Epigenetic Switching by the Metabolism-sensing Factors in the Generation of Orexin Neurons from Mouse Embryonic Stem Cells*

Background: Orexin plays a central role in the integration of sleep/wake states and feeding behaviors. Results: Orexin neurons were induced from pluripotent stem cells by supplementation of ManNAc. Conclusion: ManNAc induced switching of epigenetic factors from Sirt1/Ogt to Mgea5 at Hcrt gene locus. Significance: This study will be useful to investigate molecular mechanism in the orexin system and development of regenerative medicine. The orexin system plays a central role in the integration of sleep/wake and feeding behaviors in a broad spectrum of neural-metabolic physiology. Orexin-A and orexin-B are produced by the cleavage of prepro-orexin, which is encoded on the Hcrt gene. To date, methods for generating other peptide neurons could not induce orexin neurons from pluripotent stem cells. Considering that the metabolic status affects orexin expression, we supplemented the culture medium with a nutrient factor, ManNAc, and succeeded in generating functional orexin neurons from mouse ES cells. Because DNA methylation inhibitors and histone deacetylase inhibitors could induce Hcrt expression in mouse ES cells, the epigenetic mechanism may be involved in this orexin neurogenesis. DNA methylation analysis showed the presence of a tissue-dependent differentially methylated region (T-DMR) around the transcription start site of the Hcrt gene. In the orexin neurons induced by supplementation of ManNAc, the T-DMR of the Hcrt gene was hypomethylated in association with higher H3/H4 acetylation. Concomitantly, the histone acetyltransferases p300, CREB-binding protein (CBP), and Mgea5 (also called O-GlcNAcase) were localized to the T-DMR in the orexin neurons. In non-orexin-expressing cells, H3/H4 hypoacetylation and hyper-O-GlcNAc modification were observed at the T-DMRs occupied by O-GlcNAc transferase and Sirt1. Therefore, the results of the present study suggest that the glucose metabolite, ManNAc, induces switching from the inactive state by Ogt-Sirt1 to the active state by Mgea5, p300, and CBP at the Hcrt gene locus.

. These metabolites are integrated into numerous cellular functions as regulators of gene expression.
Acetyl-CoA is a donor of protein acetylation. NADϩ is a critical regulator of Sirtuins (15). Sirt1, a member of the sirtuin family, functions as histone deacetylase and is recognized as a nutrient sensor because a fasting condition or reduced calorie intake up-regulates its expression and activity (16).
UDP-GlcNAc is a donor of O-GlcNAcylation of cytoplasmic as well as nuclear proteins, including transcription factors, epigenetic factors such as polycomb group, and core histones (17)(18)(19)(20). O-GlcNAc transferase (Ogt) catalyzes the addition of O-GlcNAc to Ser or Thr residues of target proteins, and O-GlcNAcase (Oga) removes O-GlcNAc (11,21). Ogt is known to interact with other nuclear proteins such as polycomb group through TRP domain (21). The Oga gene is annotated as meningioma-expressed antigen 5 (Mgea5) and contains a putative histone acetyltransferase (HAT) domain (22), suggesting its involvement in the epigenetic system. Therefore, we hypothesized that glucose metabolites may have an impact on orexin neurogenesis, which is mediated by the epigenetic system.
The epigenetic system underlies not only the in vivo development but also the in vitro differentiation of pluripotent stem cells to various-type cells (23)(24)(25). Epigenetic alterations such as changes in the DNA methylation status and histone modifications result in chromatin remodeling of strictly regulated developmental genes (26 -29). Numerous tissue-dependent differentially methylated regions (T-DMRs) have been identified in the mammalian genome (23,25,30). Hypermethylated T-DMRs associate with silent loci, whereas hypo-methylated T-DMRs associate with active loci (30,31). In combination with the DNA methylation status of T-DMRs, histone modifications create the multilayered epigenetic control of long term gene activity (27,28,(32)(33)(34). The epigenetic system regulates the metabolism as shown by our previous finding, i.e. there are numerous T-DMRs at loci of nuclear-encoded mitochondrial proteins (31).
In the present study, by using a neural cell culture protocol, we found that the addition of ManNAc promotes the expression of the Hcrt gene and demonstrated the epigenetic regulation of the expression of the Hcrt gene by Sirt1, Ogt, and Mgea5. Thus, we successfully generated functional orexin neurons from mouse ESCs (mESCs).
Neural Differentiation from mESCs-Neural differentiation by using the SDIA and SDIAϩBMP4 methods was carried out as described in previous reports (36). We cultured mESCs (1.7 ϫ 10 3 cells/cm 2 ) on PA6 feeder cells in Glasgow MEM (Invitrogen) supplemented with 10% KnockOUT Serum Replacement, 0.1 mM nonessential amino acid, and 0.1 mM ␤-mercaptoethanol. PA6 cells were provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan. The culture medium was changed on day 4 and every 2 days thereafter. In the case of the SDIAϩBMP4 method, 5 nM BMP4 (Wako) was added to the medium from day 4. The gfCDM/SFEBq differentiation culture was performed as previously reported but with minor modifications (9). mESCs were dissociated to a single cell solution in 0.25% trypsin-EDTA and quickly re-aggregated in growth factor-free CDM (3000 cells per 200 l per well), which contained Iscove's modified Dulbecco's medium/Ham's F-12 1:1 (Invitrogen), 1 ϫ chemically defined lipid concentrate (Invitrogen), 450 M monothioglycerol (Wako), and purified BSA (Sigma) using 96-well low celladhesion plates (NUNC). After day 7 of culture, spheres were dissociated by using 0.25% trypsin-EDTA, quickly re-aggregated using low cell-adhesion 96-well culture plates (5000 cells per well), and cultured in DMEM/F-12 supplemented with 38.8 mM glucose and 10% KnockOUT Serum Replacement. On day 10, half of the medium was replaced with DFBN medium, which contained DMEM/F-12 supplemented with 38.8 mM glucose, N2 (Wako), B27 (Invitrogen), and 10 ng/ml ciliary neurotrophic factor (Wako). On day 13, spheres were dissociated by using 0.25% trypsin-EDTA and plated onto poly-D-lysine/laminincoated dishes (BD Biosciences) at a density of 8.5 ϫ 10 4 cells/ cm 2 in DFNB supplemented 50 ng/ml BDNF (Wako) and 50 ng/ml NT3 (Wako) until day 25.
Neurosphere Culture-Pregnant C57BL/6N mice were euthanized, and fetuses at embryonic day 14.5 were recovered in ice-cold PBS containing 0.6% glucose. For neurosphere culture, cells derived from telencephalons were suspended in DMEM/F-12 (1:1) supplemented with 5.5 mM HEPES, 2 mM L-glutamine, B27, 20 ng/ml EGF (Sigma), 20 ng/ml basic FGF (PeproTech), and 5 g/ml heparin (Sigma). Next, 3 ϫ 10 4 cells were seeded onto a low cell binding dish (NUNC) and cultured for 10 days, replacing half of the medium with fresh medium at every 3 day. To induce differentiation, cells were dispersed, suspended in the absence of growth factors, and seeded onto poly-L-lysine-and laminin-coated dishes (BD Biosciences).
Tissue Collection-Adult mice (C57BL/6N) were purchased from Charles River Japan and maintained on a 12-h light/12-h dark schedule with free access to food and water. The hypothalamus was recovered by separation from the whole brain of 13-week-old male mice using fine forceps. The collected tissues were stored at Ϫ80°C until use for RNA and DNA extraction. All experiments using mice were carried out according to the institutional guidelines for the care and use of laboratory animals (Graduate School of Agriculture and Life Sciences, The University of Tokyo).
RT-PCR and Quantitative RT-PCR-Total RNA was isolated from cells and tissues with the RNeasy Plus Mini kit (Qiagen) according to the manufacturer's instructions. First-strand cDNA was synthesized from 3 g of total RNA by using oligo(dT) 20 primers and the SuperScript III First-Strand Synthesis System (Invitrogen). RT-PCR was conducted with the LA TaqDNA polymerase (Takara) using 10 ng of cDNA per reaction. PCR reactions were performed under the following conditions; denaturation at 95°C for 3 min and the appropriate number of cycles, with each cycle consisting of 95°C for 30 s, 60°C for 30 s, and 72°C for 15 s. PCR products were subjected to agarose gel electrophoresis and stained using GelRed (Biotium). The primer sequences used are listed in supplemental Table 1. Each qPCR was performed with 10 ng of cDNA and Thunderbird Syber qPCR Mix (Toyobo) using the ABI7500 thermal cycler (Applied Biosystems). PCR was performed with the following thermocycling conditions; denaturation at 95°C for 1 min and 40 cycles, with each cycle consisting of incubation at 95°C for 10 s and 60°C for 1 min. Data were normalized to the expression of Actb.
Immunofluorescence Assay-Cells cultured in 4-well dishes were fixed with 4% paraformaldehyde (Wako) and permeabilized with 0.2% Triton X-100 (Wako) followed by blocking with 5% BSA, 0.1% Tween 20/PBS (Sigma) for 1 h at room temperature and incubating with the primary antibody overnight at 4°C. The secondary antibody was added, and the incubation was continued for 1 h at room temperature. Nuclei were stained with DAPI (1 g/ml; Dojindo). The primary antibodies used are listed in supplemental Table 2. The following secondary antibodies were used: donkey anti-goat Alex-Flour 488, rabbit antimouse Alexa Flour 594, and chicken anti-rabbit Alexa-Flour 594 (1:1000; Invitrogen). Fluorescence images were acquired with a microscope (BZ-8000; KEYENCE). Immuno-and DAPIstained cells were counted in at least 30 randomly chosen areas by using the Image J software (rsb.info.nih.gov). Percentages indicate the mean of a ratio of an Orexin-A or -B-positive area to a DAPI-positive area in three independent cultures.
Orexin-A-releasing Assay by Using ELISA-mESCs were cultured under SDIAϩBMP4 conditions for 10 days in 4-well dishes and were subjected to the following analyses. For the KCl assay, cells were incubated in 500 l of aCSF medium (124 mM NaCl, 3 mM KCl, 26 mM NaHCO 3 , 2 mM CaCl 2 , 1 mM MgSO 4 , 1.25 mM KH 2 PO 4 , and 10 mM D-glucose, pH 7.4) for 10 min at 37°C followed by stimulation with aCSF plus 100 mM KCl medium for an additional 10 min. For measurement of neural peptide sensitivity, neural differentiated mESCs were incubated in 500 l of medium of the SDIA condition supplemented with leptin (Wako), ghrelin (Wako), or TRH (Wako) at the appropriate concentrations at 37°C. After 3 h of incubation, the supernatants were collected, and the Orexin-A concentration was measured using the Orexin-A Fluorescent EIA kit (Phoenix Pharmaceuticals) according to the manufacturer's instructions.
DNA Methylation Analysis Using the Bisulfite Method-Genomic DNA was extracted from cells and tissues as described previously (30). Bisulfite conversion was performed using the EZ DNA Methylation-Gold kit (Zymo Research). The EZ DNA Direct-Methylation kit (Zymo Research) was used to analysis single colony under SDIAϩBMP4 conditions with 1 mM ManNAc. The Orexin-A-positive and -negative colonies were picked up using fine pipette after immunostaining by using an anti-Orexin-A antibody. For each bisulfite PCR, BIOTAQ HS DNA polymerase (Bioline) was used to catalyze the amplification. PCR was performed with the following thermocycling conditions; denaturation at 95°C for 10 min and 40 cycles, with each cycle consisting of incubation at 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s followed by a final extension for 5 min at 72°C. For sequencing, the PCR fragments were cloned into the pGEM-T Easy vector (Promega). The vectors were sequenced by BigDye sequencing (Applied Biosystems).
Chromatin Immunoprecipitation Assay-The ChIP assay was performed with 1 ϫ 10 6 cells per assay using the ChIP-IT Express Enzymatic kit (Active Motif) according to the manufacturer's instructions. Briefly, fixed cells were lysed and mixed with an enzymatic shearing mixture for 10 min. Antibodies, which were used for immunoprecipitation, are listed in supplemental Table 2. After immunoprecipitation, DNA was recovered by using an elution buffer (10% SDS, 300 mM NaCl, 10 mM Tris-HCl, and 5 mM EDTA, pH 8.0) at 65°C for 6 h and then collected using the Chromatin IP DNA Purification kit (Active Motif). PCR reactions with LA TaqDNA polymerase were performed under the following conditions; denaturation at 95°C for 3 min and 32 cycles, with each cycle consisting of 95°C for 30 s, 60°C for 30 s, and 72°C for 15 s. PCR products were subjected to agarose gel electrophoresis and stained using GelRed.
Western Blotting-Nuclear and cytoplasmic fractions of each sample were collected using the Nuclear Extract Kit (Active Motif) according to the manufacturer's protocols. The proteins were fractionated by 5-20% SDS-PAGE (XV PANTERA Gel; DRC), blotted onto nitrocellulose membranes (Millipore), and incubated at 4°C overnight with the Primary antibody diluted in 5% BSA, 0.1% Tween 20/TBS (supplemental Table 2). Protein bands were detected using secondary antibody conjugated with horseradish peroxidase (Jackson ImmunoResearch) and Super-Signal West Pico or Femto (Thermo).

Construction of Overexpression Vectors and Transfection-
The sequences of all primers used for plasmid construction are listed in supplemental Table 1. DNA of 3ϫFLAG-fused mouse Ogt and Mgea5 were generated by PCR amplification from cDNA of mESCs using PrimeSTAR HS DNA Polymerase (Takara) and cloned into the pENTR/D-TOPO vector (Invitrogen). Point mutations of Mgea5 at position 175 (Asp 3 Ala) or 891 (Tyr 3 Phe) of the amino acid sequence were generated by using the PrimeSTAR mutagenesis Basal kit (Takara) and the pENTR/D-TOPO vector-cloned 3ϫFLAG-Mgea5 as template. The resulting constructs were confirmed by BigDye sequencing. 3ϫFLAG-fused genes were subcloned into a pCAG-DEST vector, which was generated by using a combination of the Gateway Vector Conversion System (Invitrogen) and pCAGEN (Addgene) and Gateway LR Clonase (Invitrogen). For transient overexpression using these vectors, mESCs were cultured in 10-cm dishes under SDIAϩBMP4 conditions with 1 mM Man-NAc. At day 7 of culture, the cells were then transfected with 24 g of plasmid and 30 l of Lipofectamine 2000 (Invitrogen) per dish. Twenty-four hours after transfection, medium was changed, and transfected cells were collected at day 10 for the subsequent experiments.
The experiments described in the present study were repeated at least three times with similar results in each case. The results shown are representative for all repeated experiments.

RESULTS
ManNAc Treatment Promotes Generation of Orexin Neurons from Mouse ES Cells-The gene expression of hypothalamic peptides and transcription factors was induced in mESCs by employing two known in vitro differentiation methods, i.e. SDIAϩBMP4 and gfCDM/SFEBq (9, 36) (Fig. 1A). The Hcrt gene, however, could not be induced by using either of these methods. We then investigated the effect of supplementation with GlcN, GlcNAc, ManNAc, and Neu5Ac at a concentration of 1 mM and found that only ManNAc could induce Hcrt expression in cells produced by either in vitro differentiation methods (Fig. 1, B and D). On the other hand, other hypothalamic peptide genes, Npy and Gnrh1, were not affected by Man-NAc supplementation (Fig. 1B). By using an immunofluorescence assay, we detected Orexin-A-and Orexin-B-positive cells (6.2 Ϯ 1.3 and 7.1 Ϯ 1.9%, respectively, of total cells) in the colonies of Tubb3-and Ncam-positive neural cells differentiated in the presence of ManNAc (Fig. 1C). In addition, Dynor-phin-A, which is known to be expressed with orexin neuron (37), was co-localized with Orexin-A signals (Fig. 1C). Furthermore, we examine the expression profiles of eight orexin neural markers by RT-PCR on the basis of recent reports (38). All of the marker genes were detected in ManNAc-treated cells (Fig.  1E).
At the early differentiation stage (days 0 -4 and 0 -7), Man-NAc supplementation was less effective on the induction of orexin neurons than at the late stage (days 4 -10 and 7-10) (Fig.  1F). Considering that neural progenitor cells (NPCs) first appear around day 4 under the culture conditions employed (36), we conclude that ManNAc acts at the later stage of neural differentiation, i.e. after the neuronal fate commitment of progenitor cells. Indeed, ManNAc supplementation could induce Hcrt expression in neurospheres (Nsph) derived from fetal mouse telencephalons (embryonic day 14.5) (Fig. 1G).
Orexin neurons respond physiologically to leptin and ghrelin (39,40). Secretion of Orexin-A in the presence of high levels of KCl suggested the involvement of leaky K ϩ channels in glucose sensing of induced orexin neurons (41) (Fig. 1H). Physiological stimulants such as ghrelin and TRH dose-dependently stimulated Orexin-A secretion (Fig. 1I) in ManNAc-induced cells, which was inhibited by leptin. These data provide further support for the conclusion that supplementation of an intermediate metabolite of glucose, ManNAc, enables the generation of functional orexin neurons from mESCs.

DNA Methylation Status of T-DMRs at the Hcrt Gene Locus-
To explore epigenetic mechanisms that underlie the differentiation of orexin neurons, we treated mESCs with 5 M 5-aza-2Јdeoxycytidine or 100 M Zebularine, inhibitors of DNA methyltransferase, and/or 200 nM trichostatin A, an inhibitor of histone deacetylase. Both inhibitors and their combination induced Hcrt expression ( Fig. 2A). These data suggested that DNA methylation and histone deacetylation are involved in Hcrt gene silencing. Orexin-A and -B are produced by cleavage of a polypeptide, prepro-orexin, encoded by the Hcrt gene, whose promoter has two putative orexin regulatory elements, ORE1 and ORE2, conserved among animal species (42). We, therefore, examined the DNA methylation status around the transcription start site containing ORE1.
Bisulfite sequencing identified a T-DMR upstream (T-DMR-U, Ϫ778 ϳ Ϫ10 bp) and downstream (T-DMR-D, ϩ5 ϳ ϩ665 bp) of the transcription start site (Fig. 2B). T-DMR-U, which includes ORE1, was heavily methylated in mESCs (98%), whereas the methylation status was 84 ϳ 89% in mESC-derived neural cells produced by employing the SDIAϩBMP4 culture method. In contrast, ManNAc treatment caused a decrease in the CpG methylation status to 78% (Fig.  2B). Furthermore, the colony of Orexin-A-positive cells showed hypomethylation at T-DMR-U (40%) compared with the negative colony (81%). ORE1 contained binding sites for Nr6a1 and Ebf2 (so called O/E3), which are important for Hcrt expression (43,44). Because these binding sites show no CpG sequences, methylation of T-DMR-U would not directly inhibit the binding of these factors. However, the accessibility of these factors could be restricted by condensed chromatin, which is induced by DNA methylation of T-DMR-U.
T-DMR-D, which is located in the 1st intron of the Hcrt gene, was 46% methylated in ManNAc-treated cells and 71% in mESCs (Fig. 2B). Similarly, cells differentiated for 4 days in vitro showed hypomethylation at T-DMR-D (Fig. 2B). In contrast, Neu5Ac treatment promoted methylation at T-DMR-D in mESC-derived neurons. These data indicated that there are T-DMRs that are affected by supplementation with intermediates of HBP.
ManNAc Causes Hyperacetylation of H3 and H4 at T-DMRs of Hcrt-We examined the epigenetic status at these T-DMRs by ChIP analysis for molecules related to histone acetylation. ChIP analysis of region 1 in T-DMR-U and region 2 in T-DMR-D revealed that acetylation of H3K9, H3K14, H3K27, H3K56, H4K8, and H4K16 as well as trimethylation of H3K4 were markedly elevated by ManNAc treatment (Fig. 2C). In accordance with the acetylation status, HAT p300 and CBP, which are responsible for pan-H3Ac and -H4Ac, are increased at the T-DMRs. Furthermore, Mgea5, which has putative histone acetyltransferase activity responsible for H3K14Ac and H4K8Ac, accumulated at the T-DMR (Fig. 2C).
In repressive factors, Sin3A, and Ezh2 were decreased at the regions in ManNAc-treated cells by ChIP analysis (Fig. 2C), corresponding with the decrease in repressive histone marks, H3K9me3 and H3K27me3. An increase in histone deacetylases in Neu5Ac-treated cells suggested that Neu5Ac might induce hyper-repression of Hcrt in combination with hypermethylation of DNA.
Western blot analysis of nuclear and cytoplasmic proteins of treated cells indicated that ManNAc induced subcellular de-localization of Sirt1 from the nucleus to the cytoplasm (Fig. 2D). In contrast, p300, CBP, and Mgea5 were increased in the nucleus of ManNAc-treated cells. Concomitantly, levels of histone acetylation were also increased (Fig. 2E). Immunofluores-cence assay, focusing on orexin neuron, confirmed that Mgea5 was mainly located in nuclear, whereas Sirt1 was in cytoplasm (Fig. 2F). These data suggested that with regard to the epigenetic control of the Hcrt gene, p300, CBP, and Mgea5 contributed in the active state, whereas Sirt1, Sin3A, and Ezh2 contributed in the inactive state.

Generation of Orexin Neurons by Using a Sirt1
Inhibitor-The involvement of Sirt1 in orexin neural differentiation was investigated using Sirt1-knock-out (Sirt1 Ϫ/Ϫ ) mESCs. RT-PCR revealed Hcrt expression in differentiated Sirt1 Ϫ/Ϫ mESCs when using the SDIAϩBMP4 method, even in the absence of ManNAc (Fig. 3A). Orexin-A-and orexin-B-positive cells were observed among differentiated Sirt1 Ϫ/Ϫ mESCs (Fig. 3B). Acetylation levels of H3K9, K14, K27, K56, H4K8, and H4K16 at regions 1 and 2 were higher in differentiated Sirt1 Ϫ/Ϫ cells compared with wild type cells (Fig. 3C). Furthermore, treatment with EX-527, a Sirt1 inhibitor, resulted in Hcrt expression in differentiated wild type cells (Fig. 3A) and an increase in histone acetylation of the same residues of neural differentiated Sirt1 Ϫ/Ϫ cells (Fig. 3C). Therefore, inhibition of Sirt1 at the Hcrt locus, which was induced by ManNAc supplementation, is a key event in the differentiation of orexin neurons.
Another mechanism of the effect of ManNAc on differentiation of orexin neurons was suggested by the observation that ManNAc treatment caused a further increase in Hcrt expres-sion in cells derived from Sirt1 Ϫ/Ϫ mESCs as well as in EX-527treated neural cells derived from mESCs (Fig. 3A). The increased expression of Hcrt by ManNAc treatment was associated with an elevated acetylation at H3K9, K14, K27, K56, H4K8, and K16 and ManNAc-induced accumulation of Mgea5, p300, and CBP at T-DMRs of Hcrt (Fig. 2C). Therefore, both steps, i.e. deletion of Sirt1 and accumulation of Mgea5, p300, and CBP, could be responsible for Hcrt gene activation during orexin neurogenesis.

Loss of O-GlcNAcylation at T-DMRs Promotes Hcrt
Expression-Mgea5 has dual enzymatic activity, i.e. HAT and Oga activities (22). Thus, in addition to histone acetylation, O-GlcNAcylation might also be involved in the regulation of the Hcrt gene. ChIP analysis using an RL2 antibody, which recognizes O-GlcNAc modifications, revealed that there are  Treatment with Thiamet-G, an Oga inhibitor, diminished, and treatment with benzyl 2-acetamido-2-deoxy-␣-d-galactopyranoside, an Ogt inhibitor, augmented the expression of ManNAc-induced Hcrt expression (Fig. 4B). Therefore, O-GlcNAcylation plays a suppressive role in Hcrt gene activation. Overexpression of Ogt inhibited Hcrt gene expression, and that of Mgea5 increased Hcrt gene expression in cultures with ManNAc supplementation (Fig. 4C). In these experiments, increased or decreased O-GlcNAc modification levels were observed at T-DMRs in cells overexpressing Ogt and Mgea5, respectively (Fig. 4D) Re-ChIP analysis using RL2 as the first antibody showed co-localization of Sirt1, Sin3A, and Ezh2 with O-GlcNAc modification at regions 1 and 2 in Hcrt-non-expressing cells (Fig. 4E). This is in contrast to the results of ManNAc-treated cells in which the signal for these repressive molecules was substantially reduced (Fig. 4E). Again, in the Re-ChIP analysis with an anti-Ogt antibody, the RL2 signal as well as those of Sirt1, Sin3A, and Ezh2 was strongly detected in the Hcrt inactive state, especially in Neu5Ac-treated cells.
Mgea5/Oga Activity Links to Histone H3 and H4 Acetylation-Mgea5 accumulated by ManNAc treatment at T-DMRs, and Thiamet-G treatment not only decreased the acetylation levels of H3K14 and H4K8 and other histone acetylation levels but also increased the levels of Ogt and O-GlcNAc (Fig. 5A). These findings suggested that reciprocal modifications between O-GlcNAc modification and histone acetylation could be important for the regulation of the Hcrt gene. Under these circumstances, the absence of increase in Sirt1 suggested that this reciprocal change occurs after clearance of Sirt1 at regions 1 and 2.
We hypothesized that Mgae5 could play a role in both Oga and HAT processes. Therefore, we prepared constructs for 3ϫFLAG-fused Mgae5; wild type (WT), a mutation at the Oga domain (D175A), and a mutation at HAT domain (Tyr-891), respectively (22,45) (Fig. 5B), and introduced them in neural differentiated mESCs. In Western blot analyses of whole nuclear extracts, WT-overexpressing neural differentiated mESCs showed an increase in H3K14Ac and H4K8Ac, indicating HAT activity of Mgea5 (Fig. 5C). More importantly, acetylated histone levels were diminished in both mutants (D175A and Y891F) compared with the WT (Fig. 5C), suggesting that O-GlcNase activity is important for HAT activity of Mgea5. As indicated by the above findings, at regions 1 and 2 of the T-DMRs, H3K14Ac and H4K8Ac were increased in the WT and decreased in both mutants (D175A and Y891F) (Fig. 5D).
In the D175A mutant, O-GlcNAc modification was increased, whereas the WT and Y891F mutant showed lower O-GlcNAc modification levels. The expression of the Hcrt gene was increased in WT-Mgea5 overexpressing cells, and both mutants did not show any activity (Fig. 5E). Based on these results, we conclude that a dual function of Mgae5, consisting of Oga and HAT, plays an important role in the activation of the Hcrt gene.

DISCUSSION
This is the first report on the generation of orexin neurons from ESCs. The epigenetic status at T-DMRs of the Hcrt locus exhibited a unique feature, i.e. involvement of DNA methylation, histone acetylation, and O-GlcNAcylation. We found that treatment with ManNAc induced hypo-CpG methylation and hyperacetylation of H3 and H4 at Hcrt T-DMRs by regulating the localization of the metabolic sensing molecules Sirt1, Ogt, and Mgea5. Thus, an epigenetic switch on histones from a hypoacetylated state with unidentified O-GlcNAcylated nuclear proteins to a hyperacetylated state at T-DMRs is the key event in the generation of orexin neurons. Furthermore, the hyperacetylation of the T-DMRs was associated with trimethylation of H3K4. The mESC-derived orexin neurons are equipped with the physiological response to neurotransmitter peptides such as TRH, ghrelin, and leptin, as demonstrated in previous findings by using brain tissue slices of experimental animals (39,40). Thus, our findings provide novel and important information for research into the orexin system.
Among the glucose metabolites involved in the HBP that we examined, it is interesting that only ManNAc, but not GlcN, GlcNAc, and Neu5Ac, was effective in generating orexin neurons. In PC12 cells, both Neu5Ac and ManNAc stimulate the differentiation of neurons, and overexpression of UDP-GlcNAc 2-epimerase/ManNAc kinase (GNE), which immediately converts ManNAc into Neu5Ac, causes a similar effect on PC12 cells, prompting us to consider ManNAc as a precursor for sialic acids (46). In the present study, however, only ManNAc but not Neu5Ac could induce orexine neurons, suggesting that ManNAc exerts its effect other than Neu5Ac production. Indeed, ManNAc causes dislocation of Ogt/Sirt1/Sin3A/Ezh2 from T-DMRs of the Hcrt gene locus and recruitment of p300, CBP, and Mgea5 in the present study. These data suggest a unique role for ManNAc as an element of a signaling pathway that contributes to modifications of the epigenetic status at the Hcrt gene locus to induce orexin neurons.
UDP-GlcNAc is a precursor for ManNAc and is synthesized via the HBP from ϳ3% of the total glucose (47,48). It is reported that GlcN enters the HBP and subsequently causes an increase in UDP-GlcNAc (12,13), and then UDP-GlcNAc regulates Ogt activity by increasing the availability of the substrate (49,50).
Between the inactive and active status of the Hcrt gene, a dynamic change was observed in the histone acetylation level at T-DMRs. In the active state of Hcrt gene expression, histone H3 and H4 were hyperacetylated. Mgea5, p300, and CBP were localized at Hcrt T-DMRs, which were occupied by Sirt1 and Ogt in the inactive state. Considering the dynamic changes that occur in conversion from a hypoacetylated to a hyperacetylated status, these are the molecules believed to be responsible for histone acetylation at T-DMRs. Importantly, Mgea5 has both Oga and HAT activities (22,45). The zinc finger-like motif of Mgea5 is responsible for the recognition of the acetyl substrate (55). Mgea5 showed HAT activity and the capability to modify all core histones in vitro, which include H3K14 and H4K8. In the present study, overexpression of Mgea5 induced the hyperacetylation of H3K14 and H4K8, whereas the mutant Mgea5 (Y891F) did not show this activity, indicating that Mgea5 works as HAT intracellularly. Treatment with an Oga inhibitor (Thiamet-G) and overexpression of both Mgea5 mutants (D175A and Y891F) did not induce histone acetylation and Hcrt gene expression. Thus, Mgea5 must exert its dual function of Oga and HAT in the establishment of an active epigenetic state on the Hcrt gene locus. We conclude that Mgea5 is the primary factor for the epigenetic change at the T-DMR to express the Hcrt gene and induce the differentiation of orexin neurons.
The epigenetic status of each T-DMR is regulated by the interplay between DNA methyltransferases, histone modification enzymes, histone subtypes, non-histone nuclear proteins, non-coding RNAs, and other factors (28). Many non-histone proteins such as Sin3A and Ezh2 function as the links between DNA methylation and histone modification. At each T-DMR, the DNA methylation status locally correlates with the histone modification status and vice versa (28). Therefore, a decrease in the levels of Sin3A and Ezh2 at the T-DMR of Hcrt after Man-NAc treatment may be a molecular link between histone modification and induction of DNA demethylation.
Re-ChIP experiments using antibodies against O-GlcNAc modification and Ogt revealed a protein complex (O-GlcNAc complex) consisting of Sirt1, Ogt, Sin3A, and Ezh2 at the T-DMRs. Considering that there are hundreds of O-GlcNAcylated proteins (10), the components of the O-GlcNAc complex are also likely to be O-GlcNAcylated. Sin3A, which is a core component of several transcriptional co-repressor complexes, including Sin3A/histone deacetylase, has been shown to recruit Ogt to promoters to repress the transcription (56), and Sin3A itself is modified by O-GlcNAc (57). Although there is no report on such a modification of Sirt1 and Ezh2, localization of the Polycomb-repressive complex 2, including Ezh2, has significant similarity to the O-GlcNAc modification on the Drosophila melanogaster genome (17). The remaining strong O-GlcNAc signal at T-DMRs after treatment with ManNAc and an Oga inhibitor (Thiamet-G) prompted us to consider another unidentified O-GlcNAcylated protein involved in the repression mechanisms of Hcrt, because this treatment removes the O-GlcNAc complex from T-DMRs. Possible other targets of O-GlcNAcylation are core histone proteins. To date, O-GlcNAcylated H2B, H3, and H4 have been reported by MASS analysis (18 -20). The O-GlcNAcylation sites of core histones are also assumed on the basis of analogy to non-histone proteins of which Thr/Ser residues are modified by mutually exclusive phosphorylation and O-GlcNAcylation (58). Future studies using induced orexin neurons will allow us to address these issues.
The induced orexin neurons will also be useful to investigate molecular mechanisms in the orexin system. Because Sirt1 has been recognized as neuro-protective molecule (59), the inhibitory role in the differentiation of orexin neurons was unexpected. This finding is contradictory to the previous findings demonstrating co-localization of orexin and Sirt1 in mouse lateral hypothalamus and decreased orexin levels in adult Sirt1 KO mice (60), indicating that Sirt1 is necessary for the orexin expression. We believe that the molecular mechanism of differentiation should be considered separately from the responses observed in differentiated neurons. Induced orexin neurons in particular will provide a strong tool for the development of medical applications; molecules identified in this study could be target molecules for the evaluation and screening for diseases related to the orexin system.
ManNAc exhibited novel effects on epigenetic processes, including DNA demethylation, histone acetylation, and O-GlcNAcylation (Fig. 5F). In mESCs and neural precursor cells, T-DMRs are hypermethylated, and H3/H4 are hypoacetylated. Hypoacetylation is established by Sirt1. Ogt, Ezh2, and Sin3A are also co-localized with the silencing complex. Loss of O-GlcNAcylation is a pivotal step in the transformation from the silent state with H3/H4 hypoacetylation to the active state with hyperacetylation at T-DMRs of the Hcrt gene. ManNAc treatment caused de-localization of Sirt1, Ogt, Ezh2, and Sin3A, and recruitment of Mgea5, which have Oga and HAT activities. In this active state, other HATs such as p300 and CBP are also involved. We find it intriguing that the process of generation of orexin neurons, central regulators of the whole body metabolism, comprises of an epigenetic mechanism consisting of nutrient-sensing molecules.
Here, we have demonstrated the multilayered epigenetic regulation by Sirt1, Ogt, and Mgea5 in orexin neurogenesis. We propose that induced orexin neurons will provide a valuable tool in development of regenerative medicine applications.