Integrated control of hepatic lipogenesis versus glucose production requires FoxO transcription factors

Abstract

Insulin integrates hepatic glucose and lipid metabolism, directing nutrients to storage as glycogen and triglyceride. In type 2 diabetes, levels of the former are low and the latter are exaggerated, posing a pathophysiologic and therapeutic conundrum. A branching model of insulin signalling, with FoxO1 presiding over glucose production and Srebp-1c regulating lipogenesis, provides a potential explanation. Here we illustrate an alternative mechanism that integrates glucose production and lipogenesis under the unifying control of FoxO. Liver-specific ablation of three FoxOs (L-FoxO1,3,4) prevents the induction of glucose-6-phosphatase and the repression of glucokinase during fasting, thus increasing lipogenesis at the expense of glucose production. We document a similar pattern in the early phases of diet-induced insulin resistance, and propose that FoxOs are required to enable the liver to direct nutritionally derived carbons to glucose versus lipid metabolism. Our data underscore the heterogeneity of hepatic insulin resistance during progression from the metabolic syndrome to overt diabetes, and the conceptual challenge of designing therapies that curtail glucose production without promoting hepatic lipid accumulation.

Figure 1

Glucose parameters in L–FoxO1,3,4 mice. (a) Percentage of pups surviving to weaning at 21 days. n = 70 control and 24 L–FoxO1,3,4 pups. ***P < 0.001 by Fisher’s exact test. These numbers may underestimate the true mortality of L–FoxO1,3,4 pups; at genotyping (day 9) they are already present at less than Mendelian ratios. (b–d) During hyperinsulinemic–euglycemic clamp (n = 8 controls, 6 L–FoxO 1,3,4): (b) glucose infusion rate (GIR); (c) rate of glucose disposal (Rd); (d) glucose production (GP) **P < 0.01 by Student’s t test (2–tailed). (e–i) Liver gene expression and glycogen content during F–RF time course (n = 4–7, exact n for each time point and genotype listed in materials & methods): (e) G6pc expression; (f) liver glycogen; (g) Slc37a4, encoding the glucose 6–phosphate transporter; (h) Gck; (i) G6pc/Gckratio. *** P < 0.001, **P < 0.01, *P< 0.05 for control vs. L–FoxO1,3,4 mice by Student’s t test (2–tailed). Black and white bars indicate the dark/light cycle. (j) Correlation between glucose levels and the G6pc/Gck ratio in pups at P2. Data are mean ± s.e.m.

Figure 2

Glucose utilization and lipogenesis. (a) Plasma levels of [2–²H₁]–glucose after injection. (b) Plasma levels of [6,6–²H₂]–glucose after injection. (c) Fractional difference in labeled glucose enrichment. (n = 9 control, 11 L–FoxO1,3,4). ***P < 0.001, **P < 0.01, *P < 0.05 for control vs. L–FoxO1,3,4 mice, by Student’s t–tests (2–tailed). (d) De novo lipogenesis. (n = 4 control, fasted; 5 L–FoxO1,3,4, fasted; 7 control, RF; 5 L–FoxO1,3,4, RF). **P < 0.01 for control vs. L–FoxO1,3,4 mice; ^††P < 0.01 for RF vs. fasted control mice; ^§§P < 0.01 for RF vs. fasted L–FoxO1,3,4 mice by 2–way ANOVA. (e) Fasting gene expression (n = 7 control; 5 L–FoxO1,3,4). *P < 0.05. (f) Western blot of glucokinase in liver lysates from fasting mice (n = 4 control; 5–L FoxO1,3,4; representative blot shown). Numerical values above the blot are mean ±s.e.m. of Gck/actin, relative to controls. *P < 0.05. (g) Liver metabolites measured by GC–MS, relative to controls (n = 4 control; 5 L–FoxO1,3,4). *P < 0.05. (h) Liver lipidomic analyses. Total levels of the indicated lipid species. Full data set, including individual lipid species available in Supplementary Table 1. (n = 14 control; 11 L–FoxO1,3,4). **P < 0.01 by Student’s t tests (2–tailed).

Figure 3

Lipogenic gene expression during F–RF in chow–fed C57BL/6J mice and proposed physiologic model (a) Comparison of Gck, Elovl6, and Pklr expression. For each gene, the 24–hr fasting time point is set equal to 1. Elovl6 and Pklr are plotted using the vertical axis on the left, and Gck is plotted using the vertical axis on the right. n = 5 per group. Black and white bars indicate the dark/light cycle. (b) Model. Earlier data suggested parallel action of insulin through FoxO to regulate HGP, and through Srebp–1c to regulate DNL. We propose a new model, whereby: insulin acts first at low levels and early time points through FoxOs to reduce HGP and initiate postprandial DNL by reducing the G6pc/Gck ratio; and second at high levels and late time points through Srebp–1c to amplify DNL.

Figure 4

Pathophysiology and model. (a) Insulin and (b) glucose during F–RF in C57BL/6J mice fed chow or WTD for one week. n = 5 per group. (c) G6pc expression (d) Gck expression (e) G6pc/Gck ratio from chow and WTD–fed C57BL/6J mice during F–RF. Expression in chow–fed mice at time 0 is set equal to 1.n = 5 per group. Black and white bars indicate the dark/light cycle. *P < 0.05; **P < 0.01; ***P < 0.001 for chow vs. WTD–fed mice by Student’s t tests (2–tailed). (f) Model of the progression of insulin resistance and type 2 diabetes. The current model suggests bifurcation of insulin signaling, such that the FoxO branch is resistant to insulin and the Srebp–1c branch is excessively activated by hyperinsulinemia. We propose a new model, with successive defects in hepatic insulin action. In early hyperinsulinemia, FoxOs are strongly suppressed, thereby decreasing the G6pc/Gckratio and activating lipogenesis. Later in disease progression, the Srebp–1c and Chrebp pathways are activated by insulin, whereas FoxOs may be activated by other mechanisms, e.g. oxidative stress induced acetylation or glucagon–induced alternative phosphorylation.