; catalog No. 2215; RRID:AB_331682), S6 (1:1000; catalog No. 2217; RRID:AB_331355), P-HSL (1:1000; catalog No. 4126; RRID:AB_490997), HSL (1:1000; catalog No. 4107; RRID:AB_2296900), P-Perilipin (1:5000; catalog No. 4856; RRID:AB_2909466), and Perilipin (1:1000; catalog No. 9349; RRID:AB_10829911).

Messenger RNA isolation and real-time polymerase chain reaction

Total RNA was isolated from the frozen livers samples using the RNeasy Plus kit from Qiagen. Complementary DNA was generated using M-MuLV (New England Biolabs; M0253L) reverse transcriptase and quantitated for the relative expression of genes of interest by real-time polymerase chain reaction using the SYBR Green (Fisher Scientific; 4368702) dye-based assay.

Glucose tolerance test

For the glucose tolerance test (GTT), overnight fasted mice were injected with 2 g/kg of glucose solution intraperitoneally. Blood glucose was measured at 0, 15, 30, 60, and 120 minutes after glucose injection.

Insulin tolerance test

For the insulin tolerance test (ITT), 5-hour fasted mice were injected with 0.75 U/kg of insulin intraperitoneally. Blood glucose was measured at 0, 15, 30, 45, and 60 minutes after insulin injection.

Blood Chemistry

Serum free fatty acid concentrations were measured using NEFA-HR (2) R1 and R2 reagents (Wako Chemicals; 999-34691; 991-34891) following the manufacturer’s instructions, and glycerol concentrations were measured using 5 µL of sample and 200 µL free glycerol reagent (Sigma-Aldrich; F6428). Both methods were quantified using enzymatic colorimetric analysis. Insulin levels were measured using an ultrasensitive enzyme-linked immunosorbent assay (ELISA; Crystal Chem; 90080; RRID:AB_2783626). FGF21 levels were measured using an ELISA (BioVendor; RD291108200R; RRID:AB_2909467).

Hepatic Glycogen Content

Liver samples were homogenized in 6% perchloric acid solution. Then, the samples were centrifuged at 4 °C at 16 200g for 10 minutes. Next, the supernatant was diluted and centrifuged again at the same conditions. After this, 10N KOH was added to the sample to reduce it to a 6 to 7 pH level. Next, the samples were centrifuged at the same conditions one last time, and the supernatant was saved. Finally, 100 µL amyloglucosidase (1 mg/mL stock solution) was added to 20 µL of the sample and incubated at 40 °C for 2 hours while shaking. The Glucose Assay Reagent (Sigma; G3293) was used to determine the glycogen content.

Lipolysis Ex Vivo

Fresh epididymal white adipose tissue (eWAT) depots were dissected from ad libitum–fed mice and put into FluoroBrite DMEM (Thermo Fisher Scientific; A1896701). The depots were then cut into 5 small similar-sized pieces and transferred into 150 µL FluoroBrite Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 2% fatty acid (FA)-free bovine serum albumin (BSA) in 96-well plates for 30 minutes for basal preincubation. To analyze basal lipolysis, tissues were transferred into 150 µL of fresh FluoroBrite DMEM supplemented with 2% FA-free BSA for 60 minutes. This media was collected to measure glycerol levels of basal lipolysis. Then, to analyze CL 316 243 -stimulated lipolysis, the tissues were transferred and pre-incubated for 30 min in 150 µL FluoroBrite DMEM supplemented with 2% FA-free BSA in the presence of 100 µM CL 316243 (Sigma-Aldrich; C5976-5MG). To analyze stimulated lipolysis, tissues were transferred into 150 µL fresh FluoroBrite DMEM supplemented with 2% FA-free BSA and 100 µM of CL 316 243 for another 60 minutes. This media was collected to measure glycerol levels of the stimulated lipolysis. Glycerol levels were analyzed by combining 5 µL of supernatant and 200 µL of Free Glycerol Reagent (Sigma-Aldrich; F6428) and incubating for 15 minutes at RT before measuring absorbance at 540 nm. At the end of the experiment, the tissue pieces were delipidated by CHCL₃ extraction and solubilized in 0.3 N NaOH/0.1% sodium dodecyl sulfate at 65 °C overnight. Protein content was determined using a Pierce BCA Protein assay. Results are expressed as micromole (µmol) of glycerol per milligram (mg) of tissue protein.

Quantification and Statistical Analysis

Statistical analysis was performed using 1-way analyses of variance when more than 2 groups were compared, 2-way analyses of variance when 2 conditions were analyzed, and the unpaired 2-tailed t test when 2 groups were being assayed. All data are presented as ± SEM. Asterisks in the figures are defined as follows: * indicates a P value less than .05; ** indicates a P value less than .01; *** indicates a P value less than .001, and **** indicates a P value less than .0001.

Results

Hepatic AKT-FOXO1 Axis Regulates Glucose Tolerance, Fasting Hyperinsulinemia, and Fgf21 Expression

Liver insulin signaling orchestrates systemic metabolism. In mice, deletion of both liver isoforms of AKT, Akt1 and Akt2, induced insulin resistance and glucose intolerance, in part by causing adipose tissue insulin resistance. These abnormalities are dependent on increased FOXO1 activity, as deletion of Foxo1 in the same liver restored insulin sensitivity and glucose homeostasis via both direct and indirect mechanisms (7). The direct mechanism in liver is likely due to increased expression of hepatic Gck and suppression of G6pc, which restores hepatic glucose utilization. However, controversy remains over how FOXO1 cell-nonautonomously regulates adipose tissue insulin sensitivity. To gain more mechanistic insight into the underlying extrahepatic mechanisms mediating this effect, we generated new cohorts of mice by injecting 8- to 14-week-old Akt1^loxP/loxP, Akt2^loxP/loxP mice with an AAV expressing Cre recombinase or GFP under the thyroxine-binding globulin (Tbg) promoter to make an acute liver-specific knockout (L-AktDKO) or control mice, respectively. As expected, L-AktDKO mice were glucose intolerant and hyperinsulinemic, thus serving as a suitable genetic insulin-resistant model (Fig. 1A and 1B) (7). Consistent with previous work, these metabolic phenotypes are normalized by simultaneous deletion of hepatic Foxo1 (L-AktFoxo1TKO) (see Fig. 1A and 1B) (7). To determine if deletion of liver FOXO1 alone was sufficient to improve glucose tolerance during diet-induced insulin resistance, we injected 6-week-old Foxo1^loxP/loxP mice with AAV-TBG-GFP or with AAV-TBG-CRE to induce liver-specific deletion of FOXO1 (L-Foxo1KO) and placed them on a high-fat diet (60 kcal% fat; D12492i, Research Diets) for 26 to 28 weeks after AAV injection to induce an obesity-driven insulin-resistant phenotype. Under these conditions, mice with liver-specific deletion of FOXO1 had improved glucose tolerance (Fig. 1D) and lower circulating insulin levels when compared to control mice, suggesting improved insulin sensitivity (Fig. 1E). This improvement in glucose tolerance was independent of weight loss (data not shown) and consistent with previous data using a different obesogenic diet (43).

Figure 1.

Hepatic AKT-FOXO1 axis regulates glucose tolerance, hyperinsulinemia, and Fgf21 expression. A, Glucose levels of control, L-AktDKO, and L-AktFoxo1TKO mice housed at room temperature (RT) after intraperitoneal (ip) injection of 2 mg/kg of glucose solution (n = 15 for control mice, n = 9 for L-AktDKO mice, and n = 8 for L-AktFoxo1TKO mice). B, Fasting insulin levels of control, L-AktDKO, and L-AktFoxo1TKO mice housed at RT (n = 11 for control mice, n = 7 for L-AktDKO mice, and n = 8 for L-AktFoxo1TKO mice). C, Relative messenger RNA (mRNA) levels of Fgf21 in liver from control, L-AktDKO, and L-AktFoxo1TKO mice housed at RT (n = 13 for control mice, n = 9 for L-AktDKO mice, and n = 11 for L-AktFoxo1TKO mice). D, Glucose level of control and L-Foxo1KO mice on a high-fat diet (HFD) for 26 to 28 weeks after ip injection of 2 g/kg of glucose solution (n = 5 for control HFD mice, and n = 9 for L-Foxo1KO HFD mice). E, Fasting insulin levels of control and L-Foxo1KO mice on an HFD for 26 to 28 weeks housed at RT (n = 5 for control HFD mice, and n = 9 for L-Foxo1KO HFD mice). F, Relative mRNA levels of Fgf21 in the liver from control and L-Foxo1KO mice on an HFD for 26 to 28 weeks housed at RT (n = 9 for control HFD mice, and n = 7 for L-Foxo1KO HFD mice). Data are presented as means ± SEM. *P less than .05, **P less than .01, ***P less than .001.

Given the extensive published literature, we focused on FGF21 as a potential FOXO1-dependent hepatokine mediating the cell-nonautonomous control of insulin sensitivity in following FOXO1 deletion. Our group recently reported that liver FOXO1 represses Fgf21 expression in the liver (36), an observation consistent with elevation of Fgf21 in FOXO1-deficient mouse models (34, 35). In support of this notion, L-AktDKO mice displayed decreased Fgf21 gene expression that was rescued completely in L-AktFoxo1TKO, confirming the repressive effect of FOXO1 activity on Fgf21 expression (Fig. 1C). Similarly, liver-specific deletion of hepatic FOXO1 under diet-induced obesity conditions increased hepatic Fgf21 gene expression compared to control mice (Fig. 1F). Our group also reported recently that FOXO1 alone was sufficient to cause glucose intolerance and fasting hyperinsulinemia, which correlated with decreased hepatic Fgf21 gene expression, consistent with the data presented in Fig. 1 (36). Collectively, these data show that genetic inhibition of FOXO1 improves glucose homeostasis both in genetic and dietary mouse models of insulin resistance. These improvements in glucose homeostasis following FOXO1 deletion correlate with increased hepatic Fgf21 gene expression in multiple models.

Acute Loss of Hepatic Fibroblast Growth Factor 21 Does Not Affect Glucose or Insulin Tolerance in Control or Mice Lacking Hepatic Insulin Signaling

To determine if the restoration of Fgf21 expression was required to improve glucose homeostasis following AKT and FOXO1 deletion, we generated mice lacking either FGF21 alone or mice lacking AKT1/2, FOXO1, and FGF21 (L-QKO) by injecting AAV-TBG-CRE into Fgf21^loxP/loxP or Akt1^loxP/loxP, Akt2^loxP/loxP, Foxo1^loxP/loxP, Fgf21^loxP/loxP mice. In addition, AAV-TBG-GFP–injected littermates mice served as controls. Successful KO of targeted genes was confirmed by measuring the expression of Akt2, Foxo1, and Fgf21 in the liver and by measuring circulating levels of FGF21 of control, L-Fgf21KO, or L-QKO (Fig. 2A and 2B). First, we determined if acute loss of liver FGF21 alone impaired liver insulin signaling response to refeeding by monitoring AKT signaling. Here, we performed fasting (16 hours) and refeeding (4 hours) experiments and collected the liver for Western blot analysis. We observed increased phosphorylation of AKT2 in L-Fgf21KO mice after refeeding comparable to control mice, suggesting normal postprandial AKT activation both in control and L-Fgf21KO livers (Fig. 2C). We also observed a reduction in IGFBP1 expression on refeeding similar to control mice, consistent with the typical decrease in FOXO1 activity on refeeding (see Fig. 2C). Finally, levels of hepatic GCK were unaffected in L-Fgf21KO compared to control mice (see Fig. 2C). On the other hand, L-QKO mice showed the expected loss of AKT signaling using the same feeding paradigm. In L-QKO mice, we observed a decrease in phosphorylation of AKT2 on refeeding and total AKT2 levels both in fasted and re-fed conditions of L-QKO mice compared to control mice (Fig. 2D). In addition, protein levels of IGFBP1 increased during fasting conditions but decreased on refeeding conditions of L-QKO mice (see Fig. 2D). Consistent with a mouse model of deficient liver AKT signaling (7, 15), L-QKO mice showed a significant decrease in GCK protein levels both in fasting and refeeding conditions (see Fig. 2D).

Figure 2.

Acute loss of hepatic fibroblast growth factor 21 (FGF21) alone does not affect liver insulin signaling. A, Relative messenger RNA (mRNA) levels of Akt2, Foxo1, and Fgf21 in the liver from control, L-Fgf21KO, and L-QKO mice housed at room temperature (RT) after 16 hours fast or 4 hours re-fed (n = 4-9 per group/per condition). B, FGF21 serum levels of control, L-Fgf21KO, and L-QKO mice after 16 hours fasting (n = 8 for control mice, n = 5 for L-Fgf21KO mice, and n = 6 for L-QKO mice). C, Western blot analysis of phosphorylation (p-) of AKT2 at Ser474, total AKT2, IGFBP1, and GCK of liver from control and L-Fgf21KO mice housed at RT. D, Western blot analysis of phosphorylation (p-) of AKT2 at Ser474, total AKT2, IGFBP1, and GCK of liver from control and L-QKO mice housed at RT. Data are presented as means ± SEM. *P less than .05, **P less than .01, ***P less than .001.

Next, we determined if liver-derived FGF21 was required for glucose homeostasis and asked if the normalization in glucose tolerance seen in L-AktFoxo1TKO mice is dependent on the restoration of liver Fgf21 gene expression. Therefore, we first measured glucose serum levels from mice fasted for 16 hours and/or re-fed for 4 hours. No difference in fasting or refeeding glucose levels were observed between the 3 genotypes (Fig. 3A). Next, a glucose tolerance test and ITT were performed, which did not reveal any significant abnormalities in glucose or insulin tolerance in either liver-specific FGF21 KO alone or L-QKO mice (Fig. 3B and 3C). In support of the ITT assessment, fasted and/or ad libitum levels of circulating insulin were no different between the genotypes (Fig. 3D and 3E). These results indicate that the acute deletion of liver FGF21 has minimal effects on glucose homeostasis under normal conditions and does not contribute to the insulin-sensitizing effects of inhibition of FOXO1 in AKT-deficient mice.

Figure 3.

Acute loss of hepatic fibroblast growth factor 21 (FGF21) does not affect glucose or insulin tolerance with or without liver insulin signaling. A, Blood glucose level of control, L-Fgf21KO, and L-QKO mice housed at room temperature (RT) after 16 hours fast or 4 hours re-fed (n = 5-9 per group/per condition). B, Glucose level of control, L-Fgf21KO, and L-QKO mice housed at RT after intraperitoneal (ip) injection of 2 mg/kg of glucose solution (n = 20 for control mice, n = 12 for L-Fgf21KO mice, and n = 15 for L-QKO mice). C, Glucose level of control, L-Fgf21KO, and L-QKO mice housed at RT after ip injection of 0.75 U/kg of insulin (n = 15 for control mice, n = 8 for L-Fgf21KO mice, and n = 13 for L-QKO mice). D, Insulin levels of control, L-Fgf21KO, and L-QKO mice housed at RT after 16 hours fasted (n = 18 for control mice, n = 7 for L-Fgf21KO mice, and n = 15 for L-QKO mice). E, Insulin levels of random fed control, L-Fgf21KO, and L-QKO mice housed at RT (n = 13 for control mice, n = 7 for L-Fgf21KO mice, and n = 9 for L-QKO mice). F, Liver glycogen content of control, L-Fgf21KO, and L-QKO mice housed at RT after 16 hours fast or 4 hours re-fed (n = 4-9 per group/per condition). Data are presented as means ± SEM. *P less than .05, **P less than .01, ***P less than .001.

Although FOXO1 deletion corrects many glucose abnormalities of L-AktDKO mice, it does not entirely restore the liver’s ability to store glycogen in the re-fed state, confirming that AKT is required for postprandial glycogen deposition (7, 15). Consistent with this notion, L-QKO mice showed a significant decrease in liver glycogen content compared to control mice under re-fed conditions. In contrast, no significant change was observed in L-Fgf21KO mice (Fig. 3F), confirming that liver AKT signaling is required for glycogen synthesis. This effect on glycogen content paralleled the reduction in hepatic GCK expression observed in L-QKO mice (see Fig. 2C). Collectively, these data demonstrate that acute deletion of liver-derived FGF21 has minimal effects on glucose homeostasis and insulin tolerance in mice in vivo.

Loss of Liver Fibroblast Growth Factor 21 Does not Affect Adipose Tissue Insulin Sensitivity or Lipolysis in Mice With or Without Hepatic Insulin Signaling

As previously mentioned, FOXO1 deletion improves insulin sensitivity in L-AktDKO mice in part via cell-nonautonomous control of adipocyte lipolysis in the white adipose tissue (WAT) (15-17). Therefore, we addressed if acute loss of liver FGF21 alone or the FOXO1-dependent control over FGF21 affects insulin signaling and lipolysis in the adipose tissue. To assess insulin signaling in white adipose tissue, we measured the phosphorylation and activation status of AKT and the mammalian target of rapamycin complex 1 (mTORC1)-dependent target S6 after 16-hour fasting or after 4-hour refeeding conditions in inguinal WAT (iWAT). On refeeding conditions, we observed an increase in the phosphorylation both of AKT and S6 when compared to fasted states in control mice (Fig. 4A). This increase on refeeding was similar between control, L-Fgf21KO, and L-QKO mice (see Fig. 4A). These data indicate that loss of circulating FGF21 does not affect insulin signaling in WAT in vivo.

Figure 4.

Acute loss of hepatic fibroblast growth factor 21 (FGF21) does not affect adipose tissue insulin signaling or lipolysis in mice with or without liver insulin signaling. A, Western blot analysis of phosphorylation (p-) of AKT1/2 at Ser474/474, Pan AKT, phosphorylation (p-) of S6 at Ser240/244, S6, and HSP90 of inguinal white adipose tissue (iWAT) from control, L-Fgf21KO, and L-QKO mice housed at room temperature (RT) after 16 hours fasted/4 hours re-fed. B, Western blot analysis of phosphorylation (p-) of HSL at Ser660, HSL, phosphorylation (p-) of Perilipin at Ser522, and Perilipin of iWAT and epididymal white adipose tissue (eWAT) from control and L-Fgf21KO mice housed at RT after 16 hours fasted/4 hours re-fed. C, Western blot analysis of phosphorylation (p-) of HSL at Ser660, HSL, phosphorylation (p-) of Perilipin at Ser522, and Perilipin of iWAT and eWAT from control and L-QKO mice housed at RT after 16 hours fasted/4 hours re-fed. D, Glycerol serum levels of control, L-Fgf21KO, and L-QKO mice housed at RT after 16 hours fast or 4 hours re-fed (n = 7-15 per group/per condition). E, NEFA serum levels of control, L-Fgf21KO, and L-QKO mice housed at RT after 16 hours fast or 4 hours re-fed (n = 7-15 per group/per condition). F, Glycerol levels of control, L-Fgf21KO, and L-QKO mice eWAT explants before and after 100 µM CL treatment (n = 4 for control mice and n = 3 for L-Fgf21KO and L-QKO mice). Data are presented as means ± SEM. *P less than .05, **P less than .01, ***P less than .001.

In addition to adipose tissue insulin signaling, we next probed the key lipolysis regulators in eWAT and iWAT. Here, we determined the phosphorylation status of hormone-sensitive lipase (pSer660 HSL) and Perilipin (pSer522) in response to fasting and refeeding (44-47). In all genotypes, the phosphorylation status of HSL (pSer660) and Perilipin (pSer522) was increased during fasting conditions and suppressed to a similar extent between all conditions (Fig. 4B and 4C). These effects on insulin and lipolysis signaling correlated with similar suppression of circulating glycerol and free fatty acids in all genotypes (Fig. 4D and 4E). Last, we performed ex vivo lipolysis assays in basal and in response to CL 316 243 treatment to stimulate lipolysis. Explants of eWAT from L-Fgf21KO, and L-QKO mice induced glycerol levels to a similar extent as control mice (Fig. 4F). Taken together, these data demonstrate that acute deletion of liver-derived FGF21 does not affect adipose tissue lipolysis or insulin signaling in vivo.

Metabolic Response to Feeding on Acute Loss of Liver Fibroblast Growth Factor 21

In the liver, the AKT-FOXO1 axis serves as an important signaling node downstream of insulin responsible for feeding-dependent regulation of gene expression and metabolism (7). In addition to liver AKT-FOXO1 signaling in fasting/refeeding regulation, FGF21 is suggested to affect genes involved in hepatic glucose and lipid metabolism (19). Given that liver insulin signaling via the AKT-FOXO1 axis and liver FGF21 are involved in the transcriptional response to nutrient intake, we next determined how acute loss of hepatic FGF21 contributes to gluconeogenic and lipogenic gene expression in L-Fgf21KO and L-QKO mice. G6pc and Pck1 levels in L-Fgf21KO and L-QKO mice were largely not different from control mice during fasting; however, there was a partial blunting of G6pc and Pck1 in L-QKO re-fed mice (Fig. 5). Igfbp1 and Pgc1a expression in control and L-Fgf21KO mice did not change during fasting; however, L-QKO mice partially increased basal expression compared to control mice. On refeeding, Igfbp1 and Pgc1a gene suppression were normal in control, L-QKO, and L-Fgf21KO mice (see Fig. 5). Irs2 showed no difference between the groups in both conditions (see Fig. 5). Last, the lipogenic genes, Srebp1c and Gck, were induced by refeeding in L-Fgf21KO mice while L-QKO displayed a significant reduction in these lipogenic genes (see Fig. 5). These results from the lipogenic genes (Gck, Srebp1c, and Fasn) in the L-QKO mice were expected given that AKT-dependent activation of mTORC1 is also required to drive postprandial lipogenic gene expression (15).

Figure 5.

Metabolic responsive genes to feeding in L-Fgf21KO and L-QKO mice. Relative messenger RNA levels of G6pc, Pck1, Igfbp1, Srebp1c, Fasn, Gck, Pgc1α, and Irs2 in liver from control, L-Fgf21KO, and L-QKO mice housed at room temperature after 16 hours fast or 4 hours re-fed (n = 5-12 per group/per condition). Data are presented as means ± SEM. *P less than .05, **P less than .01, ***P less than .001.

Discussion

In addition to liver-specific pathways to regulate systemic metabolism, liver-derived circulating FGF21 affects carbohydrate and lipid metabolism (19). Owing to its positive effects on systemic glucose metabolism, several pharmaceutical companies have designed and tested FGF21 analogues and derivatives in human clinical trials (48-51). Unfortunately, efficacy was quite modest in terms of glycemic control compared to rodent studies (42, 48, 52-54). Despite robust pharmacological findings in rodents, there is a dearth of knowledge related to the physiological role of FGF21 in systemic glycemic control, insulin sensitivity, and body weight regulation. To address this knowledge gap, we used an acute loss-of-function approach to determine the physiological role of hepatic FGF21 both in normal conditions and against the background of hepatic insulin signaling deficiency. Here, we show that acute loss of hepatic FGF21 does not alter systemic glucose metabolism, insulin tolerance, or adipose tissue lipolysis under either condition.

It is important to note that these results differ slightly from the 2 previously published manuscripts investigating the role of liver-derived FGF21 (18, 20). In Markan et al (20), mice with congenital liver-specific deletion of FGF21 have a very modest impairment in glucose tolerance under normal chow-fed conditions, an effect that is more pronounced under obesogenic conditions. Mechanistically, the authors argue that brown adipose tissue glucose uptake was impaired following the loss of liver FGF21 under these conditions, suggesting that brown adipose tissue response to FGF21 is required for proper systemic glucose uptake. There are 2 main differences between this study and ours: (1) as mentioned; this study was performed in mice lacking liver FGF21 from birth, raising the possibility of other signaling and compensatory metabolic mechanisms contributing to the metabolic phenotype, and (2) the most significant phenotypes were observed under obesogenic conditions, which were not explored in our manuscript. In our manuscript, we specifically decided to try to minimize any compensation that can occur from a congenital loss of function; therefore, we took advantage of the AAV-TBG system and developed an acute KO model of liver FGF21. In addition, a major goal of our study was to define the physiological role of FGF21 under nonobesogenic conditions to explore the biology of FGF21 in controlling glucose homeostasis under normal conditions in vivo.

Similar to our manuscript, Stöhr et al (18) published that inhibition of hepatic insulin action (induced by deletion of the insulin receptor substrate proteins) decreases hepatic Fgf21 expression in a FOXO1 dependent manner. The authors employed adenoviruses to overexpress FGF21 and found it increased adipose tissue glucose uptake that correlated with systemic insulin sensitivity changes. However, there are some important technical differences between this study and the one described here: (1) a congenital liver-specific Cre system was afforded the potential of compensatory or secondary mechanisms to compensate; and (2) the use of adenoviruses to overexpress FGF21 and short hairpin-expressing viruses to knock down hepatic Fgf21. Here, we performed acute loss of function to minimize artifacts due to overexpression, efficacy concerns of short hairpin RNA delivery to hepatocytes in vivo, and inflammation-induced by adenovirus delivery in vivo that is known to induce acute bouts of hyperglycemia. Using AAV methods to generate robust hepatocyte-specific KOs in vivo, we did not observe any change in glucose or insulin tolerance or adipose tissue insulin signaling and lipolysis following acute KO of FGF21 in either control mice or mice lacking hepatic AKT and FOXO1 signaling under normal chow feeding conditions.

In addition to defining the physiological role of liver-derived FGF21 in animals with intact hepatic insulin signaling, our data also showed that improvements in systemic insulin sensitivity following FOXO1 deletion are unlikely to require intact FGF21. It is important to note that these experiments do not question the robust insulin-sensitizing effects of pharmacological FGF21 administration, which are well documented (38-42, 52, 55-57). Rather, these studies were intended to clarify the physiological role of hepatocyte-derived FGF21 on systemic metabolism using acute genetic loss-of-function experiments in mouse models under nonobesogenic conditions. Moreover, these experiments do not address the role of adipose tissue FGF21 signaling in the acute insulin-sensitizing effects of FGF21, which were explored in detail in BonDurant et al (56). Here, adipocyte-specific deletion of the FGF21 obligate coreceptor, β-klotho, prevented the acute insulin-sensitizing effect of FGF21 while also indicating that insulin signaling in the adipocytes is required for the acute insulin-sensitizing effect of FGF21. Moreover, this study does not address how regulation of adipocyte lipolysis controls hepatic FGF21 production, as recently demonstrated by Abu-Odeh and colleagues (58). Here, the authors show that mice stimulated with CL had an increased liver-derived FGF21 serum level effect that was blunted when treated with a lipolysis inhibitor (atglistatin).

In summary, hepatocyte-derived FGF21 is not required to maintain glucose tolerance and insulin tolerance in mice under normal chow conditions. Furthermore, FOXO1-dependent control of FGF21 is not required for adipose tissue insulin sensitivity and the subsequent cell-nonautonomous control of hepatic insulin sensitivity.

Glossary

Abbreviations

AAV

adeno-associated virus

BSA

bovine serum albumin

DMEM

Dulbecco’s modified Eagle’s medium

eWAT

epididymal white adipose tissue

FA

fatty acid

FGF21

fibroblast growth factor 21

FOXO1

factor forkhead box O1

GCK

glucokinase

GTT

glucose tolerance test

HFD

high fat diet

HSL

hormone sensitive lipase

ITT

insulin tolerance test

iWAT

inguinal white adipose tissue

KO

knockout

mRNA

messenger RNA

mTORC1

mammalian target of rapamycin complex 1

RT

room temperature

TBG

thyroxine-binding globulin

WAT

white adipose tissue

Financial Support

This work was supported by the National Institutes of Health (NIH; grant Nos. R01DK125497 and R01DK123252 to P.M.T., (grant No. T32DK007314 to J.S.C.), and institutional startup funds from the University of Pennsylvania, Cox Research Institute, and the Samuel and Josephine Chiaffa Memorial Fund.

Author Contributions

J.S.C. and P.M.T. were responsible for conceptualization, data analysis, and manuscript preparation. J.S.C. conducted most of the experiments. M.J.G. and D.S. assisted with technical support. P.M.T. directed the project. J.S.C. and P.M.T. wrote the manuscript with suggestions provided by all authors.

Disclosures

The authors have nothing to disclose.

Data Availability

Data sharing is not applicable to this article because no data sets were generated or analyzed during the present study.