Nutrient Sensor mTORC1 Regulates Insulin Secretion by Modulating β-Cell Autophagy

Autophagy is a mechanism essential for cell survival, especially under stressful conditions (1). Consistently, in β-cells, disruption of autophagy induces cellular stress and degeneration, leading to diabetes (2–8). It is logical to expect that autophagy is also important for physiological control of insulin secretion; however, its housekeeping role has hindered evaluation of its moment-to-moment regulatory effect on β-cell function.

Knockdown of autophagy genes in β-cell lines enhances insulin release, suggesting that autophagy negatively regulates insulin secretion (9–11); this could help lower insulin in fasting. By contrast, Goginashvili et al. (12) found that in β-cells, nutrient deprivation paradoxically inhibits autophagy as a result of starvation-induced crinophagy (autophagy-independent lysosomal degradation of secretory granules), leading to accumulation of amino acids (AAs) adjacent to lysosomes, thereby stimulating mechanistic target of rapamycin complex 1 (mTORC1).

mTORC1 functions as a hub, integrating signals from growth factors, nutrients, and energy (ATP) to regulate cellular functions (13,14). AA sensing by mTORC1 has been extensively studied (15–17). AA availability is conveyed to mTORC1 via Rag GTPases, which recruit mTORC1 to the lysosomal surface for activation by Rheb. mTORC1 inhibits autophagy via phosphorylation of ULK1, which initiates autophagosome generation, and TFEB, master regulator of lysosome biogenesis and autophagy (18,19). The regulation of mTORC1-autophagy signaling in β-cells under different nutritional states and its impact on insulin secretion are not clear.

Herein, we show that in β-cells, mTORC1 is inhibited under fasting, while it is rapidly stimulated by acute exposure to nutrients, mainly leucine and glucose acting in concert to stimulate both mTORC1 and insulin secretion. Nutrient stimulation of mTORC1 is associated with inhibition of ULK1 and TFEB, along with decrease in autophagy-regulating genes. Inhibition of mTORC1 by Raptor knockout (KO), mimicking the effects of fasting, stimulated autophagy and inhibited nutrient-stimulated insulin secretion, whereas inhibition of autophagy under these conditions rescued insulin release. Thus, mTORC1 is an in vivo physiological regulator of insulin secretion that exerts its action by modulating β-cell autophagy.

The experimental protocol was approved by the Institutional Animal Care and Use Committee of the Hebrew University of Jerusalem. Mouse strains used in this study included C57BL/6 (Harlan Laboratories, Jerusalem, Israel), Tg(Ins1-Cre/ERT)1Lphi (MIP-CreER) (20), Rptor^tm1.1Dmsa (Raptor^fl/fl) (The Jackson Laboratory, Bar Harbor, ME) (21), and B6.Cg-Atg7<tm1Tchi> (ATG7^fl/fl) (RIKEN BioResource Research Center, Tsukuba, Japan) (22). β-Cell–specific KO of Raptor and Atg7 was generated by two consecutive subcutaneous daily injections of 8 mg tamoxifen (Sigma-Aldrich, Cleveland, OH) (20 mg/mL in corn oil). We performed all the experiments in male mice to avoid leakiness of the system resulting from activation of Cre by circulating estrogen in females. Mice were housed at the Hebrew University animal care unit under a 12-h light/dark cycle with free access to food and water.

The β-cell line INS-1 832/13 (23) was provided by Dr. Ángela M. Valverde (Instituto de Investigaciones Biomédicas “Alberto Sols,” Madrid, Spain). Cells were cultured in RPMI medium containing 11 mmol/L glucose and supplemented with 10% heat-inactivated FCS, 100 IU/mL penicillin, 100 μg/mL streptomycin, 10 mmol/L HEPES, 2 mmol/L L-glutamine, 1 mmol/L sodium pyruvate (Biological Industries, Beit Haemek, Israel), and 0.05 mmol/L 2-mercaptoethanol (Sigma-Aldrich) in a humidified 37°C, 5% carbon dioxide, incubator.

Monomeric red fluorescent protein (RFP)–green fluorescent protein (GFP) tandem fluorescent-tagged LC3 plasmid (24) was provided by Dr. Adi Kimchi (Weizmann Institute of Science, Rehovot, Israel). pEGFP-N1-TFEB plasmid was a gift from Shawn Ferguson (Addgene plasmid 38119; https://n2t.net/addgene:38119; RRID:Addgene_38119) (25). INS-1 cells were transfected with TransIt LT1 (Mirus Bio, Madison, WI).

Mice were fasted overnight and then given 2 g/kg glucose and/or 0.39 g/kg leucine by oral gavage. Blood samples were drawn before and 30 min after gavage. Glucose tolerance was assessed by i.p. glucose tolerance test; mice were fasted overnight and then injected i.p. with 2 g/kg glucose. Insulin sensitivity was assessed by insulin tolerance test; mice were fasted overnight and then given 0.75 units/kg insulin i.p. Glucose was monitored and blood samples were obtained at the indicated time points. Plasma insulin was determined using an ultrasensitive mouse insulin ELISA kit (Crystal Chem, Elk Grove Village, IL).

Islets were isolated by collagenase P (Roche Diagnostics GmbH, Mannheim, Germany) injection into the bile duct followed by separation by Histopaque gradient (Sigma-Aldrich). Islets were handpicked and cultured overnight in complete RPMI as described above.

Islets or INS-1 cells were preincubated for 1 h in Krebs-Ringer modified buffer containing 3.3 mmol/L glucose, then consecutively incubated for 1 h in Krebs-Ringer modified buffer containing 3.3 or 16.7 mmol/L glucose with or without AAs. Insulin and proinsulin were extracted with GB/NP40 buffer and analyzed by mouse insulin ELISA: Crystal Chem Ultra-Sensitive Insulin ELISA Kit for islet experiments and Mercodia Insulin ELISA Kit for experiments in INS-1 cells. Proinsulin was measured in islet extracts using the Mercodia Rat/Mouse Proinsulin ELISA Kit.

Pancreases were fixed with 4% formaldehyde overnight. Paraffin-embedded sections were rehydrated, and antigen retrieval was performed in citrate buffer (pH 6). Sections were blocked with CAS-Block (Thermo Fisher Scientific, Waltham, MA). Islets were dispersed with trypsin, fixed with 3.7% paraformaldehyde, permeabilized with 0.1% Triton X, and blocked with 1% BSA buffer. Primary and secondary antibodies used are listed in Supplementary Table 1. Digital images were obtained with an A1R confocal microscope (Nikon Corporation, Tokyo, Japan) using a ×60 or ×40 oil objective. Image analysis was performed with NIS-Elements AR v4.20.003 (Nikon Corporation), ImageJ 1.53c (National Institutes of Health, Bethesda, MD), and Photoshop CC 22.1.0 (Adobe, Inc., San Jose, CA).

Mouse islets were fixed using a cold solution of 2.5% glutaraldehyde and 2% formaldehyde in 0.1 mol/L cacodylate buffer (pH 7.4) for 2 h at room temperature. Samples were postfixed and stained with 1% (w/v) osmium tetroxide and 1.5% (w/v) potassium ferricyanide in 0.1 mol/L cacodylate buffer for 1 h, followed by dehydration in increasing concentrations of ethanol and embedding in araldite. Ultrathin sections were collected onto copper grids and sequentially stained with uranyl acetate and lead citrate. Sections were imaged with a Tecnai 12 transmission electron microscope (Philips, Amsterdam, the Netherlands) equipped with a MegaView-II CCD camera and Analysis software version 3.0 (Soft Imaging System GmbH, Münster, Germany).

Lysates were separated by electrophoresis and then transferred to a nitrocellulose membrane. Membranes were probed using appropriate primary and secondary antibodies (Supplementary Table 1) and developed with Clarity Western ECL Substrate (Bio-Rad Laboratories, Hercules, CA) using the Chemidoc Touch Imaging System. Image analysis and quantifications were performed with Image Lab Software version 6.1 (Bio-Rad Laboratories).

RNA was extracted using Bio Tri RNA (Bio-Lab, Jerusalem, Israel) and an RNeasy Micro Kit (Qiagen, Venlo, the Netherlands). Samples were reverse transcribed using the qScript cDNA Synthesis Kit (Quantabio, Beverly, MA). Quantitative PCR was performed on a StepOnePlus system using Fast SYBR Green Master Mix (Thermo Fisher Scientific). Gene expression was normalized to 18S ribosomal RNA and analyzed with StepOnePlus software version 2.3. The primers used for quantitative PCR are shown in Supplementary Table 2.

Islets were isolated and incubated overnight in complete RPMI growth medium containing 11 mmol/L glucose. The next day, islets were preincubated for 1 h in starvation medium (RPMI medium lacking glucose and AAs) and then incubated for 30 min in RPMI including all AAs supplemented with 2 mmol/L L-glutamine and 2.8 or 16.7 mmol/L glucose. Islets were then dispersed into single-cell suspension using Accutase (STEMCELL Technologies, Inc., Vancouver, BC, Canada), fixed and washed with Cytofix/Cytoperm solution (BD Biosciences, Franklin Lakes, NJ), and incubated overnight at 4°C with primary antibodies (Supplementary Table 1). Cells were then washed and incubated for 2 h with species-specific secondary antibodies (Supplementary Table 1). Analysis and quantifications were performed by the LSRFortessa flow cytometer and FlowJo software version 10.5.3 (BD Biosciences).

Islets were derived from three brain-dead donors (see Human Islet checklist), incubated overnight in RPMI medium supplemented with 10% FBS at 3.3 or 16.7 mmol/L glucose with or without 10 and 100 nmol/L rapamycin and/or 10 μmol/L chloroquine (CQ). Static incubations were performed on 15 islets in replicates that were preincubated at 3.3 mmol/L glucose for 1 h and then at 3.3 and 16.7 mmol/L for 45 min each. Secreted insulin and islet insulin content were analyzed by ELISA (Mercodia).

Data are presented as mean ± SEM. Unpaired two-tailed Student t test with Welch correction was used to compare differences between two groups. Differences between multiple groups were analyzed by Brown-Forsythe and Welch one-way ANOVA with Dunnet T3 or Games-Howell post hoc test or with two-way ANOVA with Dunnet or Šídák post hoc test. All statistical analyses were performed using Prism 9.0.0 (GraphPad Software, San Diego, CA).

All relevant data are available from the authors upon request.

The β-cell line INS-1 823/12 and islets were cultured under starvation conditions (3.3 mmol/L glucose in absence of serum and AAs) for 30 min, then at low (3.3 mmol/L) or high glucose (HG) (16.7 mmol/L) in presence of serum with different AAs for 30 min. Using phosphorylated S6 as readout, detailed analysis of mTORC1 activation in INS-1 β-cells showed that the branched-chain AA (BCAA) leucine rapidly stimulated mTORC1, whereas other AAs known to regulate mTORC1 had either modest or no effect (Supplementary Fig. 1A). In islets, leucine stimulated mTORC1 dose dependently (Supplementary Fig. 1B). In INS-1 cells, HG failed to activate mTORC1 in absence of AAs; however, it amplified the AA effect. As expected, rapamycin inhibited the stimulation of mTORC1 by AAs (Supplementary Fig. 1C). In islets, HG moderately increased S6 phosphorylation (not significant), whereas AAs increased the number of pS6⁺ β-cells and fluorescence intensity at both low glucose and HG, indicating enhanced mTORC1 activity (Supplementary Fig. 1D). RAPTOR levels remained unchanged by AA starvation and following AA supplementation (Supplementary Fig. 1E). These findings show that in vitro exposure of β-cells to starvation inactivates, rather than stimulates, mTORC1; a single AA, leucine, effectively activates mTORC1 and may act in concert with glucose.

We next corroborated the effects of fed-fasted states on β-cell mTORC1 activity in vivo. Mice were fasted overnight with free access to water with or without added BCAA (13.3 g/L each leucine/isoleucine/valine). Fasting led to inhibition of mTORC1, evident by decrease in β-cells expressing pS6 and by reduction of total pS6 fluorescence intensity (Fig. 1A). BCAA in drinking water increased mTORC1 activity in fasted animals, proving response to AAs in vivo. We then assessed acute effects of leucine and glucose on mTORC1 activity in β-cells. Leucine (0.39 g/kg), glucose (2 g/kg), or both were administered by oral gavage after 16-h fast; the pancreases harvested 1 h later, and sections or dispersed isolated islets were stained for pS6 and insulin. Both leucine and glucose increased β-cell mTORC1 activity in β-cells; feeding them together stimulated mTORC1 in an additive manner (Fig. 1B).

Thus, in β-cells, mTORC1 responds to external nutritional cues as expected, with no evidence of paradoxical stimulation by starvation in vitro or in vivo. Leucine is a potent activator of mTORC1, and glucose further amplifies it in β-cells.

mTORC1 is important in postnatal β-cell development and functional maturation (26–28); however, its role in mature β-cells has been less studied. To this end, we generated MIP-CreER; Raptor^fl/fl mice and induced Raptor KO in adult animals by tamoxifen. We assessed insulin secretion early (2 weeks) after tamoxifen injection to minimize deleterious effects of prolonged mTORC1 deficiency. At this time point, mTORC1 was indeed inhibited (Supplementary Fig. 2A and B); the presence of the MIP-Cre construct and tamoxifen injection did not affect glucose tolerance (Supplementary Fig. 2C and E) or insulin sensitivity (Supplementary Fig. 2D and F). Relevant to the current study, we further show that the MIP-CreER construct did not affect autophagy in islets (Supplementary Fig. 2H). In agreement with a previous study (26), we found that βRaptor KO mice developed mild glucose intolerance (Supplementary Fig. 2G). At 30 min after oral gavage of glucose and leucine, glucose levels were higher in βRaptor KO than in control mice; serum insulin levels were also decreased and insulinogenic index markedly reduced (Fig. 2A–C).

Leucine stimulated mTORC1 concentration dependently with maximal effect at 4–10 mmol/L (Supplementary Fig. 1B). Consistent with the hypothesis that leucine amplifies insulin secretion in an mTORC1-dependent manner, it dose dependently increased secretion in islets treated with the metabolic fuels glutamine and glucose (Supplementary Fig. 3). Maximal stimulation of mTORC1 and insulin secretion was observed at 10 mmol/L leucine, which was used in additional in vitro studies. To decipher the role of mTORC1 in mediating the leucine effects on autophagy and insulin secretion, we performed static incubations. The nonmetabolizable leucine analog 2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid and the AA combination of leucine plus glutamine enhance insulin secretion by increasing glutamate oxidation in the tricarboxylic acid (TCA) cycle (29). While at 3.3 mmol/L glucose, AA-stimulated insulin secretion was similar in βRaptor KO and control islets, the response to HG with or without AAs was decreased by ∼50% (Fig. 2D). In INS-1 cells, the mTORC1 inhibitor rapamycin inhibited HG plus AA–stimulated insulin secretion (Fig. 2E). Nutrients regulate insulin secretion by plasma membrane depolarization, with subsequent opening of voltage-dependent calcium channels (triggering pathway), and via metabolic signals that amplify secretion independently of KATP channels (amplifying pathway). We studied KATP-independent amplification of secretion by glucose and AAs (leucine plus glutamine) in islets and INS-1 cells treated with or without rapamycin. Islets or INS-1 cells were exposed to nutrients in presence of diazoxide (KATP channel opener) together with 30 mmol/L potassium chloride (membrane depolarizer) with or without 100 nmol/L rapamycin (Fig. 2F and G). In islets, glucose and AAs (leucine plus glutamine) amplified insulin secretion by 2.5- and 4-fold, respectively; their combination further augmented secretion. In both islets and INS-1 cells, rapamycin abrogated the amplification of insulin secretion by AAs at HG (Fig. 2F and G). In INS-1 cells, rapamycin also inhibited amplification of insulin secretion by glucose alone (Fig. 2G).

In summary, glucose and AAs act in concert to stimulate mTORC1 and increase insulin secretion in an mTORC1-dependent manner.

mTORC1 is a central regulator of autophagy; we therefore studied β-cell autophagy in the postabsorptive and fed states. Autophagosome maturation involves lipidation of LC3 recruited to the autophagosome membrane, which is used to quantify autophagosomes. Staining of pancreatic sections of fasted mice showed reduced mTORC1 activity associated with markedly increased autophagosome number, whereas refeeding for 4 h or adding BCAA to drinking water inhibited autophagy (Fig. 3A and B). We then studied the early effects of nutrient administration on autophagy. Surprisingly, gavage of leucine and glucose alone or together markedly increased autophagosomes 1 h following nutrient administration, despite the simultaneous stimulation of mTORC1 (Fig. 3C). Accumulation of autophagosomes may result from increased generation (stimulated autophagy) or from disruption of autophagosome-lysosome fusion (inhibited autophagy). To distinguish, we isolated islets followed by incubation with glucose and leucine with and without bafilomycin A1, which alkalinizes lysosomes, thereby inhibiting autophagosome-lysosome fusion and degradation. Bafilomycin A1 increased autophagosome number at both 3.3 mmol/L glucose and 16.7 mmol/L glucose plus leucine, indicating autophagy was indeed enhanced (Fig. 4A). By contrast, prolonged incubation of islets in presence of HG and leucine for 16 h inhibited autophagy, as evident by lower number of autophagosomes without and with bafilomycin A1 (Fig. 4B). These findings suggest that glucose and AAs, while rapidly stimulating mTORC1, induce a biphasic response of autophagy: an early burst of autophagosome generation followed by late inhibition of autophagy.

To obtain further insight into nutrient regulation of mTORC1-autophagy interaction, we studied ULK1 and TFEB activities. AA stimulation of mTORC1 in INS-1 cells (Supplementary Fig. 1C) was associated with inhibition of autophagy, evident by decreased LC3-ΙΙ/Ι ratio by Western blotting (Fig. 5A) and reduced LC3⁺ puncta (Fig. 5B), mimicking the autophagy inhibition by prolonged islet exposure to nutrients (Fig. 4B). Accumulation of P62/SQTM1, which is degraded by lysosomes, indicates inhibited autophagic flux. Consistently, P62/SQSTM1 was increased in cells treated with HG plus AAs (Fig. 5C). Transfection with an LC3-GFP-RFP construct enables assessment of the autophagic flux: high autophagosome turnover in stimulated autophagy leads to enrichment of red RFP⁺ puncta as GFP fluorescence undergoes quenching in autolysosomes; in contrast, inhibition of autophagosome-lysosome fusion is associated with accumulation of yellow RFP⁺/GFP⁺ puncta. Incubation of INS-1 cells with AAs and glucose for 4 h increased the number of GFP⁺/RFP⁺ puncta (Fig. 5D), suggesting that nutrients inhibited both autophagosome generation and autophagosome-lysosome fusion.

mTORC1 directly phosphorylates ULK1 at serine 757, which prevents autophagosome initiation, and TFEB at serine122. We incubated islets and INS-1 cells with glucose and AAs for different periods of time and analyzed mTORC1 activity and ULK1 and TFEB phosphorylation (Fig. 5E and F). Islet incubation at HG plus leucine increased both total and phosphorylated S6, whereas the protein levels of TSC1 and S6K1 and the housekeeping GAPDH remained unchanged. S6 is a ribosomal protein involved in protein synthesis, suggesting that in addition to 4EBP1 and S6 phosphorylation, mTORC1 may promote protein synthesis by increasing S6 biosynthesis; this was observed in islets, but not in INS-1 cells (Supplementary Fig. 1A). Incubation of islets with HG or leucine alone showed that the leucine effect on S6 activity is indeed greater than that of HG; this was associated with increased ULK1 phosphorylation (Supplementary Fig. 4). Importantly, incubation of islets and INS-1 cells at HG plus AAs for 1 h increased both ULK1 and TFEB phosphorylation in parallel with the activation of mTORC1 (Fig. 5E and F).

We studied the effects of nutrients on TFEB localization in INS-1 cells and islets. We transfected INS-1 cells with TFEB-GFP construct and incubated the cells at 3.3 mmol/L glucose or HG plus AAs. Nutrients induced nuclear exclusion of TFEB (Fig. 5G).

This could also be demonstrated in vivo in pancreases removed 1 h after gavage from mice administered glucose and leucine. By contrast, in βRaptor KO pancreases, the nuclear localization of TFEB was increased and was not affected by nutrients (Fig. 6A). Ex vivo exposure of dispersed islets to HG plus leucine showed two patterns of intracellular TFEB distribution: small puncta diffusely spread in the cytosol, and cytosolic aggregates (Fig. 6B). At low glucose, a majority of cells (∼80%) showed diffuse staining, whereas exposure to HG and leucine resulted in a mixed phenotype, with ∼50% containing cytosolic aggregates (Fig. 6B). βRaptor KO partially prevented TFEB aggregation in islets treated with HG plus leucine. These alterations in TFEB localization in βRaptor KO islets were associated with increased expression of TFEB-regulated genes, including cathepsin D (CtsD), H⁺/Cl⁻ exchanger, transporter 7 (Clcn7), and WIPI1 (Atg18 homolog), along with nonsignificant increases in CtsA and CtsB (Fig. 6C). Overnight incubation of wild-type (WT) islets with HG and leucine decreased the expression of TFEB-regulated genes (Fig. 6D). TFEB also regulates lysosome biogenesis; consistently, we found higher numbers of lysosomes (LAMP1⁺ vesicles) in βRaptor KO mice treated with or without glucose and leucine compared with controls (Supplementary Fig. 5).

We further assessed how the mTORC1-autophagy crosstalk regulates insulin secretion by inhibiting autophagy in βRaptor KO mice. Induced Raptor KO in β-cells stimulated autophagy along with inhibition of insulin secretion, hence mimicking fasting (Fig. 2). To further assess the role of autophagy in regulating insulin secretion, we generated double-transgenic mice, MIP-CreER; Raptor^fl/fl; Atg7^+/fl, in which autophagy is moderately inhibited in β-cells in the presence of mTORC1 deficiency. We first tested the effects of βRaptor KO with and without Atg7 haploinsufficiency on autophagy. Indeed, Atg7 haploinsufficiency effectively reduced stimulation of autophagy by βRaptor KO in β-cells (Supplementary Fig. 6A and B). Transmission electron microscopy showed increased autophagosomes and peroxisomes in βRaptor KO mice that was prevented by Atg7 haploinsufficiency (Supplementary Fig. 6C). Secretory granules were reduced in the βRaptor KO mice without affecting the proportion of immature granules (Supplementary Fig. 6D).

Next, we studied the effects of Atg7 haploinsufficiency on insulin secretion in control mice and in βRaptor KO mice. Heterozygous Atg7 KO in mTORC1-competent β-cells (MIP-creER; Atg7^fl/+) modestly increased glucose-stimulated insulin secretion (GSIS) along with mild improvement in glucose tolerance (Fig. 7A–C). Static incubations showed similar GSIS in control and βAtg7^+/− islets (Fig. 7D). Thus, the overall effect of heterozygous Atg7 KO on insulin secretion was relatively small.

We then studied the effects of heterozygous Atg7 KO on insulin secretion in mTORC1-deficient β-cells (MIP-creER; Raptor^fl/fl; Atg7^fl/+). Prevention by Atg7 KO of the autophagy stimulation induced by mTORC1 deficiency was accompanied by notable decrease in postprandial (glucose/leucine gavage) glucose excursions and increase in the insulinogenic index (Fig. 7E–G). Atg7 heterozygous mice failed to correct the reduced islet insulin content of βRaptor KO islets (Fig. 7H); however, it completely prevented the decline in insulin secretion (Fig. 7I). In fact, normalized for the reduced content, insulin secretion was greater than in WT islets. The proinsulin/insulin ratio was not modified in the transgenic islets (Fig. 7J). We reiterated these findings by injecting 5 μg/kg CQ, an inhibitor of lysosomal function, for 3 days into WT mice; this resulted in increased fasting insulin levels, along with amplification of GSIS and improvement of glucose tolerance (Fig. 7K and L). Collectively, these findings show that autophagy restrains insulin secretion in the postabsorptive and fed states. Inhibiting autophagy may enhance insulin secretion in vivo.

We tested whether the above conclusions apply to human islets by incubating at 3.3 or 16.7 mmol/L glucose and 10 mmol/L leucine with or without rapamycin and/or CQ. Acute exposure to leucine amplified insulin secretion in islets treated with HG (Supplementary Fig. 7). Rapamycin modestly decreased GSIS (not significant) and significantly reduced exocytosis at HG plus leucine. CQ partially prevented the inhibition of insulin secretion by rapamycin (Supplementary Fig. 7). Our studies in mouse islets showed that the inhibition of autophagy by nutrients evolves over time. We therefore tested the effects of overnight treatment with HG plus leucine with or without rapamycin and/or CQ on mTORC1-autophagy signaling and insulin secretion. HG plus leucine markedly increased mTORC1 activity, including S6 protein level and phosphorylation and 4EBP1, ULK1, and TFEB phosphorylation. Treatment with rapamycin prevented nutrient stimulation of mTORC1, whereas treatment with CQ did not prevent the inhibition of S6 phosphorylation by rapamycin (Fig. 8A). As expected, HG increased the accumulation of insulin in the medium without affecting islet insulin content (Fig. 8B and C). Of note, leucine amplified insulin release at HG, whereas rapamycin decreased secretion in islets incubated at HG with and without leucine. Importantly, treatment with CQ completely prevented the inhibition of insulin secretion by rapamycin. We then tested the islet response to glucose after overnight exposure to nutrients and pharmacological inhibitors of mTORC1 and autophagy. Overnight incubation at HG enhanced insulin secretion at low glucose and HG concentrations (Fig. 8D). Overnight incubation at HG plus leucine further enhanced GSIS, whereas pretreatment with rapamycin, including at a low concentration of 10 nmol/L, markedly attenuated the acute response to glucose. CQ alone did not affect GSIS in islets that were incubated overnight with HG plus leucine; however, it completely prevented its inhibition by rapamycin (Fig. 8D).

Collectively, these findings show that exposure to HG not only increases the chronic secretion to the medium, but also potentiates GSIS. Leucine further increased chronic and acute GSIS, whereas rapamycin markedly inhibited insulin secretion in islets incubated with HG and leucine. Strikingly, treatment with CQ completely prevented the inhibition of insulin secretion by rapamycin.

We conclude that leucine amplifies insulin secretion in adult mouse and human islets through modulation of mTORC1 and autophagy (Fig. 8E).

mTORC1 governs the functional maturation of β-cells during postnatal development and its adaptation to nutrition (26–28,30). In a seminal article, Blandino-Rosano et al. (26) showed that constitutive disruption of mTORC1 in β-cells by conditional Raptor KO induced apoptosis, decreased proliferation and mass, and impaired function through the S6K and 4EBP2-eIF4E pathways. These alterations were partially prevented by inhibiting lysosomal hydrolases with CQ, which rescued β-cell mass and increased serum insulin (26). The multiple effects of prolonged mTORC1 deficiency on β-cell mass and function render it difficult to decipher the physiological role of mTORC1-autophagy crosstalk in nutrient sensing and regulation of insulin secretion. To address this fundamental question, we first studied mTORC1 and autophagy during cycles of fasting and feeding and performed fine-tuned modulation of mTORC1 activity by short-term inducible Raptor KO in adult β-cells and of autophagy by heterozygous Atg7 KO; with these tools at hand, we tested the effects on glucose and AA stimulation of insulin secretion in vivo and ex vivo.

We show here that mTORC1 is a bona fide nutrient sensor and important regulator of insulin secretion in adult rodent and human β-cells. Furthermore, we demonstrate that the mTORC1 effect on insulin secretion in the postabsorptive period is mediated by autophagy. Nutrient sensors that are well adjusted to cope with fast-feed cycles are expected to exhibit low activity during nutrient deprivation and be rapidly and effectively stimulated when food is available. Consistently, we found that β-cell mTORC1 activity was barely detectable under starvation conditions both ex vivo and in vivo fasting, while it was rapidly stimulated by refeeding or following acute administration of glucose and AAs.

We and others have shown that mTORC1 plays a developmental role in the functional maturation of β-cells, later being downregulated during adulthood (28,31). Furthermore, it has been suggested that postweaning nutrient sensing is switched from mTORC1 to AMPK (31). However, decline in mTORC1 after weaning, rather than developmental involution, may correspond to changes in feeding behavior and diet composition resulting from the switch from suckling to periodic feeding, with mTORC1 now assuming a new, oscillating role in nutrient sensing. Our findings are in marked disagreement with the suggestion that mTORC1 is activated during fasting through enhanced proteolysis of secretory granules (12).

In agreement with previous studies (32–35), we show here that in β-cells, AAs, mainly the BCAA leucine, are potent activators of mTORC1. We further show that in vivo glucose enhances mTORC1 activity and amplifies the leucine effect, perhaps via dihydroxyacetone phosphate as suggested (36).

It has been recently suggested that in the postweaning period, islets shift from primary AA-stimulated insulin secretion to GSIS (32). On the contrary, we show here that in adult β-cells, mTORC1 exquisitely senses both AAs (leucine) and glucose, and that intermittent stimulation of mTORC1 is required for insulin secretion in response to these nutrients.

Leucine acutely stimulates insulin secretion through deamination to α-ketoisocaproic acid and enhances glutaminolysis by allosterically activating glutamate dehydrogenase, with subsequent glutamate oxidation in the TCA cycle (29,37,38). At HG, glutamate dehydrogenase promotes nutrient export from the TCA cycle to the cytosol (cataplerosis) rather than glutamate oxidation. We show that under these conditions, AAs (leucine) stimulate secretion at least in part through mTORC1. Inhibition of mTORC1 reduced AA-stimulated insulin secretion from depolarized islets and INS-1 cells, suggesting that mTORC1 is required for the amplification of insulin secretion by glucose/AAs downstream to the KATP channel.

Modulation of mTORC1 by the nutritional state was accompanied by inverse modulation of autophagy; fasting was associated with enhanced autophagy, and this was reversed by refeeding, supplementation of BCAA in drinking water, or overnight exposure to HG and leucine in vitro. mTORC1 deficiency induced by βRaptor KO also stimulated autophagy, thus mimicking the effects of fasting. Mechanistically, we show that inhibition of mTORC1 either by fasting or by genetic manipulation decreased the inhibitory phosphorylation of ULK1 and TFEB and modulated the intracellular localization of the transcription factor TFEB in islets. This was accompanied by increased expression of the autophagy-regulating gene WIPI1 and of genes involved in lysosomal function and biogenesis. These findings are consistent with recent reports showing that TFEB plays an important role in regulating autophagy in β-cells (39,40).

A striking and unexpected finding was the paradoxical enhancement of autophagy following acute administration of nutrients despite simultaneous stimulation of mTORC1 and inhibition of its downstream targets TFEB and ULK1, suggesting that it is most likely independent of mTORC1-ULK1-TFEB signaling. We speculate that in β-cells, nutrient-derived signals may promote a rapid and general mobilization of membranes and granules, including autophagosomes. Additional studies are required to clarify the mechanisms for the initial stimulation of autophagy by nutrients and assess whether a similar biphasic response is observed in other metabolic or secretory tissues.

Importantly, we found that the mTORC1-autophagy signaling regulates insulin secretion. Inhibiting autophagy by heterozygous Atg7 KO amplified GSIS. Short-term treatment with CQ augmented both fasting and stimulated insulin secretion. Furthermore, mTORC1 deficiency impaired insulin secretion in an autophagy-dependent manner, with partial inhibition of autophagy being sufficient to rescue secretion. Islet proinsulin and insulin content was reduced in βRaptor KO mice; inhibiting autophagy did not rescue hormone stores, suggesting the restoration of insulin secretion is not due to enhancement of insulin synthesis.

How autophagy regulates insulin secretion is unknown. Based on studies in a β-cell line, we suggested that secretory granules are degraded by autophagy (10). However, electron microscopy of βRaptor KO islets showed that secretory granule localization within autophagosomes and lysosomes is relatively rare (data not shown), suggesting that in primary β-cells, stimulation of autophagy does not affect exocytosis through major effects on the secretory granule pool. The possibility that autophagy selectively affects a small, readily releasable pool of secretory granules cannot be excluded. Inhibition of autophagy by Atg7 KO in mice fed a high-fat diet enhanced peroxisome function and modulated lipid composition with increased ether lipids and prevented n-3 polyunsaturated fatty acid depletion; this was associated with a transient increase in insulin secretion (41). Interestingly, we found that the number of peroxisomes was modulated by mTORC1 deficiency and autophagy. Ongoing unbiased metabolomic, lipidomic, and proteomic analyses will shed light on how autophagy modulates insulin secretion.

Autophagy-restrained insulin secretion may have important implications for obesity and diabetes. Elevated serum BCAA along with hyperinsulinemia characterizes obesity-associated insulin resistance (42). It is possible that increased β-cell exposure to BCAA inhibits autophagy, which in turn increases basal insulin secretion. We have previously shown that in neonatal diabetes induced by β-cell endoplasmic reticulum stress, inhibition of mTORC1 precedes the development of diabetes, resulting in marked impairment of β-cell growth and function (28). Dysregulation of mTORC1 in response to endoplasmic reticulum stress might impair insulin secretion in part through sustained stimulation of autophagy.

Finally, our findings may have therapeutic implications. CQ and hydroxychloroquine are FDA-approved drugs. They alkalinize endosomal organelles and inhibit lysosomal cargo degradation. There is a consistent literature, including a randomized placebo-controlled trial on uncontrolled T2D, showing hydroxychloroquine treatment is associated with improved metabolism by reducing insulin resistance and improving β-cell function (43–45). Therefore, moderate inhibition of autophagy might become a novel therapeutic approach for the treatment of type 2 diabetes.

In summary, we show here that in adult β-cells, mTORC1 is a central and dynamic sensor responding to rapid changes in nutrient availability. mTORC1 inhibits autophagy via ULK1 and TFEB, with subsequent inhibition of the genetic program regulating autophagy and lysosome biogenesis and activity. As illustrated in Fig. 8E, we propose that the mTORC1-autophagy crosstalk forms a central node that modulates insulin secretion in accordance with nutrient availability over the diurnal variations. When prevented from oscillating and stimulated continuously by overnutrition, the ensuing inhibition of autophagy and increased insulin secretion constitute a plausible pathogenic mechanism for the insulin resistance of obesity and the β-cell failure of type 2 diabetes.

Funding. This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, grant R01-DK073716 to E.B.-M. and G.L. and by Israel Science Foundation grant ISF-398/20 and German-Israeli Foundation for Scientific Research and Development grant I-429-201.2/2017 to G.L.

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

Author Contributions. T.I., Y.R., P.G., R.A.L., J.P.W.-d.-C., M.B.-R., R.Y.-S., L.K., S.T.-B., G.H., N.I., and O.A. performed experiments and analyzed results. T.I., Y.R., E.C., B.T., E.B.-M., and G.L. designed experiments. T.I. and G.L. wrote the paper. E.C., E.B.-M., and G.L. conceived the study. G.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.