PRMT1 Is Required for the Maintenance of Mature β-Cell Identity

Maintaining the functional β-cell mass is crucial for preventing diabetes, which develops when β-cells fail to meet the insulin demand (1,2). Although β-cell death is thought to be the major mechanism of β-cell failure (3), recent studies indicate that β-cell dedifferentiation can decrease the functional β-cell mass and thereby deteriorate systemic glucose homeostasis (4,5). Maintaining mature β-cell identity is also important for maintaining β-cell function (6,7). A hierarchy of transcription factor (TF) cascades directs β-cell differentiation, and β-cells require continuous activation of these TFs to maintain their function and identity (8–10). The genetic identity of a differentiated cell is generally controlled by the chromatin state, which is overall stable and has limited epigenomic flexibility (11,12). Likewise, epigenetic regulation plays an essential role in the postnatal maturation of β-cells and the maintenance of mature β-cell identity (13–16).

Histone arginine methylation, which is regulated by protein arginine methyltransferase (PRMT), can affect chromatin structures to facilitate the recruitment of protein complexes that regulate gene transcription (17,18). PRMT4-dependent histone H3 arginine 17 asymmetric dimethylation (H3R17me2a) in β-cells has been reported to regulate glucose-stimulated insulin secretion (GSIS) (19). However, the role of PRMT-induced histone arginine methylation in regulating β-cell identity has not yet been elucidated. Among the nine members of the PRMT family, PRMT1 predominates in mammalian cells (20). It appears to be associated with diabetes, as its catalytic activity is decreased in the liver and pancreas of diabetic Goto-Kakizaki rats (21). PRMT1 has also been shown to specifically induce the active histone code, histone H4 arginine 3 asymmetric dimethylation (H4R3me2a), which potentiates subsequent histone acetylation and contributes to establishing euchromatin structure (22,23). Based on these previous findings, we herein explored the role of PRMT1-dependent H4R3me2a in mature β-cells.

Prmt1-floxed (Prmt1^fl/fl) (Mouse Genome Informatics [MGI]: 4432476) mice were crossed with Rip2-Cre (MGI: 2387567) and Pdx1-Cre^ERT2 (MGI: 2684321) mice to generate Prmt1 β-cell knockout (βKO) and inducible β-cell–specific Prmt1 KO (Prmt1 βiKO) mice, respectively. R26-eYFP (MGI: 2449038) mice were crossed for lineage-tracing experiments and β-cell sorting. All mice were backcrossed and maintained on a C57BL/6J background. Cre recombination for Cre^ERT2 was induced by a total of five intraperitoneal injections of corn oil–dissolved tamoxifen (75 mg/kg) over 2 weeks. Mice were housed in climate-controlled, specific pathogen-free barrier facilities under a 12-h light/dark cycle, and chow and water were provided ad libitum. Mice were fed either a standard chow diet or high-fat diet (HFD) (60% kcal fat). The animal experiment protocols for this study were approved by the Institutional Animal Care and Use Committee at the Korea Advanced Institute of Science and Technology. All experiments were performed in accordance with the relevant guidelines and regulations.

Body weight and random blood glucose levels were measured in the afternoon hours each day. The glucose tolerance test and insulin tolerance test were performed as previously described (24).

For histological analyses, formalin-fixed paraffin-embedded pancreatic slides were prepared, stained, and analyzed as described in Supplementary Data.

Pancreatic tissues were dissected, placed in acid-ethanol (1.5% HCl in 70% ethanol), homogenized, and incubated at 4°C for 16 h. The aqueous phase of pancreatic insulin extract was neutralized with an equal amount of 1 mol/L Tris-Cl buffer (pH 7.5). The pancreatic insulin content was calculated by dividing the total pancreatic insulin by the weight of the pancreas.

For the in vivo GSIS assay, mice were fasted for 16 h and then given an intraperitoneal injection of d-glucose in PBS (2 g/kg). For the ex vivo islet GSIS assay, pancreatic islets were isolated from mice as described previously (25), and the assay was performed as described in Supplementary Data.

Pancreatic islets were isolated from mice as described previously (25), and the oxygen consumption rate (OCR) assay was performed as described in Supplementary Data.

Total RNA was extracted from mouse tissues, and quantitative RT-PCR (qRT-PCR) was performed as described in Supplementary Data. The sequences of the primers used are listed in Supplementary Table 1.

Chromatin immunoprecipitation (ChIP) experiments were performed in MIN6 cells as previously described (26) with modifications. RNA sequencing (RNA-seq) experiments were performed using wild-type (WT) and Prmt1-null islets. Assay for transposase-accessible chromatin (ATAC) experiments were performed as previously described (27), using MIN6 cells and FACS WT and Prmt1-null β-cells. ChIP sequencing (ChIP-seq), RNA-seq, and ATAC sequencing (ATAC-seq) analyses were performed as described in Supplementary Data.

Chromosome conformation capture (3C) experiments were performed in MIN6 cells as previously described (28) with modifications. The data were normalized with respect to those obtained using internal primers that recognized sequences within the Gapdh gene. At least three independent biological replicates were included for each 3C-PCR assay. The sequences of the primers used are listed in Supplementary Table 1.

All values are expressed as mean ± SEM. The two-tailed Student t test or one-way ANOVA followed by post hoc Tukey test was used to compare groups. P values <0.05 were considered statistically significant.

The ChIP-seq, RNA-seq, and ATAC-seq data that support the findings of this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) under accession code GSE117100. All data that support the findings of this study are available from the authors on reasonable request. No applicable resources were generated or analyzed during the current study.

We first checked the gene expression of the Prmt family genes in pancreatic islets and confirmed that Prmt1 exhibited the highest expression level among them (Supplementary Fig. 1A and B). The expression level of Prmt1 was higher in pancreatic islets than in liver or brain, and PRMT1 and H4R3me2a were enriched in pancreatic islets of both mice and humans (Supplementary Fig. 1C and D). To test the possible role of H4R3me2a in β-cells, we performed ChIP-seq for H4R3me2a and ATAC-seq in MIN6 cells. Motif analysis of the H4R3me2a ChIP-seq data suggested that H4R3me2a was significantly associated with the β-cell TFs: MAFA, NEUROD1, and FOXA1 (Supplementary Fig. 1E). Intriguingly, the most significantly associated TF was CCCTC-binding factor (CTCF), which is known to play a crucial role in regulating the chromatin architecture (29,30). In order to assess the association of H4R3me2a with CTCF and β-cell TF, we performed ChIP-seq for CTCF in MIN6 cells. Analyses of ChIP-seq data obtained for H4R3me2a and CTCF, combined with publicly available ChIP-seq data for MAFA, indicated that CTCF and MAFA bind near H4R3me2a-occupied chromatin regions (Supplementary Fig. 1F). The association of H4R3me2a with CTCF and MAFA, together with the enrichment of PRMT1 and H4R3me2a in adult β-cells, suggests that PRMT1 and H4R3me2a may play roles in β-cells.

To further investigate the physiological role of PRMT1 and H4R3me2a in β-cells, we generated β-cell–specific Prmt1 KO (Rip2-Cre; Prmt1^fl/fl, herein called Prmt1 βKO) mice. Immunofluorescence staining confirmed the deletion of PRMT1 at postnatal day 7 and the subsequent removal of H4R3me2a in the β-cells of these mice around weaning at 3 weeks of age (Supplementary Fig. 2). In WT control mice, the fluorescence signals of both PRMT1 and H4R3me2a became enriched in pancreatic islets after weaning at 3 weeks of age, when the β-cells become mature. In Prmt1 βKO mice, PRMT1 was nearly undetectable at postnatal day 7, whereas H4R3me2a remained detectable in a substantial number of β-cells until 3 weeks of age, suggesting the more important role of PRMT1-dependent H4R3me2a in mature β-cells. Indeed, the pancreatic islets of Prmt1 βKO mice developed normally and did not show any abnormality in the markers of β-cell development (Supplementary Fig. 3A and B). Prmt1 βKO mice grew normally and showed normal glucose tolerance until they developed glucose intolerance at 12 weeks of age (Fig. 1A and B and Supplementary Fig. 3C). Despite this glucose intolerance, Prmt1 βKO mice showed no defects in insulin sensitivity and insulin production (Supplementary Fig. 4A–C). Instead, GSIS was impaired in Prmt1 βKO islets (Fig. 1C). Basal insulin secretion was increased when Prmt1 βKO islets were treated with 2.8 mmol/L glucose, and those treated with 20 mmol/L glucose failed to show any further increase of insulin secretion. The impairment of GSIS was further confirmed by analyses of plasma insulin levels and mitochondrial OCRs in the isolated islets (Fig. 1D and E). To investigate whether a transcriptional change was responsible for the defect of GSIS in Prmt1 βKO mice, we performed RNA-seq analysis in the islets of Prmt1 βKO mice at 12 weeks of age (Supplementary Fig. 5A). Although there was no robust change in global gene expression, Prmt1 βKO islets exhibited downregulation of mature β-cell genes that are involved in GSIS and misexpression of genes that are disallowed to be expressed in mature β-cells (31–33) (Supplementary Fig. 5B). These gene expression changes were further confirmed by qRT-PCR analysis (Supplementary Fig. 5C–F). Pathway analysis showed that most of the genes downregulated in Prmt1 βKO islets were involved in pancreas development and maturity-onset diabetes of the young (Supplementary Fig. 5G). This notion was further supported by electron microscopic analysis (Fig. 1F), which showed that Prmt1-null β-cells exhibited ultrastructural changes resembling those found in the β-cells of patients with type 2 diabetes (34,35). The volume and density of insulin granules were reduced, the endoplasmic reticulum was dilated, and the mitochondria appeared round and swollen. Condensed chromatin was observed in the nuclei of Prmt1-null β-cells (Fig. 1F), but apoptosis was not observed in the islets of Prmt1 βKO mice (Supplementary Fig. 5H). In addition, islet hormones and β-cell TFs were normally expressed in these mice (Supplementary Fig. 5I and J). These data indicate that PRMT1 is needed to maintain the function, not the survival, of β-cells.

As the phenotypes of Prmt1 βKO mice resembled the early features of type 2 diabetes (34–36), we fed Prmt1 βKO mice with an HFD from 8 weeks of age to test how these mice respond to metabolic stress (Supplementary Fig. 6A). HFD exacerbated the glucose intolerance in Prmt1 βKO mice without perturbing compensatory β-cell expansion (Fig. 1G and Supplementary Fig. 6B). Interestingly, the islets of HFD-fed Prmt1 βKO mice contained polyhormonal cells that coexpressed insulin and glucagon (GCG) (Fig. 1H). A lineage-tracing analysis revealed that these polyhormonal cells were originated from insulin-producing β-cells (GCG, ∼0.4%; somatostatin [SST], ∼2.4%; and pancreatic polypeptide [PPY], ∼0.19%), suggesting that the β-cells of HFD-fed Prmt1 βKO mice had undergone some changes in their differentiated states (Fig. 1I–M). Immunofluorescence staining of mature β-cell markers further confirmed the loss of mature β-cell identity in HFD-fed Prmt1 βKO mice; such cells showed loss of MAFA and SLC2A2, cytoplasmic localization of NKX6.1, and blockage of HFD-induced FOXO1 nuclear translocation (4,37,38) (Fig. 1N and O). These phenotypes of HFD-fed Prmt1 βKO mice suggest that PRMT1 plays an essential role in maintaining the mature β-cell identity.

Although Prmt1 βKO mice presented the features of loss of β-cell identity, these phenotypes were weak. This could reflect the presence of metabolic compensation, which often comes into play in genetic KO models. To minimize the involvement of any compensatory mechanism and further confirm the role of PRMT1 in mature β-cells, we generated an inducible β-cell–specific Prmt1 KO mouse model by crossing Prmt1^fl/fl mice with Pdx1 promoter-driven Cre^ERT2 (Pdx1-Cre^ERT2) mice (hereafter called Prmt1 βiKO). Prmt1 KO was induced in adult β-cells by intraperitoneally injecting the mice with tamoxifen (75 mg/kg) five times over 2 weeks, beginning at 6 weeks of age (Supplementary Fig. 7A). At 8 weeks of age, these mice exhibited deletion of PRMT1 in β-cells and subsequent removal of H4R3me2a, but maintained normoglycemia (Fig. 2A and Supplementary Fig. 7B and C). At 12 weeks of age, Prmt1 βiKO mice developed glucose intolerance and exhibited elevated random glucose levels due to impaired GSIS (Supplementary Fig. 7C–E). Thus, acute loss of PRMT1 in adult β-cells is sufficient to induce the loss of mature β-cell function.

Further immunofluorescence staining revealed more severe defects in the β-cells of Prmt1 βiKO mice (Fig. 2A). At 8 weeks of age, insulin expression was robustly reduced in the β-cells of Prmt1 βiKO mice, and substantial numbers of INS⁻/PDX1⁺, INS⁻/NKX6.1⁺, and INS⁻/MAFA⁺ cells were observed in the islets. At 12 weeks of age, the insulin signals were slightly recovered in Prmt1 βiKO mice, but their intensities were still low, and most of the β-cells had lost MAFA and SLC2A2. A lineage-tracing analysis showed that a number of β-cells lost their identity and became INS⁻ cells (∼30%), urocortin 3 (UCN3)⁻ cells (∼33%), polyhormonal cells, or other endocrine cells (GCG, ∼1.5%; SST, ∼3.5%; and PPY, ∼0.24%) after induction of Prmt1 KO (Fig. 2B–G and Supplementary Fig. 8A and B). Electron microscopic analyses showed similar ultrastructural changes in Prmt1 βiKO and Prmt1 βKO mice (Fig. 1F and Supplementary Fig. 8C).

Furthermore, HFD aggravated the metabolic phenotypes of Prmt1 βiKO mice (Supplementary Fig. 9A–D). Random blood glucose levels were continuously elevated, and glucose intolerance became more severe in HFD-fed Prmt1 βiKO mice, but the insulin sensitivity and β-cell mass were comparable to those of the WT mice (Supplementary Fig. 9C–F). Consistent with these findings, β-cells lost their identity in HFD-fed Prmt1 βiKO mice: loss of insulin (∼30%) and expression of other hormones (GCG, ∼1.7%; SST, ∼3.8%; and PPY, ∼0.37%), presence of INS⁻/PDX1⁺ cells, cytoplasmic localization of NKX6.1, and loss of MAFA (Fig. 3A–F). In addition, the HFD-induced nuclear translocation of FOXO1 was blocked in Prmt1 βiKO mice (Fig. 3G). These data indicate that β-cells require PRMT1 to maintain their mature identity and that the loss of PRMT1 leads to the aberrant reprogramming of β-cells to express other hormones.

In Prmt1 βiKO mice, β-cells lost their identity soon after Prmt1 was ablated and before the glucose homeostasis deteriorated. To examine the molecular mechanism underlying the observed loss of mature β-cell identity, we explored the global gene expression pattern in islets of Prmt1 βiKO mice at two different stages: the early stage at 8 weeks of age and the late stage at 12 weeks of age (Fig. 4A). RNA-seq analysis revealed robust changes in gene expression at both stages (Supplementary Fig. 10A and B). In particular, genes involved in cellular energy production processes, such as oxidation reduction and the electron transport chain, were commonly downregulated at both stages (Fig. 4B). Our stage-specific gene expression analysis showed that the gene expression patterns differed between the early and late stages (Fig. 4C and Supplementary Table 2). The mature β-cell genes showed various changes at the early stage, whereas most of these genes were downregulated at the late stage. In particular, the expression levels of Ins1, Ins2, Ucn3, and Pdx1 were decreased in Prmt1 βiKO mice at both stages, whereas those of Mafa and Slc2a2 were downregulated only at the late stage. These results were further confirmed by qRT-PCR analysis (Fig. 4D–I). The notable novel features in the β-cells of Prmt1 βiKO mice included the robust downregulation of Ins1 at the early stage and its recovery at the late stage and the downregulations of Ins2 and Ucn3 at the early stage (Fig. 4D–F). These findings correlated with our immunofluorescence staining observations in Prmt1 βiKO mice (Fig. 2B). Meanwhile, RNA-seq and qRT-PCR analyses revealed that most of the genes related to oxidative phosphorylation (OXPHOS) were downregulated in Prmt1 βiKO mice at both stages, suggesting that mitochondrial dysfunction could be a feature of the loss of mature β-cell identity (Fig. 4C and Supplementary Fig. 10C). Stage-specific gene ontology and pathway analyses revealed that most of the genes downregulated at the early stage were linked with the electron transport chain and maturity-onset diabetes of the young, whereas genes related to GSIS function (e.g., those related to vesicle transport and intracellular protein trafficking) were downregulated at the late stage (Fig. 4J and K). Taken together, these data indicate that PRMT1 is required to maintain the transcriptional program of mature β-cells.

Cell type–specific chromatin state is essential for maintaining the transcriptional program and identity of a differentiated cell (39). Given the extensive gene expression changes in Prmt1 βiKO islets and the association of H4R3me2a with CTCF and β-cell TFs in MIN6 cells, we speculated that PRMT1-dependent H4R3me2a could regulate gene transcription through the actions on the chromatin status of mature β-cells. To test our hypothesis, we performed ATAC-seq with β-cells purified from Prmt1 βiKO and WT mice at 8 weeks of age (Fig. 5A). Unlike MIN6 cells that showed clear and distinctive ATAC-seq peaks, ATAC-seq with purified β-cells showed some background noise in the peaks. Therefore, we analyzed ATAC-seq data from WT and Prmt1-null β-cells along with ATAC-seq data from MIN6 cells and identified 5,044 peaks corresponding to the PRMT1-dependent open chromatin regions in mature β-cells (Fig. 5B). The average ChIP-seq peak intensities of H4R3me2a and H3K27ac (active enhancer marks) were higher in PRMT1-dependent open chromatin regions compared with PRMT1-independent open chromatin regions (Fig. 5C and D). However, the average ChIP-seq peak intensity of H3K4me1 (a poised enhancer mark) was similar in PRMT1-dependent and -independent open chromatin regions (Fig. 5E). These data indicate that H4R3me2a is responsible for the PRMT1-dependent chromatin openings in mature β-cells. The PRMT1-dependent open chromatin regions were also correlated with the binding sites of CTCF and β-cell TFs, including NKX6.1, NKX2.2, MAFA, NEUROD1, and PDX1 (Fig. 5F). Further motif analysis showed that the DNA-binding motifs of CTCF and β-cell TFs were highly associated with PRMT1-dependent open chromatin regions (Fig. 5G and Supplementary Fig. 11). Intriguingly, genes near PRMT1-dependent open chromatin regions were significantly associated with mature β-cell function and identity (Fig. 5H and I). These data indicate that PRMT1-dependent H4R3me2a is needed to maintain the unique chromatin architecture of mature β-cells and that the loss of mature β-cell identity in Prmt1 βiKO mice is associated with widespread alterations of the chromatin landscape.

To delineate how PRMT1-dependent chromatin openings relate to the transcriptional changes observed in Prmt1-null β-cells at the early stage, we closely examined the regulatory regions of genes downregulated in the islets of Prmt1 βiKO mice at 8 weeks of age in parallel with RNA-seq and ChIP-seq data of β-cell TFs. Indeed, the PRMT1-dependent open chromatin regions included multiple promoter or enhancer regions of β-cell and OXPHOS genes that were downregulated in the islets of Prmt1 βiKO mice. A comparative analysis of data from RNA-seq and ATAC-seq showed a correlation between the gene expression changes of β-cell genes and OXPHOS genes and the chromatin accessibility in the promoters of these genes (Fig. 5J and K and Supplementary Table 3). Moreover, the average ATAC-seq peak intensities of Prmt1-null β-cells were decreased in the promoters of β-cell genes and OXPHOS genes, indicating that the decreased expression of these genes at the early stage could be attributed to the loss of PRMT1-dependent H4R3me2a (Supplementary Fig. 12A and B).

We also identified PRMT1-dependent open chromatin regions at 12 kb upstream of the transcription start site (TSS) for the Ins1 gene and at 40 kb (E1) and 60 kb (E2) upstream of the TSS for the Ucn3 gene (Fig. 5L and N). These regions contained binding sites for β-cell TFs, including PDX1, MAFA, NKX2.2, and NKX6.1. Although the expression levels of Ins1 and Ucn3 were robustly and rapidly downregulated in Prmt1 βiKO mice at the early stage, the ATAC-seq peaks at the promoters of both genes were not significantly reduced at this point. Instead, 3C-PCR experiments showed the long-range interactions between the upstream enhancer elements and the promoters of the Ins1 and Ucn3 genes, indicating that the loss of chromatin accessibility for the β-cell TFs at the upstream enhancer elements had reduced the promoter activities of these genes (Fig. 5M and O). We also identified a PRMT1-dependent open chromatin region at 5 kb upstream of the TSS for the Pdx1 gene (Fig. 5P); this region, which is called area IV, was recently reported to play an essential role in β-cell maturation during the weaning period (40). PDX1 has also been shown to directly regulate the gene expression of numerous mitochondrial genes that were downregulated in Prmt1-null β-cells (41–44). These data suggest that PRMT1-dependent H4R3me2a plays a critical role in maintaining the unique chromatin architecture of mature β-cells and that the alteration of this chromatin architecture can result in the loss of mature β-cell identity.

In an effort to test the possible implication of PRMT1-dependent H4R3me2a in human diabetes, we performed sequence-alignment analysis of PRMT1-dependent open chromatin regions in the human genome and searched for conserved regulatory elements in these regions. The E2 element of the mouse Ucn3 gene and area IV of the mouse Pdx1 gene were highly conserved in the human UCN3 and PDX1 genes, which are found at similar genomic locations (Fig. 6A and B). A type 2 diabetes–associated locus was found near area IV of PDX1, and mice lacking endogenous area IV showed the impairment of β-cell maturation (40,45). We also found a highly conserved PRMT1-dependent open chromatin region in the human SLC30A8 gene, which is strongly associated with type 2 diabetes (45–48) (Fig. 6C). These findings prompted us to examine the association of human orthologous sequences of PRMT1-dependent open chromatin regions with diabetes genome-wide association study single nucleotide polymorphisms. Interestingly, the human diabetes-associated loci were more closely localized with PRMT1-dependent open chromatin regions than with PRMT1-independent open chromatin regions (Fig. 6D). This suggests that there may be a link between PRMT1-dependent H4R3me2a and the susceptibility to type 2 diabetes in humans.

Epigenetic regulation is crucial for β-cell maturation and the maintenance of mature β-cell identity (13–16). As one of the major mechanisms of epigenetic regulation, histone arginine methylation plays important roles in transcriptional regulation (17,18). However, its role in β-cells has not yet been explored. Here, we demonstrate a novel function of PRMT1-dependent H4R3me2a in maintaining mature β-cell identity. Both Prmt1 βKO and Prmt1 βiKO mice developed diabetes, which was aggravated by HFD-induced metabolic stress (Supplementary Fig. 13). Deletion of Prmt1 in adult β-cells resulted in the immediate loss of H4R3me2a, which induced robust changes in the transcriptions of genes necessary for the maintenance of mature β-cell function and identity. PRMT1-dependent H4R3me2a worked as an active histone code that increased chromatin accessibility at the binding sites for CTCF and β-cell TFs, including NKX6.1, MAFA, PDX1, and NEUROD1 (Fig. 6E). Furthermore, PRMT1-dependent open chromatin regions appear to be associated with genes that have been associated with diabetes susceptibility in humans.

Genome-wide association studies have indicated that most diabetes-susceptibility genes are related to β-cells, and thus, β-cell failure is thought to be an essential feature of diabetes (49,50). The dysfunction of β-cells occurs long before hyperglycemia develops in humans (51), suggesting that β-cell dysfunction is an early feature of diabetic β-cell failure. However, due to the lack of an appropriate animal model, researchers have been limited in their ability to study how β-cell dysfunction begins in response to metabolic stress and how β-cell failure progresses before the apoptosis or dedifferentiation of β-cells is observed. In this regard, Prmt1 βiKO mice provide the following useful features: initial events of loss of β-cell identity, which has not previously been described in an animal model; loss of INS and UCN3 prior to the loss of other mature β-cell TFs; decreased expression of β-cell genes; decreased expression of mitochondrial genes; and subsequent mitochondrial dysfunction. Despite these extensive changes of gene expression in the β-cells of Prmt1 βiKO mice, their metabolic phenotype was unexpectedly mild and became severe upon HFD feeding. This discrepancy prompted us to propose the following explanations: 1) since β-cells are highly dedicated to insulin production and secretion, even though the β-cells of Prmt1 βiKO mice are not fully functional, they can maintain glycemic control as long as mice are insulin sensitive; and 2) there may be compensatory mechanisms to maintain glycemic control in Prmt1 βiKO mice. The restoration of Ins1 expression in the late stage of Prmt1 βiKO mice (12 weeks of age) supports the existence of these compensatory mechanisms.

Since the phenotype of Prmt1 βiKO mice resembles the natural history of type 2 diabetes, this mouse model may be useful for studying how β-cells behave in response to metabolic stress during the development of type 2 diabetes. Prmt1 βiKO mice showed the following features in the progression of β-cell failure: loss of β-cell identity, aberrant expression of β-cell TFs, and cell type change of β-cells to other endocrine cells. We speculate that Prmt1 deletion resulted in the loss of H4R3me2a and that this causes β-cells to lose their cell-specific chromatin architecture and gene expression program and thereby lose their identity. These β-cells that lose their identity then undergo different changes based on their genetic heterogeneity in response to metabolic stress. However, further study will be needed to elucidate the precise mechanisms underlying the severe β-cell phenotypes of HFD-fed Prmt1 βKO and Prmt1 βiKO mice. The chromatin changes driven by HFD feeding together with the losses of H4R3me2a and arginine methylation in nonhistone substrates of PRMT1 may have affected these phenotypes. FOXO1 and HNF4α are PRMT1’s nonhistone targets that are also known to play roles in β-cell function and identity (4,52–54). However, the phenotypes of β-cell–specific Foxo1 or Hnf4α KO mice differed from those of Prmt1 βKO mice (4,54). In this regard, we think that the phenotypes of Prmt1 βKO and Prmt1 βiKO mice may be largely attributed to the loss of H4R3me2a in β-cells.

Here, we provide novel insight into the importance of epigenetic control of PRMT1-dependent H4R3me2a in maintaining mature β-cell identity. Taken together with the associations seen among CTCF, β-cell TFs, and H4R3me2a, our work reveals previously unknown functions of PRMT1-dependent open chromatin regions that govern mature β-cell identity. Thus, our phenotypic, transcriptomic, and epigenomic analyses of stage-specific Prmt1 KO in β-cells provide a new mechanistic insight into the regulation of mature β-cell identity.

Acknowledgments. The authors thank Hee-Saeng Jung, Jueun Kim, and Hanna Jung (Korea Advanced Institute of Science and Technology) for technical support, Dr. Dahee Choi (Korea University) and Hyun Jung Hong (Chungnam National University) for technical advice and support, Drs. Yong-Ho Ahn and Joo-Man Park (Yonsei University) for help with electron microscopic analysis, Jeong-Hwan Kim (Korea Research Institute of Bioscience and Biotechnology) for technical assistance with next-generation sequencing library preparation, Drs. Kyong Soo Park (Seoul National University), Kun-Ho Yoon (Catholic University of Korea), Soo Heon Kwak (Seoul National University), and Kyoung-Jae Won (University of Copenhagen) for helpful discussions, and Drs. Mark O. Huising (University of California, Davis) and Paul E. Sawchenko (Salk Institute) for the gift of the UCN3 antibody.

Funding. This work was supported by grants from the National Research Foundation funded by the Ministry of Science, ICT and Future Planning, Republic of Korea (grants NRF-2013M3A9D5072550 to J.K.S., NRF-2017M3C9A5028693 to M.K., and NRF-2014M3A9D5A01073546, NRF-2018R1A2A3074646, and NRF-2015M3A9B3028218 to Hai. Kim), the Korea Research Institute of Bioscience and Biotechnology Research Initiative (to M.K.), and the Korea Advanced Institute of Science and Technology Institute for the BioCentury (grant N10180027 to Hai. Kim).

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

Author Contributions. Hyu. Kim, B.-H.Y., C.-M.O., Jo. Lee, M.K., and Hai. Kim generated the hypothesis, designed the experiments, and analyzed the results. Hyu. Kim, C.-M.O., Jo. Lee, K.L., H.S., M.-Y.K., Hye. Kim, Y.K.K., and J.K.S. performed the animal experiments. Hyu. Kim, C.-M.O., E.K., E.-H.S., H.H., H.-J.K., Ju. Lee, J.M.S., S.-H.K., S.K., and M.S. performed the cell and in vitro experiments. Hyu. Kim, B.-H.Y., C.-M.O., K.Y., Y.S.J., M.K., and Hai. Kim analyzed the next-generation sequencing data. Hyu. Kim, B.-H.Y., C.-M.O., Jo. Lee, M.K., and Hai. Kim wrote the manuscript. M.K. and Hai. Kim supervised the research. M.K. and Hai. Kim are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Drivers of Type 2 Diabetes: From Genes to Environment (S1) conference of the Keystone Symposia, Seoul, Republic of Korea, 7–11 October 2018, and at the Third Joint EASD Islet Study Group and Beta Cell Workshop, Oxford, U.K., 1–3 April 2019.