Dietary protein governs the role of insulin signaling in the postprandial regulation of hepatic mTORC1

Abstract

The nutrient-sensing mechanistic target of rapamycin complex 1 (mTORC1) signaling pathway controls cellular and organismal growth and metabolism, and aberrant activation is linked to human disease, including metabolic disease. Cellular studies have established regulatory mechanisms influencing mTORC1 activation, but the physiological signals that control mTORC1 at the organismal and tissue levels are less well-defined. mTORC1 is dynamically regulated by fasting and feeding in metabolic tissues, with both nutrients and insulin proposed to activate mTORC1 in response to feeding. Here, studies employing a liver-specific genetic mouse model demonstrate that AKT-mediated TSC2 phosphorylation is the predominant mechanism of hepatic mTORC1 induction by insulin but is dispensable for activation by feeding. Furthermore, postprandial activation of hepatic mTORC1 requires dietary protein, which dictates the insulin-responsiveness of the pathway. Contrary to dogma, hepatic mTORC1 signaling was not elevated in response to diet-induced obesity, despite overt impairments in insulin and glucose homeostasis, and blocking hepatic AKT-TSC-mTORC1 signaling did not affect these metabolic phenotypes. Evidence is also provided supporting a role for glucagon in hepatic mTORC1 suppression during fasting. This study reveals a hierarchy of physiological signals regulating hepatic mTORC1.

The nutrient-sensing kinase complex mTORC1 regulates metabolism, yet the physiological signals controlling its activity in the liver are incompletely defined. Here, the authors show that dietary protein is required for postprandial hepatic mTORC1 activation and determines its responsiveness to insulin.

Introduction

The liver is a central hub for macronutrient metabolism that coordinates physiological responses to dynamic nutrient and energy states, such as across physiological fasting and feeding cycles^1,2. To coordinate these metabolic responses, the liver must be able to both monitor and adapt to acute changes in nutrient availability. A key function of the liver is the regulation of glucose and lipid homeostasis in response to hormonal cues, such as those from insulin^3,4. Importantly, hepatic insulin resistance is implicated in metabolic diseases, including obesity and type 2 diabetes⁵. However, the relevant targets of insulin signaling in the liver and the associated contribution to hepatic or systemic metabolism remain incompletely understood.

At the cellular level, mechanistic target of rapamycin complex 1 (mTORC1) signaling controls the balance between anabolic and catabolic metabolism^6,7. mTORC1 activation promotes anabolic processes and biosynthetic pathways in response to a variety of nutrients and signals propagated by growth factors and hormones, tailoring metabolism to local and systemic nutrient status^6–8. In vitro studies have established insulin, as well as glucose and amino acids, as robust stimuli of mTORC1 activation⁹. In the liver, mTORC1 activation is sensitive to fluctuations in nutrients, with acute activation upon feeding and suppression during fasting^10,11, and aberrant hepatic mTORC1 activation has been linked to metabolic diseases^4,12,13. However, it is unclear how specific upstream inputs cooperate to regulate hepatic mTORC1 in physiological or pathological settings. Therefore, dissecting the mechanisms of hepatic mTORC1 regulation and its consequences for systemic metabolism is central to understanding metabolic control in health and disease.

Insulin activates mTORC1 via PI3K-AKT signaling, and several mechanisms of AKT-dependent mTORC1 activation have been identified in cell culture models^14–19. The most well-characterized mechanism involves AKT-mediated inhibitory phosphorylation of the TSC2 component of the TSC complex, a key regulatory hub that inhibits mTORC1^20–22. TSC2 is a GTPase-activating protein (GAP) for the small GTPase Rheb, which facilitates maximal activation of the mTOR protein kinase by engaging mTORC1 at the lysosomal surface specifically when GTP-bound^23–27. Therefore, the TSC complex inhibits mTORC1 activation by promoting the conversion of GTP- to GDP-bound Rheb. In response to insulin, AKT phosphorylates TSC2 at five distinct residues, corresponding to S939, S981, S1130, S1132, and T1462 of the full-length human TSC2 protein, and this regulation is required for full activation of mTORC1 in cell culture models^{20–22,26,28}. AKT-mediated phosphorylation of TSC2 inhibits the TSC complex by causing its dissociation from the lysosome, where a subpopulation of Rheb resides, thereby allowing Rheb-GTP to locally accumulate and activate mTORC1²⁶. The five AKT-mediated phosphorylation sites are conserved among vertebrate TSC2 orthologs, suggesting their physiological significance in complex multicellular organisms, but this has not been directly tested in mammalian models²².

The TSC complex is essential for the dynamic regulation of mTORC1 in the liver, as mice with liver-specific ablation of TSC complex components fail to suppress mTORC1 activation during fasting^10,11,29,30. However, multiple signaling pathways converge on the TSC complex to either promote or suppress mTORC1 activity²², thus it is not entirely clear which signal is disrupted in such models. Like genetic loss of TSC complex components, other liver-specific mouse models used to study mTORC1 regulation in vivo, including genetic ablation of AKT or of mTORC1 components^{10,11,29,31–38}, disrupt the entire, highly branched signaling network and may lead to compensatory rewiring not reflective of physiological signaling^6,8. Thus, more precise models are needed to understand the physiological regulation of mTORC1 and its implications for health and disease.

In addition to regulation by growth factors through the TSC complex and Rheb, mTORC1 activation is also sensitive to cellular amino acid levels through a pathway regulating a separate set of small G-proteins, the Rag GTPases, that control mTORC1 lysosomal localization^39,40. In nutrient-replete conditions, Rag heterodimers consisting of RagA- or RagB-GTP (RagA/B-GTP) complexed with RagC- or RagD-GDP (RagC/D-GDP) recruit mTORC1 to the lysosome, where it can access Rheb-GTP^39,41. Insight from cell culture studies supports a model whereby maximal mTORC1 activation requires the presence of both extracellular growth signals and sufficient intracellular nutrients to convert the Rheb and Rag GTPases, respectively, to the correct guanylate binding state to engage mTORC1 at the lysosomal surface⁴⁰. Importantly, like liver-specific loss of the TSC complex, mice that express a GTP-bound mutant of RagA exhibit aberrant hepatic mTORC1 signaling even during fasting⁴², demonstrating that signals through both the TSC-Rheb and Rag pathways are critical for the proper dynamic regulation of mTORC1 with fasting and feeding.

Given its role in metabolic regulation, hepatic mTORC1 signaling has been proposed to contribute to impaired glucose and lipid homeostasis in obesity and diabetes^8,9,12,13,43. Some studies have concluded that hepatic mTORC1 is chronically elevated in obese mice, and mouse models characterized by repression of hepatic mTORC1 signaling exhibit protection from obesity-associated metabolic impairments^{4,12,13,38,44}. However, the mechanisms driving hepatic mTORC1 activation in obesity and the causal link of hepatic mTORC1 to metabolic dysfunction have not been clearly defined. It has been widely proposed that chronically active mTORC1 can contribute to insulin resistance through negative feedback regulation of insulin signaling to AKT, with several distinct mechanisms described^13,45–50. Furthermore, studies have shown that liver AKT activation in response to feeding or insulin is severely blunted in mice with chronic activation of hepatic mTORC1 and enhanced in mice lacking hepatic mTORC1 or its direct downstream substrate S6K1^{4,10,12,29,51}. Importantly, enhanced liver AKT activation in obese S6k1−/−¹² or liver-specific S6K1 knockdown mice⁴ is accompanied by improved systemic insulin tolerance and insulin-mediated suppression of hepatic glucose production^4,12. Thus, defining the mechanisms underlying hepatic mTORC1 regulation by pathological signals, such as hyperinsulinemia and overnutrition, will be necessary to understand its putative role in metabolic diseases.

Here, we describe a genetic mouse model that disconnects hepatic insulin signaling from mTORC1 activation. We utilize mice that conditionally express wild-type human TSC2 (TSC2-WT) or a TSC2 phospho-mutant that cannot be phosphorylated by AKT (TSC2-5A) in place of endogenous TSC2. Hepatic mTORC1 signaling is unresponsive to insulin but activated normally by feeding in liver-specific TSC2-5A (L-TSC2-5A) mice. We find that dietary protein is the dominant factor driving hepatic mTORC1 activation by feeding and that dietary protein content influences the role of insulin-AKT-TSC signaling in mTORC1 activation—a hierarchy that may be unique to the liver. In contrast to previous literature^12,13, high fat diet-induced obesity, hyperinsulinemia, and diabetes were not accompanied by elevated mTORC1 signaling in liver tissue. Furthermore, L-TSC2-5A mice were similar to L-TSC2-WT mice in the development of obesity-associated impairments in glucose homeostasis, despite mTORC1 signaling remaining unresponsive to insulin in this model. We also investigate glucagon as a negative regulator of hepatic mTORC1 and provide genetic evidence that glucagon signaling contributes to mTORC1 suppression during fasting. These data provide key insights into dietary influences on hepatic mTORC1 signaling and highlight the importance of understanding distinct mechanisms of nutrient-sensing across metabolic tissues.

Results

Generation of a mouse model decoupling AKT-TSC signaling from hepatic mTORC1 activation

The TSC complex is required for the dynamic regulation of hepatic mTORC1 in response to fasting and feeding^10,11,29. As expected, refeeding mice after an overnight fast induced phosphorylation of AKT, as well as the mTORC1 targets S6K1 and 4EBP1 (indicated by mobility shift), in the liver (Fig. 1a). Importantly, feeding also induced phosphorylation of TSC2 at T1462, one of the established sites regulated by AKT (Fig. 1a)^20,21, indicating that AKT-mediated TSC2 phosphorylation coincides with mTORC1 activation in this physiological setting. To determine the causal role of this phosphorylation in the control of

Fig. 1. Generation of a mouse model decoupling AKT-TSC signaling from hepatic mTORC1 activation.

a Eight-week-old male C57BL/6J mice were fasted overnight for 16 hours and, where indicated, refed for one hour prior to immunoblot analyses of whole liver extracts; n = 4 per group. b Schematic depicting activation of mTORC1 by insulin-AKT signaling, highlighting the TSC2-5A phospho-mutant characterized in this study. c Schematic of the conditional L-TSC2-WT and -5A genetic mouse models. The human TSC2-WT or -5A transgene expressed from the Rosa26 (R26) locus replaces endogenous mouse TSC2 expression following Cre-mediated recombination. Liver-specific TSC2-WT or -5A expression was achieved through Cre-driven expression from the Albumin (Alb) promoter. LSL: loxP-STOP-loxP. d Recombination at the R26-LSL-TSC2 and Tsc2^fl/fl gene loci was validated by RT-qPCR analysis of human versus mouse Tsc2 expression in liver tissue in mice with or without expression of Albumin-Cre. e Immunoblots of livers from 12-week-old L-TSC2-WT or -5A, or age-matched R26-LSL-TSC2^WT/5A; TSC2^fl/fl Cre-negative, ad libitum-fed male mice; n = 2 per group. f Immunoblots of liver lysates from 12-week-old L-TSC2-WT or -5A male mice fed ad libitum with accompanying bar graphs of the mean ± SD phospho-to-total protein signal quantifications normalized to the L-TSC2-WT group; n = 6 per group. g Body weights of male mice graphed as mean ± SD; n = 8 per genotype. Fat (h) and lean (i) mass of L-TSC2-WT and -5A male mice measured by echo-MRI at indicated ages graphed as mean ± SD; n = 8 WT, 10 5A (8 and 12 wks) or 13 5A (6 mos). j Glucose tolerance test was performed on 6-month-old male mice. Blood glucose measurements over time are plotted as mean ± SD; n = 8 WT, 9 5A. k Insulin tolerance test was performed on 6-month-old male mice. Blood glucose measurements over time are plotted as mean ± SD; n = 7 WT, 10 5A. l Plasma insulin following a 6-hour daytime fast in 6-month-old male mice graphed as mean ± SD; n = 8 WT, 9 5A. Statistical analyses: f, h, i and l Welch’s t-test (two-tailed); g, j and k Two-way ANOVA with Šidák correction. ns p ≥ 0.05. Source data are provided as a Source Data file.

mTORC1 signaling in mammalian tissues, we have generated a knock-in mouse model expressing wild-type TSC2 (TSC2-WT) or TSC2 with alanine mutations at the five AKT-regulated phosphorylation sites (TSC2-5A)⁵² (Fig. 1b). The transgenes are expressed conditionally from the Rosa26 (R26) locus through the use of a transcriptional and translational stop cassette flanked by LoxP sites (LSL)⁵² and encode Flag-tagged TSC2-WT or -5A human cDNAs^52,53(Fig. 1c). Thus, expression from the ubiquitous R26 promoter is controlled by Cre recombinase, allowing transgene expression with concomitant deletion of endogenous Tsc2 when bred to Tsc2^fl/fl mice (Fig. 1c). R26-LSL-Tsc2^WT/WT; Tsc2^fl/fl and R26-LSL-Tsc2^5A/5A; Tsc2^fl/fl mice were crossed to mice expressing Cre recombinase from the hepatocyte-specific Albumin promoter (Alb-Cre^+/−) to generate mice with liver-specific expression of the TSC2-WT or -5A transgenes coupled to liver-specific deletion of endogenous Tsc2 (referred to throughout as L-TSC2-WT or L-TSC2-5A) (Fig. 1c). Cre-dependent recombination at both the R26-LSL and endogenous Tsc2 loci in liver tissue was confirmed by RT-qPCR analysis of mouse versus human Tsc2 expression (Fig. 1d).

We next compared liver TSC2 protein expression between the transgenic L-TSC2-WT and -5A mice and R26-LSL-Tsc2^WT/5A; Tsc2^fl/fl Cre-negative mice. Free pools of TSC2 and TBC1D7 exist, whereas TSC1 appears to only be present in the TSC complex^26,54. While TSC2 expression was modestly higher in the transgenic mice, levels of TSC1 were comparable, indicating similar levels of TSC complex across genotypes (Fig. 1e). Furthermore, levels of AKT and mTORC1 pathway activation were unchanged in the livers from mice fed ad libitum (Fig. 1e). We also assessed TSC complex formation through size-exclusion chromatography, as performed previously^26,55, on liver tissue. This experiment demonstrated that the amount of TSC2 that overlaps with the peak TSC1-containing fractions is similar (40–45%) in mice expressing the TSC2-WT and -5A transgene (Fig. S1a). These data are in line with observations from whole-body (CMV-Cre) TSC2-WT and -5A embryos⁵².

An analysis of signaling in liver tissue from L-TSC2-WT and -5A mice under ad libitum-fed conditions confirmed complete loss of phosphorylation of TSC2 by AKT at T1462 in the L-TSC2-5A livers. However, there were no significant differences in the activating phosphorylation of AKT or in markers of mTORC1 signaling, including phosphorylation of S6K1, S6, and 4EBP1 between L-TSC2-WT and -5A livers (Fig. 1f). Liver tissue weight, which is changed in mice with liver-specific disruption of the TSC complex or mTORC1¹¹, was similar between L-TSC2-WT and -5A mice (Figure S1b (males), S1c (females)). Unlike the whole-body TSC2-WT and -5A mice⁵², there was no difference between the L-TSC2-WT and -5A mice in body weight and body composition measured at regular intervals from 6 weeks to 6 months of age, for males (Fig. 2g–i) or females (Fig. S1d–f). Furthermore, there was no difference in glucose or insulin tolerance or plasma insulin levels between age-matched L-TSC2-WT and L-TSC2-5A mice in either males (Fig. 1j–l) or females (Fig. S1g–i) at 6 months of age. These results indicate that AKT-mediated TSC2 phosphorylation is not required for basal mTORC1 activation in the liver or for the maintenance of glucose homeostasis on a normal chow diet.

Fig. 2. Hepatic mTORC1 activation by insulin and glucose, but not feeding, requires AKT-mediated TSC2 phosphorylation.

a Immunoblots of liver lysates from 11-week-old L-TSC2-WT or -5A male mice fasted for 14 hours overnight and injected intraperitoneally with vehicle (saline) or insulin (0.5 U/kg) for 20 minutes with accompanying bar graphs of the mean ± SD phospho-to-total protein signal quantifications normalized to the L-TSC2-WT vehicle-treated group; n = 5 per group. b Immunoblots of liver lysates from 11-week-old L-TSC2-WT or -5A male mice fasted for 14 hours overnight prior to administration of glucose (3 g/kg) or vehicle (saline) by oral gavage for 25 minutes with accompanying bar graphs of the mean ± SD phospho-to-total protein signal quantifications normalized to the L-TSC2-WT vehicle-treated group; n = 4 per group. c Immunoblots of liver lysates from 10-week-old L-TSC2-WT or -5A male mice fasted for 14 hours overnight and, where indicated, refed for one hour with accompanying bar graphs of the mean ± SD phospho-to-total protein signal quantifications normalized to the L-TSC2-WT fasted group; n = 4 WT per group, 4 5A (fasted), 5 5A (refed). d Representative immunohistochemistry images of L-TSC2-WT and -5A liver sections from (b) and (c) stained with an antibody to p-S6 (S235/236). Statistical analyses: a–c two-way ANOVA with Šidák correction. Source data are provided as a Source Data file.

Hepatic mTORC1 activation by insulin and glucose, but not feeding, requires AKT-mediated TSC2 phosphorylation

AKT-mediated TSC2 phosphorylation is required for mTORC1 induction by insulin in cells²⁶. To evaluate the role of this regulation in the intact liver, fasted L-TSC2-WT and L-TSC2-5A mice were treated with insulin or vehicle for 20 minutes, and AKT and mTORC1 signaling were assessed in liver tissue lysates. AKT was robustly activated by insulin in both L-TSC2-WT and -5A livers (Fig. 2a). This was also reflected in phosphorylation of TSC2 at T1462 in L-TSC2-WT, but not L-TSC2-5A, livers. Importantly, mTORC1 stimulation by insulin was impaired in L-TSC2-5A compared to L-TSC2-WT livers (Fig. 2a). As a control, signaling was also analyzed in gastrocnemius muscle from the same mice, where mTORC1 was stimulated by insulin to a similar extent in L-TSC2-WT and -5A mice (Fig. S2a), confirming that the impaired mTORC1 induction in L-TSC2-5A mice is a liver-intrinsic effect. To determine whether L-TSC2-5A mice have deficient hepatic mTORC1 induction in response to endogenous insulin secretion, fasted L-TSC2-WT and L-TSC2-5A male mice were administered glucose or vehicle via oral gavage and signaling was assessed in liver tissue collected after 25 minutes. Oral glucose administration resulted in significant elevation in blood glucose and insulin compared to vehicle treatment, with similar levels in both L-TSC2-WT and -5A mice (Figure S2b, c). AKT was activated by glucose in both L-TSC2-WT and -5A livers, but the induction of mTORC1 signaling was blunted in L-TSC2-5A compared to L-TSC2-WT livers (Fig. 2b). It is worth noting that, despite this decrease in mTORC1 signaling in the L-TSC2-5A livers, AKT activation was enhanced in response to glucose, perhaps reflecting partial relief of mTORC1-dependent feedback inhibition of insulin signaling in the L-TSC2-5A livers. These results confirm that TSC2 phosphorylation is necessary for proper hepatic mTORC1 induction by both exogenous and endogenous insulin and suggest that insulin signaling through AKT and TSC2 is the primary mechanism by which glucose acutely stimulates hepatic mTORC1.

The above experiments establish the L-TSC2-5A mouse model as a unique genetic setting in which to evaluate the contribution of insulin signaling to hepatic mTORC1 activation by a variety of physiological signals. To test its role in response to feeding, L-TSC2-WT and -5A mice were fasted or fasted and refed for one hour. Plasma insulin was significantly elevated in refed compared to fasted mice, with no differences between L-TSC2-WT and -5A mice (Fig. S2d). Somewhat surprisingly, mTORC1 signaling was induced by refeeding to a similar extent in both L-TSC2-WT and L-TSC2-5A livers (Fig. 2c). Since hepatic mTORC1 is disconnected from insulin regulation in L-TSC2-5A mice, these results indicate that insulin signaling is not essential for mTORC1 activation by feeding in the liver. The differential response of the L-TSC2-5A livers to glucose and refeeding was also evident by immunohistochemical staining of p-S6 in liver sections (Fig. 2d).

Transcription Factor EB (TFEB) is directly phosphorylated by mTORC1 in a manner that is sensitive to nutrients but is independent of growth-factor-mediated regulation of the TSC complex and Rheb^56–60. As expected, TFEB was phosphorylated in response to refeeding, but not insulin injection, similarly in L-TSC2-WT and -5A livers (Fig. S2e, f). These data also provide evidence that nutrient-sensitive mechanisms dictating mTORC1 recruitment to the lysosome, where it phosphorylates TFEB^56,57,60, are intact in these mouse models.

Given that TSC2 is also phosphorylated by AMPK, which is sensitive to energy and glucose levels, we evaluated whether differences in AMPK activation could contribute to hepatic mTORC1 regulation in L-TSC2-WT and -5A mice. Phosphorylation of AMPK and its downstream targets acetyl CoA carboxylase (ACC) and Raptor was unchanged in liver lysates in response to glucose or refeeding and was not different between L-TSC2-WT and -5A mice in any condition (Fig. S2g, h). These data are also consistent with previous findings that fasting is not sufficient to activate AMPK in the liver³⁰.

TSC2 phosphorylation by AKT releases the TSC complex into the cytosol in cultured cells, while the TSC2-5A mutant is resistant to this spatial regulation downstream of insulin^26,51. To test whether TSC complex release is similarly impaired in L-TSC2-5A liver tissue, TSC1 and TSC2 abundance in the soluble pool following permeabilization with digitonin of L-TSC2-WT and -5A liver tissue was examined. Both insulin treatment and refeeding induced cytosolic accumulation of TSC1 and TSC2 in L-TSC2-WT, but not -5A livers (Fig. S2i, j), suggesting that the inability of AKT to phosphorylate the TSC2-5A protein prevents the dynamic localization of the TSC complex in liver tissue. These experiments also suggest that phosphorylation-induced cytosolic localization of the TSC complex in response to refeeding does not underlie the induction of hepatic mTORC1 activation by feeding in L-TSC2-5A mice.

The above results are consistent with fasting and refeeding studies in whole-body TSC2-WT and -5A mice, in which mTORC1 activation by feeding was retained in TSC2-5A livers, despite a clear defect in feeding-induced activation of mTORC1 in skeletal muscle⁵². Together, these data support a model whereby nutrients or nutrient-derived signals independent of insulin are sufficient for hepatic mTORC1 activation in response to feeding a normal chow diet.

Amino acids are necessary and sufficient for near-complete hepatic mTORC1 activation

Given that mTORC1 activation in cell culture models is dependent on amino acids^8,40,61, we hypothesized that dietary protein content might override the insulin response in the activation of hepatic mTORC1 signaling by feeding. Indeed, quantitative metabolomics revealed that plasma concentrations of nearly all amino acids were elevated upon two hours of feeding relative to fasted mice, with concentrations of alanine, glutamine, and valine being significantly increased with feeding in this analysis (Fig. 3a (histidine not detected)). To test whether dietary protein is required for feeding to induce hepatic mTORC1 activation in L-TSC2-5A mice, mice were fasted and refed with a protein-free (PF) diet or an isocaloric semi-purified control diet (CTL) (see Fig. 3b for diet compositions) for one hour. Interestingly, mTORC1 activation was significantly blunted by refeeding the PF diet in both L-TSC2-WT and -5A livers (Fig. 3c), despite significant increases in blood glucose (Fig. S3a) and plasma insulin (Fig. S3b) in both genotypes. Plasma insulin was lower in response to the PF diet relative to the control diet, which likely reflects the established contribution of amino acids to pancreatic insulin secretion^62–64. However, insulin levels were still significantly induced over fasted (~five-fold) by this diet, and upstream activation of AKT was enhanced relative to the CTL diet. These data demonstrate that dietary protein is essential for hepatic mTORC1 activation by feeding. Furthermore, since insulin or AKT failed to compensate in its absence, dietary protein is the dominant signal contributing to the activation of hepatic mTORC1 by feeding. Relative to L-TSC2-WT livers, AKT activation was enhanced in the L-TSC2-5A livers under all conditions. mTORC1 signaling was significantly reduced in the CTL diet-induced L-TSC2-5A livers (Fig. 3c, see graphs for quantification), a finding distinct from the response to chow diet-refeeding (Fig. 2c) and likely reflecting the difference in macronutrient content between these two diets (Fig. 3b; see below).

Fig. 3. Amino acids are necessary and sufficient for near-complete hepatic mTORC1 activation.

a Plasma amino acid concentrations in blood collected from male TSC2-WT (CMV-Cre) mice prior to and following 2-hour refeeding graphed as mean ± SD; n = 3 fasted, 4 refed. b Macronutrient composition of diets used in this study, with caloric density of diet (kcal/g) depicted at the top of each bar. c Immunoblots of liver lysates from 9-week-old male mice fasted for 14 hours overnight and refed for one hour with control (CTL) or protein-free (PF) diets. (right) Phospho-to-total protein signal normalized to the L-TSC2-WT fasted group plotted as mean ± SD; n = 5 per group. Notations directly above the bars indicate whether the signal was induced from its respective fasted control. L-TSC2-WT and -5A lysates were analyzed on separate gels. Differences in gel loading were accounted for by normalizing to total protein signals. d Immunoblot analysis of lysates from primary mouse hepatocytes cultured in standard or low amino acid (low AA) DMEM and stimulated with insulin (10 nM). Phospho-to-total S6K signals (low AA condition) are depicted as mean ± SD fold induction from the vehicle-treated L-TSC2-WT (calculated separately for each replicate). Notations directly above the bars indicate whether the signal was induced from its respective vehicle control; n = 3 biological replicates. e Immunoblot analysis of lysates from primary mouse hepatocytes stimulated with insulin (10 nM) for 15 minutes in DMEM or human plasma-like media (HPLM); n = 1 biological replicate. f Immunoblot analysis of lysates from primary mouse hepatocytes stimulated with amino acids and/or insulin (10 nM) for 15 minutes, with accompanying graphs of the mean ± SD phospho-to-total S6K signal; n = 5 WT, 4 5A (biological replicates). Statistical analyses: a Mixed-effects analysis with Šidák correction (ns P > 0.9999); c, d, f Two-way ANOVA with Šidák correction (c, p-AKT/AKT: ns P = 0.3285, *P = 00301, *P = 0.0387, **P = 0.0052; c p-S6K/S6K: ****P = 3.2 × 10⁻⁹, ns P = 0.8523, *** P = 3.8 × 10⁻⁴, ns P = 0.8535; d: ns P > 0.9999, ns P = 0.8167, ns P = 0.9986, ns P > 0.9999, ns P = 0.6351, ns P = 0.1967, ns P = 0.4814; f: ns P > 0.9999). Source data are provided as a Source Data file.

To further determine the interaction between the insulin response and amino acid concentrations in the regulation of hepatic mTORC1 signaling, primary L-TSC2-WT and -5A hepatocytes were treated with a time course of insulin in the presence of high or low amino acids, while keeping all other media components constant. Insulin acutely stimulated AKT activation and TSC2 phosphorylation similarly in high and low amino acid media (Fig. 3d). However, basal mTORC1 signaling was greatly elevated in the high amino acid media, with only modest further induction by insulin (Figs. 3d and S3c). Along with lower basal mTORC1 activity, mTORC1 signaling was more overtly stimulated by insulin under low amino acids conditions in L-TSC2-WT hepatocytes, and this induction was almost entirely abolished in L-TSC2-5A hepatocytes (Fig. 3d). As an orthogonal approach to using more physiological levels of amino acids, the response to insulin was compared in L-TSC2-WT and -5A hepatocytes cultured in DMEM versus human plasma-like medium (HPLM)⁶⁵, which has amino acid concentrations similar to that in fasted mouse plasma (Fig. 3a). As seen in the low amino acid media, insulin-stimulated mTORC1 signaling was impaired in L-TSC2-5A hepatocytes in HPLM (Fig. 3e). These results indicate that amino acid levels dictate basal hepatic mTORC1 activation and its dynamic sensitivity to insulin stimulation, with insulin signaling to mTORC1 being dependent on AKT-mediated phosphorylation of TSC2.

Outside of transformed cells, such as HEK-293Ts, that have constitutive PI3K-AKT signaling, full activation of mTORC1 is thought to require signals from both nutrients and exogenous growth factors or hormones⁶. However, the results above suggest that dietary amino acids are dominant over insulin signaling for full activation of mTORC1 in the liver. To test this hierarchy further, L-TSC2-WT and -5A primary mouse hepatocytes were depleted of amino acids for one hour, followed by stimulation with amino acids and insulin alone or in combination. Amino acids were sufficient to robustly activate mTORC1 in both L-TSC2-WT and -5A hepatocytes, while insulin alone did not stimulate mTORC1 in either genotype (Fig. 3f). As expected, neither AKT activation nor TSC2 phosphorylation were induced by amino acids alone but both were stimulated by insulin. The addition of insulin with amino acids modestly increased mTORC1 induction compared to amino acids alone, but this effect was not significant. These results indicate that amino acids are both necessary and sufficient for near-maximal mTORC1 activation in hepatocytes, providing an insulin-independent mechanism capable of activating hepatic mTORC1 in L-TSC2-5A mice upon refeeding a normal chow diet.

Dietary carbohydrate-to-protein ratio governs the contribution of insulin-AKT-TSC2 signaling to hepatic mTORC1 activation

The results above provide evidence that amino acid concentrations dictate the role of insulin signaling in the activation of hepatic mTORC1. Given the macronutrient differences between the CTL and chow diets (Fig. 3b), we sought to determine whether the dietary carbohydrate-to-protein ratio dictates the dominant route to mTORC1 activation in liver tissue by refeeding a high carbohydrate, low protein (HC-LP) diet. The formulated HC-LP diet derives 80% calories from carbohydrate and 10% from protein (Fig. 3b), as previous studies have demonstrated that short-term feeding of a 10% protein diet does not elicit markers of protein restriction⁶⁶. Following one-hour refeeding with the HC-LP diet, hepatic mTORC1 signaling was induced by approximately seven-fold in L-TSC2-WT mice, with only modest and significantly reduced induction in L-TSC-5A livers (Fig. 4a). Importantly, insulin levels in response to acute feeding were similar between the HC-LP (Fig. 4b) and chow diets (Fig. S2d). Thus, taken together with the primary hepatocyte experiments, these results demonstrate that lower dietary protein content renders AKT-mediated phosphorylation of TSC2 the primary route to feeding-induced hepatic mTORC1 signaling.

Fig. 4. Dietary protein-to-carbohydrate ratio governs the contribution of insulin-AKT-TSC2 signaling to hepatic mTORC1 activation.

a Immunoblots of liver lysates from 12-week-old male mice fasted for 14 hours overnight and, where indicated, refed for one hour with a high carbohydrate, low protein (HC-LP) diet. Accompanying bar graphs of phospho-to-total protein signal quantifications normalized to the L-TSC2-WT fasted group, plotted as the mean ± SD, are shown to the right; n = 4 WT per group, 3 (fasted) or 4 (refed) 5A. b Plasma insulin concentration in fasted and refed mice from the refed group in a graphed as mean ± SD. Samples were collected from fasted mice prior to refeeding and 30 minutes after refeeding; n = 4 per group. c Timeline of diet study and phenotypic data reported in (d–g). d Pyruvate-lactate tolerance test was performed after a 4 weeks diet. Mice were fasted for 16 hours overnight, followed by i.p. injection with lactate (1.5 mg/kg) and pyruvate (150 μg/kg). Blood glucose measurements are plotted over time as mean ± SD; n = 11 WT, 12 5A. e Oral glucose tolerance test performed after 5 weeks of diet. Mice were fasted for 16 hours overnight, followed by oral gavage of glucose (2 g/kg). Blood glucose measurements are plotted over time as mean ± SD; n = 11 WT, 10 5A. f Insulin tolerance test was performed after 6 weeks diet. Blood glucose measurements are plotted over time as mean ± SD; n = 8 WT, 9 5A. g Immunoblots of liver lysates from L-TSC2-WT or -5A male mice fed HC-LP diet for 7 weeks, fasted for 14 hours overnight, and where indicated, refed for one hour with the HC-LP diet. Accompanying bar graphs of phospho-to-total protein signal quantifications normalized to the L-TSC2-WT fasted group, plotted as the mean ± SD, are shown to the right; n = 3 (fasted) or 5 (refed) WT, 3 (fasted) or 4 (refed) 5A. Statistical analyses: a, b, d–g Two-way ANOVA with Šidák correction (d, g: ns P > 0.9999). Source data are provided as a Source Data file.

While there are no known direct effects of mTORC1 signaling on hepatic or systemic glucose metabolism, mTORC1-mediated feedback mechanisms can theoretically influence the insulin responsiveness of a given tissue^{10,11,29,67–70}. Interestingly, accompanying the decrease in mTORC1 stimulation, the induction of AKT activation was more pronounced in the L-TSC2-5A livers with HC-LP refeeding (Fig. 4a), despite insulin levels being equivalent to those in L-TSC2-WT mice (Fig. 4b). Measurements of direct targets of AKT, including PRAS40 and GSK3α/β phosphorylation, further demonstrated pronounced AKT activation in the L-TSC2-5A livers (Fig. S4a), indicative of enhanced hepatic insulin sensitivity. While in vivo evidence of purported mTORC1-dependent feedback mechanisms influencing hepatic insulin signaling is limited, it is likely that several such mechanisms function in parallel. Indeed, S6K1-mediated phosphorylation of the mTORC2 component Rictor on T1135, which is one established feedback mechanism downstream of mTORC1 that influences AKT activation^49,50, was significantly induced by HC-LP feeding in TSC2-WT livers but not TSC2-5A livers (Fig. S4a). Therefore, mTORC1-mediated negative feedback mechanisms are reduced in L-TSC2-5A compared to L-TSC2-WT livers in response to the HC-LP diet.

To test the effects of the HC-LP diet on systemic glucose homeostasis, we assayed pyruvate, glucose and insulin tolerance in L-TSC2-WT and -5A mice during prolonged HC-LP diet feeding (Fig. 4c). As on a normal chow diet, no differences in body weight, lean mass, or fat mass were observed between L-TSC2-WT and -5A mice fed this diet for six weeks (Fig. S4b–d). Additionally, fasting plasma insulin was not significantly different between L-TSC2-WT and −5A mice (Fig. S4e). After four weeks on the HC-LP diet, a pyruvate-lactate tolerance test was performed on fasted L-TSC2-WT and -5A mice to measure gluconeogenic capacity. While there was a trend towards reduced endogenous glucose production in L-TSC2-5A compared to L-TSC2-WT mice, the values were not significantly different (Fig. 4d). In response to an oral glucose tolerance test performed after five weeks on the HC-LP diet, L-TSC2-5A mice exhibited a modest, but significant, improvement in glucose clearance (Fig. 4e). However, systemic insulin sensitivity was not significantly different between L-TSC2-WT and -5A mice after 6 weeks HC-LP diet (Fig. 4f). After 7 weeks on the HC-LP diet, L-TSC2-WT and -5A mice were fasted and refed the HC-LP diet for one hour, as in Fig. 4a. Again, feeding-induced AKT activation was significantly higher, while mTORC1 activation was diminished, in L-TSC2-5A mice (Fig. 4g). Together, these data demonstrate that AKT-mediated TSC2 phosphorylation regulates mTORC1 activation in the context of a low protein diet, which exerts moderate effects on systemic glucose homeostasis, likely through feedback regulation of the hepatic insulin response.

Hyperinsulinemia accompanying diet-induced obesity does not drive chronic upregulation of hepatic mTORC1

Next, we used a parallel approach to evaluate hepatic mTORC1 activation in response to the pathological states of obesity and hyperinsulinemia. C57BL/6J male mice were fed a high-fat diet (HFD) containing 60% kilocalories from fat or a low-fat control diet (LFD) containing 10% fat beginning at eight weeks of age. Dietary protein was constant between the HFD and LFD. Systemic glucose homeostasis was evaluated by glucose and insulin tolerance tests in cohorts fed HFD or LFD diets for 4, 8, or 16 weeks (Fig. 5a). mTORC1 signaling was measured in liver tissue collected from these same mice following a daytime fast (Fig. 5a). As expected, HFD-fed mice weighed significantly more than LFD-fed mice starting after just one week of diet (Fig. 5b). This elevated body weight was associated with increased fat mass in HFD versus LFD mice after 4 and 8 weeks of diet (Fig. S5a), as well as significantly higher lean mass at 8 weeks of HFD (Fig. S5b). Plasma insulin levels were also elevated in HFD compared to LFD mice at 4, 8, and 15 weeks, and were significantly higher when measured at 8 and 15 weeks (Fig. 5c). At 4 weeks, systemic insulin sensitivity was significantly impaired in HFD compared to LFD mice (Fig. 5d), with progressive impairments observed at longer durations of HFD feeding (Fig. 5f, h). Glucose tolerance was similarly impaired in HFD-fed mice (Figs. S5c, S5d). Strikingly, mTORC1 signaling was not significantly elevated in the liver tissue of HFD compared to LFD mice, even after 16 weeks, despite the overt obesity, hyperinsulinemia, and insulin resistance phenotypes of these mice (Fig. 5e, g, i). Accompanying the chronic hyperinsulinemia was significantly elevated hepatic AKT activation in the HFD compared to LFD mice at 16 weeks diet, while total IRS1 protein levels, an indicator of insulin sensitivity, were reduced in these livers (Figs. 5i and S5e). Collectively, these data are in opposition to the purported dogma that hepatic mTORC1 is chronically elevated in diet-induced obesity and contributes to systemic insulin resistance. However, they are in alignment with our results in L-TSC2-5A mice, demonstrating that insulin is not the predominant signal driving mTORC1 activation in the liver.

Fig. 5. Hyperinsulinemia accompanying diet-induced obesity does not drive chronic upregulation of hepatic mTORC1.

a Timeline of HFD diet study and phenotypic data reported in (b–i and S5a-F). b Body weight measurements of male C57Bl/6J mice fed HFD or LFD graphed as mean ± SD; n = 18 (0–3 weeks), 12 (4–8 weeks), 6 (9–10 weeks). c Plasma insulin from 6-hour fasted mice fed HFD or LFD diets for 4, 8, or 15 weeks graphed as mean ± SD; n = 6 per group. Insulin tolerance test performed after 3.5 (d), 7.5 (f), and 12.5 (h) weeks of diet. Blood glucose measurements over time are plotted as mean ± SD; n = 6 per group (d), 5 LFD and 6 HFD (f), 11 LFD and 10 HFD (h). e Immunoblots of liver lysates from mice in d fed HFD or LFD for 4 weeks and fasted for 6 h prior to tissue collection. Accompanying bar graphs of phospho-to-total protein signal quantifications normalized to the LFD group, plotted as the mean ± SD, are shown to the right; n = 6 per group. g Immunoblots of liver lysates from mice in f fed HFD or LFD for 8 weeks and fasted for 6 hours prior to tissue collection. Accompanying bar graphs of phospho-to-total protein signal quantifications normalized to the LFD group, plotted as the mean ± SD are shown to the right; n = 6 per group. i Immunoblots of liver lysates from mice fed HFD or LFD for 16 weeks and fasted for 12 h (daytime) prior to tissue collection. Samples from both groups were analyzed on the same gel with the same exposure, but images were cropped to remove unrelated lanes. (Right) accompanying graphs of the mean ± SD phospho-to-total protein signal normalized to the LFD group; n = 4 per group. Statistical analyses: b, d Mixed-effects analysis with Šidák’s correction (b: *P = 0.0455, ***P = 2.6 × 10⁻⁵, ****P < 0.0001); c, f, h Two-way ANOVA with Šidák correction; e Mann–Whitney test (two-sided); g, i Welch’s t-test (two-sided). Source data are provided as a Source Data file.

We previously demonstrated that, in contrast to the liver, muscle mTORC1 activation in response to feeding is largely dependent on AKT-mediated TSC2 phosphorylation⁵², suggesting that insulin may play a more dominant regulatory role for mTORC1 in muscle tissue. In agreement with this, activation of mTORC1 signaling was higher in the gastrocnemius muscle of HFD- versus LFD-fed mice after 15 weeks (Fig. S5f). These data highlight that different regulatory mechanisms contribute to mTORC1 activation across metabolic tissues. Whether the elevated mTORC1 activation in the muscle of HFD-fed mice contributes to systemic insulin resistance or impaired glucose clearance requires further study.

Obese L-TSC2-5A mice are not protected from metabolic impairment but maintain defects in insulin-stimulated hepatic mTORC1

We next evaluated whether chronic hyperinsulinemia driven by diet-induced obesity could accentuate differences between L-TSC2-WT and -5A mice, as seen with the HC-LP diet. We hypothesized that if hyperinsulinemia contributes to the pathological activation of hepatic mTORC1, it would be lower in L-TSC2-5A mice following HFD feeding and potentially accompanied by improvements in glucose homeostasis. To test this, L-TSC2-WT and -5A male and female mice were placed on a high-fat diet (HFD) at 8 weeks of age, with glucose and insulin tolerance tests performed after 8–10 weeks and again after 15–16 weeks of diet (Fig. 6a). There were no significant differences in body weight or body composition between L-TSC2-WT and -5A mice across high fat diet feeding (Fig. 6b–d (males), Fig. S6d–f (females)). Plasma insulin levels increased progressively and to a similar extent in both L-TSC2-WT and -5A mice during HFD feeding (Fig. S6a (males), Fig. S6g (females)). Systemic glucose tolerance and insulin sensitivity were not significantly different between L-TSC2-WT and -5A mice after 16 weeks of HFD (Fig. 6e and f (males), Fig. S6h and i (females)). Additionally, hepatic glycogen levels were similar between L-TSC2-WT and -5A mice (Fig. S6b). Plasma albumin, a marker of liver function, was also similar between L-TSC2-WT and -5A mice fed HFD (Fig. S6c). These data indicate that L-TSC2-5A mice are not protected from systemic metabolic impairments associated with diet-induced obesity.

Fig. 6. Obese L-TSC2-5A mice are not protected from metabolic impairment but maintain defects in insulin-stimulated hepatic mTORC1.

L-TSC2-WT and -5A male mice were fed a high-fat diet (HFD) beginning at 8 weeks of age. a Schematic of the HFD study with timeline of phenotypic data reported in (b–g and S6a-j). b Body weight measurements of L-TSC2-WT and -5A male mice fed HFD for the indicated duration graphed as mean ± SD; n = 8 WT, 12 5 A. Fat (c) and lean (d) mass of L-TSC2-WT and -5A male mice at baseline (eight weeks of age) and after 8 or 16 weeks HFD, measured by echo-MRI and graphed as mean ± SD; n = 8 WT, 12 5A. e Oral glucose tolerance test performed after 15 weeks HFD. Mice were fasted for 6 hours followed by oral gavage of glucose (1 g/kg). Blood glucose measurements over time are plotted as mean ± SD; n = 8 WT, 12 5A. f Insulin tolerance test performed after 16 weeks HFD. Mice were fasted for 6 hours followed by i.p. injection with insulin (0.5 U/kg). Blood glucose measurements over time are plotted as mean ± SD ; n = 8 WT, 11 5A. g Immunoblots of liver lysates from L-TSC2-WT or -5A mice fed HFD for 10 weeks and fasted for 14 hours overnight, then injected i.p. with vehicle (saline) or insulin (0.5 U/kg) for 20 minutes. Accompanying bar graphs of phospho-to-total protein signal quantifications normalized to the L-TSC2-WT vehicle-treated group plotted as the mean ± SD are shown to the right; n = 4 (vehicle) and 5 (insulin) WT, 4 5A per group. Statistical analyses: b–d, f, g Two-way ANOVA with Šidák’s correction (c, d: ns P > 0.05); e Mixed-effects analysis with Šidák correction. Source data are provided as a Source Data file.

To test whether fasted or insulin-stimulated hepatic mTORC1 activation was altered in L-TSC2−5A compared to -WT mice under conditions of obesity and hyperinsulinemia, we assessed liver signaling in fasted L-TSC2-WT and −5A males treated with vehicle or insulin after 10 weeks on an HFD. Fasting hepatic mTORC1 activation was not significantly different between L-TSC2-WT and -5A mice (Fig. 6g). Importantly, as detected in lean mice (Fig. 2a), mTORC1 induction by exogenous administration of insulin was impaired in diet-induced obese L-TSC2-5A compared to -WT mice, but this was not associated with significantly increased AKT activation (Fig. 6g). Given that skeletal muscle mTORC1 signaling was significantly higher in obese compared to lean mice (Fig. S5f), we measured skeletal muscle signaling effects in the L-TSC2-WT and -5A mice fed a HFD. L-TSC2-WT and -5A skeletal muscle remained insulin sensitive following 10 weeks of HFD feeding, and, as expected, there was no difference in AKT or mTORC1 activation between L-TSC2-WT and -5A muscle (Fig. S6j). Thus, any potential effects of skeletal muscle mTORC1 on systemic insulin sensitivity would affect L-TSC2-WT and -5A mice to the same extent.

These data further support that obesity-induced hyperinsulinemia is not a primary driver of hepatic mTORC1 activation and that an impairment in insulin-stimulated hepatic mTORC1 signaling does not grossly affect systemic glucose homeostasis under such conditions.

Glucagon is a suppressive signal of hepatic mTORC1

The present study has focused on the interplay between insulin and nutrients in the induction of hepatic mTORC1 upon feeding. However, whether there are specific signals that suppress hepatic mTORC1 activation during fasting, outside of a reduction in feeding-induced stimulatory signals, is unknown. To this end, the fasting hormone glucagon has been reported to inhibit mTORC1 in perfused rat liver and primary hepatocytes^71–73, and forskolin, which recapitulates one aspect of glucagon signaling by increasing cAMP levels and activating protein kinase A (PKA), also suppresses mTORC1 in non-hepatic cell lines^72,74. However, the potential contribution of glucagon to the suppression of hepatic mTORC1 during fasting has not been directly tested.

In a time course experiment, we found that glucagon acutely suppresses mTORC1 signaling in primary mouse hepatocytes, with maximal inhibition at 30 minutes and sustained inhibition for at least six hours (Fig. 7a). The induction of canonical glucagon signaling was confirmed by the PKA-mediated phosphorylation of CREB at S133. Additionally, acute glucagon treatment attenuated mTORC1 activation induced by insulin (Fig. 7b) or by siRNA-mediated knockdown of Tsc2 (Fig. 7c) in primary hepatocytes, suggesting a TSC complex-independent mechanism of suppression.

Fig. 7. Glucagon is a suppressive signal of hepatic mTORC1.

a Immunoblots of lysates from C57BL6/J primary mouse hepatocytes serum-starved overnight, then treated with glucagon (10 nM) for the indicated times. The phospho-to-total S6K signal quantifications for the vehicle and 30-minute glucagon treatment, normalized to the vehicle-treated group, are plotted as the mean ± SD; n = 4 biological replicates. b Immunoblots of lysates from C57BL/6J primary mouse hepatocytes, serum starved overnight, then pre-treated with insulin (100 nM) for 30 minutes followed by treatment with glucagon (10 nM) for 30 minutes. The experiment was performed a total of two times with similar results. c Immunoblots of lysates from C57BL/6J primary mouse hepatocytes transiently transfected with siRNA targeting Tsc2 or with a scrambled control siRNA for 72 hours followed by overnight serum deprivation and treatment with glucagon (10 nM) for 30 minutes. Three technical replicates from a single experiment are shown. Experiment was performed a total of three times with similar results. d Immunoblots of lysates from Gcgr^fl/fl or L-Ggcr-KO primary mouse hepatocytes serum-starved overnight and treated with glucagon (10 nM) or forskolin (10 μM) for 30 minutes. These results are of a single experimental replicate. e mRNA expression of Gcgr in liver tissue of Gcgr^fl/fl and L-Gcgr-KO mice was analyzed by RT-qPCR and graphed as mean ± SD. Top: males; n = 9 Gcgr^fl/fl, 12 L-Gcgr-KO. Bottom: females; n = 11 Gcgr^fl/fl, 12 L-Gcgr-KO. Immunoblots of liver lysates from male (f) and female (g) Gcgr^fl/fl and L-Gcgr-KO mice fasted for 12 hours (daytime) prior to tissue collection. Accompanying bar graphs of phospho-to-total protein signal quantifications normalized to the Gcgr^fl/fl group plotted as the mean ± SD are shown to the right. f n = 4 Gcgr^fl/fl, 6 L-Gcgr-KO. g n = 5 Gcgr^fl/fl, 6 L-Gcgr-KO. Red asterisk indicates a sample that was removed from analysis for both S6K and S6 signal quantification, as it was found to be a statistical outlier by the Grubb’s test. Statistical analyses: a Mann–Whitney test (two-sided); f, g Welch’s t-test (two-sided). Source data are provided as a Source Data file.

We next took a genetic approach to evaluate the role of glucagon signaling in the regulation of hepatic mTORC1. Gcgr^fl/fl mice were bred with mice expressing Albumin-Cre to generate mice with liver-specific knockout of the glucagon receptor (L-Gcgr-KO), as previously described⁷⁵. We first assessed whether the glucagon receptor (GCGR) is required for the suppression of mTORC1 in response to exogenous glucagon treatment using hepatocytes from L-Gcgr-KO or Gcgr^fl/fl mice. Glucagon treatment had no effect on mTORC1 signaling in Gcgr^−/− hepatocytes, while forskolin, which activates the cAMP signaling cascade downstream of GCGR, suppressed mTORC1 signaling similarly in Gcgr^fl/fl and L-Gcgr-KO hepatocytes, with corresponding effects on CREB phosphorylation (Fig. 7d). These data indicate that suppression of mTORC1 by glucagon is mediated through the hepatic GCGR.

To assess whether hepatic glucagon signaling is required for the suppression of hepatic mTORC1 during fasting in vivo, mTORC1 signaling was analyzed in liver tissue of fasted Gcgr^fl/fl and L-Gcgr-KO mice. Loss of Gcgr expression was verified in liver tissue of male and female mice by RT-qPCR (Fig. 7e). Interestingly, phosphorylation of markers of mTORC1 signaling was modestly elevated in the livers of both male and female L-Gcgr-KO mice compared to Gcgr^fl/fl mice (Fig. 7f (males), Fig. 7g (females)). Together, these data provide support that glucagon can inhibit hepatic mTORC1 signaling and may contribute to its suppression in the fasted state.

Discussion

The liver coordinates systemic metabolism in response to physiological nutrient intake. The mTORC1 signaling network controls cellular metabolism and is sensitive to local and systemic nutrient status, with low pathway activation in the liver during fasting and acute induction in response to feeding. By mutating TSC2 at the five AKT-mediated phosphosites that are required for mTORC1 activation by insulin in cells^26,28,52, we generated the L-TSC2-5A mouse model as a genetic tool to define the specific contribution of insulin-AKT-TSC signaling to the physiological regulation of mTORC1 in the liver.

Studies in mice with liver-specific TSC1 or TSC2 deletion demonstrate that the TSC complex is required for the dynamic regulation of hepatic mTORC1 by fasting and feeding, resulting in constitutively elevated mTORC1 signaling, even in the fasted state^10,11,29. However, these models do not provide insight into the relevant physiological inputs integrated to control mTORC1 activation in this setting. AKT1/2 knockout mice have demonstrated that AKT is required for hepatic mTORC1 activation in response to insulin^34,36. The findings described here provide mechanistic insights into this liver-intrinsic mechanism, demonstrating that insulin activates mTORC1 through AKT-mediated phosphorylation of the five sites on TSC2. Our data also showed that AKT-mediated TSC2 phosphorylation underlies hepatic mTORC1 activation by glucose, which is likely to occur through the glucose-stimulated increase in insulin. As the mTORC1 response to insulin and glucose is significantly decreased, but not entirely absent, from L-TSC2−5A livers, it is likely that other AKT-dependent mechanisms, such as phosphorylation of PRAS40^16,17,76 or of other sites on TSC2, play an additional, more minor, role in this induction. Furthermore, other reported mechanisms of glucose sensing by mTORC1, such as Rag- and TSC complex-mediated mechanisms^42,77–81 or through the glycolytic intermediate dihydroxyacetone phosphate (or DHAP)⁸², may also contribute in some settings.

We confirmed that feeding also robustly activates hepatic AKT, the AKT-mediated phosphorylation of TSC2, and mTORC1 signaling. However, our genetic data demonstrate that the AKT-TSC2 circuit is dispensable for the induction of hepatic mTORC1 signaling by feeding a normal chow diet. This finding, while surprising, is supported by previous studies showing that hepatic mTORC1 activation by feeding is intact in mice with liver-specific genetic deletion of the insulin receptor, which lack AKT activation^35,83. We found that dietary protein is essential for hepatic mTORC1 induction by feeding regardless of TSC2 phosphorylation status. This finding is consistent with a previous report that the induction of hepatic mTORC1 signaling upon refeeding requires dietary leucine⁸⁴. Furthermore, studies in mice expressing constitutively active RagA or lacking the leucine sensors Sestrin1 and Sestrin2 have provided genetic evidence that the amino acid sensing branch plays a critical role in hepatic mTORC1 regulation^42,84–86.

While many studies in immortalized cell lines have established that maximal mTORC1 activation requires both nutrient and growth factor signaling^6,8, our observations in the L-TSC2−5A mice suggest that insulin signaling may be dispensable for the activation of hepatic mTORC1, particularly in response to protein-rich diets. In fact, amino acids appear to be both necessary and sufficient for near full mTORC1 activation in primary mouse hepatocytes, while the impact of insulin signaling through AKT and TSC2 phosphorylation was only observed under lower amino acid concentrations or, in the intact liver, when feeding a low protein diet. Given that plasma insulin levels were similar following chow and HC-LP diet feeding, we conclude that dietary protein is the primary factor determining the impact of insulin-AKT-TSC signaling in postprandial hepatic mTORC1 activation (Fig. 8).

Fig. 8. Model of physiological regulation of hepatic mTORC1 signaling by nutritional and hormonal cues during fasting and feeding.

During feeding, dietary protein is the dominant signal required for hepatic mTORC1 activation, perhaps through cellular amino acid sensing. Hepatic mTORC1 is also activated by insulin signaling, which is mediated through AKT-mediated phosphorylation of the TSC complex. TSC complex phosphorylation by AKT relieves its suppression of mTORC1, leading to pathway activation. However, insulin-AKT-TSC signaling is a secondary mode of hepatic mTORC1 activation that is dispensable for activation in response to feeding of a balanced diet containing both protein and carbohydrates. Additionally, dietary carbohydrates activate hepatic mTORC1 primarily through the stimulation of insulin secretion and hepatic insulin signaling through AKT and the TSC complex. During fasting, in addition to a decrease in the above stimulatory signals, increased glucagon signaling suppresses mTORC1 activity through currently undefined mechanisms.

This hierarchical regulation of mTORC1 activation in the liver is in contrast to the model of signal integration established predominantly from cell culture models, where maximal mTORC1 activation appears to require both nutrients and growth factors (e.g., insulin)^6,8. However, previous work found that feeding-induced mTORC1 activation is dependent on AKT-mediated TSC2 phosphorylation⁵² in muscle tissue. Thus, our data do not overturn this established model but indicate that not all tissues will completely fit the existing model, with distinct tissues being adapted to the relevant physiological signals and unique metabolic functions of their respective organ system. For instance, the liver is particularly unusual in that it receives both insulin and nutrient signals directly from portal circulation and plays a major role in clearing both.

It is important to note that our conclusions are largely limited to acute activation of hepatic mTORC1 and acknowledge that patterns of regulation may shift across a broader time course of feeding. One potential discrepancy raised from our experiments is that oral delivery of glucose activated mTORC1 signaling in the liver, but refeeding with a protein-free diet high in carbohydrate (48% sucrose) did not. This inconsistency may be due to differences in timing (25 minutes glucose gavage versus one-hour feeding). The distinct modalities of nutrient delivery—gavage versus voluntary feeding—may also lead to differences in nutrient absorption, stress responses, or induction of gut- or nervous system-derived signals. Interestingly, POMC neuron-mediated activation of the sympathetic nervous system reportedly induces hepatic mTORC1 in response to food perception in mice⁸⁷, which may not occur upon oral gavage. Additionally, dietary sucrose, which is metabolized into glucose and fructose, may be less potent for mTORC1 induction than a bolus of pure glucose. Indeed, fructose was found in an independent study to acutely suppress hepatic mTORC1 in mice⁸⁸.

Published studies suggest that the dynamic nature of mTORC1 signaling in insulin-responsive tissues, including the liver, is critical to whole-body metabolic homeostasis^12,29,38. It has been proposed that liver mTORC1 activation becomes elevated in obesity^4,13,89 and contributes to hepatic insulin resistance, with in vitro studies implicating a variety of negative feedback mechanisms on insulin signaling^{13,45–47,49,50}. However, the present study found no differences in hepatic mTORC1 activation between mice fed a high-fat versus a low-fat diet, despite overt impairments in glucose homeostasis and hyperinsulinemia. Interestingly, we did find mTORC1 activation to be elevated in skeletal muscle tissue of diet-induced obese mice. One potential explanation for the discrepancy is that some previous studies utilized a chow diet to control for HFD^12,13. Chow diets contain minimal sucrose and various whole food ingredients that semi-purified diets do not, confounding the comparison between obese and lean mice in these studies. Interestingly, obesity in humans is associated with elevated levels of branched-chain amino acids (BCAAs)⁹⁰, which have been proposed to contribute to insulin resistance, at least in part, through liver and/or muscle mTORC1 activation^90–92. Given our findings that dietary protein and amino acids are dominant for the activation of hepatic mTORC1, it is possible that a BCAA–mTORC1 axis drives metabolic impairments in obesity without the necessity of broad, persistent upregulation. Another possibility is that zonal regulation of hepatic mTORC1, which is not reflected in immunoblots of whole liver lysates, is impaired under conditions of obesity^84,93,94. Our findings contradict the purported dogma that hepatic mTORC1 is chronically activated in obesity, emphasizing the need for further investigation into the causal role of mTORC1 in tissue insulin resistance and systemic metabolic dysfunction.

In addition to investigating mechanisms of hepatic mTORC1 activation, we also evaluated glucagon, which inhibits mTORC1 in hepatocytes and perfused liver tissue^71–73, as a potential mediator of hepatic mTORC1 suppression during fasting. We provide genetic evidence in support of hepatic mTORC1 regulation by glucagon in vivo. The mechanism of mTORC1 regulation by glucagon remains unclear but is dependent on the canonical glucagon receptor and appears to rely on cAMP signaling, as it is recapitulated by forskolin, and seems at least partly independent of the TSC complex^72,74. Glucagon promotes amino acid catabolism in hepatocytes and liver tissue⁹⁵, which may contribute to its inhibition of mTORC1. However, it is likely that several mechanisms contribute to mTORC1 suppression during fasting, including a decrease in stimulatory signals from circulating amino acids and insulin. The metabolic response to fasting, including upregulation of glycogenolysis, gluconeogenesis, fatty acid oxidation, and ketogenesis, is impaired in L-Gcgr-KO mice^96–99. Given that several of these processes are opposed by mTORC1^10,11,85, it is possible that aberrant mTORC1 signaling contributes to the impaired fasting response in L-Gcgr-KO mice, but this remains to be directly tested. Importantly, the regulation of glucagon secretion is defective in obese mice¹⁰⁰, leading to diminished dynamic shifts in plasma ratios of glucagon to insulin, which could also contribute to aberrant mTORC1 regulation in metabolic disease.

This study demonstrates that macronutrient composition influences hepatic mTORC1 activation, independent of insulin induction, which could have important implications for energy balance and metabolic health. Important follow-up studies include evaluating how dietary insights gleaned from this study can be translated into insights for human health. Interestingly, restriction of dietary protein or select amino acids is sufficient to promote longevity and metabolic health^101–104, and lifespan extension is maximal in mice fed a low protein, high carbohydrate diet¹⁰⁵. There is some evidence that these benefits of low protein diets are mediated by reduced mTORC1 signaling, including in the liver^67,106. Additionally, not only total dietary protein content, but protein source has also been found to influence hepatic mTORC1 activation and metabolic dysfunction in rats¹⁰⁷. Importantly, low protein consumption is linked to health benefits in humans, including reduced cancer risk and all-cause mortality¹⁰⁸. Understanding how physiological mTORC1 is regulated and functions within each tissue in response to nutrient and hormonal cues will clarify its role in mediating the metabolic and anti-aging effects of dietary interventions.

Methods

Animals

All animal studies were reviewed and approved by the Harvard Medical Area Standing Committee on Animal Care and Use (IACUC), an AAALAC International-accredited facility. Mice were group-housed (3–4 mice per cage) at the Harvard T.H. Chan School of Public Health in standard static microisolator top cages and maintained in a temperature- (22 °C) and humidity- (40–60%) controlled, pathogen-free facility with 12-h light:dark cycles (beginning at 7:00 AM). Mice were provided ad libitum access to autoclaved food and water. Trained animal care personnel performed cage changes and routine husbandry according to institutional standard operating procedures. Researchers, animal care technicians, and veterinary staff routinely monitored animals for general health and well-being throughout the study. All euthanasia procedures were performed by trained personnel in accordance with the approved animal-use protocol. For routine colony maintenance and tissue collection, mice were euthanized by CO₂ inhalation followed by cervical dislocation. For primary hepatocyte isolation experiments, animals were anesthetized by injection of a non-fatal dose of ketamine/xylazine (20 mg/ml/2 mg/ml) prior to tissue perfusion, with euthanasia by exsanguination and removal of an essential organ (liver).

C57BL/6 J mice were purchased from The Jackson Laboratory (Strain #000664). Rosa26-LSL-TSC2-WT; Tsc2^fl/fl and Rosa26-LSL-TSC2-5A; Tsc2^fl/fl mice were generated as previously described^52,109. For liver-specific transgene expression, these lines were bred to commercially available mice expressing Albumin-Cre (Jackson Laboratory, Strain #003574)¹¹⁰. To prevent genetic drift, L-TSC2-WT and −5A colonies were maintained via heterozygous crosses of Rosa26^{LSL-TSC2-WT/LSL-TSC2-5A}; Tsc2^fl/fl; Alb-Cre^+/− (L-TSC2-WT/5 A) mice. To increase yield for experimental cohorts, homozygous L-TSC2-WT and -5A offspring generated from such heterozygous crosses were used as a parental generation to set up homozygous crosses. Age-matched littermates from these homozygous crosses were used for all study groups, ensuring that L-TSC2-WT and -5A lines were not bred separately for more than one generation. The Gcgr^fl/fl;Alb-Cre (L-GCGR-KO) and Gcgr^fl/fl mice were provided by Dr. Daniel Drucker and shared by the lab of Dr. Gökhan Hotamışlıgil⁷⁵. L-GCGR-KO experiments utilized littermates obtained from hemizygous crosses.

Detailed methods for genotyping TSC2-WT and -5A mice were previously described⁵².

Mice were fed PicoLab Rodent Diet 20 (LabDiet, 5053) chow diet from weaning at three weeks of age. For high-fat diet studies, mice were started on a 60% fat diet (Research Diets, D12492i) or a sucrose-matched control diet containing 10% fat (Research Diets, D12450Ji) beginning at 8 weeks of age. The protein-free diet was prepared using a semi-purified water-soluble diet (Research Diets, D12450BSpxi) in 1% agar (Sigma-Aldrich, #A1296), as previously described⁶⁶. This diet mix was supplemented with casein hydrolysate (Thermo Scientific, #J12855.P5) and L-cystine (Sigma-Aldrich) for the isocaloric control diet (18% kcal protein). The high-carbohydrate diet was a variation of D12450BSpx that contained 10% kcal protein (Research Diets, D12450BSpx5i). For the protein-free or high carbohydrate, low protein diet fasting and refeeding studies, mice were acclimated to the control, semi-purified diet for 5–7 days prior to the experiment.

For animal treatments, mice were fasted overnight for 14 hours followed by intraperitoneal (i.p.) injection or oral gavage of vehicle (sterile 0.9% saline), insulin (Eli Lily, HumulinR) dissolved in sterile saline with protease-free BSA at 0.5 U/kg, glucose dissolved in sterile saline at 3 g/kg, or feeding of the given diet, as indicated. Mice were humanely euthanized immediately prior to tissue collection, and tissues were snap-frozen in liquid nitrogen and stored at −80 °C until analysis.

Glucose, insulin, and pyruvate tolerance tests

For glucose tolerance tests, mice were fasted for 6 hours during the day and injected i.p. with glucose dissolved in sterile saline at 2 g/kg (unless otherwise noted). For oral glucose tolerance tests, mice were fasted for either 6 hours during the day or 16 hours overnight, followed by oral gavage with glucose dissolved in sterile saline at 1–2 g/kg (as indicated in the figure legend). For insulin tolerance tests, mice were fasted for 6 hours during the day and then injected i.p. with insulin (Eli Lily, HumulinR) in sterile saline containing 0.1% protease-free BSA at 0.5 U/kg. For the pyruvate-lactate tolerance test, mice were fasted for 16 hours overnight, followed by i.p. injection with lactate (1.5 mg/kg; Sigma-Aldrich, L7022) and pyruvate (150μg/kg; Sigma-Aldrich, P5280) dissolved together in sterile saline. Blood glucose was monitored using the Contour Next EZ glucometer (Bayer).

Body composition

Body composition was measured in awake mice using the EchoMRI analyzer system (EchoMRI LLC).

Primary Hepatocyte Isolation and Culture

Mice were anesthetized using a ketamine/xylazine solution at a dose of 100 µg/10 µg per gram body weight. The liver was perfused through the portal vein with perfusion buffer (10 mM HEPES, 150 mM NaCl, 5 mM KCl, 5 mM glucose, 2.5 mM sodium bicarbonate, and 0.5 mM EDTA) for 5 min at a rate of 3–5 ml/min, followed by digestion buffer (10 mM HEPES, 150 mM NaCl, 5 mM KCl, 5 mM glucose, 2.5 mM sodium bicarbonate, 35 mM CaCl₂, and 30 µg/ml Liberase TM (Sigma Aldrich)) for 5–7 min at a rate of 3–5 ml/min. Perfusion and digestion buffers were pre-warmed to 37 °C prior to perfusion. Livers were excised and placed in DMEM (4.5 g/L glucose, without glutamine or sodium pyruvate) supplemented with 2.5% FBS, 2 mM GlutaMAX, and 1% penicillin–streptomycin. Digested liver tissue was agitated to release hepatocytes, and the cell suspension was gently pelleted. Hepatocytes were purified through Percoll density separation, washed in DMEM, and seeded on collagen-coated plates (BioCoat, Corning) at a density of 1.25 × 10⁶ cells for each well of a six-well plate. Media was replaced 4 hours after seeding.

For insulin stimulation experiments, human insulin (Sigma-Aldrich I9278) was dissolved in water and used at a concentration of 10 or 100 nM, glucagon (Sigma-Aldrich, G2044) was dissolved in 1% acetic acid and used at a concentration of 10 nM, and forskolin (Tocris, 1099) was dissolved in DMSO and used at a concentration of 10 μM. Hepatocytes were deprived of serum overnight, followed by treatment with insulin, glucagon, or forskolin for the indicated durations.

For experiments in low amino acid (Low AA) media, standard DMEM was diluted 1:10 with filter-sterilized DMEM without amino acids (U.S. Biologicals, D-9800-27; prepared according to manufacturer’s instructions and supplemented with 4.5 g/L glucose). Hepatocytes were cultured overnight in serum-free DMEM or serum-free Low AA prior to insulin stimulation.

For amino acid stimulation, amino acid deprivation was performed after two washes with PBS by incubating hepatocytes in serum-free, amino-acid-free DMEM for one hour. To stimulate with amino acids, the media was replaced with standard serum-free DMEM with or without insulin for 15 minutes.

For siRNA transfection experiments, control nontargeting pool (D-001810-10-50) and ON-Target Plus SMARTpool siRNAs against mouse Tsc2 (L-047050-00-0020) were purchased from Horizon Discovery. Hepatocytes were transfected 4 h after seeding with 100 nM siRNA and Lipofectamine RNAiMax (6 μl/well, six-well plate) and cultured in DMEM supplemented with 1% FBS, 1% penicillin–streptomycin, and 1 nM dexamethasone. After 72 h, the media was changed to serum-free DMEM (without dexamethasone) for insulin stimulation the following day.

Plasma and tissue analysis

Plasma insulin (Crystal Chem, #90080) and albumin (Crystal Chem, #80630) were measured by ELISA following the manufacturer’s instructions. Liver glycogen was measured using the Glycogen-Glo assay (Promega, #J5051) according to the manufacturer’s instructions.

Immunoblotting and antibodies

Primary cultured hepatocytes and liver tissue were lysed in RIPA buffer (150 mM NaCl, 5 mM EDTA, 50 mM Tris–HCl, 1% NP-40, 0.5% sodium deoxycholate, 10% SDS, 50 mM β-glycerophosphate, 50 mM NaF, 10 mM pyrophosphate tetrabasic, and 0.5 mM sodium orthovanadate) with protease and phosphatase inhibitors (Sigma Aldrich). Muscle tissue was lysed in NP-40 buffer (20 mM Tris, 150 mM NaCl, 10% glycerol, 1% NP-40) with protease and phosphatase inhibitors. Liver tissue was homogenized using the Red Lysis Kit (Next Advance) and Next Advance Bullet Blender (speed 8, for 3 min). Muscle tissue was homogenized manually using a handheld homogenizer (Cole-Parmer). Lysates were centrifuged at 20,000×g for 15 min at 4 °C and protein concentrations of the supernatant were determined using the Pierce Detergent-Compatible Bradford assay (Thermo Fisher Scientific). Equal amounts of protein were loaded onto polyacrylamide gels (Criterion, Bio-Rad), separated by electrophoresis, and transferred to nitrocellulose membranes. After transfer, membranes were blocked with 5% milk in Tris-buffered saline plus Tween-20 (TBST; 25 mM Tris, 150 mM NaCl, 2.5 mM KCl, 0.1% Tween-20) for 1 h at room temperature. For immunoblotting, membranes were incubated with primary antibodies (1:1000) in 5% BSA at 4 °C overnight. Membranes were washed three times with TBST and then incubated with IRDye 800CW anti-Rabbit (LICORbio #926-32211) or anti-Mouse (LICORbio #926-32210) secondary antibodies at 1:5000 in 5% milk for 1 h at room temperature. Membranes were washed three times in TBST and imaged using the LI-COR Odyssey CLx imaging system. LI-COR images were quantified using Image Studio Software. For immunoblots shown in Fig. 7, as well as Figs. 1A and S1A, membranes were developed by ECL (West Pico or Femto, Thermo Scientific) and imaged on autoradiography film. Immunoblot images on film were scanned and quantified using ImageJ software. The following primary antibodies were purchased from Cell Signaling Technologies: p-S6K1 (T389, #9234), S6K1 (#2708), p-rpS6 (S240/44; #2215), rpS6 (#2217), 4EBP1 (#9644), p-AKT (S473; #4060), AKT (#4691), p-TSC2 (T1462; #3617), TSC2 (#3612), p-Rictor (T1135, #3806), Rictor (#9476), p-AMPK (T172, #2535), AMPK (#2603), p-Raptor (S792 #2083), Raptor (#2280), p-ACC (S79 #3661), ACC (#3662), p-TFEB (S122, #87932) TFEB (#83010), p-PRAS40 (T246, #2997), PRAS40 (#2610), p- GSK3α/β (S21/9, #9331), GSK3α/β (#5676), GAPDH (#2118), IRS1 (#2382), p-CREB (S133, #9198), CREB (#9197), and vinculin (#4650). The β-actin antibody was purchased from Sigma-Aldrich (#A5316).

Immunohistochemistry (IHC)

Freshly resected liver lobes were fixed in 10% formalin at 4 °C for 24 h, then embedded in paraffin and sectioned at the Harvard Medical School Rodent Pathology Core. For staining, sections were rehydrated by three washes in Histoclear (National Diagnostics #HS-200), followed by two washes in 100% ethanol, two washes in 95% ethanol, and two washes in water. Sections were then boiled in 1x Sodium Citrate Antigen Retrieval Buffer for 10 min (Proteintech; PR30001, pH 6.0), washed in water, then incubated in 3% hydrogen peroxide (Sigma-Aldrich; H3410) for 10 min and washed again. Sections were blocked in Tris-buffered saline with Tween-20 (TBST; 25 mM Tris–hydroxymethyl Aminomethane, 137 mM NaCl, 2.7 mM KCl, 0.1% Tween-20) plus 5% normal goat serum (NGS; CST #5425) for 1 h at room temperature. p-S6 (S235/236;CST #4858) primary antibody was diluted 1:400 in SignalStain Antibody Diluent (CST #8112) and slides were incubated overnight at 4 °C. The next day, slides were washed three times in TBST, and signal was detected using the Boost IHC Detection Reagent (CST #8114) and SignalStain DAB Substrate Kit (CST #8059) according to the manufacturer’s instructions. Slides were counterstained with hematoxylin (Sigma #GHS132), washed, and dehydrated prior to mounting. Slides were imaged at ×20 magnification using an inverted Olympus CKX41 microscope and a Lumenera Infinity 2 camera.

Quantification of plasma amino acids by liquid chromatography–mass spectrometry

The method used for quantifying the absolute concentrations of plasma amino acids in mice following feeding was adapted from a previously described protocol¹¹¹. Blood was collected from the tail vein in EDTA tubes (Microvette® CB 300 EDTA K2E, Sarstedt) and centrifuged to separate plasma. Plasma was mixed at a 1:10 volume ratio with a −20 °C mixture of methanol:acetonitrile:water (40:40:20) and a stable isotope–labeled internal standard^13c labeled yeast extract (Cambridge Isotope Laboratory, Andover, MA, ISO1). These internal standards, together with a library of external metabolite standards, enabled absolute quantification across a wide dynamic range. The sample was vigorously vortexed and then centrifuged at 4 °C at 13,000×g for 10 min. 30 μl of the resulting supernatant was transferred to an HPLC vial, from which 5 μl was injected into the LC–MS system for analysis.

Chromatographic separation was achieved using an XBridge BEH Amide XP Column (2.5 μm, 2.1 mm × 150 mm) with a guard column (2.5 μm, 2.1 mm × 5 mm) (Waters). The mobile phase consisted of water:acetonitrile 95:5 for mobile phase A and water:acetonitrile 20:80 for mobile phase B, both containing 10 mM ammonium acetate and 10 mM ammonium hydroxide. The elution linear gradient was as follows: 0–3 min, 100% B; 3.2–6.2 min, 90% B; 6.5–10.5 min, 80% B; 10.7–13.5 min, 70% B; 13.7–16 min, 45% B; and 16.5–22 min, 100% B, with a flow rate of 0.3 ml/min. The autosampler temperature was maintained at 4 °C, and the injection volume was 5 μl. Needle wash was performed between samples using methanol:acetonitrile:water. The mass spectrometer used was a Q Exactive HF (Thermo Fisher), which scanned from 70 to 1000m/z with switching polarity and a resolution of 120,000. Metabolite identification was based on accurate mass and retention time, confirmed against external standard pools, using El-MAVEN (Elucidata). Quantification was performed either by stable isotope dilution, where a labeled internal standard was available, or by external calibration curves for metabolites lacking an internal standard. This approach enabled robust and sensitive quantification of amino acids in plasma, resolving concentrations from low nanomolar to high millimolar levels.

Gene expression analysis

Liver tissue was lysed in cold TRIzol (Invitrogen) and RNA was isolated following the manufacturer’s instructions. RNA was reverse transcribed using the Advanced cDNA Synthesis Kit (Bio-Rad). qPCR analysis was performed with duplicate technical replicates using iTaq SYBR green (Bio-Rad) and the CFX96 Real-Time PCR Detection System (Bio-Rad). Samples were normalized to actin (human) or 36b4 (mouse) for DDCt analysis.

Liver FPLC

Liver tissue was flash frozen and then homogenized in CHAPS lysis buffer (40 mM HEPES, pH 7.4, 120 mM NaCl, 1 mM EDTA, 0.3% w/v CHAPS, 10 mM pyrophosphate, 10 mM glycerol-2-phosphate, 50 mM NaF) containing freshly added protease inhibitors at 1:100 (Sigma, P8340) using the Red Lysis Kit (Next Advance) and Next Advance Bullet Blender (speed 8, for 3 min). Lysates were centrifuged at 16,000×g for 15 min and supernatants were normalized for protein concentration. 700 l of each sample was injected into a size-exclusion column (HiLoad 16/600 Superdex 200, Cytiva) attached to a Fast Protein Liquid Chromatography (FPLC) system consisting of an AKTA PURIFIER 10 core system and a Frac-900 fraction collector, flushing loop with 10 ml CHAPS lysis buffer between samples. Samples were eluted with CHAPS lysis buffer and collected as 0.5 ml fractions. For SDS–PAGE, Laemmli SDS sample buffer was added to the fractions and samples were boiled at 95 °C for 10 min. Equal volumes were loaded on a 4–15% gel and analyzed as above.

Liver permeabilization for analysis of soluble fraction

Flash-frozen livers were subjected to differential detergent fractionation¹¹². Briefly, liver tissue was pulverized using a mortar and pestle on dry ice and transferred to the pre-chilled vessel of 2 ml dounce homogenizer. Tissue was homogenized in 500 μl digitonin buffer (0.015% (w/v) digitonin, 10 mM PIPES (pH 6.8), 300 mM sucrose, 10 mM NaCl, 3 mM MgCl, 5 mM EDTA, and 1 mM PMSF) containing freshly added protease and phosphatase inhibitors with 10 strokes with the A pestle, followed by a short ~ 30 s break on ice and 8 strokes with the B pestle. The samples were gently rotated at 4 °C for 30 min, then centrifuged at 16,000×g at 4 °C for 15 min. The supernatant rich in cytosolic proteins was removed, and equal amounts of protein were analyzed by immunoblot as above.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 10.0, and details about statistical tests applied for each panel are provided in the corresponding figure legends. For experiments comparing two groups, an unpaired two-tailed Student’s t-test was used. Experiments comparing two or more groups across conditions were analyzed by two-way ANOVA with Sidak’s multiple comparison test, unless otherwise noted in the figure legend. No assumptions were made regarding sample distribution or standard deviation, and non-parametric tests were used in cases where data did not pass one or more normality tests applied by GraphPad Prism. Each data point represents an individual value, and graphical data are presented as mean ± SD, unless otherwise noted in the figure legend. In cases where exact values are not provided in figure panels or legends, asterisks above comparisons indicate the degree of significance as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: K.C.K., Y.C., and B.D.M.; Experimentation: Y.C., M.Y.C., and S.C.L. assisted with in vivo experiments; M.Y.C. performed and analyzed plasma metabolomics; K.E.I., with the support of G.S.H., assisted with metabolic phenotyping follow-up studies and facilitated acquisition of Gcgr^fl/fl and L-GCGR-KO mice; K.C.K. performed all other experiments and analysis; Writing: K.C.K. wrote this article with editing from B.D.M.; Funding acquisition: K.C.K., Y.C., M.Y.C., and B.D.M.; Resources: G.S.H. and B.D.M.; Supervision: B.D.M.

Peer review

Peer review information

Nature Communications thanks Jason Kim and other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was supported by a Glenn Foundation for Medical Research Postdoctoral Fellowship (Y.C.); NIH grants T32-DK128781 (K.C.K. and S.C.L.), F31-DK128873 (K.C.K.), R35-CA197459 (B.D.M.), P01-CA120964 (B.D.M.); and Damon Runyon Cancer Research Foundation Merck Fellowship DRG-#2443-21 (M.Y.C.). These funders were not involved in the design, execution, or interpretation of this study.

Data availability

All pertinent data are available within the paper and its Supplementary Information. The metabolomics data pertaining to Fig. 3 generated in this study are provided in the file titled Supplementary Data 1. Source data are provided with this paper. The raw data (and associated statistical analyses) generated in this study have been deposited in the Figshare database under accession code https://doi.org/10.6084/m9.figshare.33019523. Source data are provided with this paper.

Competing interests

B.D.M. is a consultant for Gondola Bio, independent of the contents of this article. G.S.H. is a member of the Scientific Advisory Board and holds equity in Crescenta Pharmaceuticals (not related to the contents of this article). Additionally, technology from the Hotamışlıgil Lab is licensed to Iş Private Equity, from which G.S.H. also receives research funding. This is also not related to the contents of the article. All the other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77216-6.