Raptor regulates functional maturation of murine beta cells

Diabetes is associated with beta cell mass loss and islet dysfunctions. mTORC1 regulates beta cell survival, proliferation and function in physiological and pathological conditions, such as pregnancy and pancreatectomy. Here we show that deletion of Raptor, which is an essential component of mTORC1, in insulin-expressing cells promotes hypoinsulinemia and glucose intolerance. Raptor-deficient beta cells display reduced glucose responsiveness and exhibit a glucose metabolic profile resembling fetal beta cells. Knockout islets have decreased expression of key factors of functional maturation and upregulation of neonatal markers and beta cell disallowed genes, resulting in loss of functional maturity. Mechanistically, Raptor-deficient beta cells show reduced expression of DNA-methyltransferase 3a and altered patterns of DNA methylation at loci that are involved in the repression of disallowed genes. The present findings highlight a novel role of mTORC1 as a core mechanism governing postnatal beta cell maturation and physiologic beta cell mass during adulthood.

T ype 2 diabetes is characterized by hyperglycaemia due to insulin resistance in peripheral tissues and a defect of insulin secretion resulting from continuous loss of functional beta cell mass 1 . In the adult pancreas, functional beta cell mass is not only determined by the number of beta cells, but also by their functional maturation status, including the ability to produce and secrete mature insulin in response to elevated blood glucose level 2,3 . Dedifferentiation of mature beta cell to a progenitor-like state under metabolic stress is one of the mechanisms of beta cell failure in type 2 diabetes 4 . Several transcription factors (Pdx1, Nkx6.1, NeuroD1, MafA, Islet-1) 5-9 , cellular signals (thyroid hormone, CTGF) 10,11 and nuclear receptor ERRg (ref. 12) have been reported to be crucial for maintaining beta cell identity and functional maturation. Yet, it remains unclear which signalling pathway determines beta cell functional maturation and mass expansion after birth. Fetal beta cells are poorly responsive to glucose 13 , but undergo a dramatic process of functional maturation in the early postnatal period, between day 2-8 (ref. 14), characterized by acquisition of the ability to secrete insulin in response to glucose 15 . mTOR signalling is composed of a number of highly conserved Ser/Thr protein kinases, functioning in the form of at least two large protein complexes, mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2) 16 . mTORC1 consists of RAPTOR (regulatory associated protein of mTOR), mLST8, PRAS40, DEPTOR and mTOR, which is sensitive to rapamycin 17 . mTORC1 pathway is recognized as a key factor to integrate signals from nutrients, growth factors and hormones, and controls cellular processes, including protein synthesis, ribosome biogenesis and lipogenesis by modulating eukaryotic translation initiation factor 4E binding protein 1 (4E-BP1) and ribosomal protein S6 kinase 1 (S6K1) (ref. 18). Multiple studies have clarified the role of mTORC1 signalling in beta cell function and mass. First, in vitro studies using rapamycin demonstrated that mTORC1 regulates beta cell function and survival in beta cell lines and murine/human islets 19 . Second, we and others have reported that rapamycin in vivo prevented beta cell adaptation to challenges, such as pregnancy 20 and 60% partialpancreatectomy 21 . Third, stimulation of mTORC1 activity in beta cells by overexpression of Rheb (ref. 22) or disruption of Tsc1/Tsc2 (refs 23,24) increased beta cell mass and improved glucose tolerance. Fourth, global S6k1 deficient and Rps6 knockout mice exhibited hypoinsulinemia and glucose intolerance, with diminished beta cell size [25][26][27] . Observations from the genetic models targeting component of the mTORC1 pathway suggest that mTORC1 is a key signal to regulate beta cell mass; nevertheless, its effect on beta cell proliferation and apoptosis remains controversial. Recent work using conditional Raptor knockout mice demonstrated tissue-specific mTORC1 functions in controlling whole-body metabolism [28][29][30] . Currently, the role of Raptor in beta cells remains unknown.
In the present study, we use beta cell specific Raptor knockout mice and report a direct link between mTORC1 signalling and beta cell functional maturation, which is an important and novel field of beta cell research. There exist multiple layers of regulation, including protein/insulin synthesis, translational capacity, cell size, mitochondria metabolism and DNA methylation. Raptorinduced beta cell functional immaturity precedes the onset of hyperglycaemia, which leads to beta cell failure.

Results
Impaired glucose tolerance and insulin secretion in bRapKO mice. In preliminary studies, we examined the expression of mTORC1 target PS6 in pancreas sections during the transition from early neonatal life to post-weaning (Fig. 1a). IHC analysis of pancreatic sections showed that PS6 protein could be detected in some insulin þ cells from P1 pups, but strongly increased in nearly all insulin þ cells between P4 and P8 (Fig. 1a). By P11 just after the end of the maturation window, fewer beta cells showed staining for PS6 (Fig. 1a). Thus, expression of PS6 coincides with the transition from 'immature' to 'mature' glucose-stimulated insulin secretion (GSIS) response between 2 and 8 postnatal days (ref. 14), raising the possibility that the mTORC1 pathway plays a role in beta cell functional maturation.
To investigate the role of mTORC1 in mature beta cells, we generated mice lacking the key mTORC1 component Raptor specifically in beta cells (bRapKO). Successful knockout of Raptor was confirmed by western blot: RAPTOR was selectively absent in islets from 8-week-old bRapKO mice ( Supplementary Fig. 1a). In addition, the mutant islets showed reduced phosphorylation of mTORC1 targets 4E-BP1 and PS6 (Ser240/244) (Fig. 1b). 4E-BP1 dephosphorylation was reflected in the shift from the highly phosphorylated g-band to the nonphosphorylated a-band and an intermediate b-band (Fig. 1b). Thus, bRapKO mice are specifically defective in mTORC1 signalling in beta cells.
Raptor heterozygous mutant mice (bRapHET) exhibited similar weight, blood glucose levels, plasma insulin concentrations and survival rates as their littermate controls carrying the floxed allele of Raptor (WT) ( Fig. 1c-g,i). bRapKO mice were born in the expected Mendelian ratio and did not differ in body weight from WT (Fig. 1c). However, the mutant mice started to display elevated random-fed glucose and 6-h fasted glucose level at the age of 4 weeks, and their glycemic control worsened with age (Fig. 1d,e). This rise was associated with significantly lower 6-h fasted plasma insulin levels in mutant animals, as early as 8 weeks after birth (Fig. 1f). We next measured blood glucose and plasma insulin levels after intraperitoneal glucose injection in 8-week-old bRapKO and WT: there was no significant difference in fasting glucose concentration, but a dramatic increase in glycaemia was observed in bRapKO mice following glucose challenge (Fig. 1g). As expected, these mutant mice exhibited lower basal insulin concentrations and mounted a poor insulin response when challenged with glucose ( Fig. 1h). bRapKO mice showed a significant decrease in body weight at the age of 16 weeks compared with their age-matched littermates (Fig. 1c), and eventually died between 14 and 36 weeks after birth (mean life span 18 weeks, Fig. 1i) with severe and sustained hyperglycaemia (430 mmol l À 1 ). Female bRapKO mice also became diabetic, but the phenotype developed more slowly and was less severe than in males ( Supplementary Fig. 1b,c).
Decreased beta cell mass in bRapKO mice. To understand if diabetes in bRapKO mice was caused by reduced beta cell mass, we examined islets morphology in 8-week-old WT and mutant mice. The 8-week-old bRapKO mice did not display disrupted islet structure, and their alpha cells were still residing at the periphery (Fig. 2a). Notably, the adjusted beta cell mass of bRapKO was 49.8% lower than WT (Fig. 2c). It is known that mTORC1 regulates beta cell growth 17 . To evaluate beta cell size, we used insulin staining to mark beta cells and b-catenin staining to highlight cell boundaries (Fig. 2b): a 27% reduction in beta cell size was observed in bRapKO mice (Fig. 2d). We detected a threefold increase in the percentage of apoptotic Tunel þ insulin þ cells (Fig. 2e), with comparable proportions of Ki67 þ insulin þ cells in mutant islets (Fig. 2f). These changes resulted in a significant decrease in the number of insulin þ cells per islet (Fig. 2g). Therefore, ablation of Raptor affected beta cell mass and number, possibly due to defects in beta cell growth and survival. Moreover, a striking reduction in pancreatic insulin content (Fig. 2h) in bRapKO mice further demonstrated that loss of functional beta cell mass was attributed to the poor glycemic control. We also checked the pancreata from 16-week-old bRapKO mice with prolonged severe hyperglycaemia (6-h fasted glucose, 23.8±1.4 mmol l À 1 ). In this case, we found a change in alpha cell distribution, with some located in the centre of islets ( Supplementary Fig. 2).
Gene-expression profile of bRapKO islets. To understand the global molecular basis of the phenotype caused by the loss of Raptor in beta cells, we isolated pancreatic islets from 8-week-old bRapKO and WT mice. At this time point, bRapKO mice did not show weight loss and their overnight fasting glucose was below 8 mmol l À 1 . We compared gene-expression profile by microarray of islets between the two groups and detected significant differences in the expression levels of 1,150 genes (fold change41.5), among which 786 were increased and 364 were decreased. To define the cellular processes affected in bRapKO islets, we performed Gene Ontology (GO) analysis of differentially expressed genes. This analysis revealed alterations of several biological processes that are crucial to beta cell function, that is, insulin secretion, cell growth and apoptotic process, carbohydrate metabolic process, ATP biosynthetic process and ion transport following Raptor ablation (Fig. 3a). We further used the ingenuity pathway analysis to investigate alterations of known pathways and listed the top 10 increased and decreased pathways in bRapKO islets (Fig. 3b). Metabolic pathways, calcium signalling and insulin secretion were among the top decreased cellular biological processes, while PI3K-AKT, RAS and MAPK signalling related to canonical mitogenic and survival pathways were preferentially increased in mutant islets (Fig. 3b).
Defects in GSIS and glucose metabolism in bRapKO islets. The observation that crucial pathways required for physiologic insulin secretion were preferentially decreased in bRapKO transcriptome led us to evaluate beta cell function ex vivo. In parallel to decreased pancreatic insulin content, a dramatic reduction in both insulin (58.9% of WT, Supplementary Fig. 3a) and protein content (71.8% of WT, Supplementary Fig. 3b) was observed in islets from 8-week-old bRapKO mice. By calculating insulin secretion as a percentage of total islet insulin content, we found that 8-week-old mutant islets secreted nearly 40% less of their total insulin than WT at 16.7 mmol l À 1 glucose, whereas they had comparable basal insulin secretion at 2.8 mM glucose (Fig. 3c). Impaired GSIS was also observed when insulin secretion was corrected by total DNA content ( Supplementary Fig. 3c). We then evaluated whether Raptor ablation also affects formation of insulin secretory vesicles. We analysed beta cell ultrastructure in 8-week-old bRapKO and WT using electron microscopy ( Fig. 3g) and found a significant decline in the overall number (Fig. 3h) and mean size of secretory vesicles (Fig. 3i). Moreover, the proportion of immature vesicles was significantly increased in Raptor-deficient beta cells (Fig. 3j). We conclude that loss of Raptor results in a defect in protein/insulin production, GSIS response and secretory vesicle formation/maturation. Interestingly, we did not observe differences in mRNA expression levels of Ins1, Ins2 and proinsulin/insulin convertases (Pcsk1, Pcsk2) in 8-week-old bRapKO and WT islets (Fig. 3d). It is known that mTORC1 regulates translational initiation, which occurs primarily at the level of translation of pre-transcribed Insulin mRNA 31 . Indeed, deletion of Raptor induced dephohsphorylation of 4E-BP1 and S6 (Fig. 1b) and reduced the abundance of ribosomal proteins, that is, RPL7 and RPL26 ( Supplementary Fig. 3e), indicating a diminished translational capacity. We then interrogated expression of genes implicated in insulin release, and found that most of them were not changed, including Kcnj11, Abcc8, Pclo, Rims2 (Fig. 3d). In contrast, genes encoding proteins participating in granule trafficking and membrane exocytosis such as Vamp2, Sytl4, Gjd2 were significantly decreased (Fig. 3d), as was the type 2 diabetesassociated gene Slc30a8, encoding zinc transporter-8 (ZnT8), which is important for insulin secretion 32 (Fig. 3d).
In Raptor-deficient islets, metabolic pathway was the top decreased biological process identified by microarray analysis (Fig. 3b). To determine whether the failure to respond to glucose occurs at the level of glucose transport across the membrane, we examined Glut2 levels in 8-week-old bRapKO mice and found a dramatic decrease in Glut2 mRNA expression and protein levels, as determined by RT-PCR (Fig. 3f) and immunostaining ( Supplementary Fig. 3f). Moreover, the key anaplerotic enzyme in mitochondria, pyruvate carboxylase (Pcx), as well as several other genes encoding subunits of electron transport chain complexes I-V including Ndufs2, Ndufb8, Cox6a2, Cox6c, Cox8a, Cox8c, ATP5a1, ATP5j2 ( Fig. 3f and Supplementary Table 1) were significantly decreased. Of interest, several glycolytic enzymes, including lactate dehydrogenase A (Ldha), aldolase B (AldoB) and hexokinase 1 (Hk1) that are minimally expressed in mature beta cells were significantly upregulated (Fig. 3f). We further     confirmed that LDHA protein level was 4.25-fold increased by western blot (Supplementary Fig. 3g). These characteristics of enhanced glycolytic gene expression in bRapKO islets resemble those of poorly responsive neonatal islets 33 . Next, we measured intracellular ATP content and found that ATP level in 8-week-old mutant islets was decreased by 65% compared to WT (Fig. 3e). mTORC1 has been suggested to control mitochondrial oxidative function via its target, proliferator-activated receptor gamma coactivator 1 (PGC1a) 34 . Accordingly, we also detected a strong decrease in PGC1a protein abundance in the mutants ( Supplementary Fig. 3e).
Loss of beta cell functional maturity in bRapKO mice. Decreased expression of key metabolic genes Glut2 and Pcx, derepression of glycolytic gene Ldha, AldoB and Hk1, as well as reduced ATP production and GSIS, are properties shared by 8-week-old bRapKO islets and fetal beta cells [35][36][37][38] . Importantly, Raptor ablation significantly reduced Pdx1, Nkx6.1, NeuroD1 and MafA expression (Fig. 4a,b and Supplementary Fig. 3d), which are essential transcription factors for maintaining beta cell functional maturity [5][6][7][8] . Urocortin 3 (UCN3) is a peptide hormone expressed exclusively in mature beta cells and was recently identified as a molecular marker of functional beta cell maturation 14 . The level of Ucn3 mRNA decreased 450% in islets from 8-week-old bRapKO mice (Fig. 4b). Accordingly, double immunofluorescence labelling demonstrated strong Ucn3 staining in nearly all insulin þ cells in 8week-old WT, but remarkably decreased staining in mutant beta cells of age-matched animals (Fig. 4c). We found an increase in neonatal markers 36 delta-like 1 homologue (Dlk1) and MafB expression in 8-week-old Raptor-deficient islets, with comparable transcriptional level of Npy (Fig. 4b). Moreover, we found that 3.8% of cells were insulin þ MafB þ in 8-week-old bRapKO pancreas, but not in WT controls, as determined by morphometric analysis of immunostaining (Fig. 4d). Interestingly, expression of the endocrine progenitor marker Ngn3 doubled in mutant islets. (Fig. 4b).
Ectopic expression of genes that are disallowed in beta cells is also crucial to maintaining the mature phenotype of beta cells 39 . Pullen et al. 40 and Thorrez et al. 41 have identified 39 and 33 disallowed genes, respectively, that are downregulated in islets relative to all other mouse tissues studied (red circles in Supplementary Fig. 5a). Remarkably, we found 31 out of 65 genes in 8-week-old bRapKO islets were significantly upregulated based on microarray data (blue circle in Supplementary Fig. 5a and listed in Supplementary Fig. 5b). Next, we confirmed the upregulation of platelet-derived growth factor receptor a (Pdgfra), insulin-like growth factor binding protein 4 (Igfbp4), ornithine aminotransferase (Oat) and myosin light chain kinase (Mylk) in 8-week-old Raptor-deficient islets by RT-PCR (Fig. 4b,  Supplementary Fig 3d).
To exclude the possibility that the observed immature phenotype is merely caused by hyperglycaemia, we performed experiments on 2-week-old euglycemic bRapKO mice (Fig. 5a) and WT with comparable beta cell mass (Fig. 5b). These young bRapKO mice exhibited similar beta cell proliferation rates and number of insulin þ cells per islet, but a significant decreased beta cell size, insulin content and increased beta cell apoptosis ( Supplementary Fig. 4). We then performed immunostaining with antibodies against insulin, glucagon and MafA (or MafB) (Fig. 5e,f). Quantification revealed that the percentage of beta cells coexpressing MafA was decreased (Fig. 5c), while the proportion of insulin þ MafB þ cells was significantly increased (Fig. 5d) in mutant islets. Moreover, a significantly lower level of GLUT2 expression was found in mutant beta cells, as determined by immunostaining (Fig. 5g). Importantly, isolated islets from 2-week-old mutant mice displayed an 'immature' response to basal and stimulated glucose: they had elevated basal insulin secretion at 2.8 mM glucose, but reduced GSIS response under high glucose (16.7 mM) stimulation (Fig. 5h). Then we calculated their stimulation index (fold change in GSIS), and found that WT and mutant islets can be significantly distinguished (Fig. 5i). Thus, the changes in beta cells antedate the onset of diabetes and are likely a cause, rather than a result of it.
Altered DNA methylation and disallowed genes in bRapKO islets. DNA methylation has been proposed to drive functional maturation of pancreatic beta cells 42 . To evaluate the possibility that altered methylation is involved in mTORC1-dependent maturation, DNA immunoprecipitation sequencing (MeDIP-seq) was carried out on islets isolated from 8-week-old bRapKO and WT mice. We showed distribution frequencies in violin plots, and we could clearly find that median, 25% quantile and 75% quantile of methylation values (RPKM) in bRapKO mice were all lower than WT islets (Fig. 6a). A total of 22,046 differential methylation regions (DMRs) were significantly changed (fold-change41.5). Among them, 17,144 regions corresponding to 5,713 genes were hypo-methylated; while 4,902 regions corresponding to 1,783 genes were hyper-methylated. Decrease levels of DNA methylation have previously been associated with transcription upregulation 43 . We therefore sought to correlate changes in DNA methylation with changes in mRNA expression (fold-change41.5) of the same genes and identified 1,224 DMRs related to 459 genes whose changes were consistent with changes in gene expression (Fig. 6b). A total of 306 genes had an inverse relation, with 90% exhibiting a decrease in DNA methylation and an increased mRNA expression (Fig. 6b), including many pancreatic beta cell disallowed genes (Fig. 6i).
It has been reported that loss of Dnmt3a in pancreatic beta cells induced Ldha, Hk1, AldoB demethylation and upregulation, thus prevents developmental metabolic reprogramming, resulting in loss of GSIS 42 . We found a comparable Dnmt3a mRNA expression, but a significant decrease in DNMT3A protein abundance in Raptordeficient islets (Fig. 6c,d), as well as in intact islets and INS-1 cell in vitro treated with mTORC1 inhibitor rapamycin (Fig. 6e-g). As shown in Fig. 6h, 12 h treatment with protein translation inhibitor Cycloheximide (CHX) significantly reduced DNMT3A protein expression, to a similar extent as 48 h rapamycin did. Importantly, a further suppression of DNMT3A abundance did not occur when rapamycin treatment occurred in the presence of Cycloheximide (Fig. 6h). These results suggest that mTORC1 directly regulates DNMT3A levels via translational control.
Notably, our data showed that B40% of the upregulated disallowed genes in Raptor-deficient islets were found to be preferentially hypo-methylated (Fig. 6i). We confirmed that genes expressed in the neonatal period and beta cell disallowed genes (Hk1, Dlk1, Pdgfra, Oat and Mylk) were hypo-methylated (Fig. 6j) and transcriptionally activated (Fig. 4b) in Raptor-deficient islets. These results indicate that DNA methylation might participate in the control of repression of disallowed genes in bRapKO mice.

Discussion
The surprising finding of our work is that ablation of Raptor in adult beta cells resulted in a loss of functional maturation. Although it is well known that functional maturation is important for proper beta cell function, pathways that regulate this process are virtually unknown. By demonstrating a lag in functional maturation in Raptor-deficient beta cells, our data not only begin to provide insight into this complex process, but also raise the possibility of engaging the TOR pathway to induce beta cell maturation in vivo. Moreover, lack of functional maturation has consistently plagued efforts to develop glucose-responsive beta cell from ES or iPS cells. Our data provide a new mechanism that can be leveraged to induce maturation in vitro.
mTORC1 is known to be essential for beta cell function and mass 44 ; however, which aspects of the maintenance of functional beta cell mass require mTORC1 remains poorly understood. In the present study, we generated bRapKO mice in which mTORC1 signalling is inactivated in beta cells at the onset of insulin gene expression. These mice displayed defective insulin secretion in response to glucose and overt diabetes starting at the age of 4 weeks. Morphometry showed that the decreased GSIS capacity could be partly attributed to a significant reduction in beta cell mass, accompanied by diminished beta cell size and number. Another remarkable finding of the present study is that pancreas and islet from bRapKO mice contained less than half the normal amount of insulin. The total protein concentration in islets was also reduced by 30%. Impaired efficacy of protein synthesis is also found in Raptor-deficient skeletal muscle, resulting in muscle dystrophy 28 . It is known that mTORC1 regulates protein synthesis through translational mechanisms 31 . Indeed, the reduced insulin production of Raptor-deficient islets was mainly due to post-transcriptional regulation, as evidenced by limited translational initiation and ribosomal biogenesis. Importantly, as early as 14 days after birth, mutant mice showed defects in insulin production, beta cell growth and survival before the onset of diabetes, indicating that mTORC1 directly regulates these functional characteristics, eventually leading to a loss of functional beta cell mass.
Our RNA profiling analysis identified signature abnormalities of insulin secretion and metabolic pathways in bRapKO islets. Islets from bRapKO mice secreted similar amount of insulin as WT at 2.8 mM glucose, but their GSIS was significantly lower. In mature beta cells, glucose is transported via GLUT2 and generates ATP in mitochondria, thus causing closure of ATP-sensitive K þ channels, ultimately leading to Ca 2 þ influx and insulin exocytosis 45 . We showed that the reduction of GSIS capacity in Raptor-ablated islets was not only caused by beta cell loss, but might also due to loss of GLUT2 expression, reduced expression of key cataplerotic enzyme Pcx and several components of OxPhos, impaired cellular ATP production, and defects in densecore granule formation and maturity. Interestingly, mutant islets exhibited a metabolic profile resembling that of neonatal islets, in which glycolysis predominates and results in impaired GSIS 46 . Likewise, we found an induction of a number of glycolytic genes, including AldoB, Hk1 and Ldha. The ectopic expression of Hk1 and Ldha in beta cells has been reported to impair glucose  47,48 . In addition, aberrant expression of Hk1 in human beta cells has been associated with a form of congenital hyperinsulinemia 49 .
Evidence for a role of Raptor in functional maturation is manifold. First, in Raptor-deficient islets we found a significant decrease in NeuroD1 and MafA, loss of which has been reported to reverse beta cell maturation 7,8 . Second, Urocortin3 (Ucn3), a molecular marker for functional beta cell maturation 14 , was significantly decreased at mRNA levels in Raptor-deficient islets. Immunofluorescence staining further confirmed that Ucn3 was highly expressed in all adult beta cells in WT, whereas the signal intensity of this protein dramatically declined in insulin þ cells from bRapKO mice. Third, transcripts encoding Ngn3 and Dlk1, which are abundantly expressed in fetal and neonatal beta cells 36  of insulin þ cells maintained MafB expression, a feature of neonatal beta cells 50,52 . During embryonic stem cell differentiation, the transition from MafB to MafA is critical for beta cell functional maturation 52 . Notably, the decrease of insulin þ MafA þ cells and the increase of insulin þ MafB þ cells preceded the onset of diabetes, consistent with a causative role in hyperglycaemia, rather than a consequence thereof. Likewise, Glut2 was dramatically lowered in beta cells from euglycemic mutant animals. Moreover, P14 Raptor-deficient islets had a lower threshold for GSIS, secreting insulin in response to a lower glucose concentration than their age-matched WT islets, which is another feature of functional immaturity 14 . These progenitor-like beta cells showed a three-fold increase in apoptotic rate, implying that they underwent programmed cell death before achieving full maturation.
Equally important for normal beta cell function is the selective repression of a number of disallowed genes compared to other tissues 40,41 . The repression of these genes is critical in the context of maintaining mature beta cell phenotype, and expression levels of a number of these genes increase in animal models of type 2 diabetes [53][54][55] . Strikingly, our RNA profiling data showed that nearly 50% of the disallowed genes identified by Pullen et al. 40 and Thorrez et al. 41 were strongly induced in Raptor-deficient beta cells. Recent evidence suggests that DNA methylation directs beta cell functional maturation via modulating Hk1 and Ldha expression 42 . In bRapKO islets, we detected a dramatic decrease in DNMT3A protein abundance, with comparable Dnmt3a mRNA expression. Rapamycin treatment also reduced DNMT3A protein abundance in isolated islet and INS-1 cells, which is primarily due to translational modulation. MeDIP-seq analysis revealed that Raptor-ablated islets had a significant decline in the average methylation level aggregated over all covered CpG sites. A decrease level of DNA methylation has previously been associated with transcriptional activation 43 . Approximately 40% of disallowed genes that were preferentially upregulated in bRapKO islets displayed hypo-methylation. Several of these demethylated genes, such as Oat 56 , acyl-CoA thioesterase 7(Acot7) 57 and Mylk 58 have the potential to interfere with glucose metabolism and insulin secretion. Demethylation and upregulation of Pdgfra, Hk1 and Dlk1 in bRapKO islets are associated with the process of beta cell maturation. Recent work showed that the progressive downregulation of Pdgfra during islet maturation is responsible for the age-dependent decrease in the proliferation capacity of islet cells 59 , whereas upregulation of Pdgfra was observed in islets in human type 2 diabetes and in a rat model of obesity-induced diabetes 54,55 . Moreover, Hk1 is selectively excluded from beta cells and liver 40 and its repression allows beta cell to sense glucose concentrations in the physiological range (44 mM). The transient overexpression of Hk1 in Min6 cells 60 and primary beta cells 61 yields an immaturelike beta cell phenotype with enhanced glucose threshold for insulin secretion, increased basal glycolytic flux, elevated basal insulin release, and poor responsiveness to elevated glucose 14 . In the present study, we did not detect significant differences in the methylation levels of Ldha and AldoB in bRapKO islets, indicating that other regulatory mechanisms, such as microRNAs and histone modifications [62][63][64] , may be responsible for their induction. Currently, the function and the regulation of most of the disallowed genes are not fully understood. It is also not clear how much these deeply repressed genes contribute to the acquisition of properties required for beta cell functional maturation. Targeted in vivo rescue studies are needed to understand the extent and specific aspects of DNA methylation regulation that contribute to Raptor ablation-induced immature phenotype in the future. This work provides support for the evolving concept that mTORC1 signalling not only controls beta cell mass, but also is critical for achieving beta cell functional maturation (Fig. 7). Our findings show that Raptor coordinates increased oxidative metabolism and translational capacity to ensure that beta cells acquire the mature GSIS phenotype. Epigenetic regulation, including DNA methylation, might participate in derepression of disallowed gene, which play a role in mTORC1-induced immaturity. Modulation of mTORC1 activity, or its regulatory pathways in beta cells might serve as therapeutic targets to protect beta cell function and create functional mature beta cells suitable for transplantation.

Methods
Animal. We used Cre/loxP system to conditionally delete Raptor in pancreatic beta cells. Beta cell specific Raptor knockout mice (bRapKO) were generated by crossing Raptor flox/flox mice (purchased from the Jackson Laboratory, C57BL/6J) with mice expressing Cre recombinase driven by the rat insulin promoter (RIP-Cre, mixed C57BL/6J:129/SvJ), as described before 29,65 . bRapKO Metabolic studies. Weight and blood glucose of bRapKO, bRapHET and WT were monitored biweekly from 2 to 18 weeks. Fasting plasma insulin was measured at the age of 4, 8, 12 and 16 weeks. Intraperitoneal glucose tolerance test (IPGTT) and GSIS in vivo were performed on 8-week-old mice. Mice were injected intraperitoneally with 2 g dextrose/kg body weight after an overnight fasting. Blood glucose concentrations were measured at various time points (0, 15, 30, 60, 120 min) by glucometers (Accu-Chek, Roche, Mannheim, Germany). Blood samples were taken at 0, 15, 30 min after dextrose injection, and serum insulin concentrations were determined by using ELISA assay (Mouse Ultrasensitive Insulin ELISA kit, Alpco) 21 .
Islet isolation and ex vivo experiments. Pancreatic islets were isolated, as described before 66 . Briefly, mice were injected with Collagenase P (Roche) solution into the common hepatic bile duct to perfuse the pancreas. Pancreas were then removed and digested at 37°C. Islets were handpicked after several purification steps under a microscope. The ATP content of isolated islets was measured using an assay kit (FLASC; Sigma) and expressed as relative light units per islet equivalent using a luminometer (Promega, Madison, WI).
For ex vivo GSIS, isolated islets were recovered in 1640 RPMI supplemented with 10% serum, 11 mM glucose (islets from 8-week-old mice), or 2.8 mM glucose (islets from 2-week-old mice) for 4-h, followed by incubation with Krebs Ringer Bicarbonate HEPES solution (KRB) containing 2.8 mM glucose for 1-h. A total of 30 islets of similar size were randomly chosen and incubated at 2.8 or 16.7 mM glucose for 1-h. At the end of incubation, the insulin concentrations of the supernatant and insulin content were calculated. Islet insulin content was measured by insulin ELISA assay (Mouse Insulin ELISA kit, Alpco) and normalized by total DNA extracted using DNeasy Micro kit (Qiagen, GmBH, Germany).
After passaging, cells were synchronized in a medium containing 11.1 mmol l À 1 glucose with or without 25 nM rapamycin (Cell Signaling Technology, Boston, MA, USA) for 48 h, 5 mM CHX (Sigma-Aldrich, St. Louis, MO, USA) was added in the last 12 h incubation. At the end of culture, cells were collected for protein or RNA extraction.
To measure beta cell mass, pancreas was removed from mice, weighed, and then embedded and sectioned continuously. At least 10 evenly 200 mm apart sections throughout the entire pancreas were picked to perform immunostaining of insulin, followed by peroxidase conjugated secondary antibody, visualized using a DAB Peroxidase Substrate Kit (Fuzhou Maixin Biotech, China), and counterstained with eosin 66 . Digital images of whole pancreas were captured by a Nikon MZ 100 microscope. Total pancreatic and insulin positive areas of each section were measured using Meta-Morph version 6.1 (Molecular Devices). Beta cell mass was obtained by multiplying the ratio of total insulin positive area to total pancreatic area with the pancreas weight.
RT-PCR. Total islet RNA was extracted using RNeasy Plus Micro kit (Qiagen, GmBH, Germany) according to the manufacturer's protocol and used for reverse transcription with Superscript III (Life technologies, Carlsbad, CA, USA) with random hexamer primers. Quantitative PCR amplification and detection were performed with SYBR Premix Ex Taq Mixes (TaKaRa, Kyoto, Japan) on a Light-Cycler 480 instrument (Roche Applied Science, Mannheim, Germany). The gene expression levels were normalized to individual b-actin. Primer sequences can be obtained in Supplementary Table 2.
Microarray analysis. Total RNA was extracted from freshly isolated islets of 8-week-old bRapKO and WT mice using RNeasy Micro kit (Qiagen, GmBH, Germany) and sample qualities were assessed using an Agilent Bioanalyzer 2100 (Agilent technologies, Santa Clara, CA, USA), taking a minimal cut-off RIN 4 ¼ 7.
We included at least 4-5 independent biological replicates, each composed of RNA isolated from islets of 5 animals with the same genotype. 2 mg of total RNA samples of each group were then used to generate biotinylated cRNA targets for the Affymetrix Gene Chip Mouse Genome 430.2.0 Array. Data were analysed using Genespring software (Agilent).
MeDIP sequencing. Genomic DNA from freshly isolated islets was extracted using DNeasy Blood Tissue Kit (Qiagen, GmBH, Germany). MeDIP was performed using a MagMedIP kit (Diagenode, Denville, NJ, USA), as previously described 67 . Libraries were prepared following the manufacture's instruction and were sequenced by Illumina HiSeq 2500 with 50 cycles. Reads were aligned to reference mouse genome (mm10) by Bowtie2 v2.2.5 after QC and further pre-procession. Peaks were called by MACS v0.14.2, and DMRs (Differential methylation region) were analysed by MEDIPS v1.12.0R Bioconductor package. DMRs were calculated using a window size of 500 bp, extend by 100 bp, increasing in 500 bp increments, a minimum number of 10 reads in a window was required prior to inclusion in the DMRs. Additionally, DMRs were filtered out if it has no overlap with peaks called by MACS.
Electron microscopy. Samples were fixed, dehydrated, stained and embedded as previously described 68 . The density and subtype of beta cell vesical were determined by counting images captured at Â 7,400 magnification 68 . The vesical size was measured by analysing images captured at Â 16,500 magnification 68 . Each analysis was based on at least eight images per mice.

Statistics.
No randomization was used in this study, and the investigators were not blinded to group allocation during experiments or outcome assessments. No statistical method was used to predetermine sample size. Sample sizes were estimated from previous experience and common knowledge of animal studies. All data are presented as mean ± s.e.m. Statistical analysis was performed with unpaired Student's t test with a two-tailed distribution for two groups or ANOVA for multiple groups. Po0.05 was considered as statistically significant.
Data availability. Gene expression and MeDIP-seq data that support the findings of this study have been deposited in NCBI GEO (Gene Expression Omnibus) under accession code GSE84404 and GSE95775. All data supporting the findings of this study are available from the corresponding author on reasonable request.