Nardilysin Is Required for Maintaining Pancreatic β-Cell Function

Type 2 diabetes (T2D) is a common metabolic disorder that afflicts more than 300 million people globally (1). It is characterized by impaired insulin secretion from pancreatic β-cells and insulin resistance of the target tissues, such as liver, adipose tissue, and muscle. Although both factors play important roles in the pathogenesis of the disease, β-cell dysfunction, mainly manifested by deteriorated glucose-stimulated insulin secretion (GSIS), seems to be predominant for the transition from simple obesity to T2D (2). The observation that impaired GSIS in islets from patients with T2D cannot be accounted for by reduced insulin content also supports this concept (1). Several biological processes, including endoplasmic reticulum stress, inflammation, and oxidative stress, are suggested to impair GSIS in T2D (3,4).

The pancreas comprises an exocrine compartment, consisting of acinar and ductal cells, and an endocrine compartment, consisting of α, β, δ, ε, and pancreatic polypeptide cells. These different types of endocrine cells express and secrete the hormones glucagon, insulin, somatostatin, ghrelin, and pancreatic polypeptide, respectively. Sequential expression of pancreatic transcription factors determines cell fate and pancreatic development. Among them, several transcriptional factors enriched in β-cells, such as Pdx-1, Nkx6.1, and MafA, play essential roles in the maintenance of β-cell identity and function. β-Cell-specific ablation of Pdx-1 results in hyperglycemia with fewer insulin-positive cells and more glucagon-positive cells (5,6), whereas conditional Nkx6.1 inactivation in adult β-cells causes reduced insulin production (7). In contrast to the earlier expression of Pdx-1 and Nkx6.1 in endocrine progenitor cells, MafA is first produced at embryonic day 13.5 and is observed only in insulin-positive mature β-cells. MafA binds to an insulin promoter, like Pdx-1, and regulates β-cell-selective insulin transcription (8). Mafa-deficient mice consistently develop diabetes because of impaired GSIS without a defect of islet cell development (9). Moreover, β-cell-specific ablation of pan-endocrine transcription factors such as NeuroD1 and Islet-1 demonstrated their critical roles in β-cell maturation and function (10,11). In particular, postnatal ablation of Islet-1 in β-cells resulted in impaired GSIS without significantly reducing β-cell mass (11). Collectively, these findings demonstrated that transcription factors involved in the determination of endocrine or β-cell identity also play essential roles in β-cell functions.

A regulatory network of β-cell-enriched transcriptional factors (e.g., Pdx-1, NeuroD1, NKx6.1, and MafA) critically controls GSIS (5–10). These transcriptional factors also connect oxidative stress with impaired GSIS in diabetic mouse models and human T2D (4). Among them, several lines of evidence have suggested that the Islet-1–MafA pathway critically controls GSIS in adults. First, ablation of these genes in β-cells results in impaired GSIS without a severe defect in β-cell development, as described above (9,11). Second, MafA is a direct target gene of Islet-1 (11,12). Third, MafA directly regulates several genes critical for GSIS, such as Ins1, Ins2, and Glut2 (also known as Slc2a2).

In this study, we demonstrate a novel regulatory mechanism of GSIS by nardilysin (N-arginine dibasic convertase; Nrd1 and NRDc) through the Islet-1–MafA pathway. NRDc is a zinc peptidase of the peptidase M16 domain (M16) family that selectively cleaves dibasic sites (13,14). We rediscovered NRDc as a specific receptor for heparin-binding epidermal growth factor–like growth factor (HB-EGF) (15), and our subsequent studies showed multiple functions of NRDc, which depend on its cellular localization. In the extracellular space, NRDc enhances ectodomain shedding of HB-EGF and other membrane proteins, such as tumor necrosis factor-α, amyloid precursor protein, and neuregulin-1 (16–20). We also revealed that NRDc in the nucleus works as a transcriptional coregulator through the modulation of NCoR/SMRT corepressor or PGC-1α coactivator function (21,22). We show here that NRDc-deficient (Nrd1^−/−) mice and pancreatic β-cell-specific NRDc-deficient (Nrd1^delβ) mice show glucose intolerance and severely impaired GSIS. Islets isolated from Nrd1^−/− and Nrd1^delβ mice also show lower insulin production and GSIS than wild-type islets. Notably, NRDc and Islet-1 colocalize in the MafA enhancer, where NRDc regulates the recruitment of Islet-1 and MafA transcription, indicating the essential role of NRDc in GSIS through Islet-1–MafA regulation.

Nrd1^−/− mice (accession no. CDB0466K; http://www.clst.riken.jp/arg/mutant%20mice%20list.html) were generated as described previously (19) and backcrossed to the ICR background (>98%). Nrd1^flox/flox mice (accession no. CDB1019K; http://www.clst.riken.jp/arg/mutant%20mice%20list.html) were generated by gene targeting in TT2 (23) embryonic stem (ES) cells (http://www.clst.riken.jp/arg/Methods.html). The targeting vector was designed to insert loxP sites upstream and downstream of exon 1 of Nrd1 (Supplementary Fig. 1A). Successful homologous recombination in TT2 ES cells was confirmed by PCR (Supplementary Fig. 1A) and Southern blotting. Resulting mutant mice were genotyped with tail DNA by PCR (Supplementary Fig. 1B) and crossed with CAG-FLPe transgenic mice to remove the neomycin selection cassette surrounded by Flp recombinase target sites, then backcrossed to the C57BL/6J background (>98%). β-Cell-specific NRDc-deficient mice were generated by crossing Nrd1^flox/flox mice with rat insulin II promoter-Cre transgenic (RIP-Cre) mice (24). Male mice were used unless otherwise indicated. All animal experiments were performed according to procedures approved by the Institute of Laboratory Animals, Kyoto University. Mice were maintained on a diet of standard rodent chow in environmentally controlled rooms.

For glucose tolerance tests, mice were fasted for 16 h and intraperitoneally injected with 2 g/kg body weight of d-glucose. For insulin tolerance tests, mice were fasted for 5 h and intraperitoneally injected with insulin (Humulin R; Eli Lilly). To determine fasting plasma glucagon concentrations, mice were fasted for 16 h. To determine pancreatic insulin content and the proinsulin-to-insulin ratio, whole pancreata were homogenized in 4 mL of a 2% HCl/75% ethanol solution. After neutralization, samples were diluted 1:1,000 in PBS before measuring insulin and proinsulin. Glucose was measured using a glucometer. Insulin (MS303; Morinaga Institute of Biological Science, Yokohama, Japan) and proinsulin (AKMPI-111; Shibayagi, Shibukawa-City, Gunma, Japan) were measured by ELISA. Glucagon was measured by enzyme immunosorbant assay (YK090; Yanaihara Institute, Fujinomiya-City, Shizuoka, Japan).

MIN6 and INS832/13 cells were gifts from Dr. Miyazaki and Dr. Newgard, respectively. Nrd1^−/− mouse embryonic fibroblasts (MEFs) were prepared as described previously (19). MIN6 cells and Nrd1^−/− MEFs were grown in DMEM (4.5 g L⁻¹ glucose) supplemented with 10% FBS and antibiotics. INS832/13 cells were cultured as previously described (25). GM6001 (26) was purchased from Calbiochem (CA). For overexpression or gene knockdown of Nrd1 or Mafa in MIN6 and INS832/13, cells were infected with lentiviral vectors expressing Nrd1, Mafa, or BLOCK-iT miR RNAi (Life Technologies) targeting Nrd1 or Mafa, respectively.

Pancreatic islets were isolated, and insulin release from the islets was measured as previously described (27). The amount of immunoreactive insulin was determined by radioimmunoassay (Aloka Accuflex γ 7000; Hitachi, Tokyo, Japan).

Insulin secretion from INS832/13 cells (25) was determined as previously described, with some modifications (28). In brief, cells cultured on six-well plates were washed with Krebs-Ringer bicarbonate HEPES buffer (KRBH; 140 mmol/L NaCl, 3.6 mmol/L KCl, 0.5 mmol/L MgSO₄, 0.5 mmol/L NaH₂PO₄, 1.5 mmol/L CaCl₂, 2 mmol/L NaHCO₃, and 10 mmol/L HEPES; pH 7.4) with 0.1% BSA and 2.8 mmol/L glucose, preincubated at 37°C for 2 h in KRBH with 2.8 mmol/L glucose, and then incubated at 37°C for 2 h in KRBH with 2.8 and 15 mmol/L glucose. Insulin was measured by ELISA (MS303; Morinaga Institute of Biological Science).

Chromatin was prepared from MIN6 cells (29) as previously described (21). ChIP assays were performed using ChIP-IT Express Chromatin Immunoprecipitation Kits (Active Motif). Antibodies against NRDc (mouse monoclonal antibody 2E6), PDX-1 (sc-14664×; Santa Cruz Biotechnology, Inc.), Islet-1 (ab20670; Abcam), Neuro D1 (sc-1084×; Santa Cruz Biotechnology, Inc.), or control IgG (Santa Cruz Biotechnology, Inc.) were used. For the re-ChIP assay, Re-ChIP-IT Express Magnetic Chromatin Re-Immunoprecipitation Kits (Active Motif) were used according to the manufacturer’s protocol. The primers used are listed in Supplementary Table 1.

For the luciferase reporter assays, PicaGene Promoter Vector 2 (PGV-P2; Toyo Ink, Tokyo, Japan) with or without the insertion of MafA R3 (−8,118 to −7,750 relative to the transcriptional start site of Mafa) was transiently transfected into MIN6 cells. Luciferase activity was quantified 48 h after transfection using the Dual-Luciferase Reporter Assay Kit (Promega) according to the manufacturer’s protocol. pRL-TK (Promega) was cotransfected with PGV-P2 to control for transfection efficiency.

Results are expressed as mean ± SEM. The Student t test was used for comparisons between two groups. ANOVA was used for comparisons among multiple groups, with a Tukey-Kramer post hoc test.

We previously reported that Nrd1^−/− mice show a lean phenotype, which has been attributed to increased energy expenditure as a result of enhanced thermogenesis in brown adipose tissue and hyperactivity (21). To address whether NRDc is involved in glucose metabolism, we performed the glucose tolerance test and the insulin tolerance test in Nrd1^−/− mice. Although the fasting blood glucose and serum insulin concentrations remained the same as in the wild-type (Nrd1^+/+) mice, Nrd1^−/− mice showed elevated blood glucose concentrations 60 and 120 min after glucose challenge (Fig. 1A). Importantly, the serum insulin concentration 30 min after glucose injection was strikingly low in the mutant mice (Fig. 1B). At the same time, the insulin tolerance test showed increased insulin sensitivity in Nrd1^−/− mice (Fig. 1C). On the basis of these observations, the glucose metabolism of the Nrd1^−/− mice is characterized by mild glucose intolerance and significantly reduced GSIS, accompanied by insulin hypersensitivity.

Histologically, islets of Nrd1^−/− mice showed no obvious anomalies in their size and structure; there was a predominance of insulin-positive β-cells in the core, which were surrounded by glucagon-positive α-cells (Fig. 1D). Quantitative analysis confirmed no significant difference in β-cell mass and α-cell–to–β-cell ratio between Nrd1^+/+ and Nrd1^−/− mice (Fig. 1E and F). Thus, impaired GSIS in Nrd1^−/− mice is not the result of a gross developmental defect in the pancreas, but disturbed islet function was suspected.

To evaluate islet function directly, we isolated islets from Nrd1^+/+ and Nrd1^−/− mice and performed in vitro GSIS assays. Consistent with the in vivo results, Nrd1^−/− islets showed severely reduced GSIS (Fig. 1G) compared with Nrd1^+/+ islets. Because KCl-induced insulin secretion was preserved in Nrd1^−/− islets (Fig. 1G), NRDc may regulate GSIS through the processes before membrane depolarization. Insulin content was modestly but significantly reduced in Nrd1^−/− islets (Fig. 1H). Since NRDc has endopeptidase activity, we examined pancreatic proinsulin-to-insulin ratios but found no significant difference between Nrd1^+/+ and Nrd1^−/− islets (Fig. 1I).

The analysis of isolated islets strongly suggested that NRDc is primarily involved in GSIS. To confirm the β-cell-specific role of NRDc, we generated Nrd1^flox/flox mice by inserting two loxP sites around exon 1 of the mouse Nrd1 gene (Supplementary Fig. 1A and B) and crossed them with RIP-Cre transgenic mice (24) to produce β-cell-specific NRDc-deficient (Nrd1^delβ) mice. Nrd1^delβ mice were born at the expected Mendelian frequency and showed no overt phenotypes, including body weight (Supplementary Fig. 1C). We confirmed the specific decrease of the Nrd1 mRNA level in pancreatic islets, but not in other organs, including the hypothalamus, heart, white adipose tissue, liver, and skeletal muscle (Supplementary Fig. 1D). Because glucose intolerance and decreased GSIS of RIP-Cre mice has been reported (30), we set two controls, Nrd1^flox/flox mice and RIP-Cre (Nrd1^wt/wt;RIP-Cre) mice, for comparison with Nrd1^delβ mice. While the global knockout mice showed normal fasting blood glucose (Fig. 1A), Nrd1^delβ mice demonstrated markedly elevated fasting blood glucose compared with both control mice (Fig. 2A). In glucose tolerance tests, RIP-Cre mice showed glucose intolerance and reduced insulin secretion, as previously reported (30), whereas Nrd1^delβ mice displayed more severe glucose intolerance and more reduced insulin secretion compared with RIP-Cre and Nrd1^flox/flox mice (Fig. 2B and C). Insulin sensitivity was similar between Nrd1^flox/flox, RIP-Cre, and Nrd1^delβ mice (Supplementary Fig. 1E and F). These results indicated that the β-cell-specific NRDc depletion is sufficient to cause glucose intolerance and impaired GSIS.

In sharp contrast to the normal appearance of Nrd1^−/− islets, islets in Nrd1^delβ mice showed an impaired structure (Fig. 2D). Glucagon-positive α-cells increased in number, and their locations were not restricted to the periphery of the islet but intermingled with β-cells at the core. When quantified, the α-cell–to–β-cell ratio was significantly increased in Nrd1^delβ islets (Fig. 2E). In agreement with these results, fasting plasma glucagon and the mRNA level of islet Gcg in Nrd1^delβ mice were significantly higher than those of control mice (Fig. 2F and G). β-Cell mass in Nrd1^delβ mice was significantly smaller than that in Nrd1^flox/flox mice and tended to be smaller, though without a statistically significant difference, than RIP-Cre control mice (Fig. 2H). The ratio of large islets was significantly decreased, whereas that of small islets was increased in Nrd1^delβ mice compared with either Nrd1^flox/flox mice or RIP-Cre mice (Fig. 2I).

We also performed a lineage tracing study of insulin-expressing cells in Nrd1^delβ mice. By crossing Nrd1^flox/flox mice with RIP-Cre and R26ECFP mice (31), we generated Nrd1^flox/flox;RIP-Cre;R26ECFP mice, in which the simultaneous deletion of NRDc and expression of enhanced cyan fluorescent protein (ECFP) occurred specifically in β-cells, and compared them with control Nrd1^wt/wt;RIP-Cre;R26ECFP mice (Supplementary Fig. 2A). As shown in Supplementary Fig. 2B and C, several ECFP-positive cells lost insulin staining in Nrd1^flox/flox;RIP-Cre;R26ECFP mice. Furthermore, a small subset of glucagon-positive cells was clearly labeled with ECFP in Nrd1^flox/flox;RIP-Cre;R26ECFP islets (Supplementary Fig. 2D and E). When quantified, glucagon and ECFP double-positive cells were observed in 3.28% (n = 853) of the total glucagon-positive cells in Nrd1^flox/flox;RIP-Cre;R26ECFP islets but in 0.85% (n = 233) in control islets. In addition, genes transiently expressed in endocrine progenitors (e.g., Sox9, Ngn3, and Mafb) and Aldh1a3, recently reported as a marker of β-cell dedifferentiation, were significantly or tended to be upregulated in Nrd1^delβ islets (Supplementary Fig. 3A–C). These results suggested that some β-cells may lose their identity and a subset of these β-cells may be converted to glucagon-positive α-like cells in Nrd1^delβ islets.

Consistent with the results of the in vivo glucose tolerance test, islets isolated from Nrd1^delβ mice showed markedly reduced GSIS. Compared with islets isolated from Nrd1^flox/flox and RIP-Cre control mice, both low- (2.8 mmol/L) and high-glucose–stimulated (25 mmol/L) insulin secretion (Fig. 3A) was significantly reduced in Nrd1^delβ islets. Furthermore, the insulin content of the freshly isolated islets (Fig. 3B) was significantly lower in Nrd1^delβ mice. To obtain mechanistic insights that explain the reduced insulin content and impaired GSIS of the Nrd1^delβ islets, we examined the mRNA levels of genes involved in GSIS and found that Glut2, Pcx, Ins1, Ins2, and Pcsk1 are significantly decreased in Nrd1^delβ islets compared with those in RIP-Cre islets (Fig. 3C). Reduction of Glut2, Pcx, Ins1, and Ins2 levels in Nrd1^delβ islets was also confirmed when compared with Nrd1^flox/flox islets. However, Pcsk1 was increased in Nrd1^delβ islets compared with that in Nrd1^flox/flox islets, since it was markedly increased in RIP-Cre islets (Supplementary Fig. 4A). We next investigated the expression of transcription factors in β-cells that potentially regulate Glut2, Pcx, Ins1 and 2, and Pcsk1. Intriguingly, Mafa was specifically downregulated in islets from Nrd1^delβ mice, whereas the expression of Pdx1 and others, including NKx6.1, NeuroD1, and Islet-1, was not significantly changed (Fig. 3D and Supplementary Fig. 4B). Histological findings confirmed the reduced protein expression of GLUT2, MafA, and insulin in Nrd1^delβ islets (Fig. 3E and F). These results suggested that NRDc in β-cells acts upstream of MafA.

Our loss-of-function experiments using Nrd1^delβ mice/islets revealed that NRDc is essential for proper islet function, especially for GSIS, via maintaining the expression of Mafa and its downstream genes such as Ins1, Ins2, Glut2, and Pcsk1. At the same time, islet structure was impaired in Nrd1^delβ mice. Considering a previous report describing that an abrogated structure itself disturbs islet function (32), we were prompted to perform loss-of-function or gain-of-function experiments in a β-cell line to confirm the cell-autonomous role of NRDc. INS832/13 cells are an INS-1-derived rat β-cell line that is strongly responsive to glucose in terms of insulin secretion (25). Similar to Nrd1^delβ islets, Nrd1-knocked down INS832/13 (Fig. 4A) showed decreased insulin secretion (Fig. 4B), insulin content (Fig. 4C), and mRNA levels of genes involved in insulin secretion (Fig. 4D). On the other hand, INS832/13 cells, in which NRDc was overexpressed by two- to threefold (Fig. 4E), showed augmented insulin secretion in their response to glucose (Fig. 4F) and increased insulin content at a steady state (Fig. 4G). While the expression levels of genes involved in glycolysis and ATP production were not changed, the mRNA levels of Ins1 and Ins2 were significantly upregulated by the overexpression of NRDc in INS832/13 cells (Fig. 4H). Notably, MafA expression was upregulated at the mRNA level (Fig. 4H) and the protein level in the nucleus (Fig. 4E). To ascertain whether NRDc regulates insulin expression through MafA, we performed gene knockdown of Mafa in NRDc-overexpressed INS832/13 cells. Mafa knockdown did not affect NRDc expression, whereas it cancelled the augmentation of Mafa expression by NRDc overexpression (Fig. 5A). As a result, Mafa knockdown almost fully cancelled the NRDc-mediated increase of Ins2 and partially cancelled the NRDc-mediated increase of Ins1 gene expression (Fig. 5B). Thus, NRDc regulates insulin production, at least in part, by controlling MafA expression.

Previous reports showed that MafA expression is regulated by roughly 10 kg base pairs (bp) 5′ of the transcription start site, which contain six highly conserved sequence domains, called regions 1–6 (33,34). Among them, region 3 (R3; −8,118/−7,750 bp) has been reported to be essential for the expression of MafA in β-cells both in vitro and in vivo (33,34). Several β-cell-enriched transcriptional factors, such as Pdx1, NeuroD1, NKx2.2, Islet-1, HNF1a, HNF1b, FoxA2, and NKx6.1, have their consensus binding motifs in R3 (35). Our ChIP assay using antimouse NRDc monoclonal antibody showed that NRDc is recruited to R3 of the MafA promoter in MIN6 cells, a mouse β-cell line (Fig. 6A). Furthermore, luciferase reporter assay showed that the gene knockdown of NRDc reduced the enhancer activity of R3 in MIN6 cells (Fig. 6B). Because NRDc has no canonical DNA-binding motifs, we hypothesized that NRDc forms a complex with other transcription factors and exists on R3. To this end, we performed coprecipitation assays using cells overexpressing NRDc and Pdx1, NeuroD1, NKx2.2, Islet-1, HNF1a, HNF1b, FoxA2, or NKx6.1, and found that Islet-1 forms a complex with NRDc (Fig. 6C and data not shown).

Because ChIP analysis indicated that both NRDc (Fig. 6A) and Islet-1 (Supplementary Fig. 5D) exist on R3 of the MafA promoter, we performed sequential ChIP/re-ChIP analysis to determine the in vivo colocalization of NRDc and Islet-1. As shown in Fig. 6D, ChIP/re-ChIP analysis with anti-NRDc antibody and anti-Islet-1 antibody demonstrated that NRDc and Islet-1 are colocalized in R3 of the MafA promoter. To determine whether NRDc affects the recruitment of Islet-1 to R3, we performed ChIP analysis using MIN6 cells in which NRDc was knocked down (Fig. 6E and F). Consistent with results in Nrd1^delβ islets, gene knockdown of NRDc resulted in the decrease of MafA at the protein and mRNA levels (Fig. 6E and F). Intriguingly, gene knockdown of NRDc significantly reduced the recruitment of Islet-1, but not Pdx-1 and NeuroD1, to R3 (Fig. 6G). Of note, gene knockdown of NRDc did not change the nuclear expression level of Islet-1 (Fig. 6E). Given the critical roles of Islet-1 in MafA regulation and β-cell function, our results suggest that NRDc regulates β-cell function, at least in part, through controlling the recruitment of Islet-1 to R3 of the MafA promoter.

In addition to its transcriptional coregulator function in the nucleus, NRDc enhances ectodomain shedding of membrane proteins (e.g., HB-EGF, tumor necrosis factor-α) through the activation of ADAMs (A disintegrin and metalloproteinase) (16–20). By immunoblotting and immunostaining in pancreatic β-cell lines (Figs. 4A and E and 6E, and Supplementary Fig. 5A) and immunostaining in the islets of wild-type mice and humans (Supplementary Fig. 5B and C), we demonstrate the intra- and extranuclear localization of NRDc in pancreatic β-cells. Broad-spectrum ADAM and matrix metalloproteinase inhibitor GM6001 blocks the activity of extracellular NRDc from enhancing ectodomain shedding (26). As shown in Supplementary Fig. 5E, GM6001 had no effect on the augmentation of Mafa expression by NRDc overexpression, suggesting that NRDc in the extracellular space has little, if any, effect on the transcription of MafA.

One unique characteristic of NRDc among the M16 family of metalloendopeptidases is a highly acidic domain next to the catalytic domain (Fig. 7A), which is required for its binding to several proteins, including HB-EGF, but is dispensable for its metalloendopeptidase activity (36,37). We transfected the mutant form of NRDc lacking this acidic domain (dAcD) into Nrd1^−/− MEF cells and tested whether the mutant interacts with Islet-1. Coimmunoprecipitation assays revealed that the acidic domain is responsible for the interaction between NRDc and Islet-1. On the other hand, the enzymatic activity of NRDc was dispensable for its binding to Islet-1 (Fig. 7A and B).

Islet-1 is a member of the LIM-homeodomain protein family, which is characterized by two LIM domains arrayed in tandem at the N terminus, a DNA-binding homeodomain, and a LIM homeobox protein 3-binding domain (LBD) at the C terminus (Fig. 7C). To examine which domain of Islet-1 is required for the interaction with NRDc, we performed coprecipitation assays of NRDc and two deletion mutant forms of Islet-1 (Fig. 7C). The mutant Islet-1 lacking the N-terminal LIM domain (Islet-1^delLIM) showed markedly reduced coprecipitation with NRDc, whereas the mutant without the COOH-terminal LBD domain (Islet-1^delC) showed results similar to the wild type of Islet-1 (Fig. 7D). Collectively, these results indicated that the acidic domain of NRDc and the LIM domain of Islet-1 are responsible for their interaction.

To assess the functional significance of NRDc and Islet-1 interaction in pancreatic β-cells, we overexpressed the wild-type NRDc or the mutants of NRDc in INS832/13 cells and examined NRDc-mediated augmentation of Mafa gene expression. Consistent with the results of coprecipitation assays, mRNA levels of Mafa were augmented by the wild type of NRDc, but not by dAcD (Fig. 7E). Furthermore, induction of the Ins1 mRNA level by NRDc expression was significantly reduced by the coexpression of dAcD. On the other hand, Islet1-induced upregulation of Mafa and Ins1 mRNA levels were decreased by the expression of Islet-1^delLIM (Fig. 7F). These results suggest that the interaction of NRDc and Islet-1 is required for the proper transcriptional regulation of Mafa and Ins1.

Although the enzymatically inactive E>A mutant bound to Islet-1 as the wild type of NRDc did (Fig. 7B), transcriptional induction of Mafa was not recapitulated by the E>A mutant (Fig. 7E). These results indicate that not only physical association between NRDc and Islet-1 but also enzymatic activity of NRDc seem to be involved in Mafa, Ins1, and Ins2 regulation.

In this report, we present the first in vivo evidence that NRDc is essential for proper β-cell function, especially for GSIS. Nrd1^−/− mice develop glucose intolerance and severely impaired GSIS. Although the expression of NRDc is widespread in the whole body, the analysis of islets isolated from Nrd1^−/− mice and β-cell-specific Nrd1-deficient mice clearly indicates that NRDc in β-cells is critical for insulin production and secretion. Accumulating evidence supports the notion that several β-cell-enriched transcription factors, such as Pdx-1, Nkx6.1, and MafA, play essential roles in the maintenance of β-cell function. Among them, MafA was solely downregulated in islets from Nrd1^delβ mice. Furthermore, the Mafa mRNA level was increased by the overexpression of NRDc, whereas it was decreased by the gene knockdown of NRDc in β-cell lines, indicating that NRDc is a cell-autonomous regulator of MafA transcription. Importantly, NRDc overexpression in INS832/13 cells was accompanied by increased mRNA levels and the secretion of insulin, which was partially inhibited by the simultaneous gene knockdown of Mafa. These results demonstrated that NRDc regulates insulin production and GSIS through the control of MafA expression.

MafA is first produced at embryonic day 13.5 exclusively in insulin-positive cells, which is unusually late compared with other islet-enriched transcription factors (38). Moreover, Mafa mRNA levels show an age-dependent increase from neonates to adults, which is similar to the expression pattern of insulin and other putative target genes of MafA (39). This unique spatial and temporal expression pattern of MafA is regulated by its well-established 5′ flanking enhancer region, called R3 (−8,118/−7,750 bp) (34,40). We show here that NRDc binds to the MafA enhancer R3. We also demonstrated that the enhancer activity of MafA R3 is reduced by the gene knockdown of NRDc. In addition, we performed unbiased screening of β-cell-enriched transcriptional factors for binding to NRDc and found that Islet-1 associates with NRDc. Notably, ChIP/re-ChIP assay clearly demonstrated the association of NRDc and Islet-1 in the MafA enhancer R3. Furthermore, gene knockdown of NRDc was accompanied by a decrease of Islet-1 recruitment to R3. Given that MafA is a direct target of Islet-1, MafA regulation by NRDc is, at least in part, attributed to NRDc-mediated Islet-1 recruitment to the MafA enhancer.

We show that the acidic domain of NRDc and the LIM domain of Islet-1 are required not only for complex formation but also for the full enhancement of Mafa and Ins1 mRNA. Islet-1 is a member of the LIM-homeodomain protein family, characterized by two tandemly repeated LIM domains fused to a conserved homeodomain. The LIM domain consists of two tandemly repeated zinc fingers, structurally similar to GATA-type zinc fingers (41). However, unlike GATA-type zinc fingers, the LIM domain in mammalian LIM proteins does not seem to bind to DNA. Instead, the domain seems to be important for diverse protein–protein interactions (42). Two transcription factors, LIM domain-binding protein 1 (43) and NeuroD (44), were previously reported to bind Islet-1 via the LIM domain and activate the transcriptional activity of Islet-1. Meanwhile, the deletion of the LIM domains from Islet-1 was reported to enhance insulin promoter–driven luciferase activity in HEK-293 cells (44). As opposed to that result, the LIM domain–deleted mutant of Islet-1, overexpressed in INS832/13 cells, failed to elevate the mRNA levels of Mafa and Ins1. Although these findings are inconsistent, our results in β-cells are more likely to recapitulate the physiological regulation of insulin promoter activity. Some factors in β-cells, such as NeuroD and NRDc, may activate the transcription of insulin through interacting with the LIM domains.

Although both Nrd1^−/− and Nrd1^delβ mice show glucose intolerance and severely impaired GSIS, there are some differences between these two mouse lines. First, fasting blood glucose was normal in Nrd1^−/− mice, whereas it was elevated in Nrd1^delβ mice. Second, β-cell mass was unaltered in Nrd1^−/− mice but reduced in Nrd1^delβ mice. Third, the α-cell–to–β-cell ratio was unchanged in Nrd1^−/− mice but significantly increased in Nrd1^delβ mice. Nrd1^delβ mice also showed the dysregulated distribution patterning of glucagon-positive α-cells, a phenotype that is similar to that of Mafa-deficient mice (9). Consistently, Nrd1^delβ mice showed hyperglucagonemia. Considering that insulin sensitivity was enhanced in Nrd1^−/− mice, hyperglucagonemia and unaltered insulin sensitivity in Nrd1^delβ mice may account for their severe glucose intolerance. Insulin hypersensitivity in Nrd1^−/− mice may arise from an increase in energy expenditure caused by enhanced thermogenesis in brown adipose tissue and increased physical activity (21). Our ongoing studies using organ-specific (e.g., liver, adipose tissue, and skeletal muscle) NRDc-deficient mice will further clarify the role of NRDc in the regulation of insulin sensitivity.

Recent reports have indicated that several transcription factors, such as Pdx1 (6), Nkx6.1 (7), and FoxO1 (45), are critical for β-cells to maintain their differentiated states. Loss of FoxO1 in mature β-cells leads to the dedifferentiation of β-cells, characterized by upregulation of Ngn3, an endocrine progenitor cell marker. A subset of dedifferentiated cells was shown to adopt an α-cell fate (45). Importantly, possible dedifferentiation and the conversion of β-cells to α-cells have also been reported in Mafa-deficient mice (46). Our results of lineage tracing experiments demonstrated that a small portion of glucagon-positive cells in Nrd1^delβ islets originated from β-cells. Moreover, endocrine progenitor cell markers such as Sox9, Ngn3, Mafb, and Aldh1a3 (47) were upregulated in Nrd1^delβ islets. Together, these results suggest that dedifferentiation of β-cells and redifferentiation to the glucagon-expressing α-like cells might occur in a subset of β-cells in an islet environment where NRDc-positive α-cells and NRDc-negative β-cells are intermingled. As the β-to-α conversion in Nrd1^delβ islets was observed in only a limited number of cells (3.28% of the total glucagon-positive cells), reduced β-cell proliferation, a phenotype observed in MafA-depleted β-cells (48,49), may account for the increased α-cell–to–β-cell ratio in Nrd1^delβ islets. Meanwhile, Glut2-deficient mice also display the increased α-cell–to–β-cell ratio, although the mechanism is elusive (50). Because mRNA and protein expressions of GLUT2 were clearly reduced in Nrd1^delβ islets, NRDc might modulate the α-cell–to–β-cell ratio via the regulation of Glut2 in Nrd1^delβ islets. Interestingly, Glut2-deficient mice (50,51) and/or transgenic mice expressing the glucose sensor–dead mutant of GLUT2 (52) share several phenotypes with Nrd1^delβ mice: a reduction in pancreatic insulin content and basal and GSIS, an increased α-cell–to–β-cell ratio, and an increased ratio of small islets. Future studies using Glut2-rescued Nrd1^delβ mice might clarify the precise role of Glut2 in the regulation of β-cell function and proliferation by NRDc.

NRDc in the nucleus has been identified as a histone-binding protein, specifically recognizing dimethyl-H3K4 (22). The roles of NRDc in transcriptional regulation have been illustrated by the interaction of NRDc with various nuclear proteins such as HDAC3 in the NCoR/SMRT complex and PGC-1α (21,22). We demonstrate here that Islet-1 is a novel binding partner of NRDc in β-cells. Considering the context-dependent multiple binding partners and wide range of phenotypes of Nrd1^−/− mice (19,21,53), NRDc may function as a transcriptional coregulator in general. Since NRDc has its own peptidase activity, which seems to be involved in MafA regulation, further study using peptidase-dead mutant knockin mice will be required to clarify the link between the biological role and the protein function of NRDc.

Acknowledgments. The authors are grateful to H. Iwai, K. Shimada, and N. Nishimoto (Kyoto University) for technical assistance and to C.B. Newgard (Duke University), J. Miyazaki (Osaka University), and T. Masui and A. Sato (Kyoto University) for materials.

Funding. This study was supported by KAKENHI Grants-in-Aid (26293068, 26670139, 26116715, 15K19376, and 15K19513) and a research grant of the Project for Development of Innovative Research on Cancer Therapeutics (P-Direct) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. It was also supported by a Health Labour Sciences Research Grant from the Ministry of Health, Labour and Welfare of Japan (Comprehensive Research on Lifestyle-Related Diseases including Cardiovascular Diseases and Diabetes Mellitus), the Takeda Science Foundation, the Mitsui Sumitomo Insurance Welfare Foundation, the Kao Research Council for the Study of Healthcare Science, the Uehara Memorial Foundation, and Otsuka Pharmaceutical Co., Ltd. Sponsored Research Program.

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

Author Contributions. K.N. and E.N. designed experiments and wrote the manuscript. K.N., M.O., Y.H., S.S., J.S., P.-M.C., Y.Mo., and S.M. performed experiments and analyzed data. Y.S., K.I., K.S., N.H., and N.I. isolated islets and performed the insulin secretion assay. Y.Mu. and H.K. generated Nrd1^flox/flox mice. K.F., Y.K., and S.U. carried out the histological procedures and provided human samples. Y.K, N.I., T.Kit., and T.Kim. supervised the work. E.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.