Dysregulation of OGT and mTORC1 in models of β cell dysfunction in mice and humans. β Cell mass and function are affected by both genetic and environmental factors (e.g., fetal-origins predisposition of T2D due to β cell dysfunction; ref. 8) in part by reducing key proteins such as Pdx1 and mTOR (9). For example, the offspring of dams fed a low-protein diet (LPD) throughout pregnancy (a model of maternal malnutrition) (9) or individuals born during the Dutch famine (ca. 1944–1945) have increased susceptibility to T2D (10). Islets of LPD offspring exhibited β cell dysfunction and mechanistically, in part, through microRNAs alterations targeting mTOR(9). Here, we report that OGT and its activity, O-GlcNAcylation (assessed by using RL2 ]pan-O-GlcNAc] antibody), are also reduced in LPD islets (Figure 1, A–C), in addition to loss of mTORC1 signaling (Figure 1D). OGA was reduced, and no changes in total S6 were observed between control diet (CtrlD) and LPD (Supplemental Figure 1, A–C; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.183033DS1). A downstream arm of the mTORC1 pathway is the phosphorylation and inactivation of eIF4EBP to increase protein translation. Deletion of mTOR target eIF4EBP2 (eIF4EBP2-KO) protected loss of β cell area in LPD offspring (Figure 1E), showing mechanistically the importance of mTORC1-eIF4EBP2 in maintaining β cell health.

Aberrant mTORC1 and O-GlcNAcylation signaling in models of islet stress. (A–D) OGT, pS6 S240, S6, and O-GlcNAcylation (RL2 antibody) levels from WT mouse islets from the offspring born to dam fed control diet (CtrlD) or low protein diet during pregnancy (LPD) (n = 4–7). (E) β Cell area to pancreas area ratio from control and eIF4EBP2-KO (BP2-KO) neonatal mice born to either dam fed CtrlD or low protein diet (n = 8–11). (F–I) Representative immunoblot and analysis of pS6 S240, S6, RL2 (pan-O-GlcNAc), and LC3 from islets from patients who are lean or have obesity (n = 4–5). (J–O) Islets from patients who are lean or have obesity were treated with low or high glucose (L, 3 mM; H, 16.7 mM), amino acid (A), and/or palmitate/BSA (P; 16:4; 100 μM; palmitate [Palm]) for 6 hours and assessed in immunoblot for pS6 S240, S6 and RL2 (pan-O-GlcNAc) (n = 3–4 donors). One of 4 obese donor islets did not show any RL2 signaling. Statistical analysis by Student’s t test (B–E, and G–I) and 1-way ANOVA (K, L, N, and O). *P ≤ 0.05, ***P < 0.001.

Obesity is a risk factor for T2D and is associated with β cell dysfunction. In islets from patients who are lean or have obesity, we analyzed the baseline protein levels of OGT and mTORC1, along with their activity. We previously reported that donors with obesity exhibited lower O-GlcNAcylation (11). In this new cohort of human islet samples, we show that the reduction O-GlcNAcylation seen in islets from patients with obesity was associated with lower mTORC1 activity, measured by the phosphorylation of the downstream target, S6 at Ser240 (Figure 1, F–H). We observed comparable values in the total protein levels of OGT, OGA, and mTOR (Supplemental Figure 1, D–G) among these islets. Total S6 protein was, however, reduced significantly in islets from patients with obesity versus patients who are lean (Supplemental Figure 1H). During the early phase of obesity in mice (6 weeks in HFD treatment), O-GlcNAcylation increases. However, following sustained obesity (18 weeks after HFD), OGT protein level decreases below normal levels (11). mTORC1 regulates a myriad of cellular signaling pathways, including inhibition of autophagy. With lower mTORC1 activity, we observed higher autophagy, show by increased LC3, a marker of autophagy (lipidation in response to autophagosome maturation) (Figure 1, F and I). Next, we assessed the human donor islets’ ability to engage nutrient sensor protein activities, in response to increased nutrient signals from glucose, amino acid, and lipids. In lean donor islets, we observed a 1.3-fold increase in mTORC1 activity with amino acid and a 6-fold increase with glucose stimulation (Figure 1, J and L). This change in phospho-S6 was without any changes to the total level of S6 protein (Supplemental Figure 1I). Concurrent analysis of RL2 in the same samples revealed a 1.4-fold increase in O-GlcNAcylation with amino acid and a 1.7-fold increase with glucose treatment (Figure 1, J and K). However, in obese donor islets, amino acids did not alter mTORC1 activity, but a 1.3-fold increase was observed in response to glucose (Figure 1, M and O). No alterations to O-GlcNAcylation levels in response to amino acids and glucose in obese donor islets (Figure 1, M and N). Palmitate alone had no effect on mTORC1 activity or O-GlcNAcylation level in either lean or obese donor islets. Altogether, these preliminary data demonstrate a positive correlation between OGT and mTORC1 activity in islets from individuals who are lean in response to environmental stressors, whereas this response is altered in islets from patients with obesity.

OGT positively regulates mTORC1 activity in primary β cells. To delineate the potential relationship between OGT and mTOR, we performed RNA-Seq and proteomics on islets from β cell–specific OGT-deficient mice (βOGT-KO) (7, 11). Both transcriptomics and proteomics analysis from primary islets of βOGT-KO show TSC2, a negative regulator of mTORC1, as an upstream regulator of differentially expressed genes (DEGs) in βOGT-KO islets (Supplemental Figure 2A). DEGs of these studies were previously reported by Lockridge et al. and Mohan et al. (7, 11). Here we observed an increase in Tsc2 mRNA transcript as well as protein levels in isolated islets from βOGT-KO mice (Figure 2, A–C). As expected, increased TSC2 protein levels reduce mTORC1 activity in βOGT-KO islets (Figure 2, B and D). If OGT regulates mTORC1 positively via O-GlcNAc, then β cells with OGA deletion (hyper–O-GlcNAcylation) should show increased mTORC1 activity. Indeed, islets from the βOGA-KO show decreased Tsc2 mRNA or protein levels (Figure 2, E–G) and increased levels of phosphorylated S6 at Ser240, indicating increased mTORC1 activity (Figure 2, F and H). We tested this relationship in islets from patients who are lean, and treatment with Thiamet-G (TMG), an OGA inhibitor, led to increased intensity of RL2 and mTORC1 activity (Figure 2I). These data suggest that the OGT/mTOR relationship is intact in both murine and human islets. Next, we investigated whether mTORC1 may reciprocally modulate O-GlcNAcylation. However, increasing mTORC1 signaling did not alter O-GlcNAcylation levels (Supplemental Figure 2B). Conversely, partial Raptor (a critical component of mTORC1) deletion did not change O-GlcNAcylation in β cells (Supplemental Figure 2C), suggesting that modulation of mTORC1 levels does not alter O-GlcNAcylation levels, and thus, OGT may act upstream of mTORC1 in β cells.

OGT modulates mTORC1 signaling in pancreatic β cells. (A) qPCR analysis of Tsc2 mRNA from male and female, control and βOGT-KO islets; normalized to 36B4 (n = 3-4). (B–D) Representative immunoblot and analysis of TSC2, pS6 S240, and S6 from control and βOGT-KO islets (n = 3-4). (E) Quantatitative PCR analysis of Tsc2 mRNA from male control and βOGA-KO islets; normalized to 36B4 (n = 4). (F–H) Representative immunoblot and analysis of TSC2 and pS6 S240 from control and βOGA-KO islets (n = 3–5). (I) Representative immunofluorescence imaging and quantification of insulin, pS6 S240, and RL2 from histogel embedded human donor (lean) islets treated with vehicle/control or TMG (30 μM) for 12 hours. Magnification, ×600. Scale bar: 100 μm. Statistical analysis by Student’s t test. *P ≤ 0.05, **P < 0.01, ***P < 0.001.

OGT negatively regulates autophagy in β cells. mTORC1 acts on various signaling nodes to modulate β cell growth and function, including inhibition of autophagy. However, the role of OGT in autophagy has not been explored in β cells. Analysis of ultraresolution structure by transmission electron microscopy (TEM) on βOGT-KO β cells revealed an increased number of phagophore-like, double-membraned structures (Figure 3A). We tested whether the autophagic process is dysregulated in an O-GlcNAc–dependent manner. Immunofluorescence imaging of OGT deficient β cells revealed increased LC3 puncta (Figure 3, B and C). We corroborated these data with an immunoblot of control and βOGT-KO islets, showing increased lipidation of LC3, indicative of autophagosome formation. A decrease in p62 expression also indicated increased protein turnover via autophagy (Figure 3, D–F). Using a pharmacological inhibitor of OGT, OSMI-1, we detected increased LC3 lipidation, which is further augmented in the presence of chloroquine, an inhibitor of autophagy flux (Figure 3G). These data suggest chloroquine can still block autophagy flux in cells with reduced OSMI-1. Therefore, with reduced OGT, autophagic flux was increased.

O-GlcNAcylation regulates autophagy in pancreatic β cells. (A) Transmission electron microscopy (TEM) image of control and βOGT-KO β cells. Red arrows point to phagophore like structures. Magnification, ×25,000 (small image, right) and ×10,000 (large images, left). Scale bar: 600 nm. (B and C) Representative immunofluorescence imaging of insulin, LC3 from control, and βOGT-KO pancreas. Magnification, ×400. Scale bar: 25 μm. LC3 puncta quantitation in insulin+ cell populations (n = 3 mice). (D–F) Representative immunoblot and analysis of LC3-II, LC3-I, and p62 from control and βOGT-KO islets (n = 4–5). (G) Representative immunoblot and analysis of LC3-II and LC3-I from INS-1 cells treated with vehicle or OSMI-1 (OGT inhibitor; 50 μM) for 24 hours. Cells treated with chloroquine (CQ; 10 μM) were treated 2 hours prior to collection (n = 3). (H–J) Representative immunoblot and analysis of LC3-II, LC3-I, and p62 from control and βOGA-KO islets (n = 4–5). Statistical analysis by Student’s t test (C, E, F, I, and J) or 1-way ANOVA (G). *P ≤ 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To strengthen the evidence that inhibiting O-GlcNAcylation promotes autophagy, we assessed LC3-II levels under conditions of increased O-GlcNAcylation. In βOGA-KO islets, where O-GlcNAcylation was elevated, we observed reduced LC3 lipidation and increased p62 expression (Figure 3, H–J), indicating that autophagy was inhibited in these islets. Overall, our data suggest that O-GlcNAcylation negatively regulates autophagy in β cells, as OGT deletion increased autophagy while OGA deletion decreased it.

mTORC1 restoration ameliorates hyperglycemia and glucose intolerance in βOGT-KO mice. We tested whether restoration of upstream mTORC1 signaling is sufficient to rescue the diabetes observed in the βOGT-KO mice. We deleted TSC2 in the background of βOGT-KO (βOGT/TSC2-KO), and βTSC2-KO was also used as a control in all the in vivo studies. We show efficient deletion of TSC2 and restoration of mTORC1 activity in the βOGT/TSC2-KO islets (Figure 4, A–C). By rescuing mTORC1 activity in βOGT-KO mice, we delayed the development of hyperglycemia in these mice (Figure 4D). While we observed a hyperglycemia-associated decrease in body weight in βOGT-KO after 80 days of age, we observed normalization of body weight in the βOGT/TSC2-KO (Supplemental Figure 3A). Next, we tested their glucose tolerance, insulin tolerance, and glucose-stimulated insulin secretion (GSIS) prior to when the βOGT-KO mice developed hyperglycemia and reduced body weight. At 7–8 weeks of age, we show that βOGT/TSC2-KO mice had improved glucose tolerance in both males and females (Figure 4, E and F). We observed normal insulin tolerance between βOGT-KO, βOGT/TSC2-KO, and respective controls (Figure 4, G and H). An in vivo GSIS assay revealed full rescue in insulin secretion in βOGT/TSC2-KO (Figure 4I and Supplemental Figure 3B), prompting us to assess β cell mass and islet insulin secretion in these models.

mTORC1 restoration ameliorates β cell dysfunction in βOGT-KO mice. (A–C) Representative immunoblot and analysis of TSC2, pS6 S240, and S6 from control, βOGT-KO, βOGT/TSC2-KO, and βTSC2-KO islets (n = 4–5). (D–H) Nonfasted blood glucose over time, i.p. glucose tolerance (glucose 2 g/kg i.p.), and insulin tolerance test (insulin 0.75U/kg i.p.) from male and female control, βOGT-KO, βOGT/TSC2-KO, and βTSC2-KO mice (n = 3–8). (I) In vivo GSIS assay from male control, βOGT-KO, and βOGT/TSC2-KO mice (n = 4–7). *P ≤ 0.05, Ctrl versus βOGT-KO or βOGT/TSC2-KO, **P < 0.01, ***P < 0.001, ****P < 0.0001. Statistical analysis by 1-way (B and C, AUC for E–H) and 2-way ANOVA (D and I).

mTORC1 preferentially affects β cell mass, but not insulin biosynthesis, without OGT. mTORC1 signaling is a critical regulator for both β cell mass and function. To assess the contributing factors to glucose tolerance and in vivo insulin secretion observed in βOGT/TSC2-KO mice, we assessed islet insulin secretion and β cell mass ex vivo. First, we found that islet insulin content deficit is not rescued in βOGT/TSC2-KO, compared with βOGT-KO islets (Figure 5A and Supplemental Figure 4A), suggesting defects in insulin biosynthesis. Also, proinsulin-to-insulin ratio and carboxypeptidase E (CPE) expression are significantly increased and reduced, respectively, in βOGT/TSC2-KO islets compared with βOGT-KO islets (Figure 5B and Supplemental Figure 4, B and C), suggesting that the insulin processing in OGT-deficient β cells is not rescued by increasing mTORC1 signaling. In addition to insulin content, glucose-stimulus coupling mechanisms such as glucose metabolism and ATP synthesis from the mitochondria are important factors promoting insulin secretion, and mTORC1 plays an important role in mitochondrial biogenesis and function. Despite rescuing mTORC1 signaling, we observed defects in mitochondrial response to glucose and ATP-linked respiration in both βOGT/TSC2-KO and βOGT-KO islets (Supplemental Figure 4, D–I). Additionally, the extracellular acidification rate, a proxy for glycolysis, was reduced in βOGT/TSC2-KO islets (Supplemental Figure 4, J and K). This was consistent with defects in islet GSIS, still observed in βOGT/TSC2-KO islets (Figure 5C). Next, we assessed β cell mass in βOGT/TSC2-KO mice. Though no differences in pancreas mass were detected among genotypes (Supplemental Figure 5A), we observed normalized β cell mass in βOGT/TSC2-KO mice (Figure 6A and Supplemental Figure 5B), suggesting that improved glucose tolerance is largely due to rescue in the β cell abundance. mTORC1 signaling regulates biomass through modulation of cell proliferation and apoptosis. Mechanistically, while we observed no rescue in β cell apoptosis (TUNEL), we observed an increase in β cell proliferation via Ki67 staining in βOGT/TSC2-KO mice (Figure 6, B–D), suggesting that mTORC1 can drive β cell proliferation and set β cell mass in the absence of O-GlcNAcylation.

mTORC1 is insufficient modulate insulin biosynthesis in the absence of O-GlcNAcylation. (A–C) Islet insulin content, proinsulin-to-insulin content ratio, and in vitro islet GSIS assay from control, βOGT-KO, and βOGT/TSC2-KO islets (n = 5–8). Statistical analysis by 1-way (A and B) and 2-way ANOVA (C). *P < 0.05, ***P < 0.001, ****P < 0.0001, control or βOGT-KO or βOGT/TSC2-KO. ##P ≤ 0.01, ###P < 0.001, LG vs HG.

Pancreatic β cell proliferation is primarily driven by mTORC1, independently of O-GlcNAcylation. (A–D) Ex vivo β cell mass and immunofluorescence-based analysis of apoptosis (TUNEL) and proliferation (Ki67) of insulin+ β cells from control, βOGT-KO, and βOGT/TSC2-KO pancreas (n = 4–6). Scale bar: 25 um. Statistical analysis by 1-way ANOVA. Total original magnification, ×200. *P ≤ 0.05, ***P < 0.001. (E) Proposed model. OGT positively and negatively modulates mTORC1 and autophagy, respectively, in pancreatic β cells. Mechanistically, OGT regulate apoptosis and proliferation to maintain proper β cell mass, and it modulates insulin biosynthesis and processing to control insulin content and secretion. While mTORC1 can gate apoptosis and positively signal insulin biosynthesis and processing, without OGT, it is unable to do so, suggesting that OGT is a dominant signal for these pathways in β cells (light blue line). Conversely, while OGT has been shown to modulate cell proliferation, mTORC1 can induce proliferation in the absence of O-GlcNAcylation (light blue line). Future direction includes studying the downstream effect of altered autophagy (e.g., vesicophagy, mitophagy) that may regulate β cell function. Statistical analysis by Student’s t test and 1-way ANOVA. Significance P ≤ 0.05.

Overlapping and distinct pathways regulated by mTORC1 and OGT. As key nutrient-sensing proteins, OGT and mTORC1 regulate multiple signaling nodes to orchestrate β cell cellular function and health. To investigate the differential signaling between these 2 major pathways, we performed a phospho-protein antibody array to assess and understand changes in the phosphorylation signaling cascade in the islets of βOGT-KO, βTSC2-KO, and βOGT/TSC2-KO. This approach helps delineate the signaling relationship between OGT and mTORC1, particularly since some mTORC1 targets are themselves kinases. Using a thresholding fold change of 20% to control, we identified 67, 121, and 103 differential phospho-proteins in βOGT-KO, βOGT/TSC2-KO, and βTSC2-KO, respectively. KEGG (https://www.genome.jp/kegg/) and GO-term (https://geneontology.org/) analyses of these differential phospho-protein expressions showed converging pathways in cell cycle, apoptosis, MAPK, and Akt signaling pathways (Supplemental Figure 6A). To identify common or differentially regulated signaling, we overlapped these changes between βOGT-KO, βOGT/TSC2-KO, and βTSC2-KO (Supplemental Figure 6B). The 12 overlaps with the same directionality of change between βOGT-KO and βOGT/TSC2-KO represent phosphorylation changes that are OGT dependent. Seventeen of 26 overlaps with the same directionality between βOGT/TSC2-KO and βTSC2-KO are identified as phosphorylation changes that are mTORC1 dependent. We focused on the 9 phospho-proteins that are commonly altered in all 3 genotypes, and of the 9, we identified p-MKK4 (S80), p-calmodulin (T79), and p-4E-BP (T45) as most substantially changed (Supplemental Figure 6C). Consistent with the proposed model, phosphorylation of 4EBP, a known downstream target of mTORC1, is reduced in βOGT-KO islets but is increased in βOGT/TSC2-KO and βTSC2-KO islets. p-MKK4 S80 is reduced in both βOGT-KO and βOGT/TSC2-KO and increased in βTSC2-KO islets, suggesting that while this pathway can be mTORC1 regulated, it is critically dependent on OGT signaling. Calmodulin phosphorylation at T79 is increased in βOGT-KO islets with relative reduction observed in βOGT/TSC2-KO islets, suggesting a potential intersection of OGT and mTORC1 in this pathway. These data show that OGT and mTORC1 have both overlapping and distinct pathways in regulating molecular signaling cascades to orchestrate β cell function.